Composite Structures 95 (2013) 453–463

Contents lists available at SciVerse ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Dynamic response under pedestrian load of a GFRP–SFRSCC hybrid footbridge

prototype: Experimental tests and numerical simulation
José A. Gonilha ⇑, João R. Correia, Fernando A. Branco
Department of Civil Engineering, Architecture and Georesources, Instituto Superior Técnico/ICIST, Technical University of Lisbon, Av., Rovisco Pais, 1049-001 Lisboa, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: The use of glass fibre reinforced polymer (GFRP) pultruded profiles in civil engineering structures has
Available online 31 July 2012 known considerable growth in recent years due to their high strength, low self-weight and corrosion
resistance when compared to traditional materials. However, the high deformability, the susceptibility
Keywords: to instability phenomena and the brittle failure are delaying the widespread use of this structural mate-
Dynamic behaviour rial. GFRP–concrete hybrid systems have been proposed as an alternative to fully composite structures in
Experimental tests order to overcome some of those drawbacks, namely the deformability and instability problems. These
Footbridge
hybrid solutions are especially attractive for footbridge structures whose design is often governed by ser-
Glass fibre reinforced polymer (GFRP)
Steel fibre reinforced self-compacting
viceability requirements. Nevertheless, in order to make these structural systems standard solutions for
concrete (SFRSCC) civil engineering practice it is necessary to gather in-depth understanding about their structural behav-
Hybrid structures iour, namely under dynamic loads, and to assess the ability of current design tools to predict their
response. This paper presents experimental and numerical investigations concerning the dynamic behav-
iour under pedestrian induced loads of a 6.0 m long hybrid footbridge prototype comprising two GFRP
pultruded profiles and a thin steel fibre reinforced self-compacting concrete (SFRSCC) deck. The results
of dynamic tests and respective numerical simulations are compared in order to access the ability of con-
ventional finite element (FE) models to predict the structural response of the hybrid footbridge, namely
the accelerations as a function of time for different types of pedestrian loads. The experimental data are
also compared with regulation requirements concerning human comfort. The results obtained show that
the models developed using conventional numerical tools are able to predict the dynamic response of the
footbridge prototype under pedestrian actions with fairly good accuracy. The comparison of the results
with regulation requirements also attests the feasibility of the hybrid structural solution proposed.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Among the possible applications of GFRP–concrete hybrid systems

In recent years glass fibre reinforced polymer (GFRP) pultruded
profiles have been increasingly used as structural materials in a
wide range of civil engineering applications. The high strength,
low self-weight, ease of installation, corrosion resistance and elec-
tromagnetic transparency are the main advantages of this material
[1,2]. In opposition, the brittle failure and the low Young’s and
shear moduli constitute major drawbacks regarding the structural
use of GFRP profiles, with their design being often governed by
deformability or instability phenomena, and seldom allowing the
full exploitation of the material resistance [3,4].
In this context, several authors (e.g. [5–12]) have proposed dif-
ferent hybrid structural systems, in which GFRP composites are
combined with traditional construction materials (such as con-
crete) in order to overcome the aforementioned limitations while
maintaining the attractiveness of GFRP based structural solutions.
⇑ Corresponding author.
E-mail address: jgonilha@civil.ist.utl.pt (J.A. Gonilha).
identified earlier [1] are footbridges.
The design of footbridges is often governed by their serviceabil-
ity behaviour, namely in what concerns the fulfilment of pedes-
trian comfort criteria [13,14] (cf. Section 2). Potential problems
regarding the serviceability behaviour of footbridges rise with
the slenderness of the structural solutions [14] and, therefore, with
the use of more deformable materials, such as GFRP. Thereafter, it
is particularly important to correctly assess and predict the struc-
tural behaviour of GFRP–concrete hybrid structures with available
design tools, namely commercial finite element (FE) packages.
However, there are very few published results on the dynamic
behaviour and properties of fibre reinforced polymer (FRP) bridges
(e.g. [14,16]) and, according to the authors’ best knowledge, none
about GFRP–concrete hybrid structures.
This paper presents experimental and numerical investigations
about the dynamic response under pedestrian loads of a footbridge
prototype comprising GFRP pultruded profiles and a very thin con-
crete deck made of steel fibre reinforced self-compacting concrete
(SFRSCC) pre-cast slabs. This small-scale footbridge prototype was

