Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Thin-Walled Structures 168 (2021) 108300

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Compressive behaviour of hollow box pultruded FRP columns with


continuous-wound fibres
Omar Alajarmeh a,b ,∗, Xuesen Zeng a , Thiru Aravinthan a , Tristan Shelley a ,
Mohammad Alhawamdeh a , Ali Mohammed c , Lachlan Nicol c , Alexander Vedernikov d ,
Alexander Safonov d , Peter Schubel a
a
Centre for Future Materials, University of Southern Queensland, Toowoomba Queensland, 4350, Australia
b
Department of Civil Engineering, Tafila Technical University, Tafilah, 66110, Jordan
c
Wagners Composite Fibre Technologies, Wellcamp Queensland, 4350, Australia
d
Center for Design, Manufacturing and Materials, Skolkovo Institute of Science and Technology, Moscow, 121205, Russia

ARTICLE INFO ABSTRACT


Keywords: Pultruded Fibre Reinforced Polymer (PFRP) composite profiles now have been utilized in many Civil and
Compressive strength infrastructure applications due to their low density and outstanding mechanical properties. However, different
Buckling manufacturing techniques have been used to produce these PFRP profiles, and these have an influence on the
Delamination
structural performance of the PFRP profiles. This study investigated the compressive behaviour of the PFRP
DIC
profiles manufactured using the Winder technology, allowing continuous-wound fibres to reinforce the PFRP
Pultruded composite
Pulwinding
profiles transversely. Three different cross-sections and three levels of Length-to-Depth (𝐿∕𝐷) ratios have been
tested to study the effect of these parameters on the compressive strength of the PFRP profiles. It was observed
that continuous-wound fibres led to progressive failure and prevented the splitting failure at corners. Test
results showed that rectangular profiles performed structurally better than the square profiles. Moreover, the
increase in 𝐿∕𝐷 ratio decreased both stiffness and strength but led to higher displacement buckling capacity.
It was observed that the buckling waves were governed by a polynomial function. Finally, the existing models
were examined in relation to the experimental results where the general plate theory was valid to present the
local buckling behaviour of the tested PFRP profiles.

1. Introduction low elastic properties as well as the thinner wall thickness of the man-
ufactured FPRP profiles, however, they suffer from instability problems
Increasingly, fibre reinforced polymer (FRP) composites have been before even reaching their full strength capacity, where buckling is
used widely in Civil Engineering and infrastructure applications as an shown to be the main feature of their instability.
efficient alternative to traditional construction materials. The extensive Unlike isotropic materials, with an instability governed only by
adoption of these materials is the result of their promising performance dimensions [5], composites are also affected by their anisotropic prop-
including high strength-to-weight ratio, durability, and low life-cycle
erties which can result in combined instability and fracture, adding
maintenance. Pultrusion technology is one of the most efficient man-
to the complexity of the buckling behaviour of composites [6]. Xu
ufacturing processes in the production of various thin-walled FRP
et al. [7] reviewed the possible buckling modes of composites which
composite profiles, as it is an economic, fast, and continuous process.
can occur in addition to the ordinary buckling mode that shows a grad-
Pultrusion is capable to produce various cross-section profiles with
high fibre volume fractions (typically higher than 60%) compared to ual non-linear degradation in stiffness under loading. They considered
other processes, which results in high mechanical properties [1]. In that delamination-driven buckling, initial geometrical imperfection,
this manufacturing process, the majority of fibres (mostly glass fibres) and thermomechanical buckling are other problems that composites
are aligned with the axial direction while various off-axis fibre patterns face as unstable modes of failure. Unlike the latter two modes, which
are placed through the thickness to hold the axial fibres in place and are not failures based on laminate properties, delamination-driven
provide transverse mechanical properties. Because of this layup, the buckling is considered to be the dominant failure mode of composites,
performances of the pultruded FRP composite (PFRP) profiles vary and unless they have the very low elastic properties required to initiate
are governed by their laminate properties [2–4]. Due to their inherent the ordinary buckling mode, as mentioned earlier [8,9]. These modes

∗ Correspondence to: University of Southern Queensland, West Street, Toowoomba Qld, 4350, Australia.
E-mail address: omar.alajarmeh@usq.edu.au (O. Alajarmeh).

https://doi.org/10.1016/j.tws.2021.108300
Received 25 February 2021; Received in revised form 30 June 2021; Accepted 11 August 2021
Available online 25 August 2021
0263-8231/© 2021 Elsevier Ltd. All rights reserved.
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Table 1
Stacking sequence and laminate structure of the PFRP profiles.
Profile Stacking sequence (Outside to inside) Fibre content (%) Volume fibre fraction (𝑉𝑓 ) (%) Cross-sectional area (mm2 )
[ ]
a a
R75 0, ∓47.5, 0 79.4 (0.13) 60.5 (0.19) 1585.6
𝑠
[ ] [0◦ : 80.0%/50◦ : 20.0%]
S100 0, ∓50, 0 78.5 (0.57)a 59.1 (0.81)a 1923.3
𝑠
[ ] [0◦ : 82.2%/50◦ : 17.8%]
S125 0, +50, 0, −50, 0 79.0 (0.67)a 59.8 (0.98)a 2961.4
𝑠
[0◦ : 78.1%/50◦ : 21.9%]
a
The number between parenthesis is the standard deviation.

