Composite Structures

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Composite Structures 276 (2021) 114553

Contents lists available at ScienceDirect

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

Modelling flexural performance of hollow pultruded FRP profiles


Mohammad Alhawamdeh a, *, Omar Alajarmeh a, Thiru Aravinthan a, Tristan Shelley a,
Peter Schubel a, Ali Mohammad b, Xuesen Zeng a
a
University of Southern Queensland, Centre for Future Materials, Toowoomba, Queensland 4350, Australia
b
Wagners Composite Fibre Technologies, Wellcamp, Queensland 4350, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: Hollow Pultruded Fibre-Reinforced Polymer (PFRP) profiles, as novel construction material, require further
Hollow GFRP profiles development of design tool to broaden the applications. This paper proposes a combined experimental and
Finite element analysis numerical methodology as a design tool to investigate the failure modes of these profiles under four-point
Local buckling
bending. Two different profiles, each with 10 samples, were tested until failure and were used to validate the
Flexural failure
numerical model. A finite element model was built based on a fast-convergence incremental approach that suits
flexural loading and reduces the computational cost. The validated model was used to study the failure sequence
thoroughly and perform an extensive parametric study on the design parameters. Each geometric parameter was
studied individually first to determine the relevant levels for each parameter in the full factorial study. A full
factorial design of experiment was used to capture the critical parametric interactions with over 81 numerical
models. The design rules and recommendation were established for the optimal flexural behaviour of hollow box
PFRP profiles to withstand local buckling of the top flange.

1. Introduction compression is smaller compared to the response under bending.


Smaller buckle half-wavelength means that more buckling waves will
The civil infrastructure sector witnessed an immerging use of hollow appear due to the lower restraint provided by the other walls, since they
Pultruded Fibre-Reinforced Polymer (FRP) profiles as structural mem­ are also buckled, in compression loading [14–16]. Consequently, local
bers, such as beams and girders in buildings and bridges [1,2]. One of buckling of walls is more critical under axial compression compared to
the most prominent developments on pultrusion was the introduction of bending. Therefore, it is important to investigate the design parameters
pulwinding technology, at which off-axis wound fibres are pulled along of the profile under the loading condition, which the profile is intended
with the axial fibres. The presence of wound fibres improves the for, to reach a clear understanding of the profile limitations and how to
delamination resistance, optimise the transverse properties, and alleviate them. Studying the effect, contribution, and interaction of
enhance the post-processing endurance, such as jointing and bolting these design parameters under bending is going to provide guidelines on
[3,4]. Moreover, it increases the buckling load capacity of the profiles the design for manufacturing and recommended configurations of pul­
[5,6]. Hollow box PFRP profiles manufactured by pulwinding possess wound hollow box PFRP profiles to fully utilise their potentials and
tailorable design parameters including layup parameters (wound fibre overcome local instabilities.
angle, axial-to-wound fibres ratio, and stacking sequence) and geometric The layup properties significantly affect the structural performance
parameters (wall thickness, cross-sectional aspect ratio, and corner of laminated composite beams [17,18]. Several studies were undertaken
radius). These parameters govern the profile response to local buckling, to study these properties and their effect on the local buckling capacity.
which is a dominant failure mode of the top flange of hollow PFRP The buckling load capacity of filament-wound Glass Fibre-Reinforced
beams because of their anisotropic and slender nature [7–11]. Local Polymer (GFRP) profiles subjected to compression was increased by
buckling of hollow box PFRP profiles differs depending on the loading 20% when the winding angle was increased from 30o to 60o [19]. The
condition. The four walls buckle under axial compression loading while effect of the fibres orientation on the buckling load of FRP composite
only the top flange, and partially the webs, would buckle under bending profiles made with unidirectional [20], angle-ply, and cross-ply [21,22]
loading [12,13]. Thus, the buckle half-wavelength under axial layups and subjected to axial compression was studied numerically. It

* Corresponding author: University of Southern, Queensland West Street, Toowoomba, Qld 4350, Australia.
E-mail address: Mohammad.Alhawamdeh@usq.edu.au (M. Alhawamdeh).

https://doi.org/10.1016/j.compstruct.2021.114553
Received 7 April 2021; Received in revised form 17 May 2021; Accepted 5 August 2021
Available online 11 August 2021
0263-8223/© 2021 Elsevier Ltd. All rights reserved.
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

was found that the critical buckling load decreased by 35% when the Table 1
fibre angle increased from 10◦ to 30◦ for the unidirectional composite Layup and geometric properties of hollow PFRP profiles.
profile [20]. The cross-ply laminate was observed to sustain a larger Profile label S-100 × 100 × S-125 × 125 × 6.4
buckling load than angle-ply laminate when the fibre angle is larger than 5.2
30o [21,22]. In general, balanced and symmetric laminate is preferred to Geometric Wall width (mm) 100 125
minimise the manufacturing imperfection and limit the elastic coupling properties Wall thickness 5.2 6.4
effect and warp [23–25]. It was found that adding off-axial fibres along (mm)
with axial fibres will enhance the local buckling capacity of composite Outer corner 10 10
radius (mm)
laminated beams [26]. However, no studies were found to address the Inner corner radius 4.8 4.8
optimal ratio of these fibres for pulwound profiles. (mm)
Regarding the geometric parameters, the effect of the wall thickness Layup Stacking sequence [0/+50/-50/0/- [0/+50/0/-50/0/-
of T and C shape PFRP beams on the local buckling capacity of the properties 50/+50/0] 50/0/+50/0]
Fibre percentage 0◦ : 82.2, 50◦ : 0◦ : 78.1, 50◦ : 21.9
compression flange was studied under uniformly distributed bending
(%) 17.8
load [27]. It was found that the local buckling capacity of the
compression flange is inversely related to its slenderness. The buckling
capacity was reduced by 37% when the wall thickness was decreased bending by full factorial design of experiment. These parameters are
from 12 mm to 6 mm. The buckling behaviour of hollow box pultruded three layup parameters (wound fibre angle, axial-to-wound fibres ratio,
GFRP beams subjected to four-point bending was investigated for and stacking sequence) and three geometric parameters (wall thickness,
different flange slenderness values made by maintaining the same cross-sectional aspect ratio, and corner radius). Each geometric
thickness and changing the flange width [28]. The local buckling parameter was studied individually to generate the failure map of hol­
moment of the flange was decreased by 49% due to the reduction in the low PFRP profiles and to device up the applicable levels for each
flange slenderness from 12 to 6. No limitations or inflection points of parameter in the full factorial study. The current study demonstrates a
failure modes of the top flange were reported for such profiles. comprehensive novel approach to the design process of hollow PFRP
The effect of the height-to-width ratio on the critical buckling load of profiles and maximises the material efficiency for bending applications.
hollow box PFRP profiles subjected to compression was studied [29].
The critical buckling capacity of the top flange was increased by 32% 2. Experimental program
when the height-to-width ratio was increased from 1 to 3 (maintaining
the same cross-sectional area). Moreover, the local buckling failure map An experimental program was undertaken on hollow box PFRP
of the flange and web for different cross-sectional aspect ratios of profiles to validate the numerical model that will be used to investigate
adhesively bonded laminated box beam was obtained under three-point the design parameters through a series of parametric studies.
bending [30]. However, the inflection point between the local buckling
and the compressive failure of the top flange, and the interaction be­ 2.1. Materials and structures
tween this parameter and the other geometric parameters were not
reported. The studied PFRP profiles were manufactured by Wagners CFT using
The corners are considered critical zones in PFRP beams as they need pulwinding technology. These square profiles are made of E-glass fibre
support to enhance the structural performance and avoid premature & Vinyl-Ester polymer resin with the layup and geometric properties
failure due to stress concentration [31–33]. The corner of I-shape PFRP shown in Table 1, as provided by the manufacturer. Fig. 1 depicts the
beams subjected to four-point bending was enhanced by adding hand- cross-sections of these profiles.
layup fillets (38 mm) on the top flange-web junctions [34]. The buck­
ling resistance was enhanced by 50% due to the increase in rotational 2.2. Test setup
stiffness and the failure mode was shifted from buckling of the top flange
to compressive failure of fibres. Nevertheless, no studies were found on The flexural behaviour of the hollow box PFRP profiles was exam­
the corner geometry of hollow box PFRP profiles and its configuration ined under a four-point bending test configuration. Ten specimens of
that separates the two failure modes. each profile with a length of 2435 mm and a clear span of 2235 mm were
After reviewing the relevant literature, it appears that most of the prepared to be loaded until failure. The span-to-depth (L/D) ratio was
design parameters were studied individually to obtain their effect only. 22.35 and 17.88 for the S-100 × 100 × 5.2 and S-125 × 125 × 6.4
Moreover, part of them was studied under compression, which may not profiles, respectively. Fig. 2 shows the test configuration along with the
reflect the exact behaviour under bending. In addition, their contribu­ used dimensions. These values were considered in compliance with
tion and interactions were not quantified. Practical design guidelines ASTM D7249/D7249M specifications to avoid shear effect and localised
rely on these two fundamental aspects to focus on the most significant transverse damage and to obtain pure bending behaviour [46–49].
parameters and obtain the optimal configuration of the hollow PFRP Simple (Pin-roller) steel supports were used at the ends and the load was
profiles. Furthermore, the influence of these design parameters on the applied through two steel plates, each of them covers 100 mm of the
failure sequence of these profiles under bending has not been charac­ beam length, as shown in Fig. 3.
terised, especially the corners. The finite element method (FEM) is A Linear Variable Differential Transducer (LVDT) unit was mounted
considered as an excellent option to thoroughly study the failure at the mid-span of the bottom flange to record the deflection of the
mechanism of hollow box PFRP profiles and carry out parametric studies beam. The loading cells were connected to a data acquisition system to
on the design parameters due to its capability of handling combined obtain the load values. A quasi-static loading-rate of 10 mm/min was
failure problems and complex geometries [35–45]. applied on the beam top flange through the two loading plates.
After studying the failure mechanism of hollow box PFRP profiles
under axial compression [16], this paper is going to investigate the 3. Finite element modelling
failure mechanism of these profiles under bending experimentally and
numerically. The incremental modelling approach presented by the 3.1. Background
authors [16] for compression was further developed with fast conver­
gence under flexural loading while improving the computational effi­ Abaqus/CAE 2019 was used to simulate the two hollow box PFRP
ciency. The validated model is going to be used to study the effect, profiles. The stiffness properties controlling the elastic behaviour of the
contribution, and interaction of the design parameters under four-point box profiles were defined through the elastic lamina material definition.