0263-8223/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruct.2012.07.029
454 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463
Nomenclature
Symbol Description
EL,t elasticity modulus in tension of the GFRP for the longi-
tudinal direction (GPa)
ET,c elasticity modulus in compression of the GFRP for the
transverse direction (GPa)
GLT shear modulus of the GFRP (GPa)
ftu,L tensile strength of the GFRP for the longitudinal direc-
tion (MPa)
su,LT in-plane sear strength of the GFRP (MPa)
Ec elasticity modulus in compression of the SFRSCC (GPa)
fcu compressive strength of the SFRSCC (MPa)
built in laboratory and tested in a simply supported span of 5.5 m
in order to investigate its structural behaviour and obtain in-depth
understanding about the GFRP–SFRSCC hybrid system proposed. A
full-scale footbridge, with a span of 10.5 m, will be designed and
built in the campus of Minho University (Guimarães, Portugal) over
a small river. It is worth mentioning that the prototype under
study, being a small-scale model, is not required to perform
according with design codes. One should also note that extrapolat-
ing results obtained from testing a small-scale prototype to the
full-scale structure could be potentially affected by scale effects.
Nevertheless, results obtained from the study of this prototype
are essential to understand the feasibility and dynamic perfor-
mance of the full-scale hybrid footbridge, particularly whether it
fulfils serviceability and ultimate limit states design requirements.
The experimental investigations comprised (i) material charac-
terisation tests on small-scale specimens, (ii) static flexural tests
on the footbridge prototype, (iii) the definition of pedestrian loads,
(iv) modal identification tests and (v) dynamic tests on the foot-
bridge prototype under pedestrian loads. The numerical investiga-
tions included the development of a three-dimensional FE model of
the footbridge prototype, in which the material properties derived
from testing were used as input. The model of the prototype was
first validated with the results of static flexural tests and modal
identification tests and then used to simulate the dynamic tests
under pedestrian loads.
2. Design requirements for comfort in footbridges
Regarding the design requirements for comfort in footbridges,
the Eurocode 0 [17] recommends that structural accelerations un-
der pedestrian actions shall not be greater than 0.70 m/s2 and
0.20 m/s2 for the vertical and horizontal directions, respectively,
independently of the structure’s principal frequency. Other design
codes, relate the maximum acceleration limits with the structure’s
vibration frequencies [14,18]. According to Eurocode 0 [17], the ex-
plicit verification of those accelerations is not required if the first
vertical vibration frequency of the footbridge is higher than 5 Hz.
Other standards addressing human comfort, namely BS 6472
[19] (in this case applicable for buildings), do not limit the maxi-
mum accelerations, but instead the maximum root mean square
(RMS) of the expected accelerations. Such criterion takes into ac-
count not only the peak values of the response but also the overall
structural response. Performance requirements are then set
according to the structural frequency, the type of activity and the
duration of the exposure to the vibrations. ISO 10137 [20] presents
very similar recommendations to BS 6472 [19], but it also indicates
parameters that are directly applicable to walkways or footbridges.
Many design codes, particularly the Eurocode 1 [21], do not
specify which pedestrian actions shall be considered in the
fcr tensile cracking strength of the SFRSCC (MPa)
t Poisson’s ratio of the SFRSCC (–)
Ea elasticity modulus in tension of the epoxy adhesive
(GPa)
fau tensile strength of the epoxy adhesive (MPa)
Em elasticity modulus in compression of the epoxy mortar
(GPa)
fmu compressive strength of the epoxy mortar (MPa)
fbk tensile strength of the stainless steel bolts (MPa)
nn damping of the nth vibration mode (%)
verification of design requirements, providing only frequency
ranges for the pedestrian action. The absence of specific pedestrian
actions, namely well-established load-time functions to be used in
design means that comfort verifications are performed on a case-
to-case basis, depending on the options of the designer and owner.
Nevertheless, load-time functions and other relevant parameters of
pedestrian actions are provided in the literature (e.g., Bachmann
and Ammann [22]) and may be used in comfort criteria verifica-
tions. Fig. 1 shows the load-time functions proposed in [22] for
slow, normal and fast walk, together with experimental curves de-
rived from testing within the present study (cf. Section 4.4). Bach-
mann and Ammann [22] also recommend comfort verifications to
be performed for a pedestrian weighing 70 kgf.
3. Structural concept and design of the footbridge
The main objective of hybrid structures is to fully exploit the
material properties of each one of the components, maximising
their advantages and overcoming their limitations. The hybrid
structure proposed here, comprising GFRP pultruded profiles
(200  100  10 mm, I-section) and a 40 mm thick concrete slab
made of precast segments (Fig. 2), is to be used in simply sup-
ported spans, for which bending moments are (for most actions)
positive. Thereby, the concrete elements that constitute the foot-
bridge deck, being located at the upper part of the section, are basi-
cally subjected to compressive stresses, for which cementitious
materials perform better. In opposition, the GFRP profiles under-
neath, used as main girders, are prevented to buckle by the deck
and are subjected mainly to tensile stresses, for which GFRP mate-
rials behave better.
In order to avoid local failure of the webs over the support sec-
tions, concrete jackets (Figs. 3 and 4) are provided between the
flanges, on both sides of the webs, with a length of 0.20 m, as sug-
gested in [11]. Secondary girders are provided at midspan and at
the support sections in order (i) to avoid local instability of the
webs due to shear, (ii) to increase the torsional stiffness and (iii)
to minimise the initial imperfections of the main girders. The sec-
ondary girders are made of GFRP pultruded profiles with the same
cross-section as the main girders. The main and secondary girders
are connected with small-length (160 mm) angle GFRP profiles
(50  50  8 mm) using M10  55 stainless steel bolts. Fig. 4
shows the connection details for both the support and mid-span
sections, including the cuts of secondary girders.
The connection of the SFRSCC slabs to the GFRP main girders is
provided by a 2 mm thick layer of epoxy adhesive and stainless
steel bolts. The adhesively bonded connection ensures full interac-
tion between the materials, guaranteeing maximum stiffness and
strength in the longitudinal direction [11]. The epoxy adhesive
used in the shear connections was cured at room temperature
 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463 455

Fig. 1. Load-time functions for pedestrian action (slow, normal and fast walk): tests vs. Bachmann and Ammann [22].