of failure appeared in the small-scale specimens under compression the differences in the mechanical performance of the tested box PFRP
and bending loadings [9,10], which were found to affect strength and profile previously due to the vast differences in their laminate proper-
stiffness at some loading stages. ties. Nevertheless, some parameters, including layup and geometry can
In the full-scale PFRP profiles, various modes of failure were ob- be highlighted as significantly affecting the structural performance of
served, mainly based on the layup and/or the geometry [11,12]. For these profiles.
instance, Al-Saadi et al. [11] tested short square PFRP profiles with Recently, the Winder technology has been integrated with pultru-
the same geometry and length but different layup. One specimen had
sion, which allows continuous inclined (wound) fibres as off-axis fibre
continuous-inclined fibres and unidirectional axial fibres, while the
layers to improve the structural performance of the PFRP profiles from
other had mat and non-continuous inclined fibres in addition to the
the highly efficient and consistent laminate layups [16]. To the authors’
unidirectional axial fibres. A significant difference in the mode of
knowledge, no previous study has investigated in depth the behaviour
failure was observed wherein the first specimen showed linear elastic
of PRFP profiles with continuous-wound fibres under compression.
behaviour until a load drop caused by localized failure at the corner
which initiated local buckling behaviour afterwards. However, the Therefore, this study aims to investigate experimentally the effect of
cross-section maintained the load resistance due to the existence of geometry and length on the behaviour of the box PFRP profiles with
the continuous-wound fibres until the final failure, which was denoted continuous-wound fibres. In total, 45 full-scale specimens have been
by severe delamination and fracture at corners. The second specimen tested considering three geometries: one rectangular and two square
showed also linear elastic behaviour until failure, at which major profiles with three different 𝐿∕𝐷 ratios to quantify and evaluate the
longitudinal cracks at the corners were observed. Even with lower load–displacement, strain behaviour, and failure behaviour of this type
wall slenderness, the second specimen recorded 8% and 12% lower of PFRP profiles. The test results are also compared to the available
peak strength/area and modulus, respectively. This is due to the layup theoretical models to assess the reliability of those existing models in
difference that led to lower the rotational stiffness at the corners predicting the behaviour of such PRFP profiles. The outcome of this
resulting in early splitting failure as observed by the final mode of study will provide a better understanding of the behaviour of PFRP
failure. Furthermore, Liu and Harries [12] tested pultruded GFRP box- profiles with continuous-wound fibres, and will generate additional
section beams with various thicknesses to investigate the flange local results, which will contribute to the development of a design guideline
buckling failure under bending. They confirms that the mode of failure for composite manufacturers.
is highly depending on the layup and geometry of the cross-section
which significantly affected the rotational stiffness at the web-flange
junction. This also highlights the significant role of the continuous- 2. Experimental program
inclined fibre in adding more stability to the cross-section. Cardoso
et al. [13] tested square PFRP profiles with various cross-section and
2.1. Material
layup to investigate the effect of wall slenderness and length on the
local buckling behaviour of the tested specimens. Results showed a
gradual decrease in the stiffness in the load–displacement plot indi- This study includes three box PFRP profiles with different cross-
cating an occurrence of local buckling waves at the walls of the PFRP sections provided by Wagners Composite Fibre Technology (WCFT)-
profile. Moreover, this reduction in stiffness occurred earlier when the Australia (see Fig. 1). These profiles were manufactured by pulling
wall slenderness increased, resulting in lower local buckling stress. The the unidirectional and wound EC-glass fibres impregnated in vinyl-
increase in length was found to decrease the local buckling stress as an ester resin through a heated die to obtain the final solid product with
increase in the local buckling waves was observed. At a certain length, a density around 2 g/cm3 . The layup of the cross-section consists of
the specimens started to show out-of-axis deformation indicating the unidirectional and ±50◦ continuous wound glass fibres, which provides
global buckling behaviour. It can be observing that the tested profiles the novelty of this manufacturing process. Table 1 reports the stacking
did not show equal normalized ultimate compression load even with sequence, fibre content and fibre volume fraction of the PFRP profiles.
similar slenderness ratio. However, it should be mentioned that the
These data were found by performing the burn-out test in accordance
elastic modulus of the tested profiles was in the range of 22–30 GPa,
to ISO 1172 [17].
which might be responsible for this finding. On the other hand, Godat
et al. [14] tested square and rectangular profiles with the same layup
and length under compression. They observed that the square profile 2.2. Material Characterization
failed in a brittle and sudden way due to a propagated crack initiated
at the corner. At the same time, the rectangular profile, which had a
more slender wall, failed in local buckling as demonstrated by a gradual Characterizing the mechanical properties of the PFRP profiles was
decrease in stiffness according to the load–displacement plot. Even with performed by conducting compression, tensile, in-plane shear and inter-
the 7% larger cross-sectional area of the rectangular profile, it recorded laminar shear of the small-scale coupon specimens extracted from the
46% lesser strength due to the local buckling failure. Moreover, It was PFRP profiles using a water-jet machine. Fig. 2 shows the setup of
observed that increasing the thickness of the wall in the box profiles the various coupon tests using an MTS machine with 100 kN capacity,
reduced significantly the flange-web rotation and add more stability while Table 2 lists the test results of 10 coupon replicates for each test
against local buckling failure [15]. To conclude, it is hard to quantify configuration and the relevant standards.

2
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Fig. 1. The cross-section of the PFRP profiles.

Fig. 2. The conducted coupon tests.

Table 2
Material characterization based on the coupon tests.
Property Standard Unit R75 S100 S125
Average St.D. Average St.D. Average St.D.
Strength MPa 521.1 24.2 496.2 11.0 515.1 15.2
Compression ISO 527-4
Modulus GPa 48.1 0.71 50.9 1.5 49.1 2.4
Strength MPa 918.0 29.9 922.1 38.3 857.1 34.9
Longitudinal Modulus GPa 42.8 0.9 45.1 1.1 43.2 1.1
Poisson’s ratio ASTM – 0.31 0.09 0.32 0.02 0.30 0.02
Tensile
Strength D6641 MPa 46.8 3.9 67.5 4.0 66.4 3.8
Transverse Modulus GPa 12.2 1.1 12.3 0.6 13.3 0.6
Poisson’s ratio – 0.15 0.03 0.13 0.02 0.15 0.01
Strength ASTM MPa 113.2 4.8 101.0 6.5 124.2 3.3
In-plane shear
Modulus D5379 GPa 5.42 0.3 4.70 0.6 6.71 0.4
Interlaminar shear Strength ASTM MPa 62.2 2.1 59.0 1.3 50.3 1.8
D2344

2.3. Full-scale test design Table 3


Design matrix of the tested specimens.
R75 S100 S125
Testing the 45 full-scale PFRP profiles in this study was designed 𝐿∕𝐷 ratio 2.0 3.5 5.0 2.0 3.5 5.0 2.0 3.5 4.5a
to examine the effect of geometry and the length of the short PFRP Length 200 350 500 200 350 500 200 350 560
Replicates 5 specimens for each set
columns. Therefore, three cross-sections (R75, S100 and S125) were
a The maximum 𝐿∕𝐷 ratio of the S125 was 4.5 due to the height limitation of the
selected to find the effect of the geometry. These profiles are the current
testing machine.
box profiles being pultruded using the same technology as Wagners.
Three length-to-depth (𝐿∕𝐷) ratios (2.0, 3.5, and 5.0), reflecting the
change within the short-range length, were adopted to address the
2.4. Special test fixtures
effect of the length on the short PFRP columns. The selected range of
𝐿∕𝐷 ratio was found to be able to accurately represent the behaviour
of the short PFRP columns and to avoid the effect of the global Identifying the accurate structural behaviour of the PFRP profiles
buckling [11,13,18]. Five specimens of each set of PFRP were tested requires a knowledge of the possible modes of failure and their corre-
for the validity and repeatability of the test. Table 3 shows the testing sponding strength capacities for precise and reliable design. Unfortu-
matrix of the specimens. nately, testing the PFRP profiles under compression has not yet been

3
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Fig. 3. End-fixtures.

standardized and there are no special guidelines on testing and consid- end rotation indicating a pinned-end connection. This was considered
ering the effect of boundary conditions. Most of the previous studies to test the real behaviour of the PFRP columns in this study and allowed
have performed direct compression on the PFRP profiles without any for any possible global buckling to appear. Moreover, the Digital Image
special consideration for end-fixtures. Thus, some of the studies [11,19] Correlation (DIC) instrument was used to trace the change in the
showed crushing at the end of the PFRP profiles, which indicates some distance between speckles printed at the wall surface to observe the
uncertainty as to the true failure behaviour due to the stress concen- entire behaviour of the tested specimens. It should be mentioned that
tration at the loading points, especially for the short PFRP profiles. the resolution of the camera is 2048 × 1088 pixels (2.22 megapixels).
However, other studies have successfully adopted special end-fixtures The camera is able to record a video with 1000 frames/s, and the
to overcome stress-concentration issues [20–23]. In this study, there- data acquisition frequency by default is 50 Hz. The load and vertical
fore, special steel end-fixtures have been designed and manufactured displacement were captured by the acquisition system of the SANS
to accommodate the box PFRP profiles (accounting for the possible machine coupled with the system 5000 data logger system to record
tolerance in the thickness) and to ease transfer the end stresses to the the strain values during the test. Testing videos were recorded for
testing zone, assuring precise modes of failure. Moreover, these fixtures post-processing the test in order to capture critical observations.
have been designed in a way to consider the variation in the PFRP
profile thickness. Therefore, the profile thickness can accommodate up 3. Test results
to ±0.2 mm deviation. Fig. 3 shows the drawings and the design of the
end-fixtures used in this study. Furthermore, the way of assembling this 3.1. Load–Displacement Behaviour
end-fixture is by compiling steel segments (with 20 mm height) and bolt
them in a base plate through proper drilled holes (Fig. 3.d). Fig. 5.a shows the typical load–displacement behaviour of the tested
specimens, highlighting the critical points, which differ from specimen
2.5. Full-scale test setup and instrumentation to another. Point A indicates the start of loading where the load–
displacement ascends in a linear behaviour until reaching the first axial
Testing the full-scale PFRP specimens was carried out by using a peak load (𝑁1 ) at point B. The slope of the linear load–displacement
SANS machine with a 2MN compression capacity with a loading rate of behaviour represents the axial stiffness of the tested specimen. It should
1 mm/min. The test was conducted under concentric compression until be mentioned that the DIC (as explained in the next sections) shows
failure. Prior to the test, two strain gauges with a length of 20 mm were no signs of buckling in the walls up to point B. At that level of stress,
mounted (lengthwise and crosswise directions) on the surface of two the cross-section was stable due to the corner constraints and the high
columns from each set at mid-height, to measure the elastic modulus transverse stiffness provided by the continuous-wound fibres. However,
and capture the local buckling of the walls of the PFRP profiles (see a sudden decay in the axial load resistance accompanied simultaneously
Fig. 4). It is worth mentioning that the loading platen allowed some with local buckling in the walls was observed in the zone between

4
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

highlighted the role of the wound fibres in showing high hoop stresses
to maintain structural integrity. It should be mentioned that the general
behaviour between point C and D is a brief non-linear ascending
behaviour (see Fig. 5.a) ending in the final failure.
Fig. 5.b to d present the typical load–displacement behaviour of
the various tested cross-sections with different 𝐿∕𝐷 ratio. Table 4
summarizes the results of the tested specimens including axial stiffness,
first axial peak load (𝑁1 ), second axial peak load (𝑁2 ), stress at 𝑁1 (𝜎1 ),
stress at 𝑁2 (𝜎2 ), toughness at 𝑁1 (𝑇𝑁1 ) and toughness at 𝑁2 (𝑇𝑁2 ). It is
worth mentioning that 𝜎1 and 𝜎2 are the result of dividing 𝑁1 and 𝑁2 by
the corresponding area of the various cross-sections, respectively (see
Table 1). Moreover, toughness is the area under the load–displacement
curve, which represents the stored energy in the system at any point.
Referring to Table 4 and Fig. 5, results show the clear influence of the
geometry and the 𝐿∕𝐷 ratio on the axial load and strength capacities.
Generally, the R75 profile shows higher 𝜎1 and 𝜎2 compared to the S100
and S125 profiles. Moreover, the various PFRP profiles reveal lower
axial resistance with the higher 𝐿∕𝐷 ratio. Load and strength results
show high accuracy and repeatability referring to the small coefficient
of variance values (CoV) by less than 6%. In addition, toughness results
show a slightly higher CoV than the load and strength values by a
threshold of 10% but still within the acceptable range of less than
15%. This increase in the CoV might be due to the various failure
progression of the tested specimen, which affects the area under the
Fig. 4. Compression test set-up of the PFRP profiles.
load–displacement curve.

3.2. Mode of failure


B and C. At point C, the tested profiles maintained 80% to 95% of
𝑁1 due to its unique manufacturing process using continuous-wound All tested columns showed similar and consistent final modes of
fibres resulted in delaying the failure caused by local buckling through failure by severe delamination between the layers and a final shear
showing a progressive failure until reaching point D. This behaviour failure at the mid-height where the biggest out-of-plane deformation

Fig. 5. The typical load–displacement behaviour of the tested specimens.

5
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Table 4
Test results of the tested specimens.
Profile 𝐿∕𝐷 ratio Specimen Stiffness (kN/mm) 𝑁1 (kN) 𝜎1 (MPa) 𝑇𝑃1 (kN mm) 𝑁2 (kN) 𝜎2 (MPa) 𝑇𝑃2 (kN mm)
Average 285.1 557.2 351.4 544.7 521.1 328.6 764.1
2.0 St.D. 5.3 15.7 9.9 26.0 15.0 9.5 62.4
CoV (%) 1.8 2.8 4.8 2.9 8.2
Average 180.7 474.5 299.3 624.6 441.5 278.4 981.2
R75
3.5 St.D. 6.0 12.9 8.1 43.5 14.1 8.9 23.5
CoV (%) 3.3 2.7 7.0 3.2 2.4
Average 142.2 492.0 310.3 854.3 430.4 271.5 1294.6
5.0 St.D. 5.9 27.9 17.6 92.7 8.7 5.5 62.5
CoV (%) 4.1 5.7 9.9 2.0 4.8
Average 329.1 569.8 296.3 495.4 523.2 272.0 726.9
2.0 St.D. 27.9 25.2 13.1 34.9 11.6 6.0 81.0
CoV (%) 8.5 4.4 7.0 2.2 11.1
Average 215.1 510.8 265.6 606.8 442.3 230.0 882.8
S100
3.5 St.D. 7.7 8.7 4.5 15.7 10.5 5.4 48.8
CoV (%) 3.6 1.7 2.6 2.4 5.5
Average 159.0 487.3 253.4 747.1 408.9 212.6 1031.7
5.0 St.D. 4.5 8.8 4.6 39.3 7.1 3.7 88.2
CoV (%) 2.8 1.8 5.3 1.7 8.5
Average 393.2 1001.7 338.2 1277.3 935.2 315.8 1648.3
2.0 St.D. 11.7 27.5 9.3 64.5 4.8 1.6 98.4
CoV (%) 3.0 2.7 5.0 0.5 6.0
Average 263.6 908.2 306.7 1565.3 783.6 264.6 1701.1
S125
3.5 St.D. 5.3 11.7 3.9 51.3 2.9 1.0 102.5
CoV (%) 2.0 1.3 3.3 0.4 6.0
Average 221.7 831.8 280.9 1561.5 723.7 244.4 2120.3
4.5 St.D. 6.0 24.5 8.3 70.0 16.3 5.5 102.9
CoV (%) 2.7 2.9 4.5 2.3 4.9

Fig. 6. The failure progression in the tested PFRP profiles.

Fig. 7. Microscopic investigation of the cross-section failure.

(local buckling half-wave) locates. This final failure followed a sym- Fig. 6.a). Just afterwards, the local buckling shape was noticed on the
metrical buckling shape in the walls. Describing the failure progression, walls accompanied with an increase in the previously heard audible
it is worth mentioning that the failure initiated at point B (Fig. 5.a) and unsighted cracks, as shown in Fig. 6.b. With this observation,
by hearing audible cracks without, however, any failure feature (see the load resistance dropped until point C (see Fig. 5.a). Laminated

6
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Fig. 8. Typical stress–strain behaviour of the tested specimens.

Fig. 9. Comparing the DIC results with the strain gauges and the SANS machine ones (S100 with 𝐿∕𝐷 = 2.0).

composites in general are known for their low interlaminar strength along the corners (but with partial restraint due to the presence of
capacity as it is a resin dominant property. Moreover, the failure of the wound fibres) which leads to show local buckling. In the tested
the thick wall composites is strongly influenced by material strength. columns, similar phenomena had occurred where it is believed that
Thus, when the thick composite wall tends to buckle or laterally deform delamination (interlaminar crack) was initiated at the corners by which
because of the Poisson’s ratio effect, it initiates interlaminar cracks the rotational resistance capacity was reduced. Accordingly, the wall
at the junctions (corners), as they are the most stressed zones in the becomes vulnerable to rotation and show buckling. This finding was
box cross-section. As a result, the walls behave as they were hinged supported by observing micro-cracks at the corners of the PFRP profiles

7
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

after stopping the loading machine once noticing the load drop (at point Table 5
Modulus of elastic of the full-scale PFRP profiles.
C in Fig. 5.a) (see Fig. 7.a). This phenomenon is called delamination-
driven buckling [7,9], which is common in laminated composites and 𝐿∕𝐷 ratio R75 S100 S125

indicates the toughness of the laminate [24]. It is worth highlighting 1 2 1 2 1 2


that this unique initiation of failure has not been observed before in the 2.0 55 409 56 128 52 204 50 901 54 650 55 394
PFRP profiles under compression because, to the author’s knowledge, 3.5 56 057 56 720 51 053 49 901 55 679 56 017
5.0 57 876 60 570 55 045 52 652 56 624 56 780
no study has previously considered testing these profiles with the
inclusion of continuous-wound fibres. Average 57 127 51 959 55 857
St.D. 1716 1645 727
After point C, the same shape of the local buckling was maintained
CoV (%) 3.0 3.2 1.3
between points C and D (Fig. 5.a) which presents the failure progression
stage. It is worth mentioning that the sole difference in the local
buckling behaviour between the tested columns was in the distribution
of the buckling waves along the length of the wall in the progressive It is worth mentioning that a sound agreement was noticed between
failure stage (between point C and D). Moreover, it was noticed that the axial displacement recorded by the DIC and the one extracted by
the increase in the length of the wall by increasing the 𝐿∕𝐷 ratio the displacement sensor of the SANS machine (see Fig. 9.a). Moreover,
showed more local buckling waves. However, the failure was localized the strain readings recorded by the gauges matched well with the DIC
in one wave while the other formed waves disappeared after the final readings. This finding can be used to validate the modulus of elasticity,
failure due to the release of energy (Fig. 6.c). Fig. 7.b shows the cross- peak load, and strain at peak load, as shown in Fig. 9.b. According
section after the final failure, where severe delamination between the to Fig. 9.a, DIC results confirmed the observations mentioned in the
layers and fibre debonding in the unidirectional layers can be observed. ‘‘Load–Displacement behaviour’’ section for the tested PFRP profiles
As a conclusion, due to the delamination-driven buckling, the critical including the stages described in Fig. 5.a. On the other hand, this
buckling stress of the tested PFRP profiles is considered to be the stress instrument was used to measure the generated waves and their length
at point D (𝜎2 ). as described in the next section.

3.3. Stress–Strain Behaviour 3.4.2. Wavelength observations


DIC provides an option to trace a 2D strain map setup on the
Fig. 8 presents the typical stress–strain behaviour of the tested wall perpendicular to the camera line. This option allows marking
PFRP profiles with different geometries and 𝐿∕𝐷 ratios. All of the the wrinkled waves and their boundaries based on the strain values.
PFRP profiles showed a linear ascending behaviour presenting the Using the DIC, the number and the length of the out- and in-of-plane
elastic modulus of the cross-section which confirms the linear Load– deformations can be measured for the tested profiles, as shown in
Displacement behaviour in Fig. 5. This linear behaviour resulted in Fig. 10. Table 6 reports the length of the in-plane (negative values) and
a noticeable increase in the strain reading due to the local buckling outward (positive values) buckling half-waves. Table 6 shows that the
of the walls. The extreme change in the strain reading after the peak length of the half-waves is not equal along the height of the column.
is caused by the in-plane or out-of-plane deformation of the wall Moreover, it can be clearly observed that the number of the half-waves
after buckling. At the first peak load, the axial compression strain is always odd and the longer half-wave is the middle one. This finding
of the tested specimens was in the range of 4500 𝜇𝜀 to 6000 𝜇𝜀. agrees with the polynomial local buckling theory mentioned by Cardoso
When the local buckling started, however, this value either increased et al. [27] and contradicts the assumption of having sinusoidal local
or reversed to tensile strain based on the direction of the buckling buckling behaviour [2]. Nevertheless, this finding is limited to the
half-wave (Fig. 6.b). It can be said that the out-of-plane and in-plane short PFRP columns tested in this study, where the length might affect
deformations is high for R75 then S100 and S125. This finding can be the consistency and symmetry of the local buckling half-waves [28].
demonstrated by the higher strain at failure for R75 followed by S100 This can be observed by the closer difference in the second and fourth
and S125, respectively. Interestingly, the modulus of elasticity for all half-wave length to the middle length compared to the first and fifth
the columns was recorded to be more than that reported for coupons. half-waves in the tested longer specimens (𝐿∕𝐷 ratio of 3.5 and 5.0).
The same observation was reported by Guades et al. [25] and Al-Saadi On the other hand, Table 6 shows that the half-wave length of the tested
et al. [11] where they observed a noticeable increase (around 20%) specimens with 𝐿∕𝐷 ratio of 3.5 and 5.0 is longer than the width. This
in the compressive elastic modulus compared to the tensile modulus. finding is consistent with previous recommendations on the predicted
This might be due to the nature of composites under compression in length of the half-wave to be a fraction in between the width and the
which the longitudinal compressive elastic modulus receives additional double width values [2,10]. It also can be observed generally that the
in-plane shear resistance shown by the shear failure of the samples in-plane waves are longer than the outward waves (see Table 6). This
tested under compression [26]. Table 5 reports the strain readings from observation might be related to the use of a veil at the top of the
the strain gauges attached to two profiles from each set. These values outer layer during the manufacturing process to protect and smooth
were confirmed to be similar to those recorded by the DIC, as explained the surface. This extra layer might briefly contribute to the outward
in the next section. Furthermore, it can be observed that the CoV values flexural strength, resulting in decreasing the out-of-plane deformations
for the reported modulus of elasticity values were small and the data compared to the in-plane deformations.
were consistent.
4. Discussion
3.4. Digital Image Correlation (DIC) analysis
4.1. Influence of geometry
3.4.1. Validation and reliability
The digital image correlation (DIC) instrument was used in this According to Table 4, axial stiffness of the tested specimens was
study to observe the two-dimensional (2D) changes in the wall in- triggered by the increase of the cross-sectional area. Therefore, re-
cluding strains, displacements, and the formation and extent of the gardless of the 𝐿∕𝐷 ratio value, the S125 profile recorded the highest
buckling waves on the tested profiles while applying the load. These axial stiffness followed by S100 and then the R75 profiles. However,
measurements have been traced to validate the test setup and become the normalized axial stiffness to the cross-sectional area reversed that
familiar with some features, which are hard to record during the test, order, so the R75 profile was the highest normalized stiffness followed
such as the local buckling formation, and their extension and length. by S100 and then S125 profiles because R75 has a higher fibre volume

8
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Table 6
Wavelength measurement (in mm) for the tested PFRP profiles.
𝐿∕𝐷 R75 S100 S125
2.0 +44/−119/+38 −82/+80/−42 +32/−83/+72/−15
3.5 −40/+100/−120/+95/−22 −27/+102/−118/+73/−28 +56/−79/+101/−91/+26
5.0a −14/+95/−108/+117/−44 −114/+117/−134/+102/−36 +74/−121/+165/−141/+56
a
For S125, 𝐿∕𝐷 = 4.5.

After point B in Fig. 5.a, the load capacity dropped due to the
delamination at the corner, whereas the wound fibres ruled to main-
tain the axial load capacity by showing the local buckling shape.
The amount of the load decay was based on the cross-section and
its properties. For example, S125 was the lowest affected profile by
showing an average load drop of 6.8% compared to the R75 and S100,
which showed an average load drop of 8.5% and 12.8%, respectively.
This is because S125 has the highest wound fibre percentage followed
by R75, which has also greater moment restraints at the edge of the
wall compared to S100. Moreover, it was observed that the buckling
shape (Point C to D in Fig. 5.a) maintained the axial load capacity
longer in the R75 profile than the square profiles due to the additional
edge restraint conditions. On the other hand, S125 showed a shorter
withstanding of the axial load resistance during the local buckling mode
than S100, which might be attributed to the higher storage energy in
S125 than S100 after the load drop, even with having the same wall
slenderness. This made the interlaminar failure propagation quicker
and more severe. Furthermore, the post-peak behaviour is also found
to be driven by the half-wave length as reported in Table 6. Thus, S125
showed the longest half-wave length, which caused more spreading to
Fig. 10. Wavelength measurements (R75 with 𝐿∕𝐷 of 2.0).
the interlaminar cracks because of having more in- and out-of-plane
deflections, and resulted in earlier failure. On the other hand, R75
showed the shorter half-wavelength, which ended with the axial load
maintaining resistance in the post-peak stage for longer (Point C to D).
fraction (Table 1) than the square cross-sections. Moreover, the S125
Because of this behaviour, the normalized toughness at the second peak
profile had lower normalized stiffness than S100 because of higher
load (𝑇𝑁2 ) of the cross-sectional area reveals that R75 is the toughest
wound fibre percentage.
profile, or the most able to store the energy, followed by S125 then
Based on the test results, it is clearly observed that the full-scale pro-
S100, even though Table 4 shows contradictory 𝑇𝑁2 results. These
files recorded significantly less compressive strength than the coupon
results do not consider the variation in the cross-sectional area. Just
results (Table 2) due to the failure mode of the full-scale profiles [2–
before the moment of failure, the tested profiles reached 𝜎2 , indicating
4]. This difference in the compressive strength between the coupons
the maximum strength in the local buckling mode. All profiles with
and the full-scale specimens is also observed to be controlled by the
different geometry recorded 𝜎2 as 87% to 95% of 𝜎1 . This range is
slenderness of the walls, laminate properties, moment restraints at the
higher than the load decay at point C, which confirms the brief non-
joint, and length of the full-scale specimens. The R75 profile recorded
linear increase in the axial load capacity on the tested profiles after the
the highest strength (𝜎1 ) compared to the S100 and S125 profiles, load drop. It should be mentioned that no significant difference in the
because of the lower aspect ratio effect of the cross-section. Even final mode of failure was noticed in the tested specimens.
though the slenderness of the longer wall width of R75 ( 100−2×5.0 5.0
=
18.0) is more than that of S100 ( 100−2×5.25
5.25
= 17.1), the slenderness 4.2. Influence of 𝐿∕𝐷 ratio
of the shorter dimension of R75 is 13.0 ( 75−2×5.0 5.0
). Initiating out-of-
plane deformation at the short dimension wall in R75 requires a higher According to Table 4, the increase in the 𝐿∕𝐷 ratio of the specimen
axial load compared to the tested square cross-sections. Moreover, the led to a decrease in the axial stiffness. Therefore, the stiffness in R75,
coupon-based material properties (Table 2) showed higher interlam- S100, and S125 reduced in average by 50.1%, 51.7%, and 47.6%
inar shear resistance in R75 profiles, which can add more stability when 𝐿∕𝐷 ratio was increased from 2.0 to 5.0, respectively. It can be
to the corners before initiating delamination and can thus increase noticed that the reduction in stiffness for all the tested PFRP profiles,
the moment restraint at the joint of the wall in R75 and delay the regardless the difference in geometry, is almost similar attributing to
initiation of buckling. Moreover, S125 yielded higher 𝜎1 than S100 even the close material properties of the tested profiles (see Table 2). The
with the same wall slenderness. As mentioned, when localized failure phenomenon of stiffness reduction with the increase in the length of
occurred at the junction; the walls behave as they are hinged along a profile can be explained by the volume reduction energy where the
the junction where the inclined fibres function at resisting the rotation shorter specimens require high stored energy (high-applied load) to
and separation of the adjacent walls similar to what was observed by exhibit similar deformation (reduction in the volume) to the longer
Al-saadi et al. [11]. Thus, compared to S100, the higher wound fibre specimens [29,30]. This finding explains why the longer specimens
percentage in S125 adds more stiffness in the transverse direction, recorded greater axial deformation at the first peak load compared to
resisting the out-of-plane deformation and adding more resistance to the shorter ones. This also affected the toughness capacity of the tested
the delamination at the corner, thus increasing 𝜎1 . Moreover, this profiles by having increases in the 𝑇𝑁1 values of 169%, 142%, and
behaviour is responsible for showing higher toughness at 𝑃1 (𝑇𝑁1 ) for 129% of the R75, S100, and S125 profiles, respectively, when the 𝐿∕𝐷
S125 than S100. As mentioned earlier, R75 also shows a higher 𝑇𝑁1 ratio increased from 2.0 to 5.0 (see Table 4).
than S100 due to the lower aspect ratio with the same longer wall Besides axial stiffness, the increasing 𝐿∕𝐷 ratio resulted in a re-
width, which leads to record higher strength. duction 𝜎1 of all the tested profiles. For instance, R75, S100, and

9
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

S125 recorded 11.7%, 14.5%, and 17.0% lesser 𝜎1 on average when first half-wave where the corresponding buckling strength is the local
increasing the 𝐿∕𝐷 ratio from 2.0 to 5.0, respectively. As mentioned buckling stress.
earlier, this can be explained by the size effect where the shorter To simulate the box PFRP profiles, with rotational restraint edges,
specimens require a higher applied load to reach the fracture or failure Eq. (1) was modified by [2,36,37] to accommodate a factor repre-
threshold [30]. Moreover, the stability of an edge-restrained plate is senting the moment restraints at the wall edges are generated by the
controlled by its height, slenderness (width to thickness), and the adjacent walls. Thus. Eq. (2) represents the buckling load capacity of a
number of end restraints [2,31]. However, when increasing the 𝐿∕𝐷 box PFRP profile considering the end restraints. In Eq. (2), 𝐷11 , 𝐷22 ,
ratio of a wall, the height is the only factor affecting its stability, if 𝐷12 , and 𝐷66 are the flexure stiffness coefficients, 𝐿𝑥 is the length
the imperfections presence was assumed to be negligible as they are in the loading direction (𝐿𝑥 = 𝑖.𝑙𝑥 , 𝑤ℎ𝑒𝑟𝑒 𝑖 = 1, 2, 3, …), 𝑖 is the
very minimal in pultrusion manufacturing and have low impact on number of the half-waves on the wall in the loading direction, 𝑏 is
the performance for thick-walled composites [32]. The length-based the width of the wall, and 𝜉 is the restraint factor. Appendix A details
stability threshold can be considered as a certain length of a plate (𝑙𝑥 ),
the equations’ constituents (Eq. (A.1)). It is worth mentioning that the
which can change the failure from compression to buckling. There-
material properties implemented in these equations are taken from
fore, when the length of a specimen (𝐿) is short and close to 𝑙𝑥 , the
Table 2.
compression-dominant length fraction (compared to the total length)
will have higher a contribution than the longer ones. Consequently, 𝑎1 2 1 𝑎 2 𝐿𝑥 2 𝑎
𝑁𝑡ℎ−1 = 𝐷11 + 𝐷22 3 + 2(𝐷12 + 2𝐷66 ) 4 (1)
continuous reduction in peak strength will be observed until almost 𝑎2 𝐿 𝑥 2 𝑎2 𝑏 2 𝑏2
[ ]
constant strength is shown, where the compression-dominant length 𝜋2 𝑏2 𝐿 2
fraction has an insignificant effect on the strength. This behaviour will 𝑁𝑡ℎ−2 = 𝐷11 + (1 + 4.139𝜉)𝐷22 𝑥 + (2 + 0.62𝜉)(𝐷12 + 2𝐷66 )
𝑏 2 𝐿𝑥 2 𝑏2
be explained theoretically in the next section. (2)
After point B (Fig. 5.a), the increase in the 𝐿∕𝐷 ratio increased
the amount of load drop. The load drop percentage in R75, S100, and Fig. 11 is the result of plotting Eq. (2) considering the tested profiles in
S125 is comparable to 15%–17% of 𝜎1 when the 𝐿∕𝐷 ratio is 5.0, this study by plotting the wall length with the critical buckling load. In
which is 6.8%, 8.5% and 12.8%, respectively, when the 𝐿∕𝐷 ratio Fig. 11, it should be mentioned that clamped case is when 𝜉 equals to
is 2.0. This can be explained by the larger susceptible area for crack 1, while free case is when 𝜉 equals to zero. The partially clamped with
propagation and expansion of the local buckling waves evidenced by edge restraint (restrained) case is when 0 < 𝜉 < 1. It can be noticed in
the increase in the half-wave number by increasing the 𝐿∕𝐷 ratio Fig. 11 that the restrained case significantly underestimates the 𝜎1 val-
(see Table 6). After the load drop, the tested profiles maintained the ues. However, it shows closer predictions to 𝜎2 values. This is confirms
axial load capacity during the buckling stage. It was observed through the finding where 𝜎1 values do not represent the buckling stress of the
(Fig. 5) that the increase in 𝐿∕𝐷 ratio increased the deformation tested PFPR profiles as they are localized failure peaks, as explained
capacity of the tested specimens during buckling. The increase in the earlier. It is believed that local bucking, according to Fig. 5.a, started
length results in increasing the number of half-waves from 3 to 5 after point C where the gradual decrease in the stiffness can be observed
when the 𝐿∕𝐷 ratio increased from 2.0 to 5.0. This increase in the as a nonlinear behaviour until reaching point D. Therefore, 𝜎1 can be
number of half-waves led to more stable local buckling at which these considered as the stability limit of the elastic behaviour while 𝜎2 is the
half-waves share releasing the energy instead of releasing it at once buckling stress. This finding highlights the advantages of the pultrusion
which might result in a catastrophic failure [2]. This is evidenced by with continuous-wound fibres, which allows an increase in the design
showing higher displacement capacity under buckling for the columns limit in the elastic behaviour, as no signs of failure will be observed
with higher buckling half-waves (see Fig. 5 and Table 6). Moreover, at this stage. It can be noticed that the restrained case prediction
the buckling stage in the longer specimens occurred at lower applied accurately showed the buckling stress of the square profiles with 𝐿∕𝐷
stress so that the stress distribution through the cross-section was more ratio of 3.5 and 5.0. However, it underestimated the buckling stress of
uniform and the stress concentration spots were less than the shorter
the ones with 𝐿∕𝐷 ratio of 2.0, perhaps be due to the polynomial shape
specimens. Accordingly, the increase in the 𝐿∕𝐷 ratio increased the
of the generated half-waves (not sinusoidal) (Fig. 10) and because the
toughness values at the second peak (𝑇𝑁2 ). As expected, the increase
length of the half-wave was shorter compared to the theoretical one
in the 𝐿∕𝐷 ratio decreased the 𝜎2 values. However, the reduction in
(Table 6). In contrast, the predicted buckling stress of the R75 was
𝜎2 caused by increasing the 𝐿∕𝐷 ratio was observed to be higher than
significantly less than the experimental ones, which might be due to the
the reduction in 𝜎1 for the same reason. This result confirms the effect
high end-moment restraints contributed by the shorter dimension of the
of the failure propagation in the buckling stage (point C to D), which
rectangular profile, which is not considered in the presented equations.
created many interlaminar cracks due to the buckling shape. The higher
On the other hand, there are other analytical models, which can predict
failure progression in the longer specimens can be proved by the longer
half-waves (Table 6) which are responsible for spreading the cracks one buckling stress limit based on derivatives from the plate theory [2]
along the length of the profile resulting in reducing 𝜎2 values. or simplified equations [27,33–35]. Table 7 shows the predictions of
the local buckling stress of the tested values. It can be observed that
5. Comparison of test results with theoretical models the model based on the plate theory [2] using Eq. (2) (Fig. 11) gives
closer local buckling stress predictions to the experimental results. The
In composites, many theories have been established to predict the equations used in Table 7 are presented in Appendix A.
local buckling strength considering an orthotropic plate subjected to The plate theory also suggested Eq. (3) to find the critical half-
uniaxial compression and unloaded edges with various end conditions wavelength (𝐿𝑐𝑟−ℎ𝑤 ) of the buckled plate or wall. Thus, Table 8 shows
including simply supported (free), clamped and restrained edge condi- the theoretical and experimental measurements of the half-wavelength
tions [2,27,33–36]. In the box PFRP profiles, the compressive strength of the tested profiles. Table 8 presents length values that are very close
capacity is determined by assuming restrained edges of a single wall to the ones reported in Table 6, which means the plate theory is valid
to simulate the adjacent walls [2,36,37]. Qiao et al. [36] established for predicting the local buckling behaviour of the tested PFRP profiles,
closed-form equations to analyse a discrete-orthotropic plate with sim- especially for the longer wall specimens.
ply supported edge conditions, as shown in Eq. (1). This equation deals √ √
with the various length of the plates as a function of the number of 1 𝐷
𝑙𝑥 = 4
𝑏 4 11 (3)
half waves (eigenvalues) generated at the surface of the wall. Thus, 1 + 4.139𝜉 𝐷22
it identifies the minimum height of the wall required to show the

10
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

Table 7
Theoretical buckling stresses (MPa) based on the previous models with the relevant errors.
Equation number Reference Theoretical results Error compared to experimental results
R75 S100 S125 R75 S100 S125
Regardless 𝐿∕𝐷 ratio 2.0 3.5 5.0 2.0 3.5 5.0 2.0 3.5 4.5
(A.2a) Kollar and Springer [2] 226.9 234.8 272.3 31% 23% 21% 22% 14% 11% −16% −29% −32%
(A.3) Pultex [33] and Cardoso et al. [27] 285.8 218.8 178.2 13% 3% 0% 28% 20% 18% 24% 15% 14%
(A.4) Davalos et al. [34] 185.9 195.9 204.3 43% 37% 35% 35% 28% 26% 13% 3% 1%
(A.5) Strongwell [35] 229.3 249.8 237.0 30% 22% 20% 18% 8% 6% −1% −12% −15%

Fig. 11. The 𝜎1 and 𝜎2 against the various plate end-conditions of the PFRP profiles.

Table 8 delamination cracks at corners and ended with severe buckling


Experimental and theoretical measurements of the half-wavelength (mm) of the tested
and shear failure at the corners, due to local buckling. However,
profiles.
the 𝐿∕𝐷 ratio had an influence only on increasing the number
R75 S100 S125
of the buckling half-waves appearing at the PFRP walls due to
Experimental 𝐿ℎ𝑤 120 134 165
the increase in the wall length.
Theoretical 𝐿𝑐𝑟−ℎ𝑤 121 123 156
2. Due to the existence of the continuous-wound fibres, 10%–
Error (%) 0.8 8.2 5.5
15% increased increment on the local buckling strength of the
tested PFRP profiles was noticed. Therefore, the non-linear load–
displacement response indicating the local buckling was only ob-
6. Conclusions served after the first failure took place. It should be highlighting
that the buckling behaviour was initiated by the delamination
This paper investigated and discussed the effect of the geometry cracks where the walls were freed partially to rotate around the
and 𝐿∕𝐷 ratio on the compressive behaviour of the PFRP profile with corner.
continuous-wound fibres. Special fixtures attached to the ends of the 3. All tested profiles had almost similar load–displacement be-
columns were designed and manufactured to obtain precise capturing haviour. However, R75 profiles recorded better structural per-
of the behaviour of the tested specimens. The DIC instrument is used as formance compared to the square ones, due to the lower aspect
a new way of investigating the problems of buckling instability. More- ratio that led to an increase in the moment restraints at the wall’s
over, the applicability of the existing models was examined to predict edges resulting in more stability of the cross-section. Moreover,
the performance of the tested profiles. The following conclusions are the layup and the continuous-wound fibre content played a role
drawn from this study: in the square profiles, even with different cross-sections but the
same wall slenderness. The increase in the continuous-wound
1. The geometry had no effect on either the initiation or the final fibre content increased the stability of the cross-section and
failure of the tested PFRP profiles where all profiles started with showed around 13% higher strength of S125 compared to S100

11
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

due to an increase in the lateral flexural stiffness of the walls and Appendix A. Equations
corners.
4. The increase in the 𝐿∕𝐷 ratio from 2.0 to 5.0 of R75, S100,
and S125 resulted in 50.1%, 51.7% and 47.6%, and 11.7%,
𝐿𝑥 = 𝑖.𝑙𝑥 (𝑖 = 1, 2, 3, …) (A.1a)
14.5% and 17.0% reduction in axial stiffness and 𝜎1 , respec-
tively. This decrease was due to the size effect where the shorter 𝑎1 = 𝑖𝜋 (A.1b)
specimens required higher applied load to deform and reach the 𝑎2 = 𝑖 2 𝜋 2 (A.1c)
fracture or failure threshold. Moreover, the increase in the 𝐿∕𝐷
ratio in the PFRP profiles increased the displacement capacity 𝑎3 = 𝑗𝜋 (𝑗 = 1) (A.1d)
under buckling and increased the maximum half-wave length 2 2
𝑎4 = 𝑗 𝜋 (A.1e)
because buckling in longer specimens occurred at lower applied
𝐸𝑙 𝑡3
stress, which led to more uniform stress distribution in the 𝐷11 = (A.1f)
12(1 − 𝑣12 .𝑣21 )
cross-section, resulted in better failure progression.
𝐸𝑡 𝑡3
5. The strain results showed higher compressive modulus of elas- 𝐷22 = (A.1g)
ticity in the full-scale specimens and coupons compared to the 12(1 − 𝑣12 .𝑣21 )
tensile coupons. This was due to the engagement of the shear 𝐷12 = 𝑣12 𝐷22 (A.1h)
resistance in compression configuration. G𝑡3
6. DIC proved to accurately capture the strain behaviour of the 𝐷66 = (A.1i)
12
tested PFRP profiles and trace the failure features, including
local buckling behaviour. It was observed that local buckling is
a polynomial function where the longest half-wave is generated 𝜋2 √ √ ( )
𝑁𝑡ℎ = (2 1 + 4.139𝜉 𝐷11 𝐷22 + (2 + 0.62𝜉 2 ) 𝐷12 + 2𝐷66 ) (A.2a)
at the mid-height of the specimen. It has seen that the length 𝑏2
1
of the half-wave is driven by the increase in the specimen wall 𝜉= (A.2b)
1 + 10𝜍
height for the short PFRP profiles.
𝐷
7. The existing discrete-orthotropic plate theory was able to predict 𝜍 = 22 (A.2c)
𝑘𝑏
reasonably the buckling stress of the square profiles with 𝐿∕𝐷 ra- 2𝐷66 + 𝐷12
tio of 3.5 and 5.0, while it underestimated the buckling stress of 𝑘= √ (A.2d)
the specimens with 𝐿∕𝐷 ratio of 2.0 due to the polynomial shape 𝐷11 𝐷22
of the generated half-waves (not sinusoidal) and because the ( )2
𝜋2 𝑡
length of the half-wave was shorter compared to the theoretical 𝑁𝑡ℎ = 𝑘 (A.3)
2
12(1 − 𝑣12 ) 𝑏
one. In contrast, the predicted buckling stress of the rectangular [ ] ( )2
profile was significantly less than the experimental ones, which 𝜋2 √ √ 𝑡
𝑁𝑡ℎ = 2 𝑞 𝐸𝑙 𝐸𝑡 + 𝑝(𝑣12 𝐸𝑡 + 2𝐺) (A.4)
might be due to the high end-moment restraints contributed by 12 𝑏
𝐸𝑙
the shorter dimension of the rectangular profile. 𝑁𝑡ℎ = ( )0.85 (A.5)
16 𝑏𝑡
According to the presented outcome of the tested PFRP profile with
continuous-wound fibres, it is recommended to conduct comprehensive Appendix B. Notations
investigations to consider the effect of global buckling and its interac- 𝑎1 & 𝑎2 : Coefficient of waviness 𝑗 : Number of half-waves
tion with local buckling. It is also recommended that different applied in longitudinal axis along the width (= 1)
load configurations should be commenced on such PFRP profiles. 𝑎3 & 𝑎4 : Coefficient of waviness 𝑘 : Stiffness factor
in transverse axis
CRediT authorship contribution statement 𝜉 : Restraint factor 𝐿∕𝐷 : Length-to-depth ratio
𝜍 : Restraint factor 𝑙𝑥 : Length of the half-wave
Omar Alajarmeh: Testing, Visualization, Conceptualization, Writ- (mm)
ing – original draft, Writing – review & editing. Xuesen Zeng: Super- 𝑣12 : Poisson’s ratio in 𝐿𝑥 : Length of the specimen
vision, Writing – review & editing. Thiru Aravinthan: Supervision, longitudinal direction (mm)
Writing – review & editing. Tristan Shelley: Writing – review & 𝑣21 : Poisson’s ratio in 𝑁1 : First peak load (kN)
editing. Mohammad Alhawamdeh: Testing, Visualization, Writing - transverse direction
review & editing. Ali Mohammed: Investigation, Resources. Lachlan 𝜎1 : Stress at first peak load 𝑁2 : Second peak load (kN)
Nicol: Investigation, Resources. Alexander Vedernikov: Writing - re- (MPa)
view & editing. Alexander Safonov: Writing - review & editing. Peter 𝜎2 : Stress at second peak 𝑁𝑡ℎ : Theoretical buckling
Schubel: Project administration, Management. load (MPa) load (N/mm)
𝑏 : Width of the specimen 𝑝 : Curve fitting coefficient
Declaration of competing interest (mm)
𝐷11 , 𝐷22 , : Flexural stiffness 𝑞 : Curve fitting coefficient
The authors declare that they have no known competing finan- 𝐷12 and coefficients (N mm)
cial interests or personal relationships that could have appeared to 𝐷66
influence the work reported in this paper. 𝑡 : Thickness (mm)
𝐸𝑙 : Longitudinal Modulus 𝑇𝑁1 : Toughness at first peak
Acknowledgements of elasticity load (kN mm)
𝐸𝑡 : Transverse Modulus of 𝑇𝑁2 : Toughness at second
This work was financially supported through a Cooperative Re- elasticity peak load (kN mm)
search Centres Projects (CRC-P) Grant (CRCPSIX000117) with 𝐺 : Shear modulus (MPa) 𝑉𝑓 : Fibre volume fraction
Wagners-CFT. The authors acknowledge and thank Wagners-CFT for (%)
preparing and supplying the PFRP specimens. The assistance provided 𝑖 : Number of half-waves
by colleagues in the Centre for Future Materials is also acknowledged. along the length

12
O. Alajarmeh, X. Zeng, T. Aravinthan et al. Thin-Walled Structures 168 (2021) 108300

References [19] H. Saito, R. Inai, H. Hamada, Crushing properties of pultruded glass reinforced
square tubes, Int. J. Crashworthiness 7 (1) (2002) 21–34.
[1] S. Advani, K.-T. Hsiao, Introduction to composites and manufacturing processes, [20] E.J. Barbero, I.G. Raftoyiannis, Local buckling of FRP beams and columns, J.
in: Manufacturing Techniques for Polymer Matrix Composites (PMCs), Elsevier, Mater. Civ. Eng. 5 (3) (1993) 339–355.
2012, pp. 1–12. [21] F. Nunes, J.R. Correia, N. Silvestre, Structural behaviour of hybrid FRP pultruded
[2] L.P. Kollar, G.S. Springer, Mechanics of Composite Structures, Cambridge columns, part 1: Experimental study, Compos. Struct. 139 (2016) 291–303.
University Press, 2003. [22] A.C. Monteiro, M. Malite, Behavior and design of concentric and eccentrically
[3] J.M. Whitney, Structural Analysis of Laminated Anisotropic Plates, CRC Press, loaded pultruded GFRP angle columns, Thin-Walled Struct. 161 (2021) 107428.
1987. [23] D.C. Cardoso, B.S. Togashi, Experimental investigation on the flexural–torsional
[4] S. Lekhnitskii, Anisotropic Plates, Gordon and Breach, New York, 1968. buckling behavior of pultruded GFRP angle columns, Thin-Walled Struct. 125
[5] E. Magnucka-Blandzi, K. Magnucki, Buckling and optimal design of cold-formed (2018) 269–280.
thin-walled beams: review of selected problems, Thin-Walled Struct. 49 (5) [24] G. Simitses, S. Sallam, W. Yin, Effect of delamination of axially loaded
(2011) 554–561. homogeneous laminated plates, AIAA J. 23 (9) (1985) 1437–1444.
[6] A. Teter, H. Debski, S. Samborski, On buckling collapse and failure analysis of [25] E. Guades, T. Aravinthan, Residual properties of square FRP composite tubes
thin-walled composite lipped-channel columns subjected to uniaxial compression, subjected to repeated axial impact, Compos. Struct. 95 (2013) 354–365.
Thin-Walled Struct. 85 (2014) 324–331. [26] L.A. Carlsson, D.F. Adams, R.B. Pipes, Experimental Characterization of Advanced
[7] J. Xu, Q. Zhao, P. Qiao, A critical review on buckling and post-buckling analysis Composite Materials, CRC press, 2014.
of composite structures, Front. Aerosp. Eng. 2 (3) (2013) 157–168. [27] D.C. Cardoso, K.A. Harries, E.d.M. Batista, Closed-form equations for compressive
[8] V.V. Bolotin, Delaminations in composite structures: its origin, buckling, growth local buckling of pultruded thin-walled sections, Thin-Walled Struct. 79 (2014)
and stability, Composites B 27 (2) (1996) 129–145. 16–22.
[9] S. Wang, Y. Zhang, Post-buckling and delamination propagation in debonded [28] G.G. Cintra, D.C. Cardoso, J.D. Vieira, Parameters affecting local buckling re-
composite laminates part 2: Numerical applications, Compos. Struct. 88 (1) sponse of pultruded GFRP I-columns: Experimental and numerical investigation,
(2009) 131–146. Compos. Struct. 222 (2019) 110897.
[10] L.C. Bank, J. Yin, Buckling of orthotropic plates with free and rotationally [29] S. Şener, B.I. Barr, H.F. Abusiaf, Size effect in axially loaded reinforced concrete
restrained unloaded edges, Thin-Walled Struct. 24 (1) (1996) 83–96. columns, J. Struct. Eng. 130 (4) (2004) 662–670.
[11] A.U. Al-saadi, T. Aravinthan, W. Lokuge, Effects of fibre orientation and layup on [30] Z.P. Bazant, J. Planas, Fracture and Size Effect in Concrete and Other Quasibrittle
the mechanical properties of the pultruded glass fibre reinforced polymer tubes, Materials, Vol. 16, CRC press, 1997.
Eng. Struct. 198 (2019) 109448. [31] S.T. Peters, Handbook of Composites, Springer Science & Business Media, 2013.
[12] T. Liu, K.A. Harries, Flange local buckling of pultruded GFRP box beams, [32] L. Ascione, V.P. Berardi, S. Spadea, Pre-buckling imperfection sensitivity of
Compos. Struct. 189 (2018) 463–472. pultruded FRP profiles, Composites B 72 (2015) 206–212.
[13] D.C. Cardoso, K.A. Harries, E.d.M. Batista, Compressive strength equation for [33] C.P. Inc., The Pultex Pultrusion Global Design Manual of Standard and Custom
GFRP square tube columns, Composites B 59 (2014) 1–11. Fiber Reinforced Polymer Structural Profiles, The Creative Pultrusion Inc, 2000.
[14] A. Godat, F. Légeron, B. Marmion, Use of FRP pultruded members for electricity [34] J.F. Davalos, E.J. Barbero, P. Qiao, Step-By-Step Engineering Design Equations
transmission towers, Compos. Struct. 105 (2013) 408–421. for Fiber-Reinforced Plastic Beams for Transportation Structures, Contract West
[15] F. Ascione, M. Lamberti, G. Razaqpur, Modifications of standard GFRP sections Virginia Department of Transportation, West Virginia University, Morgantown,
shape and proportions for improved stiffness and lateral–torsional stability, West Virginia, USA, Contract No: RP, 147, 2002.
Compos. Struct. 132 (2015) 265–289. [35] Strongwell design manual, Strongwell Corporation, 2007.
[16] M. Alhawamdeh, O. Alajarmeh, T. Aravinthan, T. Shelley, P. Schubel, M. Kemp, [36] P. Qiao, J.F. Davalos, J. Wang, Local buckling of composite FRP shapes by
X. Zeng, Modelling hollow pultruded FRP profiles under axial compression: Local discrete plate analysis, J. Struct. Eng. 127 (3) (2001) 245–255.
buckling and progressive failure, Compos. Struct. 262 (2021) 113650. [37] E. Barbero, I. Raftoyiannis, Buckling analysis of pultruded composite columns,
[17] I. 1172, Textile-glass reinforced plastics, 1996. Impact Buckling Struct. 20 (1990) 47–52.
[18] L. Xie, Y. Bai, Y. Qi, C. Caprani, H. Wang, Effect of width–thickness ratio
on capacity of pultruded square hollow polymer columns, Proc. Inst. Civ.
Eng.-Struct. Build. 171 (11) (2018) 842–854.

13

You might also like