2
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 1. Cross-sectional dimensions of (a) S-100 × 100 × 5.2 and (b) S-125 × 125 × 6.4 hollow box PFRP profiles.

Fig. 2. Four-point bending test configuration and dimensions of a simply supported beam.

Fig. 3. Hollow box PFRP profile loaded under four-point bending test configuration.

This definition is suitable for two-dimensional plane stress formulation transversely isotropic materials [37].
members such as laminated shells [50]. The mechanical properties of an
E-glass/Vinyl-ester lamina with fibre volume fraction (Vf ) of 0.6 (as 3.2. Local buckling and progressive failure behaviours
provided by the manufacturer) are shown in Table 2. The rule of mixture
was used to calculate the elastic modulus in the fibre direction (E1 ). The entire structural behaviour of the hollow box PFRP beams was
Whereas the empirical equations [51,52] were applied to obtain the modelled, including local buckling and progressive failure. A nonlinear
transverse elastic modulus (E2 ), the in-plane shear modulus (G12 ), and geometric analysis was performed using the Newton method in Abaqus/
the out-of-plane shear modulus (G23 ). The value of (G13 ) was assumed to Standard along with the large displacement formulation to capture the
equal the value of (G12 ) because unidirectional laminae behave as deformations accompanying the buckling behaviour and consider all the

3
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Table 2 of the four failure modes is met at any element. This algorithm is based
lamina mechanical properties of the hollow box pultruded FRP profiles. on the energy release, which equals the area under the equivalent stress-
Elastic properties Strength limits Fracture energy displacement curve of the element [50]. Thus, the maximum energy
needed to cause failure in the element; the fracture energy, should be
E1 (MPa) 45,700 T
X (MPa) 803 GLT (N/mm) 92
assigned for each failure mode. The fracture energy values of E-glass/
E2 (MPa) 12,100 Xc (MPa) 548 GLC (N/mm) 79
Ly556 epoxy lamina were used [53] since no data was found in the
v12 0.28 YT (MPa) 43 GTT (N/mm) 5
literature for each failure mode of the studied lamina. These values are
G12 = 4600 YC (MPa) 187 GTC (N/mm) 5
shown in Table 2 with double subscript notation, consisting of L and T as
G13 (MPa)
the first subscript that refers to the longitudinal and transverse di­
G23 (MPa) 4000 SL (MPa) 64
rections and the second subscript of T and C that denotes to the tension
ST (MPa) 50
and compression. The damage stabilisation scheme was used to over­
come the difficulties of results convergence when modelling material
nonlinearity effects (from geometry, material failure, and boundary failure by making the tangent stiffness matrix positive for adequate time
conditions). The adaptive automatic stabilisation scheme was used to increments. The solver introduces a viscosity coefficient to the damage
damp the severe nonlinearities, which accompany buckling and prevent evolution equations to relax the time in the vicious system [50]. The
termination errors. To eliminate the dependency of the incremental optimal viscosity coefficient values for the four failure modes was
solution on the number of increments when buckling is dominant, the specified by a sensitivity study with a range of [1 × 10− 6 − 1 × 10− 2 ] sec.
maximum increment size was reduced down to 0.35% of the total step A viscosity coefficient of 1 × 10− 3 sec was used for each failure mode
time until convergence was achieved with a tolerance of 5% between the depending on the energy balance of the models.
load capacities of the successive increment sizes. This novel modelling
approach combining the local buckling, post-buckling, and progressive 3.3. Mesh, boundary conditions, and loading condition
failure behaviours was addressed in detail and verified in previous
research under compression [16]. However, when studying the The quadrilateral (8-node) in-plane general-purpose continuum shell
convergence at the current problem under bending, the solution was (SC8R) element was used to model the hollow box PFRP profiles. This
noticed to converge faster against the number of increments compared three-dimensional element presents the best option for the studied PFRP
to compression, as shown in Fig. 4. The numerical ultimate buckling profiles and their geometric parameters. It allows geometric modifica­
stress converged at 1.33% of the total step time with the same tolerance tions through the thickness such as tapering or thicken the corner ge­
on the load capacity provided for compression. Considering this obser­ ometry. Moreover, it simulates the through-thickness behaviour more
vation, the analysis can be performed faster even with the larger number accurately than conventional shells due to its through-thickness seeding
of elements (the element size is constant under both loading conditions) feature [50]. The number of elements through the thickness was
due to the four times larger increment size compared to compression. increased by local seeding to greatly reduce hourglass modes and cap­
This finding was related to the severity of local buckling under ture the kinematic response accurately. The suitable element size
compression compared to bending. A smaller increment size was needed allowing the results to converge was obtained through a mesh sensitivity
to capture the instability on the four walls and converge the results study. For both S-100 × 100 × 5.2 and S-125 × 125 × 6.4 PFRP profiles,
under compression compared to one wall under bending. five elements through-thickness and 5 mm element edge length were
Hashin damage model [50] was utilised to simulate the progressive used. The mesh around the corners was refined by locally assigning five
failure of the lamina constituents (fibres and matrix). Four different elements to this critical zone of stresses concentration.
failure modes are considered in the model: rupture of fibre in tension, The supports were simulated by assigning boundary conditions at the
buckling and kinking of fibre in compression, cracking of matrix under ends of the beams. The translational degrees of freedom were restrained
transverse tension and shearing, and crushing of matrix under transverse at both ends with only free axial displacement (with respect to the beam
compression and shearing. Three components are required for the model axis) at one end allowed free to move to model the roller behaviour and
to be defined, which are the damage initiation criteria, damage evolu­ allow for pure axial shortening-transverse deflection deformation
tion response, and damage stabilisation scheme. For the damage initi­ without unrealistic intervention from supports in the flexural stiffness.
ation criteria, the strength limits of a unidirectional E-glass/Vinyl-Ester In addition, all the rotational degrees of freedom were free to rotate. The
lamina were extracted from literature for the same profiles [51]. These loading plates were simulated using three-dimensional discrete rigid
values are shown in Table 2. The superscripts C and T symbolise solid parts with dimensions of 150 × 100 × 20 mm. These plates were
compression and tension, and the notations X, Y, and S refer to the connected to the top flange of the hollow box beams over a 100 mm
longitudinal, transverse, and shear strength values. The damage evolu­ length using a tie constraint definition. The loading process was simu­
tion algorithm is triggered when the damage initiation criterion for any lated by imposing a vertical translational movement on the loading

Fig. 4. Ultimate buckling stress convergence when reducing the increment size for the S-100 × 100 × 5.2 profile under compression [16] and bending (cur­
rent study).