In the transverse direction, the 40 mm thick SFRSCC slab guar-

Fig. 2. Cross-section of the footbridge prototype.
(about 20 °C) for at least 2 weeks before experimental tests started.
The steel bolts are used as a redundant connection system in order
to prevent the effects of any long-term deterioration of the adhe-
sive. Since the SFRSCC slabs were precast, the bolts were grouped
and positioned in sockets provided in the slabs. Each group is com-
prised of 8 bolts (4 pairs) 50 mm apart in both longitudinal and
transverse directions. The sockets dimensions are 200  80 mm2
and their centres are approximately 32.5 cm apart (3 sockets/m).
Additionally, the transverse joints between the precast slab seg-
ments were also bonded with a 2 mm thick epoxy adhesive layer
in order to ensure the monolithic behaviour of the structure, there-
by avoiding a reduction of the longitudinal stiffness due to loss of
composite action in the transverse joints.
antees the transmission of loads to the beams by itself. Without
any bar reinforcement, the slab is able to ensure the transverse
safety of the structure owing to the high tensile strength and duc-
tility of the SFRSCC material when compared to conventional con-
crete [23]. The width of the deck was set at 2.00 m based on a
survey of existing footbridges over the Portuguese national rail
network [18], while the transverse spacing between the main gird-
ers (1.10 m) was defined in order to balance the transverse nega-
tive bending moment over the beams.
4. Experimental investigations
4.1. Test programme
The experimental programme comprised the following tests: (i)
material characterisation tests on small-scale specimens of the dif-
ferent structural materials, namely tensile, compressive and shear
tests on GFRP coupons (extracted from the web and flanges), com-
pressive and splitting tensile tests on SFRSCC cylinders, and com-
pressive tests on SFRSCC cubes; (ii) static flexural tests on the
footbridge prototype; (iii) tests for the definition of the pedestrian

Fig. 3. Footbridge prototype: side view (top) and plan view (bottom).
456 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463
Fig. 4. Connection details between main and secondary girders at support and mid-span sections.
Table 1
Main mechanical properties of the GFRP pultruded profiles.
GFRP EL,t (GPa) ET,c (GPa) GLT (GPa) ftu,L (MPa) su,LT (MPa)
Webs 32.85 ± 1.09 5.74 ± 0.73 3.51 ± 0.24 376.69 ± 7.44 24.34 ± 0.86
Flanges 35.94 ± 2.35 4.35 ± 0.93 – 410.67 ± 18.60 –
loads; (iv) modal identification tests; and (v) dynamic tests under
pedestrian actions on the footbridge prototype.
4.2. Materials
The materials used in the construction of the footbridge proto-
type and in the material characterisation tests included GFRP pul-
truded profiles with I-section (200  100  10 mm) produced by
ALTO – Perfis Pultrudidos, Lda., SFRSCC pre-cast slabs
(2.00  1.00  0.04 m), cubes and cylinders produced by CiviTest,
an epoxy resin produced by S&P, an epoxy mortar produced by
SIKA and stainless steel bolts provided by INTEC.
In what concerns the GFRP pultruded profiles, the following
material characterisation tests were performed in coupons ex-
tracted from both flanges and webs: tension (according to EN I-
SO 527 [24]); compression (ASTM D 695 [25]); and shear (by
means of 10° off-axis tension, according to the recommendations
of Hodgkinson [26]). Table 1 summarises the main mechanical
properties of GFRP derived from testing, namely the elasticity
moduli in the longitudinal direction in tension (EL,t) and in the
transverse direction in compression (ET,c), the shear modulus
(GLT), the longitudinal tensile strength (ftu,L), and the in-plane shear
strength (su,LT). The tensile modulus in the transverse direction was
not measured due to the limited geometry of the web (height) and
flanges (width) of the profile.
The SFRSCC material characterisation tests were performed in
order to measure (i) the elasticity modulus (Ec = 36.97 ± 1.94 GPa)
and the Poisson coefficient (m = 0.33 ± 0.03) in compression (tests
were performed according to LNEC E 397 [27]), (ii) the splitting
tensile strength (fcr = 9.42 ± 1.63 MPa, EN 12390-6 [28]), and (iii)
the compressive strength (fcu = 80.65 ± 2.07 MPa, EN 12390-3
[29]). The SFRSSC post-cracking strength is the property that en-
ables this material to present resisting bending moments and duc-
tility after cracking. In order to comply with design requirements,
the SFRSCC shall present a post-cracking strength of 4.00 MPa.
The epoxy adhesive (S&P Resin 220) used to bond the main gird-
ers to the deck presents elasticity modulus and strength in tension
of Ea = 8.8 GPa and fau = 17.3 MPa, respectively [30]. The epoxy
mortar (SIKA Icosit KC 220 N) used to seal the holes for the bolts
presents (according to the manufacturer) an elasticity modulus in
compression of Em = 9.9 ± 1.0 GPa and strengths in compression
and shear of fmu = 100.0 ± 5.0 MPa and smu = 19.0 ± 5.0 MPa, respec-
tively [31]. Finally, the M10  55 mm stainless steel bolts present
(according to the manufacturer) a bearing capacity of fbk = 70 MPa.
4.3. Static flexural tests
The 6.00 m long footbridge prototype was tested in a simply
supported span of 5.50 m. The supports were positioned under-
neath the bottom flanges of the profiles – the major axis flexural
rotations were free at both pairs of supports and one of the pairs
allowed also free longitudinal sliding. The fixed supports were
materialised by means of steel cylinders placed in-between carved
steel plates. For the sliding supports steel spheres were used in-be-
tween flat and polished steel plates.
The static flexural tests were performed by loading the foot-
bridge prototype with two concrete blocks, weighing 522 kgf and
524 kgf. Each block was placed (using a moving crane) on top of
the prototype’s deck at mid-span section, in a symmetric position,
each block being aligned with one of the main girders. The total
load used in the static test corresponds approximately to 1/6 of
the design load for footbridges (5 kN/m2, according to Eurocode 1
[21]), causing a bending moment at midspan that is about 1/3 of
that caused by the design load. Vertical deflections were measured
at five different positions across midspan section with displace-
ment transducers from Mitutoyo (with a stroke of 50 mm and pre-
cision of 0.01 mm). Two displacement transducers were positioned
underneath the GFRP girders and three were positioned below the
SFRSCC slab, one at the transverse mid-span and two at a distance
of 10 mm from the edges of the cantilevers.
Static flexural tests were performed according to the following
three sequential stages: (i) first block loaded over left girder (stage
1); (ii) both blocks loaded over both girders (stage 2); and (iii) right
block loaded over right girder (stage 3). Fig. 5 shows the deflections
measured at mid-span for each stage, together with results from
numeric simulation (presented and discussed in Section 5).
4.4. Definition of pedestrian induced actions
Pedestrian loads considerably depend on a large number of
variables that include the type of footwear, the type of pavement
and the type of human induced load, which naturally varies from
individual to individual [22].
Therefore, in order to ensure a proper comparison between the
results of experimental dynamic tests under pedestrian actions and
corresponding numerical simulations, the pedestrian loads in-
duced by the specific individuals who performed the experimental
tests were properly characterised, namely with respect to the fol-
lowing parameters: (i) frequency between steps; (ii) velocity; (iii)
step size; and (iv) load-time function. In the present study, the fol-
lowing three types of human motion were analysed: (i) walk slow;
(ii) walk normal; and (iii) walk fast.
The first three parameters referred earlier were measured for
each type of human motion using a chronograph and a ruler. In or-
der to measure the fourth parameter, the load-time function, the
pedestrian stepped over an SFRSCC slab (2.00  1.00  0.04 m) that
was simply supported by four load cells from Novatech, with a load
 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463 457