4
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Table 3 4.1. Local buckling load capacity of the hollow box PFRP profiles
Comparison of the probable failure modes and the current study results.
Result source Failure mode S-100 × S-125 × The studied hollow box PFRP profiles were analysed theoretically
100 × 5.2 125 × 6.4 [12,14,15,37] to define the triggering failure mode and then assign the
Theoretical Local Buckling of Walls Due to In- 30.15 58.27 appropriate theoretical equations to be compared with the current study
Plane Compression (kN) results. Table 3 shows the theoretical results compared to the experi­
[12,14,15,37] mental and numerical results. Local buckling of walls, due to in-plane
Local Buckling of Walls Due to In- 75.63 119.01 compression, controls the failure mode since its capacity is the mini­
Plane Shear (kN) [12,14,15,37]
Web Crushing in the Transverse 224.81 273.61
mum value among the theoretical results and the only capacity that is
Direction (kN) [12,14] lower than the experimental and numerical results. Nevertheless, the
Web Buckling in the Transverse 359.62 395.33 theoretical calculations proved to be conservative with a high error of
Direction (kN) [12,14] 24.9% and 36.8% between the theoretical and experimental load ca­
Flange and Web Longitudinal 64.82 127.60
pacities for S-100 × 100 × 5.2 and S-125 × 125 × 6.4 beams, respec­
Material Failure (kN) [14,37]
Flange and Web Material Shear 64.73 109.89 tively. This finding agrees with the same literature calls regarding the
Failure (kN) [14,37] inaccuracy of these empirical equations and the need to develop them to
Experimental Average experimental ultimate load 40.12 92.13 be more accurate. The buckling stress-to-longitudinal modulus ratio
(kN) ((σ cr )/(EL ))of the walls was assessed to check whether the top flange or
Numerical FEM ultimate load (kN) 44.65 97.27
the webs buckle first. For both profiles, (σcr )flange /(EL )flange <
(σ cr )web /(EL )web , which indicated that the compression flange will buckle
plates through a displacement-control loading of 10 mm/min. before the webs [14]. This finding aligns with the next section conclu­
sions regarding the failure sequence.
4. Model validation After confirming the theoretical failure mode, the numerical and
experimental buckling stresses were compared to the most cited theo­
The models’ validity was assessed by comparing the FEM results retical equations used to estimate the local buckling stress (σ)cr of an
versus the experimental results in terms of the load–deflection curves orthotropic laminated plate (the compression flange in this study). The
and the failure mode. In addition, the numerical and experimental theoretical formulas assume the top flange to be clamped along the
buckling loads were compared to the theoretical buckling loads of the beam length and subjected to uniaxial compressive loading along its
most cited equations in the literature. The validated models assisted in width. This assumption can be justified in this research by the high fibre
understanding the failure sequence and its triggering mechanism. volume fraction and the continuation of the wound fibres around the
corners [54]. The results of this comparison are shown in Table 4. The
FEM results were the highest since they reflect the behaviour of an intact
beam, while the theoretical results were lower than both experimental

Table 4
Theoretical vs experimental and FEM local buckling loads of hollow box PFRP beams.
Reference Buckling stress (σ)cr Equation Profile

S-100 × 100 × 5.2 S-125 × 125 × 6.4

(σ)cr (MPa) (σcr )EXP (MPa) (σcr )FEM (MPa) (σ)cr (MPa) (σcr )EXP (MPa) (σcr )FEM (MPa)
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
[12] 2 2
π t (ELC ETC )(1 + 4.1ξ) ETC vL GL 224.9 308.4 324.2 281.5 346.2 359.7
( + (2 + 0.62ξ2 )( + ))
b2 6 12 6
[14] π 2 √ ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
̅ 2
225.4 282.3
(2 (D L D T )(1 + 4.139ξ) + (DLT + 2D S ) + (2 + 0.62ξ ))
b2 t
[10] 24 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 274.1 325.8
(1.871 D11 .D22 + (D12 + 2D66 ))
b2
[13,15,37] π2 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 282.7 336.2
(4.6 D11 .D22 + 2.67D12 + 5.33D66 )
b2

Fig. 5. FEM vs experimental load–deflection curves of (a) S-100 × 100 × 5.2 and (b) S-125 × 125 × 6.4 beams.

5
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 6. Numerical visualisation showing (a) the deflected shape of the top flange of S-100 × 100 × 5.2 beam and the cross-sections of S-100 × 100 × 5.2 beam at the
ultimate load point (b) cross-section A-A and (c) cross-section B-B.

and numerical results as they were built on conservative assumptions 95.5% of the ultimate load (42.6 kN and 92.9 kN for S-100 × 100 × 5.2
(neglecting the shear deformations) for simple formulas and safe design and S-125 × 125 × 6.4 beams, respectively). The localised waviness
[11]. Nevertheless, a good agreement was found between these values. occurred at the mid-span of the beams, between the two loading points,
and propagated until the ultimate load was reached, as shown in Fig. 6
(a) that shows the top flange deflected shape of the S-100 × 100 × 5.2
4.2. Load-deflection curves and failure mode
beam using the numerical results. Hashin damage initiation criteria of
fibre by compression and matrix by tension were used to track the
The numerical versus experimental load–deflection curves are shown
waviness extent, as shown in Fig. 7 that presents the failure sequence of
in Fig. 5. In general, the hollow PFRP beams exhibited linear elastic
the S-100 × 100 × 5.2 beam. Moreover, Fig. 7 (b) shows the agreement
behaviour experimentally and numerically until the ultimate load. The
between the numerical and experimental buckling mode at the ultimate
agreement between the curves is very good in terms of the flexural
load point. As the top corners were resisting the load, they exhibited
stiffness (EI/L3 ) and the ultimate load capacity. The variation in the
tensile damage of matrix accompanying the buckling of the top flange.
flexural stiffness between the numerical and average experimental re­
The local buckling then extended to the webs causing more damage.
sults was 3.5% and 2.7%, with a standard deviation of 29.4 N/mm and
Nevertheless, the flange-web junction maintained its rotational stiffness
142.6 N/mm for the experimental values, for S-100 × 100 × 5.2 and S-
(evident by an angle of 90◦ ) to resist the buckling transfer from the top
125 × 125 × 6.4 beams, respectively. The numerical ultimate load
flange to the webs until the beam collapses, as shown numerically in
varied by 10.3% and 5.6% from its corresponding average experimental
Fig. 6 (b) and Fig. 6 (c). These figures show two cross-sections, A-A and
value, with a standard deviation of 4.1 kN and 1.5 kN for the experi­
B-B, taken for the S-100 × 100 × 5.2 beam at the ultimate load point, as
mental values, for S-100 × 100 × 5.2 and S-125 × 125 × 6.4 beams,
shown in Fig. 6 (a). This behaviour was attributed to the continuous
respectively.
wound fibres around the corners. These wound fibres resisted the cracks
The failure sequence started by local buckling of the top flange at

6
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 7. Failure sequence of S-100 × 100 × 5.2 beam starting by local buckling of the top flange at (a) 95.5% of ultimate load followed by localised waviness
propagation until (b) the ultimate load point.

Fig. 8. Experimental vs numerical collapse of (a) S-100 × 100 × 5.2 and (b) S-125 × 125 × 6.4 beams.

along the axial fibres and encountered delamination at the corners. flange and webs occurred at the mid-span of the beams. The output
Afterwards, the localised waviness subsided when the collapse variable (HSNMTCRT), which refers to the matrix failure by tension, was
occurred at the top flange due to compressive failure of fibres, after used to track the delamination of the beams after the collapse, as shown
tensile failure in the matrix. Fibres spalling and delamination of the top in Fig. 8. There was a concern that this experimental test setup might

7
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 9. Wall thickness effect on the flexural behaviour of hollow box PFRP profile.

potentially cause localised premature damage of the specimens at the


Table 5
loading points by the metallic plates. However, the finite element
Levels range for full factorial study on the geometric parameters of the hollow
models, with the simulation of the exactly same experimental setup,
box profile.
showed no sign of pervasive stresses concentration around the loading
Geometric Parameter Level 1 Level 2 Level 3
plates. Moreover, the buckling occurred before any sign of premature
failure, either experimentally or numerically. Thus, this test setup was Wall thickness (mm) 4.6 5.2 5.8
appropriate for the current study. Cross-sectional aspect ratio (h/b) 1.0 1.25 1.5
Corner radii ratio (r/R) 0.5 1.0 2.0

5. Design of experiments
parameters, respectively, will be discussed following.
The hollow box S-100 × 100 × 5.2 profile was selected for the
parametric studies and the geometric parameters were normalised to
produce universal results regardless of the profile’s size. Reduced Design 5.1. Level settings of geometric parameters
of experiment (DOE), such as the Taguchi method, is only valid to study
the expected interactions since they have to be assigned before the The profile’s stability and its failure mode are dominated by the
analysis is performed. Consequently, full factorial DOE was used for all geometric parameters [37,56]. Thus, each geometric parameter was
the parametric studies in this research to cover all the possible in­ investigated independently (when a parameter is studied, the others
teractions between the studied parameters [55]. The DOE and the remain constant) to locate any inflection points in the failure mode and
analysis of the numerical results were performed by Minitab 19 statis­ assess the practical levels range of each parameter to be applied in the
tical analysis software. Three parametric studies, each one has three geometric factorial parametric study. The layup properties (wound fibre
parameters, were designed to study the effect, relative contribution, and angle, axial-to-wound fibres ratio, and stacking sequence) were main­
interactions of the parameters. The first study was undertaken on the tained constant across all these individual parametric studies and in the
layup parameters (wound fibre angle, axial-to-wound fibres ratio, and geometric parametric study with the values currently used in produc­
stacking sequence), while the second parametric study was performed tion, as shown in Table 1.
on the geometric parameters (wall thickness, cross-sectional aspect The wall thickness was studied by altering its value from 4.6, 5.2,
ratio, and corner radius). The third parametric study was carried out on 5.8, 6.4, 7.0, 7.6, to 8.2 mm. The numerical results of the wall thickness
the geometric parameters of the flange-web junction (flange slender­ effect on the hollow box beam are shown in Fig. 9. These values were
ness, web slenderness, and corner radius). The studied parameters were normalised to the flange width (b) to obtain the wall slenderness (b/t).
divided into these groups to reduce the huge size of the experiment The failure mode was shifted from local buckling to compressive failure
matrix and the enormous computational cost. In all the parametric of the top flange at an inflection point of b/t = 16.4. The flexural
studies, the fibre volume fraction (Vf ) was kept constant at 0.6 with the strength is greatly reduced, and the failure occurs before reaching the
mechanical properties of an E-glass/Vinyl-ester lamina shown in ultimate material strength when local buckling controls. For instance,
Table 2. The layup parameters (wound fibre angle, axial-to-wound fibres the numerical compressive strength of the S-100 × 100 × 5.2 profile was
ratio, and stacking sequence) were studied and found to be negligible 324.2 MPa, while the ultimate material strength from the coupons test
compared to the geometric parameters. The total contribution of the reached up to 445 MPa [57]. Due to instability, the compressive strength
layup parameters on the flexural stiffness (EI/L3 ) and strength was of the top flange was reduced to 0.67 of the material failure strength.
15.7% and 7.8%, respectively. Consequently, this research will focus According to the numerical results, the compressive strength is
more on the geometric parameters and their interactions under bending increased up to 440 MPa when the wall thickness is increased up to 6.4
since they significantly control the structural performance of hollow mm, and the failure mode is shifted from local buckling to compressive
PFRP profiles and since they are recognised more internationally in failure at the top flange.
structural design approaches. Only the second and third parametric The chosen levels of the wall thickness for the geometric parametric
studies on the geometric parameters and flange-web junction study were assigned to gradually shift the failure mode from local
buckling to compressive failure to capture any differences in the

8
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 10. Effect of the cross-sectional aspect ratio on the flexural behaviour of hollow box pultruded FRP beam.

Fig. 11. Outer corner radius effect on the flexural behaviour of hollow box pultruded FRP beam.

response along with the different failure modes. Consequently, wall structural performance when moving the failure mode from local
thicknesses of 4.6, 5.2, 5.8 mm were selected, as shown in Table 5. buckling to compressive failure of top flange, as shown in Table 5.
Investing in the section height of PFRP rectangular beams is a Hollow box PFRP profiles are usually manufactured with fillet cor­
practical solution to enhance their flexural performance by increasing ners to minimise resin-rich zones, facilitate the pulling process, reduce
their moment of inertia [12,15,58,59]. The effect of changing the cross- the stresses concentration, and avoid wound fibres fracture under high
sectional height and width (while maintaining the same cross-sectional pulling force at these corners [14,37,58,60,61]. The effect of the corner
area) was studied here by introducing the height-to-width ratio (h/b) as geometry on the structural performance of hollow box PFRP beams was
a design parameter. The studied values of h/b ratio were 1, 1.25, 1.5, studied here by assessing the outer and inner corner radii. First, the
1.75, 2, 2.25 ,2.5 ,2.75 ,3, 3.5, and 4. The numerical results are shown in outer corner radius (R) was investigated while keeping the inner corner
Fig. 10. The load capacity increased when the h/b ratio was increased. radius (r) constant. The tested values of the outer corner radius were 5,
This observation agrees with the findings of [21]. It was inferred that the 10, 15, 20, 25, 30, 35, and 40 mm, and their numerical results are shown
reduction in the unsupported width of the top flange was the reason as it in Fig. 11. The section change from box to circular shape was tracked by
enhanced its buckling resistance. The inflection point between the fail­ normalising the outer corner radius with respect to the wall width (b/R).
ure modes was h/b = 1.5, at which the flexural strength of the beam was Increasing the outer corner radius resulted in a reduction of the flexural
enhanced by 30.4%. Another inflection point was found at h/b = 3, at stiffness (EI/L3 ) due to the decrease in the moment of inertia of the
which the failure mode was shifted from compressive failure of the top section. The profile lost approximately 1.2% of its flexural stiffness for
flange to local buckling of the webs. This was attributed to the increase each 1 mm increase in its outer radius. In addition, the moment capacity
in the unsupported height of the webs, which reduced their local of the beam deteriorated even though the failure mode shifted from local
buckling capacity below the material ultimate strength of the flange and buckling towards compressive failure of the top flange (due to the in­
resulted in a reduction in the overall flexural strength of the beam. crease in the circumferential confinement and the uniform distribution
The chosen h/b ratios for the geometric parametric study were 1.0, of stresses). This was attributed to the reduction in the cross-sectional
1.25, and 1.5. These values were selected to track the changes in the area of the top flange, which prevented it from attaining higher

9
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 12. Effect of the corner inner-to-outer radii ratio (r/R) on the flexural behaviour of hollow box PFRP beam.

Fig. 13. Flange slenderness effect on the flexural behaviour of hollow box PFRP profile.

moment capacity even though its compressive strength has reached the Alternatively, a proposition was introduced by keeping the outer
ultimate material strength. The top flange lost 37.5% of its cross- radius (R) at its minimum applicable value from the manufacturing
sectional area at the inflection point (b/R = 4), where the change in perspective and increasing the inner radius (r) of the corner with axial
the failure mode occurred. fibres. A new parameter of the corner geometry was presented, namely

Fig. 14. Web slenderness effect on the flexural behaviour of hollow box PFRP profile.

10
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Table 6 Table 7
Levels range for full factorial study on the flange-web parameters of the hollow Percentile contribution of each geometric parameter on the flexural behaviour of
box profile. the hollow box PFRP beam.
Geometric Parameter Level 1 Level 2 Level 3 Geometric Parameter Flexural stiffness (EI/L3 ) Flexural strength