Fig. 5. Static flexural tests – mid-span deflections for: (a) load stage 1; (b) load stage 2; and (c) load stage 3 – experimental tests vs. FE model.

Table 2
Main characteristics of pedestrian motion: experimental tests vs. Bachmann and
Ammann [22].

Motion Frequency (Hz) Velocity (m/s) Step size (m)

Test Literature Test Literature Test Literature
[22] [22] [22]
Walk Slow 1.36 ± 0.10 1.70 0.78 ± 0.06 1.10 0.55 0.60
Walk Normal 1.76 ± 0.16 2.00 1.20 ± 0.07 1.50 0.67 0.75
Walk Fast 1.81 ± 0.13 2.30 1.81 ± 0.18 2.20 0.86 1.00

load-time functions may be suitable for serviceability limit states

design, but are not recommended for direct comparison with
experimental results, as intended in the present study.

4.5. Modal identification tests

Fig. 6. Pedestrian induced actions: assessment of load-time functions.
capacity of 10 kN, as shown in Fig. 6. The load on each cell was
measured at a rate of 50 Hz using a data logger from HBM (model
Spider8) and registered in a PC. The load induced by the pedestrian
(weighing 98 kgf) was calculated from the sum of the values mea-
sured by each load cell. Fig. 1 shows the load-time functions ob-
tained for each type of motion together with the functions
proposed by Bachmann and Ammann [22]. Table 2 compares also
the other parameters obtained in these tests with the values pro-
posed by those authors.
The comparison between the results measured in the present
study and those reported by Bachmann and Ammann [22] clearly
shows that human induced actions can be highly variable for dif-
ferent individuals. In this particular case, Fig. 1 shows that
although the load-time curves have similar peak values and gen-
eral shapes, the period of contact of the foot is much longer for
the present pedestrian than that considered by Bachmann and Am-
mann [22]. Table 2 also shows that the present pedestrian walks
much more slowly (for every motion type) than what is proposed
by those authors. With this regard, one should note that general
Modal identifications tests were performed in order to gather
modal data from the footbridge prototype for FE model validation,
namely the vibration frequencies of the structure. The 6.00 m long
footbridge was tested in a 5.50 m simply supported span with the
same boundary conditions as in the static flexural tests, described
earlier.
The modal identification tests were performed by applying a
stroke, with a common rubber hammer, on the footbridge deck
and measuring the structural response in terms of vertical
accelerations.
Accelerations were measured with two accelerometers (one
from Endevco and the other from Brüel&Kjaer) connected to two
amplifiers (from Brüel&Kjaer) with precision of 0.01 m/s2. The data
acquisition was performed using a data logger from HBM (model
Spider8) at a rate of 400 Hz and registered in a PC. The stroke
was applied eccentrically in the vicinity of mid-span and the accel-
erations were measured at positions A1 and A5 (Fig. 7).
A Fast Fourier Transform (FFT) algorithm was applied to the
half-sum and to the half-difference of the acceleration measure-
ments at those positions (A1 and A5), thereby retrieving the fre-
quencies of the structural free vibration for the first flexural and
458 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463
Fig. 7. Dynamic tests: instrumentation and motion positioning (dimensions in meters).
Fig. 8. Modal identifications tests: FFT of the half-sum and half-difference of
accelerations A1 and A5.
torsional modes, respectively, as shown in Fig. 8. The frequency at-
tained for the first flexural mode is approximately 8.18 Hz, while
the one for the first torsional mode is 12.40 Hz. These modes cor-
respond, respectively, to the first and second vibration modes of
the structure. Moreover, a Stochastic Subspace Identification (SSI)
analysis was performed providing the following structural damp-
ing coefficients for the first four vibration modes: n1 = 1.3%,
n2 = 0.7%, n3 = 1.6%, n4 = 1.0%. The complete details about this anal-
ysis are reported in [32].
4.6. Dynamic tests under pedestrian actions
As already mentioned, the dynamic tests were performed above
all to assess the quality of the numerical simulation. In addition,
experimental tests were expected to provide a measure of the dy-
namic performance of the hybrid system proposed, yet being a
small-scale model, one did not expect the footbridge prototype
to necessarily comply with standard limits.
Fig. 9. Dynamic tests under pedestrian loads (one individual).
The dynamic tests on the footbridge prototype were carried out
for the same kinematic conditions adopted in the static flexural
and modal identification tests, i.e. it was tested in a simply sup-
ported span of 5.50 m. In a first stage, the tests consisted of sub-
jecting the structure to the effect of a pedestrian (weighing
98 kgf) travelling along the deck, with vertical accelerations (on
the top of the deck) being measured at predefined positions
(Fig. 7). As mentioned earlier, three types of human motion were
considered in this study: slow walk, normal walk and fast walk.
Due to the reduced length of the prototype, jogging and running
were not considered to be suitable for this study. Additionally,
two transversal motion paths were considered (Fig. 7), one along
the longitudinal axis of the prototype (centred travelling) and the
other along the centre of one of the cantilevers (eccentric travel-
ling, Fig. 9).
In a second stage of the dynamic tests, eight pedestrians
(0.67 persons/m2, corresponding to a density classification be-
tween sparse and dense crowd, according to [15]) with weights
ranging from 52.6 to 102.7 kgf were asked to cross the footbridge
prototype randomly for a period of approximately 5 min while ver-
tical accelerations were measured at points A1 and A5 (cf. Fig. 7)
and transverse (lateral, horizontal) accelerations of the deck were
measured at quarter- and mid-span. Fig. 10 shows a test with mul-
tiple pedestrians circulating along the footbridge prototype.
 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463 459