Flange thickness (mm) 4.6 5.2 5.8 Wall thickness (mm) 25.43 28.75
Web thickness (mm) 4.6 5.2 5.8 Cross-sectional aspect ratio (h/b) 49.29 18.63
Corner radii ratio (r/R) 0.5 1.0 2.0 Corner radii ratio (r/R) 25.04 49.41
Error (%) 0.24 3.21
ANOVA R2 (%) 99.76 92.79
the corner inner-to-outer radii ratio (r/R). This parameter was investi­
gated with a range of values starting at 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, to 3.5.
The numerical results of investigating these values are shown in Fig. 12. sequence) were maintained constant in this parametric study with the
This solution maintained the box shape and enhanced the flexural values currently used in production, as shown in Table 1.
stiffness (EI/L3 ) due to the increase in the cross-sectional area of the
profile. The flexural strength increased, and the failure mode shifted 6. DOE discussion
from local buckling to compressive failure of the top flange at r/R = 2.5
due to the reduction of the unsupported width of the top flange and the The effect of the design parameters on the structural behaviour of the
increase in the corners restraint. hollow box PFRP beam was assessed depending on the two main re­
The corner inner-to-outer radii ratio was nominated for the geo­ sponses in flexural design, namely, the flexural stiffness (EI/L3 ) and
metric parametric study to investigate its relative contribution and strength. The contribution (significance) of each design parameter was
interaction to the other geometric parameters. The selected levels were calculated statistically by the Two-way ANOVA (analysis of variance)
0.5, 1.0, and 2.0, as shown in Table 5. These values were chosen to track using Minitab 19 statistical analysis software. Also, the main effect and
the structural performance as the failure mode moves gradually from interaction plots were introduced to depict the effect and interaction of
local buckling to compressive failure of the top flange. each design parameter, respectively. These plots display the averaged
values for the levels of each parameter [62]. The full list of the design
5.2. Level settings of flange-web parameters matrices and numerical results of the parametric studies are presented in
Appendix A.
After studying the profile’s overall geometry, the flange-web geom­
etry is investigated here. The effect of the flange slenderness (b/tf) and 6.1. Effects of the geometric parameters
web slenderness (h/tw) on the failure mode is examined to set up their
levels in the full factorial study. The contribution of each geometric parameter on the flexural
The flange slenderness was assessed by changing its thickness from behaviour of the hollow box PFRP beam is shown in Table 7 that shows
2.8, 3.4, 4, 4.6, 5.2, 5.8, 6.4, 7, to 7.6 mm. Increasing the flange thick­ the two-way ANOVA model results. The flexural stiffness (EI/L3 )
ness increased the buckling strength since the associated slenderness increased by 17.7% per 1 mm increase in the wall thickness, by 48.1%
ratio was decreased and the flange stability increased, as shown in when h = 2b, and by 16.6% per 1.0 increment in r/R ratio due to the
Fig. 13. Similar behaviour was observed when studying the web slen­ enhancement on the moment of inertia. The cross-sectional aspect ratio
derness with the same range of thickness values, as shown in Fig. 14. was the most significant parameter affecting the flexural stiffness (EI/L3 )
However, while the failure mode was shifted from local buckling of the by contributing double of what the other two parameters contributed.
top flange to compressive failure of top flange when the flange slen­ This was attributed to its higher influence on the moment of inertia of
derness ratio was reduced (b/tf ≤ 15.6), the failure mode moved from the section. Besides that, the corner radii ratio contributed the most to
local buckling of top flange and webs towards a compressive failure of the flexural strength of the beam followed by the wall thickness then the
corners when the web slenderness ratio was reduced (h/tw ≤ 13.2). It h/b ratio. It was inferred that the increase in the corners restraint when
was inferred that the high rigidity provided by the thick webs redis­ the r/R ratio is increased presents the largest impact on the buckling
tributed the stresses from the top flange towards the top corners, thus resistance of the top flange due to the increased rigidity transferred from
preventing the stresses in the flange from reaching the buckling the web to the flange and to the effect of a large r/R ratio on reducing the
strength. This behaviour highlights the significance of the flange and effective buckling width of the flange. Moreover, the enhancement in
web interaction and the importance of their junction (top corners) in the flexural strength when the flange thickness was increased was su­
shaping the profile failure mode under bending. Moreover, the threshold perior to its counterpart resulting from reducing the unsupported width
(inflection) point of failure modes was varying depending on the of the top flange. This was attributed to the exponential effect of the wall
examined slenderness ratio. It was at h/tw = 13.2, b/tf = 15.6, or b/t = thickness on the buckling capacity of the flange [63]. For strength-
16.4 (section 5.1), which clearly shows the interaction between the controlled design, investing in r/R ratio will obtain the optimal design
flange and web slenderness in terms of the buckling performance. A compared to increasing the top flange thickness or decreasing its width
thicker web would be needed to stabilise the thin top flange and transfer (b).
the failure towards the corners. Alternatively, a thicker flange with thin The top flange slenderness was reduced and its stability was
webs can be used to eliminate local buckling in the beam. Increasing the enhanced when the wall thickness was increased, which obtained more
thickness of the flange and web would shift the failure mode at a higher strength across all h/b and r/R ratios, as shown in Fig. 15Fig. 15 (a) and
slenderness ratio compared to increasing the thickness of either the Fig. 15 (b). Alternatively, reducing the top flange width, by increasing
flange or the web. the h/b ratio, resulted in a reduction of its slenderness and an
The corner inner-to-outer radii ratio (r/R) was also nominated for the improvement of its stability and strength, as shown in Fig. 15 (c) for all
flange-web parametric study to investigate its relative contribution and r/R ratios.
interaction to the flange and web slenderness. The selected levels were The three geometric parameters exhibited significant interactions
0.5, 1.0, and 2.0, as shown in Table 6. The chosen values of the flange with each other. The first interaction was between the wall thickness
and web thickness (4.6, 5.2, and 5.8 mm) and r/R ratio were assigned to and h/b ratio, as shown in Fig. 15 (a). Increasing the h/b ratio for a
assess the flange-web junction performance as the failure mode moves thinner profile resulted in a significant enhancement of the flexural
gradually from local buckling to compressive failure of fibres. The layup strength. Whereas for a thicker profile, the effect of the h/b ratio on the
properties (wound fibre angle, axial-to-wound fibres ratio, and stacking flexural strength is diminishing. It was inferred that the effect of the wall

11
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 15. Interaction plots of the geometric parameters affecting the flexural strength of hollow box PFRP beam (a) wall thickness and h/b ratio (b) wall thickness and
corner radius and (c) h/b ratio and corner radius.

thickness on the failure mode was the reason. Local buckling is dominant 1 since they already overcome local buckling with their low flange
at thin walls, which makes the reduction of the flange width due to the slenderness. This configuration does not require large corners (large r/R
increase of h/b ratio more influential on its stability and strength. When ratio) and alleviates the manufacturing defects resulting from it such as
the wall thickness is increased, the failure mode shifts towards fibre wrinkling and resin-rich zones [64].
compressive failure of the top flange at the ultimate material strength, The third interaction was between the h/b ratio and r/R ratio, as
which is independent of the flange width. Wall slenderness of b/t ≤ 16.4 shown in Fig. 15 (c). The effect of the r/R ratio on the flexural strength of
obtains the most stable configuration for hollow square profiles, while a the square profile (h/b = 1) is notable while it is decreasing for the
hollow rectangular profile with flange slenderness of b/t > 16.4 needs rectangular profile (h/b > 1). The hollow square profile is more prone to
h/b ≥ 1.5 to attain its optimal configuration. Both these configurations local buckling of the top flange compared to its counterpart (having the
will exhibit compressive failure of the top flange with the ultimate same cross-sectional area) rectangular profile due to the wider flange
material strength. width. Thus, increasing the r/R ratio influentially increases the flexural
The second interaction was observed between the wall thickness and strength of the hollow square profile compared to the rectangular pro­
the corner radius, as shown in Fig. 15 (b). The flexural strength was file. It is recommended to use r/R ≥ 2.5 for thin square profiles while r/
effectively increased when the r/R ratio was increased for the slender R < 2 can be used for thin rectangular profiles to mitigate manufacturing
flange profile. However, when the wall thickness is increased and the defects resulting from a large r/R ratio. These two configurations will
failure mode moves from local buckling to compressive failure of the top eliminate local buckling of the thin flange (b/t > 16.4) and maximise the
flange, the corner radius effect of the strength decreasing. Similarly to flexural strength of the profile.
the first interaction, the effect of the wall thickness on the failure mode The interaction between local buckling and compression was studied
was the reason. Consequently, it is recommended to manufacture thin by introducing an adimensional parameter (γ) to reflect the combined
PFRP profiles (b/t > 16.4) with r/R ≥ 2.5 to avoid local buckling of the effect of the geometric parameters on these two failure modes. This
top flange and reach the ultimate material compressive strength. parameter was defined as the ratio of the flexural strength of the profile
Contrarily, thicker profiles (b/t ≤ 16.4) can be manufactured with r/R ≤ (σ) to the ultimate material strength of the profile (σ u ) extracted from

12
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Fig. 16. Interaction plot of failure modes controlling the flexural behaviour of hollow box PFRP beam.

FRP profile will obtain a higher buckling capacity up to 2.25 times for
Table 8
the same b/t ratio with an increase in the cross-sectional area up to
Percentile contribution of each flange-web parameter on the flexural behaviour
26.5% (at h/b = 1.5 and r/R = 2). An increase of 55.8% in the cross-
of the hollow box PFRP beam.
sectional area (twice the first value) would be needed if this buckling
Flange-web Parameter Flange-web stiffness Buckling strength capacity is to be achieved by disregarding this interaction and increasing
Flange slenderness (b/tf) 25.65 19.46 the wall thickness only.
Web slenderness (h/tw) 9.52 15.03
Corner radii ratio (r/R) 64.79 61.56
Error (%) 0.04 3.95
6.2. Effect of the flange-web parameters
ANOVA R2 (%) 99.95 92.06

The contribution of each flange-web parameter on the flexural


coupon tests [57]: behaviour of the hollow box PFRP beam is shown in Table 8 that shows
the two-way ANOVA model results. The flange-web stiffness (calculated
γ = σ /σu = f (b/t, h/b, r/R) (1) using the deflection of the corner’s centroid) increased by 16.1% and
The failure mode of the profile is a compressive failure of the top 11.9% per 1 mm increase in the flange and web thickness, respectively.
flange when γ = 1, while it is local buckling when γ < 1. The study on the Despite the significant effect of the flange slenderness and web slen­
geometric parameters was extended to plot γ versus the geometric pa­ derness on the flange-web junction performance, it was found that the
rameters, as shown in Fig. 16. The effect of the interaction of the geo­ dominant parameter was the corner geometry (corner radii ratio). It
metric parameters on the buckling capacity and the failure mode is contributed>60% to the flange-web stiffness and buckling strength. This
significant and should be considered during the design process. The was referred to the higher rigidity attained by the increased area of web-
shape of the buckling curve was changed from concave up at small r/R flange junction, which redistributes the stresses uniformly and alleviates
and h/b ratios to concave down when r/R and h/b ratios were stress concentration at this critical zone. Moreover, increasing the corner
increasing. The failure mode started to shift closer to compressive failure area reduces the unsupported length of the flange and the webs,
of the top flange as the out-of-plane waviness accompanying buckling consequently, increasing their stability. Economically, it would be better
was diminishing at h/b = 1.25 and r/R = 1 design (the red curve) and to invest in a thicker flange with thin webs to eliminate local buckling in
above. Considering larger h/b and r/R ratios when designing the hollow the profile at a lower cost than manufacturing a thin flange with thick
webs to transfer the failure towards the corners. The optimal design for

Fig. 17. Interaction plots of the flange-web parameters affecting the flexural strength of hollow box PFRP beam (a) flange thickness and web thickness and (b) flange
thickness and corner radius.

13
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

stiff corners and stable flange would have the highest investment in the reduce the effective buckling width of the wall, enhance the corners
r/R ratio. restraint, and increase the rigidity transferred between the walls.
The interactive relationship between the flange slenderness and the When compared to the wall thickness, increasing the r/R ratio ob­
web slenderness and corner geometry was studied to characterise its tains a higher buckling capacity at a lower cost (weight).
effect on the profile stability and failure mode. When monitoring the • From the structural design perspective, Fig. 16 represent a vital
failure mode, it was noticed that reducing h/tw and increasing r/R shifts design tool, which shows the importance of the interaction between
the failure mode from local buckling of the top flange to compressive the geometric parameters on the flexural strength and failure modes
failure of the top flange and corners. The interaction between the flange of hollow PFRP profiles. Considering this interaction during the
slenderness and web slenderness is shown in Fig. 17 (a). The enhance­ design stage will obtain more economical and enhanced design
ment in the buckling resistance when the web thickness is increased was configurations up to doubled flexural strength and half cost (weight).
higher for the thin flange compared to the thicker flange. This is because A hollow box profile with b/t = 21.7, h/b = 1.5, and r/R = 2 presents
of the low buckling capacity of the thin flange, which was enhanced the recommended configuration of the geometric parameters to shift
significantly by the redistribution of stresses when the web thickness the failure mode from local buckling to compressive failure of the top
was increased. However, the failure mode shifts towards compressive flange and obtain the ultimate material strength for flexural
failure of fibres when the flange thickness is increased. Thus, increasing applications.
the web thickness will have a lower influence on the high flexural • The flange-web junction plays a major role in resisting buckling of
strength of the profile as buckling is alleviated and a stable configuration top flange and webs. The top corners geometry is the main parameter
is reached. This observation was also noticed for the r/R ratio effect on controlling the rigidity, strength, and failure mode of the flange-web
the flange slenderness but on a larger scale, as shown in Fig. 17 (b), since junction. Taking the interaction between the r/R ratio and the walls
the r/R ratio is the most influencing parameter on the flange-web slenderness ratio into account can provide an enhanced design
junction. Flange slenderness of b/tf ≤ 15.6, web slenderness of 19.2 configuration that will fail at the ultimate material strength. This
≥ h/tw ≥ 16.4, and corner radii ratio of r/R ≥ 2 will provide an configuration was found to consist of wall slenderness ratios of b/tf
economical configuration for the flange-web junction and obtain a sta­ ≤ 15.6 and 19.2 ≥ h/tw ≥ 16.4 along with a corner radii ratio of r/R
ble profile that will fail at its ultimate material strength. ≥ 2.

7. Conclusions CRediT authorship contribution statement

In this research, a numerical investigation was undertaken on the Mohammad Alhawamdeh: Conceptualization, Methodology, Soft­
manufacturing design parameters controlling the flexural behaviour of ware, Data curation, Writing – original draft. Omar Alajarmeh: Su­
hollow box PFRP profiles. A reliable modelling approach was estab­ pervision, Writing - review & editing. Thiru Aravinthan: Supervision,
lished in order to obtain accurate predictions. An adjustment was per­ Writing - review & editing. Tristan Shelley: Supervision, Writing - re­
formed on the incremental approach presented by the authors [16] to view & editing. Peter Schubel: Supervision, Writing - review & editing.
suit flexural loading and reduce the computational cost. A large exper­ Ali Mohammad: Supervision, Writing - review & editing. Xuesen Zeng:
imental program comprising 20 beams was performed to validate this Supervision, Writing - review & editing.
modelling approach under four-point bending. The FEM results closely
agreed with experimental results and had good agreement with the Declaration of Competing Interest
theoretical result. Full factorial parametric studies were performed to
obtain the effect, contribution, and interaction of the manufacturing The authors declare that they have no known competing financial
parameters. The layup parameters (the wound fibre angle, axial-to- interests or personal relationships that could have appeared to influence
wound fibres ratio, and stacking sequence) were found to be insignifi­ the work reported in this paper.
cant compared to the geometric parameters (the wall thickness, cross-
sectional aspect ratio, and the corner radius). The failure map for each Acknowledgements
geometric parameter was generated to assess its applicable levels in the
factorial study. Useful guidelines on the design for manufacturing were The work was funded through a Cooperative Research Centres Pro­
concluded and recommended configurations of the design parameters jects (CRC-P) Grant (CRCPSIX000117). The authors wish to thank all the
were obtained. From this research, the following conclusions were supporters for their finance and providing the required computational
drawn: resources and testing specimens.

• Local buckling of the top flange is a critical triggering failure mode of Appendix A. .
hollow wound PFRP beams. It severely demotes the flexural strength
of the profile and restrains its potentials. Under such instability Table A.1 shows the design of experiment of the full factorial study
condition, the flexural behaviour of these profiles is governed by the on the geometric parameters of the hollow box pultruded FRP beam
geometric parameters of the hollow cross-section. This failure mode along with its numerical results.
can be eliminated and the ultimate strength of the profile can be Table A.2 shows the design of experiment of the full factorial study
attained using either a wall slenderness ratio of b/t ≤ 16.4, a cross- on the flange-web parameters of the hollow box pultruded FRP beam
sectional aspect ratio of 3 > h/b ≥ 1.5, or corner radii ratio of r/R along with its numerical results. The flange-web stiffness was calculated
≥ 2.5. using the deflection of the corner’s centroid.
• The most significant geometric parameter affecting the buckling
strength of hollow PFRP profiles is the r/R ratio. Increasing it will

14
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Table A1
Design matrix and results of the full factorial study on the geometric parameters of the hollow box beam (Wound fibre angle (Deg) = 50, Axial-to-wound fibre ratio (%)
= [82.2/17.8], Stacking sequence = [0/+θ/-θ/0/-θ/+θ/0], R = 10 mm).
Series Wall thickness Cross-sectional aspect ratio (h/ Corner radii ratio (r/ Flexural Stiffness (N/ Flexural Strength Failure mode
(mm) b) R) mm) (MPa)

S-1 4.6 1 0.5 544.4 289.7 Local buckling of top flange


S-2 4.6 1 1 568.5 327.7 Local buckling of top flange
S-3 4.6 1 2 656.9 383.7 Local buckling of top flange
S-4 4.6 1.25 0.5 630.1 335.2 Local buckling of top flange
S-5 4.6 1.25 1 663.3 369.1 Local buckling of top flange
S-6 4.6 1.25 2 773.3 431.5 Compressive failure of top
flange
S-7 4.6 1.5 0.5 700.3 364.9 Local buckling of top flange
S-8 4.6 1.5 1 739.1 396.2 Local buckling of top flange
S-9 4.6 1.5 2 868.7 437.3 Compressive failure of top
flange
S-10 5.2 1 0.5 605.3 328.2 Local buckling of top flange
S-11 5.2 1 1 632.0 379.0 Local buckling of top flange
S-12 5.2 1 2 719.5 406.7 Local buckling of top flange
S-13 5.2 1.25 0.5 705.3 391.1 Local buckling of top flange
S-14 5.2 1.25 1 737.9 409.5 Local buckling of top flange
S-15 5.2 1.25 2 847.1 433.8 Compressive failure of top
flange
S-16 5.2 1.5 0.5 784.6 410.1 Local buckling of top flange
S-17 5.2 1.5 1 822.6 416.6 Local buckling of top flange
S-18 5.2 1.5 2 951.3 428.7 Compressive failure of top
flange
S-19 5.8 1 0.5 665.3 409.4 Local buckling of top flange
S-20 5.8 1 1 693.2 410.2 Local buckling of top flange
S-21 5.8 1 2 779.1 437.1 Compressive failure of top
flange
S-22 5.8 1.25 0.5 778.1 407.6 Local buckling of top flange
S-23 5.8 1.25 1 809.9 411.4 Local buckling of top flange
S-24 5.8 1.25 2 917.1 437.1 Compressive failure of top
flange
S-25 5.8 1.5 0.5 865.9 412.9 Local buckling of top flange
S-26 5.8 1.5 1 903.2 414.9 Local buckling of top flange
S-27 5.8 1.5 2 1029.7 426.9 Compressive failure of top
flange