Accelerations were measured with the same apparatus used in

Fig. 10. Dynamic tests under pedestrian loads (multiple individuals).
Fig. 11. Accelerations attained at A3 for pedestrian walking slowly at the cantilever.
the modal identification tests (cf. Section 3.5) in paired positions.
The data acquisition was performed using data loggers from HBM
(models Spider8 and QuantumX) at rates of 400 Hz and 200 Hz.
As an illustrative example, Fig. 11 shows the vertical accelera-
tions at location A3 for slow walk eccentric travelling (one pedes-
trian), together with the results of numerical simulation. Table 3
presents a summary of the maximum accelerations (aver-
age ± standard deviation) obtained in the dynamic tests under pe-
destrian actions.
These results show that the structural accelerations vary with
the location within the structure, the motion type and the travel-
ling position (centred or eccentric). While the structural accelera-
tions rise with the walking pace, the results show that eccentric
motion increases the accelerations in eccentric positions (even if
those positions are in the opposite side of the pedestrian, e.g. A5
and A18). This effect clearly indicates that the eccentric motion ex-
cites the torsional vibration modes more than the centred motion,
as expected. Regarding the magnitudes of the accelerations at-
tained in this first stage of the dynamic tests, the results obtained
show that in some cases the limit proposed in Eurocode 0 [17] for
maximum vertical accelerations is not fulfilled by this particular
structure, although its first flexural frequency is 8.18 Hz, therefore
higher than 5.0 Hz. Eurocode 0 [17] limits, however, do not con-
sider parameters such as human sensibility, time of exposure to
the vibrations and frequency of the vibrations. Furthermore, one
should note that these acceleration should be reduced (assuming
a linear-elastic behaviour) by approximately 30% before compari-
son with regulation limits since, as stated earlier, it is recom-
mended that these limits are verified for a pedestrian weighing
70 kgf [22] and the tests were performed by a pedestrian weighing
98 kgf.
Fig. 12 illustrates the vertical accelerations measured in the test
with multiple pedestrians randomly travelling along the bridge.
Based on the above mentioned limitations of Eurocode 0 [17], the
RMS of the vertical and transverse accelerations measured in the
crowd dynamic tests were compared with the limits proposed in
ISO 10137 [20], as shown in Figs. 13 and 14, respectively. The com-
parisons were made for the comfort level applicable to footbridges

Table 3
Maximum accelerations: test (mean ± standard deviation), FE model and relative difference (D).

Location Position Walk slow (m/s2) Walk normal (m/s2) Walk fast (m/s2)
Test FE D (%) Test FE D (%) Test FE D (%)
A1 Centred 0.43 ± 0.04 0.49 14.0 0.70 ± 0.10 0.66 5.7 1.07 ± 0.25 1.00 6.5
Eccentric 0.55 ± 0.08 0.66 20.0 1.23 ± 0.14 0.88 28.5 1.43 ± 0.14 1.90 32.9
A2 Centred 0.40 ± 0.08 0.47 17.5 0.70 ± 0.12 0.57 18.6 0.69 ± 0.12 0.80 15.9
Eccentric 0.42 ± 0.07 0.52 23.8 0.77 ± 0.16 0.67 13.0 1.02 ± 0.11 1.15 12.7
A3 Centred 0.33 ± 0.09 0.47 42.4 0.75 ± 0.10 0.55 26.7 0.70 ± 0.22 0.98 40.0
Eccentric 0.30 ± 0.09 0.46 53.3 0.57 ± 0.14 0.57 0.0 0.71 ± 0.05 0.90 26.8
A4 Centred 0.39 ± 0.07 0.44 12.8 0.63 ± 0.15 0.56 11.1 0.72 ± 0.10 0.77 6.9
Eccentric 0.37 ± 0.06 0.48 29.7 0.76 ± 0.13 0.56 26.3 1.21 ± 0.13 1.27 5.0
A5 Centred 0.44 ± 0.08 0.50 13.6 0.61 ± 0.10 0.62 1.6 1.08 ± 0.30 1.02 5.6
Eccentric 0.51 ± 0.04 0.59 15.7 1.03 ± 0.17 0.71 31.1 1.48 ± 0.15 1.81 22.3
A16 Centred 0.22 ± 0.04 0.36 63.6 0.52 ± 0.05 0.41 21.2 0.59 ± 0.14 0.77 30.5
Eccentric 0.45 ± 0.06 0.41 8.9 0.84 ± 0.23 0.52 38.1 0.92 ± 0.23 1.12 21.7
A17 Centred 0.24 ± 0.07 0.30 25.0 0.48 ± 0.08 0.35 27.1 0.49 ± 0.15 0.61 24.5
Eccentric 0.19 ± 0.06 0.35 84.2 0.37 ± 0.07 0.37 0.0 0.50 ± 0.04 0.66 32.0
A18 Centred 0.19 ± 0.03 0.36 89.5 0.48 ± 0.04 0.40 16.7 0.63 ± 0.11 0.77 22.2
Eccentric 0.39 ± 0.07 0.43 10.3 0.76 ± 0.10 0.46 39.5 0.95 ± 0.26 0.93 2.1
A19 Centred 0.28 ± 0.03 0.29 3.6 0.45 ± 0.10 0.35 22.2 0.58 ± 0.17 0.51 12.1
Eccentric 0.28 ± 0.04 0.31 10.7 0.55 ± 0.13 0.40 27.3 0.84 ± 0.10 0.79 6.0
A20 Centred 0.26 ± 0.02 0.30 15.4 0.46 ± 0.07 0.35 23.9 0.56 ± 0.13 0.54 3.6
Eccentric 0.26 ± 0.03 0.35 34.6 0.56 ± 0.09 0.35 37.5 0.93 ± 0.19 0.66 29.0
460 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463
Fig. 12. Vertical accelerations attained for multiple pedestrians crossing the
footbridge randomly.
Fig. 13. Expected vertical accelerations (RMS): experimental
ISO 10137 [20].
and for periods of exposure to vibration of 16 h, 1 h, 60 s and 10 s.
For the verification of vertical accelerations, it was considered that
there could be people standing still on top of the footbridge (more
severe situation) and, therefore, according to [20] the base curve
multiplying factor was reduced to half (from 60 to 30). Results ob-
tained show that, according to [20], there is a low probability of
complaints due to discomfort for pedestrians travelling or standing
on the footbridge except for a period of exposure to vibration long-
er than 16 h, which is naturally not expectable for a footbridge.
5. Numerical modelling
5.1. Description of the FE model
The three-dimensional numerical model of the footbridge pro-
totype, illustrated in Fig. 15, was developed with the FE commer-
cial package SAP2000, using the following three types of finite
Fig. 14. Expected horizontal accelerations (RMS): experimental tests vs.
ISO 10137 [20].
elements: (i) solid brick elements; (ii) two-joint rigid link ele-
ments; and (iii) frame elements.
The solid elements were used to simulate the geometry of the
prototype, namely the GFRP profiles’ (i) flanges and (ii) webs, (iii)
the SFRSCC slabs, (iv) the 2 mm thick epoxy adhesive layer, (v)
the concrete jackets, (vi) the GFRP angle sections, and (vii) the steel
support plates. The two-joint rigid links were used to simulate the
bolts connecting the main and secondary girders and to connect
the steel support plates to the support rotation centre. The frame
elements were used in order to enable the application of the pedes-
trian loads without changing the FE mesh.
The connection between the SFRSCC slabs and the GFRP profiles
was modelled explicitly by adding a 2 mm thick epoxy adhesive
layer, as mentioned earlier. A full linear-elastic connection be-
tween these materials was considered in accordance with shear
connection tests performed earlier [1,11], which confirmed that
full connection is guaranteed for service conditions. The supports
in one side of the footbridge were defined as pinned, allowing rota-
tions in the longitudinal direction but not allowing any other dis-
placements, while the supports on the other side also allowed for
longitudinal sliding.
tests vs. The material properties adopted for the main structural ele-
ments were those derived from testing (cf. Section 4.2). The GFRP
profiles were modelled as orthotropic materials, while all other
materials were considered isotropic. All materials and analysis per-
formed were linear-elastic. The structural damping coefficients
were defined according to results of modal analysis (cf. Section 4.5).
5.2. Model validation
In order to assess the accuracy of the FE model in predicting the
structural behaviour of the footbridge prototype, the experimental
static flexural test (cf. Section 4.3) was first simulated. The 522 kgf
and 524 kgf blocks were simulated by surface pressure loads ap-
plied on top of the slab.
Fig. 5 presents for each load stage the comparison between
experimental and numerical vertical deflections at midspan. It
can be seen that results are in very good agreement, with the max-
imum relative error (at the centre of the cross-section) being only
6.7%.
The vibration frequencies computed with the FE model also
compare well with those registered experimentally (cf. Section 4.5):
 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463 461

Fig. 15. General view of the FE model.

Fig. 16. Accelerations attained at A1 for pedestrian walking fast at the centre.

Fig. 19. Maximum accelerations attained for pedestrian walking normally at the
centre.

Fig. 17. Accelerations attained at A3 for pedestrian walking fast at the centre.

Fig. 20. Maximum accelerations attained for pedestrian walking fast at the centre.

(i) 7.90 Hz ( 3.4%) for the first flexural mode; and (ii) 14.43 Hz
(+16.4%) for the first torsional mode. The higher relative error ob-
served for the first torsional vibration mode may be due to the con-
sideration of rigid links between the main and secondary girders.
The above mentioned results attest the general good agreement
between numerical results and those measured in the experimen-
tal tests. Therefore, one would expect the FE model to be able to
simulate properly the dynamic behaviour of the prototype.

5.3. Simulation of response under pedestrian actions

As mentioned earlier, the frame elements were used to apply

the pedestrian loads to the structure. To this end, two pairs of
frames, with no mass, weight or stiffness, were used at the top of
Fig. 18. Maximum accelerations attained for pedestrian walking slowly at the the deck, aligned with the longitudinal axis of the footbridge.
centre. One pair of frames was positioned centred with the longitudinal
462 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463
axis of the footbridge, while the other pair was positioned along
the centre of one of the cantilevers, simulating the centred and
eccentric motion paths (cf. Fig. 7), respectively. For each pair, the
frames were 30 cm apart themselves in the transverse direction,
each frame simulating the pathway of one foot. For each type of
motion (walk slow, walk normal and walk fast) and for each mo-
tion position (centred or eccentric), a point load of 98 kgf was as-
signed to the respective frame according to the motion
characteristics presented in Table 2 simulating each footstep. The
load cases were then combined in time history analyses with the
respective load-time function (Fig. 1).
Fig. 11 shows the comparison between vertical accelerations
measured in the dynamic tests and those determined from FE sim-
ulation at position A3, for a pedestrian walking slowly along the
cantilever. Figs. 16 and 17 show a similar comparison at positions
A1 and A3, respectively, for the pedestrian walking fast in the cen-
tred path, while Table 3 presents a detailed summary of the results
obtained in the simulation of the dynamic response under pedes-
trian actions, comparing experimental and numerical values of
maximum accelerations for all tests carried out. Such comparison
is illustrated in Figs. 18–20 for each type of motion (walk slow,
walk normal and walk fast), with the pedestrian centred with the
footbridge longitudinal axis.
The comparison shown in Figs. 16 and 17 reveals a very good
agreement between the numerical predictions and the correspond-
ing experimental data. Although the peak values may present some
differences, the curves observed in the tests are generally well sim-
ulated by the model, proving the ability of the latter to predict the
real dynamic behaviour of the structure. Such ability is attested by
the comparison of peak accelerations for the three different types
of motions presented in Figs. 18–20. In addition, these figures show
that the maximum accelerations obtained from FE simulation are
generally very similar to the corresponding average experimental
results, seldom being outside the range of maximum values mea-
sured in the dynamic tests.
6. Conclusions
The study presented herein concerns the dynamic behaviour
under pedestrian loads of a GFRP–SFRSCC hybrid footbridge and
the ability of current FE models in simulating such behaviour.
The results obtained show that literature load models concern-
ing the effects induced by pedestrians on footbridges are not suit-
able for a direct comparison with experimental data obtained from
dynamic tests, as pedestrian loads may vary considerably from
individual to individual.
The FE models developed within this study provided an excel-
lent agreement with results of static flexural tests and, when tak-
ing into account the actual loads of the pedestrian, they were
also able to predict with fairly good accuracy the dynamic response
of the structure under pedestrian loads. The models were not only
able to predict the maximum values of vertical accelerations at-
tained in the dynamic tests, but they were also capable of repro-
ducing the response of the structure over time. Therefore, these
numerical tools allow checking simple comfort requirements, such
as those of Eurocode 0 [17], but they can also be used in more com-
prehensive comfort analysis of structures using, for instance, the
criteria proposed in ISO 10137 [20].
Regarding the pedestrian comfort of this particular footbridge
prototype, both experimental and numerical results showed that
the peak values of vertical accelerations caused by a 98 kgf pedes-
trian are higher than the Eurocode 0 [17] limit of 0.70 m/s2. How-
ever, since literature [13,18,22] indicates that comfort verifications
shall be performed considering a 70 kgf pedestrian, according to
linear-elastic theory, a 29% reduction on the accelerations should
be expected compared to the present pedestrian, which would al-
low fulfilling the above mentioned limit for the vast majority of sit-
uations analysed. Furthermore, acceleration peak values are not
usually critical for the comfort of pedestrians subjected to vibra-
tion. In fact, the comparison of the experimental results with
ISO 10137 [20] limits suggests that pedestrians should not feel dis-
comfort due to the structural vibrations during the crossing of the
footbridge.
Therefore, the results presented herein show that the hybrid
structural solution proposed is feasible for footbridge applications.
In addition, the models developed using current FE commercial
packages are able to predict its structural behaviour.
Future research will address the dynamic behaviour under pe-
destrian actions of the full-scale footbridge prototype that will be
built on site comprising the same structural system, now spanning
10.5 m. In this case, the footbridge response under faster human
motions, such as jogging or running, will also be assessed. Unlike
the present prototype, the full-scale footbridge shall comply with
standard criteria for pedestrian comfort, namely that of
Eurocode 0 [17].
Acknowledgements
This work is part of the research project PONTALUMIS – Develop-
ment of a prototype of a pedestrian bridge in GFRP-ECC, involving
ICIST/Technical University of Lisbon, the company ALTO – Perfis
Pultrudidos, Lda., and ISISE/University of Minho. The authors
would like to acknowledge ICIST, FCT and ADI (Project 2009/
003456) for the financial support, ALTO Perfis Pultrudidos, Lda.
for supplying the GFRP pultruded profiles, and SIKA, S&P and INTEC
for supplying the other materials used in the experiments. The first
author is also grateful to FCT for funding his research through
scholarship SFRH/BD/70041/2010.
References
[1] Correia JR. GFRP pultruded profiles in civil engineering: hybrid solutions,
bonded connections and fire behaviour, PhD Thesis in civil engineering,
Instituto Superior Técnico, Technical University of Lisbon; 2008.
[2] Cabral-Fonseca S, Correia JR, Rodrigues MP, Branco FA. Artificial accelerated
ageing of GFRP pultruded profiles made of polyester and vinylester resins:
characterization of physical–chemical and mechanical damage. Strain
2012;48(2):162–73.
[3] Correia JR, Branco FA, Gonilha JA, Silva N, Camotim D. GFRP pultruded flexural
members: assessment of existing design methods. Struct Eng Int
2012;20(4):362–9.
[4] Correia JR, Branco FA, Silva N, Camotim D, Silvestre N. First-order, buckling and
post-buckling behaviour of GFRP pultruded beams Part 1: experimental study.
Comput Struct 2011;89(21–22):2052–64.
[5] Deskovic N, Triantafillou T, Meier U. Innovative design of FRP combined with
concrete: short-term behaviour. J Struct Eng 1995;121(7):1069–78.
[6] Hall JE, Mottram JT. Combined FRP reinforcement and permanent formwork
for concrete members. J Compos Constr 1998;2(2):78–86.
[7] Fam AZ, Rizkalla SH. Flexural behavior of concrete-filled fiber-reinforced
polymer circular tubes. J Compos Constr 2002;6(2):123–31.
[8] Hulatt J, Hollaway L, Thorne AM. The use of advanced composites to form an
economic structural unit. Constr Build Mater 2003;17(1):55–68.
[9] Keller T, Schaumann E, Vallée T. Flexural behavior of a hybrid FRP and
lightweight concrete sandwich bridge deck. Composites Part A: Appl Sci Manuf
2007;38(3):879–89.
[10] Correia JR, Branco FA, Ferreira JG. Flexural behaviour of GFRP–concrete hybrid
beams with interconnection slip. Compos Struct 2007;77(1):66–78.
[11] Correia JR, Branco FA, Ferreira JG. GFRP–concrete hybrid cross-sections for
floors of buildings. Eng Struct 2009;31(6):1331–43.
[12] Correia JR, Branco FA, Ferreira JG. Flexural behaviour of multi-span GFRP–
concrete hybrid beams. Eng Struct 2009;31(7):1369–81.
[13] SÉTRA. Assessment of vibrational behaviour of footbridges under pedestrian
loading, technical guide. Service d’études techniques des routes et autoroutes
(SÉTRA); 2006.
[14] Thambiratnam DP, Perera NJ, Abeysinghe CM, Huang M, de Silva SS. Human
activity-induced vibration in slender structural systems. Struct Eng Int
2012;22(2):238–45.
[15] Bai Y, Keller T. Modal parameter identification for a GFRP pedestrian bridge.
Compos Struct 2008;82(1):90–100.
 J.A. Gonilha et al. / Composite Structures 95 (2013) 453–463
[16] Aref AJ, Alampalli S. Vibration characteristics of a fiber-reinforced polymer
bridge superstructure. Compos Struct 2011;52(3–4):467–74.
[17] CEN. Eurocode 0 – basis of structural design. European Committee for
Standardization (CEN), Brussels; 2005.
[18] Aquino A. Design of a hybrid GFRP/ECC footbridge – first stages of deck design.
M.Sc dissertation in civil engineering. Instituto Superior Técnico, Technical
University of Lisbon; 2010.
[19] BS 6472. Guide to evaluation of human exposure to vibrations in buildings (1–
80 Hz). The British Standards Institution (BSI), London; 1999.
[20] ISO 10137. Bases for design of structures – serviceability of buildings and
walkways against vibrations. International Organization for Standardization
(ISO), Genève; 2007.
[21] CEN. Eurocode 1: actions on structures – Part 2: traffic loads on bridges.
European Committee for Standardization (CEN), Brussels; 2002.
[22] Bachmann H, Ammann W. Vibrations in structures: induced by man and
machines, structural engineering documents, vol. 3e. Zurich: IABSE; 1987. p.
176.
[23] Salehian HR, Barros JA, Taheri M. A design-based approach to estimate the
moment–curvature relationship of fiber reinforced elements failing in
bending. Pontalumis project report, Minho University; 2009.
[24] ISO 527:1997. Plastics – determination of tensile properties. International
Organization for Standardization (ISO), Genève; 1997.
463
[25] ASTM D 695. Standard test method for compressive properties of rigid plastics.
American Society for Testing and Materials (ASTMs), West Conshohocken,
Pennsylvania; 2002.
[26] Hodgkinson JM, editor. Mechanical testing of advance fiber composites. Boca
Raton, California: CRC Press; 2000. p. 362.
[27] LNEC test specification E 397. Determination of elasticity modulus in
compression. National Laboratory of Civil Engineering (LNEC), Lisbon; 1993
[in Portuguese].
[28] EN 12390-6 testing hardened concrete – Part 6: tensile splitting strength of
test specimens. European Committee for Standardization (CEN), Brussels;
2009.
[29] EN 12390-3 testing hardened concrete – Part 3: compressive strength of test
specimens. European Committee for Standardization (CEN), Brussels; 2001.
[30] Firmo JP, Correia JR, França P. Fire behaviour of reinforced concrete beams
strengthened with CFRP laminates: protection systems with insulation of the
anchorage zones. Composites Part B: Eng 2012;43(3):1545–56.
[31] SIKA Portugal, SA Icosit KC 220 N – high bearing fluid epoxy mortar. Product
data sheet; 2009 [in Portuguese].
[32] Gonilha JA, Correia JR, Sousa JB, Branco FA, Dynamic tests on a GFRP–concrete
hybrid footbridge prototype under the effects of pedestrian loads. Pontalumis
project report A5.T5.IST.1, Instituto Superior Técnico, Technical University of
Lisbon; 2012.