Table A2
Design matrix and results of the full factorial study on the flange-web parameters of the hollow box beam (Wound fibre angle (Deg) = 50, Axial-to-wound fibre ratio
(%) = [82.2/17.8], Stacking sequence = [0/+θ/-θ/0/-θ/+θ/0], R = 10 mm, h/b = 1).
Series Flange thickness Web thickness Corner radii ratio (r/ Flange-web Stiffness (N/ Flexural Strength Failure mode
(mm) (mm) R) mm) (MPa)

S-1 4.6 4.6 0.5 544.4 290.7 Local buckling of top flange and webs
S-2 4.6 4.6 1 568.1 324.7 Local buckling of top flange and webs
S-3 4.6 4.6 2 659.3 383.3 Compressive failure of top corners
S-4 4.6 5.2 0.5 564.2 317.1 Local buckling of top flange and webs
S-5 4.6 5.2 1 591.8 342.9 Local buckling of top flange and webs
S-6 4.6 5.2 2 683.9 400.1 Compressive failure of top corners
S-7 4.6 5.8 0.5 587.2 334.4 Local buckling of top flange and webs
S-8 4.6 5.8 1 614.9 360.4 Local buckling of top flange and webs
S-9 4.6 5.8 2 707.8 426.5 Compressive failure of top corners
S-10 5.2 4.6 0.5 581.1 320.8 Local buckling of top flange and webs
S-11 5.2 4.6 1 607.9 345.3 Local buckling of top flange and webs
S-12 5.2 4.6 2 696.5 396.2 Compressive failure of top corners
S-13 5.2 5.2 0.5 605.3 328.2 Local buckling of top flange and webs
S-14 5.2 5.2 1 632.0 373.5 Local buckling of top flange and webs
S-15 5.2 5.2 2 719.5 397.0 Compressive failure of top corners
S-16 5.2 5.8 0.5 627.6 366.7 Local buckling of top flange and webs
S-17 5.2 5.8 1 654.6 391.9 Local buckling of top flange and corners
failure
S-18 5.2 5.8 2 744.8 418.3 Compressive failure of top flange and
corners
S-19 5.8 4.6 0.5 620.2 351.4 Local buckling of top flange and webs
S-20 5.8 4.6 1 646.7 377.9 Compressive failure of top flange and
corners
S-21 5.8 4.6 2 732.6 401.4 Compressive failure of top corners
S-22 5.8 5.2 0.5 643.8 369.3 Local buckling of top flange and webs
S-23 5.8 5.2 1 670.0 393.8 Compressive failure of top flange and
corners
(continued on next page)

15
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

Table A2 (continued )
Series Flange thickness Web thickness Corner radii ratio (r/ Flange-web Stiffness (N/ Flexural Strength Failure mode
(mm) (mm) R) mm) (MPa)

S-24 5.8 5.2 2 757.1 410.9 Compressive failure of top corners


S-25 5.8 5.8 0.5 665.3 385.5 Local buckling of top flange and webs
S-26 5.8 5.8 1 692.9 400.1 Compressive failure of top flange and
corners
S-27 5.8 5.8 2 780.6 413.2 Compressive failure of top corners

References [24] Srinivasa CV, Prema Kumar WP, Prathap Kumar MT, Bangar AR, Kumar P,
Rudresh MS. Experimental and numerical studies on buckling of laminated
composite skew plates with circular holes under uniaxial compression. Mech Adv
[1] Gand AK, Chan T-M, Mottram JT. Civil and structural engineering applications,
Mater Struct 2017;24(4):304–17. https://doi.org/10.1080/
recent trends, research and developments on pultruded fiber reinforced polymer
15376494.2016.1142023.
closed sections: a review. Front Struct Civ Eng 2013;7(3):227–44. https://doi.org/
[25] Wagner HNR, Köke H, Dähne S, Niemann S, Hühne C, Khakimova R. Decision tree-
10.1007/s11709-013-0216-8.
based machine learning to optimize the laminate stacking of composite cylinders
[2] Vedernikov A, Safonov A, Tucci F, Carlone P, Akhatov I. Pultruded materials and
for maximum buckling load and minimum imperfection sensitivity. Compos Struct
structures: A review. J Compos Mater 2020;54(26):4081–117. https://doi.org/
2019;220:45–63. https://doi.org/10.1016/j.compstruct.2019.02.103.
10.1177/0021998320922894.
[26] Schreiber P, Mittelstedt C. A holistic approach for local buckling of composite
[3] Al-saadi AU, Aravinthan T, Lokuge W. Effects of fibre orientation and layup on the
laminated beams under compressive load. Arch Appl Mech 2019;89(7):1243–57.
mechanical properties of the pultruded glass fibre reinforced polymer tubes. Eng
https://doi.org/10.1007/s00419-018-1496-1.
Struct 2019;198:109448. https://doi.org/10.1016/j.engstruct.2019.109448.
[27] Ascione L, Berardi VP, Giordano A, Spadea S. Macro-scale analysis of local and
[4] Hizam RM, Manalo AC, Karunasena W, Bai Y. Behaviour of pultruded GFRP truss
global buckling behavior of T and C composite sections. Mech Res Commun 2014;
system connected using through-bolt with mechanical insert. Compos Part B Eng
58:105–11. https://doi.org/10.1016/j.mechrescom.2013.10.004.
2019;168:44–57. https://doi.org/10.1016/j.compositesb.2018.12.052.
[28] Liu T, Vieira JD, Harries KA. Predicting Flange Local Buckling Capacity of
[5] Muttashar M, Karunasena W, Manalo A, Lokuge W. Behaviour of hollow pultruded
Pultruded GFRP I-Sections Subject to Flexure. J Compos Constr 2020;24(4):
GFRP square beams with different shear span-to-depth ratios. J Compos Mater
04020025. https://doi.org/10.1061/(ASCE)CC.1943-5614.0001032.
2016;50(21):2925–40. https://doi.org/10.1177/0021998315614993.
[29] Cardoso DCT, Harries KA, de Batista E, M.. Closed-form equations for compressive
[6] Sayyad AS, Ghugal YM. Bending, buckling and free vibration of laminated
local buckling of pultruded thin-walled sections. Thin-Walled Struct 2014;79:
composite and sandwich beams: A critical review of literature. Compos Struct
16–22. https://doi.org/10.1016/j.tws.2014.01.013.
2017;171:486–504. https://doi.org/10.1016/j.compstruct.2017.03.053.
[30] Tang Jing, Chen Xiao, Yang Ke. Evaluating Structural Failure of Load-Carrying
[7] Awad ZK, Aravinthan T, Zhuge Y, Gonzalez F. A review of optimization techniques
Composite Box Beams with Different Geometries and Load Conditions. Appl
used in the design of fibre composite structures for civil engineering applications.
Compos Mater 2019;26(4):1151–61. https://doi.org/10.1007/s10443-019-09776-
Mater Des 2012;33:534–44. https://doi.org/10.1016/j.matdes.2011.04.061.
4.
[8] Correia JR, Branco FA, Silva NMF, Camotim D, Silvestre N. First-order, buckling
[31] Bai Yu, Keller Thomas, Wu Chao. Pre-buckling and post-buckling failure at web-
and post-buckling behaviour of GFRP pultruded beams. Part 1: Experimental study.
flange junction of pultruded GFRP beams. Mater Struct 2013;46(7):1143–54.
Comput Struct 2011;89(21-22):2052–64. https://doi.org/10.1016/j.
https://doi.org/10.1617/s11527-012-9960-9.
compstruc.2011.07.005.
[32] Liu T, Harries KA. Flange local buckling of pultruded GFRP box beams. Compos
[9] Department of Defense. Composite materials handbook, volume 3: polymer matrix
Struct 2018;189:463–72. https://doi.org/10.1016/j.compstruct.2018.01.101.
composites—materials usage, design and analysis. SAE International; 2002.
[33] Mosallam AS, Feo L, Elsadek A, Pul S, Penna R. Structural evaluation of axial and
[10] Uddin N. Developments in Fiber-Reinforced Polymer (FRP) Composites for Civil
rotational flexibility and strength of web–flange junctions of open-web pultruded
Engineering. Elsevier; 2013.
composites. Compos Part B Eng 2014;66:311–27. https://doi.org/10.1016/j.
[11] Xu J, Zhao Q, Qiao P. A critical review on buckling and post-buckling analysis of
compositesb.2014.05.018.
composite structures. Front Aerosp Eng 2013;2:157–68.
[34] Bank Lawrence C, Gentry T Russell, Nadipelli Murali. Local Buckling of Pultruded
[12] American Society of Civil Engineers. Pre-standard for load & resistance factor
FRP Beams-Analysis and Design. J Reinf Plast Compos 1996;15(3):283–94. https://
design (LRFD) of pultruded fiber reinforced polymer (FRP) structures. Arlingt VA
doi.org/10.1177/073168449601500304.
ACMA 2012:215.
[35] Aleksendric D, Carlone P. Soft Computing in the Design and Manufacturing of
[13] Ascione L, Caron J-F, Godonou P, van Ijselmuijden K, Knippers J, Mottram T, et al.
Composite Materials: Applications to Brake Friction and Thermoset Matrix
Prospect for new guidance in the design of FRP : Support to the implementation.
Composites. Woodhead Publishing; 2015.
harmonization and further development of the Eurocodes. Publications Office of
[36] Attaf B. Advances in Composite Materials: Ecodesign and Analysis. BoD – Books on
the European Union; 2016.
Demand 2011.
[14] Bank LC. Composites for Construction: Structural Design with FRP Materials. John
[37] Barbero EJ. Introduction to Composite Materials Design. CRC Press 2017. https://
Wiley & Sons; 2006.
doi.org/10.1201/9781315296494.
[15] Kollár LP, Springer GS, editors. Mechanics of Composite Structures. Cambridge
[38] Buragohain MK. Composite Structures: Design, Mechanics, Analysis,
University Press; 2003.
Manufacturing, and Testing. CRC Press; 2017.
[16] Alhawamdeh M, Alajarmeh O, Aravinthan T, Shelley T, Schubel P, Kemp M, et al.
[39] Gibson RF. Principles of Composite Material Mechanics. CRC Press; 2016.
Modelling hollow pultruded FRP profiles under axial compression: Local buckling
[40] Mallick PK. Fiber-reinforced composites: materials, manufacturing, and design.
and progressive failure. Compos Struct 2021;262:113650. https://doi.org/
CRC Press; 2007.
10.1016/j.compstruct.2021.113650.
[41] Matthews FL, Davies GAO, Hitchings D, Soutis C. Finite Element Modelling of
[17] Kalantari M, Dong C, Davies IJ. Effect of matrix voids, fibre misalignment and
Composite Materials and Structures. Elsevier; 2000.
thickness variation on multi-objective robust optimization of carbon/glass fibre-
[42] Najafi A, Rais-Rohani M. Concurrent Process-Product Design Optimization Using
reinforced hybrid composites under flexural loading. Compos Part B Eng 2017;123:
Coupled Nonlinear Finite-Element Simulations. Finite Elem Anal - Appl Mech Eng
136–47. https://doi.org/10.1016/j.compositesb.2017.05.022.
2012. https://doi.org/10.5772/47852.
[18] Nikbakt S, Kamarian S, Shakeri M. A review on optimization of composite
[43] Rammerstorfer FG. Engineering mechanics of fibre reinforced polymersand
structures Part I: Laminated composites. Compos Struct 2018;195:158–85. https://
composite structures. Springer; 1994.
doi.org/10.1016/j.compstruct.2018.03.063.
[44] Soares CAM, Soares CMM, Freitas MJM. Mechanics of Composite Materials and
[19] Betts D, Sadeghian P, Fam A. Investigation of the stress-strain constitutive behavior
Structures. Springer Science & Business Media; 2013.
of ±55◦ filament wound GFRP pipes in compression and tension. Compos Part B
[45] Vasiliev VV, Morozov EV. Advanced Mechanics of Composite Materials and
Eng 2019;172:243–52. https://doi.org/10.1016/j.compositesb.2019.05.077.
Structures. Elsevier; 2018.
[20] Akbaş ŞD. Post-buckling analysis of a fiber reinforced composite beam with crack.
[46] Abouzied A, Masmoudi R. Flexural behavior of rectangular FRP-tubes filled with
Eng Fract Mech 2019;212:70–80. https://doi.org/10.1016/j.
reinforced concrete: Experimental and theoretical studies. Eng Struct 2017;133:
engfracmech.2019.03.007.
59–73. https://doi.org/10.1016/j.engstruct.2016.12.010.
[21] Asadi A, Sheikh AH, Thomsen OT. Buckling behaviour of thin-walled laminated
[47] ASTM D7249/D7249M. Standard test method for facing properties of sandwich
composite beams having open and closed sections subjected to axial and end
constructions by long beam flexure. Am Soc Test Mater 2012.
moment loading. Thin-Walled Struct 2019;141:85–96. https://doi.org/10.1016/j.
[48] Cole Britton, Fam Amir. Flexural Load Testing of Concrete-Filled FRP Tubes with
tws.2019.04.005.
Longitudinal Steel and FRP Rebar. J Compos Constr 2006;10(2):161–71. https://
[22] Nguyen N-D, Nguyen T-K, Vo TP, Nguyen T-N, Lee S. Vibration and buckling
doi.org/10.1061/(ASCE)1090-0268(2006)10:2(161).
behaviours of thin-walled composite and functionally graded sandwich I-beams.
[49] Muttashar M, Manalo A, Karunasena W, Lokuge W. Influence of infill concrete
Compos Part B Eng 2019;166:414–27. https://doi.org/10.1016/j.
strength on the flexural behaviour of pultruded GFRP square beams. Compos Struct
compositesb.2019.02.033.
2016;145:58–67. https://doi.org/10.1016/j.compstruct.2016.02.071.
[23] Garrido M, Madeira JFA, Proença M, Correia JR. Multi-objective optimization of
[50] Simulia DS. Abaqus 6.13 Analysis User’s Guide. Dassault Syst Provid RI 2013.
pultruded composite sandwich panels for building floor rehabilitation. Constr
[51] Al-Saadi A, Aravinthan T, Lokuge W. Numerical Investigation on Hollow Pultruded
Build Mater 2019;198:465–78.
Fibre Reinforced Polymer Tube Columns. ACMSM25, Springer 2020:455–65.

16
M. Alhawamdeh et al. Composite Structures 276 (2021) 114553

[52] Daniel IM, Ishai O. Engineering Mechanics of Composite Materials. Oxford [59] Godat A, Légeron F, Gagné V, Marmion B. Use of FRP pultruded members for
University Press; 2006. electricity transmission towers. Compos Struct 2013;105:408–21. https://doi.org/
[53] Sharma AP, Khan SH, Parameswaran V. Experimental and numerical investigation 10.1016/j.compstruct.2013.05.025.
on the uni-axial tensile response and failure of fiber metal laminates. Compos Part [60] Meyer R. Handbook of Pultrusion Technology. Springer Science & Business. Media
B Eng 2017;125:259–74. https://doi.org/10.1016/j.compositesb.2017.05.072. 2012.
[54] Regel F. A modelling approach for 3D braid reinforced composites under non-axial [61] Starr T. Pultrusion for Engineers. Elsevier; 2000.
loading. PhD Thesis. University of Minho; 2014. [62] Minitab-18. Minitab Reference Manual: statistical software. Minitab Inc. State
[55] Antony J. Design of Experiments for Engineers and Scientists. 2nd Edition. Elsevier; College, PA; 2017.
2014. [63] Estep DD, GangaRao HVS, Dittenber DB, Qureshi MA. Response of pultruded glass
[56] Reddy JN. Mechanics of Laminated Composite Plates and Shells: Theory and composite box beams under bending and shear. Compos Part B Eng 2016;88:
Analysis. Second Edition. CRC Press; 2003. 150–61. https://doi.org/10.1016/j.compositesb.2015.11.008.
[57] Wagners CFT. Wagners-Composite Fibre Technologies (CFT). Wagner n.d. https:// [64] Quadrino A, Penna R, Feo L, Nisticò N. Mechanical characterization of pultruded
www.wagner.com.au/main/what-we-do/composite-fibre-technologies/cft-home/ elements: Fiber orientation influence vs web-flange junction local problem.
(accessed March 8, 2020). Experimental and numerical tests. Compos Part B Eng 2018;142:68–84. https://
[58] Clarke JL. Structural Design of Polymer Composites EUROCOMP Design Code and doi.org/10.1016/j.compositesb.2018.01.001.
Handbook. CRC Press; 2005.

17

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy