Bambach 2019
Bambach 2019
Bambach 2019
Experimental and numerical study of the bending strength of natural fibre com-
posite structural channel sections
M.R. Bambach
PII: S0263-8223(18)31508-3
DOI: https://doi.org/10.1016/j.compstruct.2018.11.031
Reference: COST 10396
Please cite this article as: Bambach, M.R., Experimental and numerical study of the bending strength of natural fibre
composite structural channel sections, Composite Structures (2018), doi: https://doi.org/10.1016/j.compstruct.
2018.11.031
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1 Experimental and numerical study of the bending strength of natural fibre
2 composite structural channel sections
3
4
5 M.R.Bambach
6 School of Civil Engineering, The University of Sydney, NSW 2006, Australia
7 mike.bambach@sydney.edu.au, T +61 2 9351 2193, F +61 2 9351 3343
8
9
10
11 ABSTRACT
13 structural materials for the built environment. Natural fibres such as flax and jute have increasingly
14 been considered for fibre-resin composites, with a major motivation for their implementation being
15 their notable sustainability attributes. This paper is part of an ongoing effort by the author to
16 demonstrate the structural properties of primary structural elements and members fabricated from
17 natural fibre composites of flax and jute. Previously the structural properties of flat plates, plain
18 channel sections and channel sections with complex stiffeners were investigated under pure
19 compression. This paper presents investigations of channel sections with complex stiffeners under
20 pure bending. A series of sixteen channels with varying geometries, complex stiffener arrangements
21 and composite thicknesses were tested in pure flexure. Material tests indicated that the mean tensile
22 elastic stiffness and strength values were 6,386 MPa and 55.1 MPa for flax, and 6,941 MPa and
23 62.1 MPa for jute. The experimental results indicated that flexural failure of the channel sections
24 was governed by tensile fracturing. The ultimate moment capacities varied from 1.043 to 1.501
25 kNm for four-layered composites, and 2.184 to 2.511 kNm for six-layered composites. The
26 analytical models predicted the experimental ultimate moment capacities well, with a mean and
27 coefficient of variation of the test to predicted ratio of 0.97 and 0.06, respectively. Finite element
28 models used progressive damage analysis via stress-based damage initiation models and damage
29 evolution laws, to replicate the tension fracture failure mode of the channels. The numerical models
1/34
30 predicted the experimental ultimate moment capacities well, with a mean and coefficient of
32
33 KEYWORDS: natural fibre composites, flax, jute, channels, bending, flexure, finite element
34
35 1. INTRODUCTION
36 Public concerns about the environment, climate change, energy consumption and greenhouse gas
37 emissions are driving demand for the use of sustainable materials in the built environment. There
38 has been substantial attention given to the use of natural fibres in fibre-reinforced plastics in recent
39 decades, where such fibres may be combined with thermoset or thermoplastic polymers to create
40 natural fibre composites. Such natural fibre composites have particularly been identified for their
41 favourable sustainability properties, including for example: renewable resource; carbon sink; short
42 growth cycle time; low herbicide requirements due to rapid growth; low energy production;
43 recyclable; biodegradable; and low hazard manufacturing and composite handling and working [1-
44 7]. Much of this research has been from a materials science standpoint, assessing materials aspects
45 such as fibre processing techniques, composite fabrication methodologies, matrix materials and
46 their effects on the mechanical properties [8-13]. This research has indicated that composites
47 consisting of natural fibres have general characteristics similar to their synthetic fibre counterparts,
48 such as glass and carbon, however have comparably low intrinsic mechanical properties [2-13]. As
49 a result, identifying structural applications such as those in civil infrastructure have thus far been
50 limited [14-22].
51
52 Recently, the author has undertaken to demonstrate the structural properties of primary structural
53 elements and members fabricated from natural fibre composites of flax and jute, including: flat
54 plates and plain channel sections in pure compression [23], and channel sections with complex
55 stiffeners in pure compression [24]. This paper extends these studies to investigate flax and jute
2/34
56 fibre composites in pure bending. It is demonstrated that while the failure modes of the previous
57 members in compression were dominated by elastic local buckling and compression matrix failure,
58 the present failure modes in pure flexure are dominated by tension fracture. The stiffened channel
59 sections were previously demonstrated to have compression capacities suitable for light structural
60 applications such as residential building framing [24]. An aim of the present study is to determine if
61 such channel sections are suitable for resisting the bending loads that such framing members
63
64 Initially a series of sixteen experiments of flax and jute fibre composite structural channel sections
65 subjected to pure (four-point) bending is described. An analytical method developed previously for
66 compression [23,24] is then extended to the case of flexure. Finally, finite element analyses are
67 described which demonstrate the suitability of progressive damage analysis to predict the bending
68 capacity.
69
70 2. EXPERIMENTAL METHODS
72 Two different natural fibres were investigated in the present study; flax and jute. The flax and jute
74 Evolution; Biotex Flax 400g/m2 2x2 Twill weave and Biotex Jute 400g/m2 2x2 Twill weave. The
75 2x2 Twill weave is a generic pattern consisting of bi-directional yarns woven over-over-under-
76 under, where the yarns are perpendicular (i.e. 0º/90º). The fabrics use a low-twist yarn. Nominal
77 density, tensile strength and modulus values for the flax were; 1.5g/cm3, 500MPa and 50GPa, and
78 for the jute were; 1.46g/cm3, 400MPa and 40GPa. The commercial bulk laminating epoxy resin
79 Kinetix R240 with H126 (fast) hardener was used for all composite fabrications, with density
80 1.1g/cm3 and measured (neat) compression ultimate stress of 105.3MPa, tension ultimate stress of
81 33.1MPa and tension ultimate strain of 0.8%. The epoxy is a room temperature out-of-autoclave
3/34
82 cure epoxy resin that does not require heat input to cure. The manufacturers’ technical data for the
83 fabrics and epoxy are summarised in Appendix A. The channels were fabricated with a hand layup
84 technique whereby each layer of fabric was wetted out with resin using a paint brush and roller. The
85 fabric 0º direction was manually aligned with the longitudinal direction of the mandrel. The fabrics
86 were laid over a mandrel and held under a full vacuum during a cure time of a minimum of 5 hours
88 laboratory scheduling arrangements, the fabricated channels were stored at room temperature for a
90
91 Fibre volume fractions were estimated using the mass of fabric prior to fabrication, the mass of
92 composite after fabrication, and the constituent densities. The fabrication technique generated
93 consistent fibre volume fractions, and for all specimens the mean and standard deviations of the
94 fibre volume fractions were; 39.6% and 1.4% for flax, and 37.0% and 2.2% for jute.
95
97 Tension material test specimens were cut from the flat portions of untested flax and jute channel
98 specimens of four fibre layers in accordance with ISO 527 [25]. Tension specimens were cut at 0º
99 and 45º to the channel longitudinal dimension so as to establish the uniaxial tension and in-plane
100 shear strengths of the [0/90] composites, respectively. In the previous study [24], compression
101 material tests of nominally identical flax and jute channel sections (using the same constituent
102 materials and fabrication procedures as the present study) were undertaken, and these values were
103 used in the present study for the uniaxial compression material properties. Mean material properties
104 are summarised in Table 1 and exemplar tension material stress-strain curves are provided in Fig 1.
105
106 It is noted that the measured tension strengths indicated in Table 1 are relatively low compared with
107 an estimate made with the rule of mixtures. For example, the jute composite has a mean tension
4/34
108 strength of 62.1MPa, while the nominal fibre and matrix strengths are 400MPa and 105.3MPa,
109 respectively. Using the rule of mixtures the estimated tension strength of the composite is 140MPa.
110 There are several possible reasons why the measured strengths are relatively low:
111 a) the twill weave and vacuum assisted fabrication technique can result in voids in the matrix
112 b) the natural fibres contain lumens and sometimes the resin cannot penetrate inside the lumens
113 c) the manufacturers’ data for the fibre strength of 400MPa was used for the rule of mixtures
114 estimate, however it is unclear if this value is for a single elementary fibre or for the yarn
115 d) while the manufacturers’ data indicates the yarn is low-twist, the twist is not zero and during
117 e) the yarn consists of elementary fibres which may move relative to each other under tension
118 While the mechanical properties of the fabricated composites are relatively low, the measured jute
119 value of 62.1MPa is reasonably consistent with the manufacturers’ typical mechanical properties,
120 being 59MPa tension strength for vacuum infused fabrication with an unsaturated polyester matrix
122
124 Flax and jute fibre-resin composite channels were fabricated with nominal geometries of web depth
125 100mm, flange width 50mm and 650mm length. Two different channel thicknesses of 4 and 6 fibre
126 layers were fabricated for each of the different fibre types. In the previous study [24], structural
127 optimisation of the channel section geometry was demonstrated with the inclusion of flange edge
128 stiffeners and intermediate flange and/or web stiffeners. Based on these previous experimental and
129 analytical optimisation results [24], several optimised shapes were considered for the present study,
130 including: one intermediate web stiffener (Figure 2a,b), two intermediate web stiffeners (Figure
131 2c,d), and two intermediate web stiffeners with one intermediate flange stiffener (Figure 2e,f). All
132 specimens contained flange edge stiffeners. For each different intermediate stiffener arrangement,
133 two different sized flange edge stiffeners were used. For all configurations (Figure 2a to f) the
5/34
134 smaller thickness was tested (4 layers of flax or jute), and for the configuration of two intermediate
135 web stiffeners with one intermediate flange stiffener (Figure 2e,f) the larger thickness was
137
138 The different stiffener configurations were fabricated by fixing 12mm diameter half-rounds to the
139 mandrel at specific locations, the mandrel being two 50mm x 50mm steel square hollow (SHS)
140 sections with external corner radius of 6mm. An example mandrel and resulting channel section is
141 exemplified in Figure 3. A split mandrel was required to facilitate extraction of the mandrel after
142 curing. It is noted that 2mm shim was placed between the SHS mandrel members to assist
143 extraction, thus the nominal channel internal web depth was 102mm. For the specimens with two
144 intermediate web stiffeners, the rounded corners of the SHS resulted in a small central stiffener, as
145 some fabric and epoxy was drawn into this space (Figure 3a). This was an unintended artefact of the
146 use of a split mandrel. The stiffeners were centrally located for all elements with one stiffener, and
147 located at the quarter points for elements with two stiffeners (Figure 3b).
148
149 The channels were fabricated with an approximate length of 650mm. The ends did not require
150 trimming due to the test setup used (described in the next section). The flange edge stiffeners were
151 fabricated with approximate length of 40mm, then following curing were trimmed to nominal
152 dimensions of 15mm or 25mm using a high speed rotary tool (Dremel brand), with a carbide cutting
153 wheel. The measured channel geometries are tabulated in Table 2. Exemplar channel specimens are
155
157 The channels were tested in pure flexure using a traditional four-point bending arrangement. The
158 distance between the support points was 600mm, the distance between the support and loading
159 points was 125mm, and the length of the pure flexure region was 350mm (Figure 5a).
6/34
160
161 A steel restraint system was clamped to the composite channels between the support and loading
162 points at both ends to prevent localised flange distortions, web crippling and/or shear failures,
163 restricting the sections to fail in the pure flexure region. The restraint system consisted of two steel
164 channel sections placed over the top and bottom flanges, connected with threaded rods to clamp
165 onto the composite channel sections. Steel rectangular hollow sections (RHS) were inserted inside
166 the composite channels. Additionally, several steel packing plates were inserted between the
167 composite channels and steel channels, while small bolts were inserted through threaded holes in
168 the steel channels and tightened against these packing plates, so as to fully restrain each of the
169 flange and web elements of the composite channels between the steel channel and the steel RHS
170 insert (Figure 5b). Failure occurred in the pure flexure region in all tested channels, demonstrating
171 that the steel restraint system was effective in precluding localised failures outside the pure flexure
172 region.
173
174 An inclinometer was clamped to each steel restraint system to measure the applied end rotations.
175 The channels were loaded in vertical displacement control at a speed of 1mm/min. The channel
176 lengths in the pure flexure region were short enough so as to preclude flexural-torsional buckling,
177 which was confirmed by the experimental results observed, such that the pure section bending
179
182 The machine vertical applied load was converted to applied moment from the geometry of the test
183 setup, and likewise the inclinometer results were converted to curvature. An exemplar moment-
184 curvature plot with photographs taken throughout the bending test is shown in Figure 6, several
7/34
185 exemplar results are plotted in Figure 7 with photographs of the specimens nearing ultimate, and
187
189 The results demonstrated a predominantly linear-elastic response, with some softening prior to
190 failure. Failure was governed by tensile fracturing, initiated in the pure flexure region immediately
191 adjacent to the load point at one end. The fracture initiated in the tension flange and progressed
192 upwards into the web, and thence into the compression flange (Figure 8a). In four jute channels the
193 fracture propagated completely through the compression flange, separating the channel into two
194 pieces (Figure 8b). These tensile fracture failures were sudden and brittle in nature, with only a
195 small amount of softening prior to failure, as evidenced in the moment-curvature plots (Figure 7).
196
197 The softening resulted from a small amount of compression flange buckling, and matrix damage as
198 the material failure stress was approached, resulting in a reduction in the bending stiffness. Small
199 compression flange buckling is evidenced in the photographs in Figure 7 for channels 1 to 4. A
200 reliable technique to measure the compression flange buckling displacement was not found, as
201 difficulties were encountered in the mounting of displacement transducers from relative movements
202 between the mount base and the flange edge that were unrelated to buckling. Generally, all four
203 layered specimens underwent some compression flange buckling, which was more pronounced for
204 the channels with smaller flange edge stiffeners and without flange intermediate stiffeners.
205
207 The moment capacities indicated that despite the jute composite having a slightly higher uniaxial
208 tension stress (Table 1), the flax channels had slightly higher moment capacities. This may result
209 from the fact that the flax composite has greater ductility than the jute (Figure 1), allowing the flax
210 channels to continue resisting load while undergoing matrix damage for longer.
8/34
211
212 For comparisons of structural efficiencies between different flange and web intermediate stiffener
213 arrangements, and different thicknesses, the average ultimate stress is a better measure than ultimate
214 moment, since the addition of stiffeners also involved the addition of extra material. The average
215 ultimate stress (fult) was calculated from the ultimate moment in the test (Mtest) and the full section
216 bending properties, Equation 1, where yfull is the full section distance from the neutral axis to the
217 extreme bending fibre, and Ifull is the full section second moment of area (where the designation of
218 ‘full’ refers to the gross section, as opposed to the effective section discussed in the next section).
219 These values are tabulated in Table 2. The full section second moments of area were calculated
221
𝑀𝑡𝑒𝑠𝑡𝑦𝑓𝑢𝑙𝑙
222 𝑓𝑢𝑙𝑡 = 𝐼𝑓𝑢𝑙𝑙
(1)
223
224 Compared with one web stiffener, the addition of a second web stiffener resulted in a very small
225 increase in average ultimate stress for both the flax and jute channels. The addition of the second
226 web stiffener was designed to increase the strength via restricting compression buckling of the web,
227 however web buckling did not occur in the single web stiffened channels, thus the effect of adding
229
230 Compared with no intermediate flange stiffeners, the addition of an intermediate flange stiffener
231 decreased the average ultimate stress for both the flax and jute channels. Since the bottom flange is
232 in pure tension, and is the location of the initiation of tension fracture, this may result from the
233 flange stiffener acting as an inclusion or disturbed region, weakening the flange and precipitating
234 earlier fracture. This was exacerbated by the fact that the stiffener was on the outside of the section,
235 such that the extreme fibre of the section (and thus the location of maximum stress during bending)
236 was the outer fibre of the stiffener. The addition of the stiffener to the flat flange element was
9/34
237 designed to increase the strength via restricting compression buckling of the flange, however it is
238 apparent that it may actually weaken the section as a result of earlier onset of tension fracture.
239
240 This moment reduction was especially notable in the four layer jute channels with two web
241 stiffeners and intermediate flange stiffeners (Table 2), to the extent that these results seemed to be
242 outliers to the rest of the data. To investigate further, tension material specimens were cut from
243 these two jute channels (from inside the steel restraints, post- bending test) and tested in uni-axial
244 tension, indicating a tension strength of 51.9MPa, notably lower than the average jute value of
245 62.1MPa. It is likely that there was a manufacturing defect in these two channels which were
246 fabricated together (for example incorrect epoxy resin mix, incomplete vacuum pressure, etc, which
248
249 While there were differences in the average ultimate stress values between different stiffener
250 arrangements as noted above, these differences were small (except for the two outlier jute channels
251 with lower material strengths). The highest average ultimate stresses were achieved in the thicker
252 channels for both flax and jute. While they may have been negatively affected by the inclusion of
253 intermediate flange stiffeners, the positive effect of the thickness in stabilising the section against
254 compression flange buckling resulted in an overall small positive effect on the strength.
255
257 The effective width analytical method was developed to predict the post-buckling strength of thin-
258 walled elements, predominantly steel, and forms the basis for the design of cold-formed steel
259 members in several international structural specifications. In the previous investigations by the
260 author, the method was applied to natural fibre composites to predict the compression capacity of
261 plates and channel sections [23,24]. Relatively good agreement with experiments was observed, due
262 to the realistic depiction of the post-buckling mechanics of thin elements; the assumption is made
10/34
263 that the buckled regions become ineffective in carrying load, redistributing the stress to the
264 unbuckled regions, which resist the load until some limit stress is reached (e.g. the yield stress for
265 steel). The limit stress in compression for the composites was taken as the compression matrix
267
268 For design purposes, the method requires a prediction of the buckling stress and an equation that
269 relates the buckling stress to the effective width that ensues at the limit state (the strength equation).
270 The strength equation for natural fibre composites was derived empirically [24] and given by
271 Equations 2 and 3, where ρ is the element effectiveness ratio, λ is the element slenderness and fcr is
273
1.28
274 𝜌=
𝜆1.5
≤1 (2)
𝑓𝑢𝐶
275 𝜆= 𝑓𝑐𝑟
(3)
276
277 The application to sections in bending is identical, except that only the regions of the section in
278 compression are susceptible to buckling. The elastic bucking and effective width equations used in
279 the cold-formed steel structural specification AS/NZS 4600 [27] were followed exactly, as was
280 done in the previous compression design analysis [24]. The exception being that the steel strength
281 equation was replaced with the natural fibre composite strength Equation 2. In pure bending, the
282 elastic buckling coefficient for the web is 24 (rather than 4 for pure compression). The compression
283 flange is in uniform compression for both load cases of pure bending and pure compression, thus
284 the flange analysis follows exactly as previously [24]. In [24] the method is fully described, and
285 several worked examples are provided in the supplementary to the article, thus readers are referred
286 to that article for further details. For the natural fibre composite channels in bending the effective
287 width concept is demonstrated in Figure 9. The compressed portion of the web is shown as fully
11/34
288 effective in Figure 9, since this was calculated to be the case for all composite channels in the
290
291 Having established the effective section (Figure 9), the effective section properties (Ieff, yeff) were
292 calculated using the cross-section analysis software ThinWall, and used in place of the full section
293 values, Equation 4, to analytically predict the moment capacity (MApred). The effective section
294 properties and the calculated moment capacities are tabulated and compared with the experimental
296
𝑓𝑢𝑇𝐼𝑒𝑓𝑓
297 𝑀𝐴𝑝𝑟𝑒𝑑 = 𝑦𝑒𝑓𝑓
(4)
298
299 While the analytical method does not incorporate the effects of matrix damage, the use of the
300 measured ultimate tensile stress (fuT, Table 1) produces good agreement with the experiments, with
301 an average test/predicted ratio of 0.97 and coefficient of variation of 0.06. It is interesting to note
302 that the flax and jute channels with 6 layers were calculated to be fully effective, which indicates
303 that the compression flange did not undergo compression buckling. This result is in agreement with
304 the experimental results, where despite the presence of flange intermediate stiffeners (which were
305 demonstrated to weaken the tension flange), these channels reached the highest average ultimate
306 stress. These channels could be considered as the most structurally efficient geometries of those
308
311 The finite element model approach did not consider the fibres and the resin as discrete and separate
312 elements, and model them separately with their own individual material properties. Rather, the
313 model used a macro-approach, whereby the fibres and resin are modelled as a composite fibre-resin
12/34
314 matrix with a single set of mechanical properties assigned to this matrix. The material properties of
315 the composite fibre-resin matrix were carefully ascertained in accordance with ISO-specified
316 procedures, and the models were prepared in accordance with typical practice for modelling fibre-
317 resin composites. The material properties of the composite fibre-resin matrix used in the models are
319
320 The commercial software ANSYS Workbench version 18.2 [28] was used for the finite element
321 modelling. The channel sections were generated from the measured geometries, and exemplar
322 models are shown in Figure 10a. Parametric studies indicated a mesh size of 5mm was suitable,
323 however in order to accurately articulate the corner radii of the internal stiffeners, a mesh size of
324 3mm was used for all models (Figure 10b). Four-node shell elements with six degrees of freedom
325 per node were used for the mesh. In accordance with the experimental setup and results, the
326 segments of the channels between the load and support points were fully restrained with the steel
327 restraint system, thus these were not required to be modelled. The full 350mm pure flexure length
328 of the channels between the loading points was modelled, and both ends were loaded with enforced
329 pure rotations about the major axis. The reaction moments at the ends were extracted from the
330 results, and the enforced rotation magnitude was increased until such time as the moment reaction
331 decreased beyond the peak value (this magnitude of rotation was defined through several trial
332 values). A large-displacement Newton-Raphson non-linear solution scheme was used, with 10 end
333 rotation steps manually defined and up to 100 sub-steps allowed within each defined step.
334
336 The material models in ANSYS are capable of defining different material properties for each
337 individual fibre layer in a composite layup. Since the present composite layups consisted of
338 multiple layers where each individual layer in the layup had the same fibre orientation [0/90], all
339 layups were modelled as a single layer with a thickness equal to the measured total thickness of the
13/34
340 layup. The material was defined as elastic orthotropic, with identical values in both in-plane
341 directions (due to the 0/90 fibre orientation) based on the measured properties (Table 1), and
342 nominal values of 10% of the in-plane values in the through-thickness direction.
343
344 In order to estimate failure and thereby predict peak moment values, a progressive damage analysis
345 via stress-based damage initiation models and damage evolution laws was used. Stress limits in
346 tension and compression were defined by specifying the measured failure stress values in tension
347 and compression (fuT and fuC in Table 1). The stress limits in shear were defined by specifying the
348 failure stress values measured in the tension material tests at 45º to the fibre direction (fuS in Table
349 1). Strain limits were set at 10% such that strain limits were not invoked in the damage initiation
350 model. The default values for the Tsai-Wu constants were used and the Hashin damage initiation
352
353 Damage evolution was defined with the continuum damage mechanics model in ANSYS. This
354 models the evolution of damage both within an element and throughout the mesh, and is based on
355 energy dissipation, as opposed to the alternative property degradation method which is an instant
356 stiffness reduction method. While this damage evolution model is more robust than stiffness
357 degradation, it requires the definition of material properties that are difficult to quantify. These
358 include the energy dissipated per unit area from: tensile fibre damage, compressive fibre damage,
359 tensile matrix damage, and compressive matrix damage. Each of these values also has an associated
360 damping constant, which dampens the damage accumulation and is designed to reduce convergence
361 issues associated with material softening. The energy dissipated per unit area (G) is defined as [28];
362
𝑈
363 𝐺 = ∫0 𝑓𝑒𝜎𝑒𝑑𝑈𝑒 (5)
364
14/34
365 where σe is the equivalent stress, Ue is the equivalent displacement, and Ufe is the ultimate
366 equivalent displacement where material stiffness is completely lost. For a simple uni-axial stress
367 state, the equivalent values are equal to the actual (uni-axial) values, and the actual displacements
370
371 where ε is the uni-axial strain and Lc is the length of the element in the direction of the stress/strain.
372 Thus for a uni-axial stress state, the energy dissipated per unit area is calculated by integrating the
373 material stress-(strain x Lc) curve up to material fracture. However, material data was not available
374 for the flax and jute composites differentiating between fibre- and matrix-dominated modes, nor
375 material compression stress-(strain x Lc) data up to complete failure. All energy dissipation
376 constants were therefore assumed to be equal to the uni-axial value for tension, which was derived
377 by integrating the flax and jute measured material stress-(strain x Lc) curves (Figure 1). Since all
378 finite element models used a constant element size of 3mm, Lc was taken as 3mm. Accordingly, the
379 calculated energy dissipated per unit area values for flax and jute were: 1.81 N/mm and 1.08 N/mm,
380 respectively. A nominal value for damping of 10% was used for all cases. The complete set of
381 values used for the flax and jute composite material definitions are tabulated in Appendix C.
382
384 The finite element models replicated the moment-curvature response well, with a linear-elastic
385 response up to the initiation of damage, some softening as damage progressed, then failure indicated
386 as a reduction in the bending moment resistance. An exemplar finite element moment-curvature
387 response is compared with the experimental curve in Figure 11. It is noted that the finite element
388 models demonstrated slightly stiffer responses than the tests, likely due to small movements of the
389 composite channels inside the steel restraints in the tests. An exemplar deformed shape at the
390 ultimate state is shown in Figure 12. It is noted that few solution steps were required in the linear-
15/34
391 elastic region, however in the highly non-linear damage phase many more steps were required to
392 correctly identify the softening and failure path (Figure 11b).
393
394 The damage evolution replicated the experimental damage progression well, where damage initiated
395 in the tension flange, then the damaged region grew across the flange and into the web, while the
396 material softened to the point that the section moment resistance began decreasing. With respect to
397 the applied end rotation, the damage progressed quickly to the point of failure (moment reduction),
398 as was found in the experiments. A typical failure progression is shown in Figure 13.
399
400 Despite using the tension value for the energy dissipated per unit area (G) in compression, the
401 progressive damage model performed reasonably well. This may have resulted from the fact that the
402 failure was in fact tension dominated, thus the compression damage values were not especially
403 relevant. The moment capacity results from the finite element models are compared with the
404 experimental moment capacities in Table 2 (Mtest/MFEpred1). The strength of the jute composite
405 channels was predicted well, with all values within 8% of the experimental ones except for the two
406 outliers with possible manufacturing defects discussed previously. For these two defect channels,
407 the measured reduced material failure stress of 51.9MPa was input into the models, which improved
409
410 The strength of the flax channels were conservatively predicted in all cases, with results up to 19%
411 lower than the experimental values (Mtest/MFEpred1). Since all material values were accurately
412 defined based on measured values except for the energy dissipated per unit area (G), this value was
413 empirically adjusted to match the experimental moment capacities. Trial and error found that the
414 energy dissipated per unit area was required to be multiplied by three in order to replicate the
415 ultimate moments. When three times the energy dissipated per unit area (i.e. Gflax = 5.43 N/mm)
16/34
416 was used for all flax channels, the moment capacities matched the experimental values well, and
417 were all within 6%. These values are referred to as Mtest/MFEpred2 in Table 2.
418
419 Using the unadjusted jute models and the adjusted flax models (Mtest/MFEpred2), the moment
420 capacities compared well with the experimental capacities, with a mean test/predicted ratio of 0.99
421 and standard deviation of 0.06 (Table 2). Except for one outlier, the maximum error was 8%. These
422 results indicate that using the measured material properties for the elastic properties and the stress
423 limits, and the calculated energy dissipated per unit area values from Equations 5 and 6 based on the
424 measured tension material properties, provides reasonable agreement with experimental values.
425 However, the energy dissipated per unit area may need to be empirically adjusted in order to obtain
426 excellent agreement. Care should be taken in extrapolating these conclusions to loading situations
427 where the failure is not tension dominated, as the use of the tension energy dissipated per unit area
428 value may not be appropriate for such situations. It is noted that the damage progression is not
429 especially sensitive to the magnitude of the energy dissipated per unit area value; an increase of 2
430 times resulted in an average increase in moment capacity of 8%, while an increase of 3 times
432
434 In the previous studies by the author on natural fibre channel sections in compression [23,24], the
435 compression capacities were noted to be only suitable for light structural applications, due to their
436 relatively modest strength. A potential light structural application identified was as stud columns in
437 residential stud walls, and comparisons with steel stud columns demonstrated suitable capacities.
438 Flax and jute channels fabricated with exactly the same dimensions and fibre thicknesses as those in
439 the present study had compression capacities between 27kN and 69kN, while a commercially
440 produced load-bearing steel stud had a compression capacity of 41kN (Rondo channel section with
441 a nominal web depth of 92mm, flange width of 33mm and thickness 1.15mm from 300MPa steel.).
17/34
442 This same steel stud has a moment capacity of 1.26kNm (as reported by the manufacturer Rondo),
443 while the flax and jute channels had moment capacities between 1.04kNm and 2.51kNm. This
444 indicates that many of the natural fibre channels had compression and bending capacities
445 comparable with, or in excess of, commercial steel channels used as residential stud wall columns.
446
447 To quantify the structural suitability of the present natural fibre channels for residential
448 applications, the Australian design criteria for timber-framed residential buildings was applied (AS
449 1720 [29]). Stud wall columns must satisfy the combined compression and bending interaction
450 Equation 7, where N* is the design compression load (from gravity loads), M* the design bending
451 moment (from lateral wind loads), ϕNu is the compression capacity and ϕMu is the bending capacity.
452 The capacity factors are derived from reliability analyses, and the compression capacity factor was
453 calculated as 0.75 in [24], and following a similar procedure was calculated as 0.8 for the moment
455
𝑁∗ 𝑀∗
456 𝑁 ‒ 𝑀𝑙𝑖𝑚𝑖𝑡 = 𝜙𝑁 + 𝜙𝑀 ≤ 1 (7)
𝑢 𝑢
457
458 To characterise the compression and bending loads the procedures in [29] were followed, with
459 several assumptions, including; a wind region of N3 (this is the second highest non-cyclonic wind
460 category in Australia), 2.7m wall height, sheet roof, plasterboard claddings, 0.6m stud spacings, etc.
461 Such assumptions result in design compression and bending loads of 8.8kN (N*) and 0.41kNm
462 (M*), respectively. Using the compression test capacities from [24] (Nu), the bending test capacities
463 in the present study (Mu), and the capacity factors of 0.75 and 0.8 respectively, the N-Mlimit values
464 for each of the flax and jute channels were calculated from Equation 7 and are tabulated in Table 2.
465 Similarly, the procedures for assessing the lateral stiffness of the stud columns were followed,
466 whereby a serviceability wind pressure was used with the measured E and I values of the channels
467 (Table 2), to calculate the lateral deflection under service wind loads. The deflection limit for a
18/34
468 2.7m high wall is 18mm, and the calculated deflections are listed in Table 2 (M defl). The results in
469 Table 2 indicate that all flax and jute channels satisfied the strength and stiffness limits for the
471
472 It should be noted that the above analyses consider only the structural suitability of natural fibre
473 channels for residential stud walls. Clearly, there are many factors that would need to be addressed
474 before natural fibre composites could be introduced into buildings, for example; fatigue-related
475 matrix damage, column buckling, other loading types, connections, fire performance, environmental
476 exposure and durability, etc., which were outside the scope of the present study.
477
478 7. CONCLUSIONS
479 Public concerns about the environment, climate change, energy consumption and greenhouse gas
480 emissions have placed increasing demands for the use of sustainable materials in the built
481 environment. Natural fibre composites such as those consisting of flax and jute fibres, may one day
482 prove a viable and environmentally sustainable alternative to traditional building materials. The
483 present study of natural fibre channels in pure bending complements previous studies of natural
484 fibre channels in pure compression, and further demonstrates the structural properties of primary
485 structural elements fabricated from flax and jute composites. Material tests indicated that the mean
486 tensile elastic stiffness and strength values were 6,386 MPa and 55.1 MPa for flax, and 6,941 MPa
487 and 62.1 MPa for jute. The ultimate moment capacities varied from 1.043 to 1.501 kNm for four-
488 layered composites, and 2.184 to 2.511 kNm for six-layered composites. While these structural
489 properties are modest, structural suitability for light structural applications such as residential load-
490 bearing stud walls has been demonstrated. Combined axial compression and bending capacities
491 were between 0.38 and 0.87, with no channels exceeding the design limit of 1.0. The suitability of
492 the effective width mechanics model for analytically predicting their compression and bending
493 strengths has been further demonstrated. The analytical models predicted the experimental ultimate
19/34
494 moment capacities with a mean and coefficient of variation of the test to predicted ratio of 0.97 and
495 0.06, respectively. Finite element procedures for predicting their strength via progressive damage
496 analysis have also been successfully demonstrated. The numerical models predicted the
497 experimental ultimate moment capacities with a mean and coefficient of variation of the test to
499
500 8. REFERENCES
501 [1] Summerscales J, Dissanayake NPJ, Virk AS, Hall W. A review of bast fibres and their composites. Part 1 – fibres as
502 reinforcements. Composites: Part A 2010;41:1329-1335.
503 [2] Summerscales J, Dissanayake NPJ, Virk AS, Hall W. A review of bast fibres and their composites. Part 2 -
504 composites. Composites: Part A 2010;41:1336-1344.
505 [3] Ku H, Wang H, Pattarachaiyakoop N, Trada M. A review on the tensile properties of natural fibre reinforced
506 polymer composites. Composites: Part B 2011;42:856-873.
507 [4] Yan LB, Chouw N, Jayaraman K. Flax fibe and its composites – a review. Composites: Part B. 2014;56:296-317.
508 [5] Pickering KL, Efendy MGA, Le TM. A review of recent developments in natural fibre composites and their
509 mechanical performance. Composites: Part A 2016;83:98-112.
510 [6] Faruk O, Bledzki A, Fink H, Sain M. Biocomposites reinforced with natural fibres: 2000 – 2010. Progress in
511 Polymer Science 2012;37:1552-1596.
512 [7] Dicker M, Duckworth P, Baker A, Francois G, Hazzard M. Green composites: A review of material attributes and
513 complementary applications. Composites Part A 2014;56:280-289.
514 [8] Weclawski BT, Fan M, Hui D. Compressive behaviour of natural fibre composite. Composites: Part B 2014;67:183-
515 191.
516 [9] Costa FHMM, D’Almeida JRM. Effect of water absorption on the mechanical properties of sisal and jute fibre
517 composites. Polym Plast Technol Eng 1999;38(5):1081-94.
518 [10] Hargitai H, Racz I, Anandjiwala RD. Development of hemp fiber reinforced polypropylene composites. J Therm
519 Comp Mat 2008;21:165-74.
520 [11] Oksman K. High quality flax fibre composites manufactured by the resin transfer moulding process. J Reinf Plast
521 Comp 2001;20(7):621-7.
522 [12] Van de Weyenberg I, Ivens J, De Coster A, Kino B, Baetens E, Verpoest. Influence of processing and chemical
523 treatment of flax fibres on their composites. Comp Sci Tech 2003;63:1241-46.
524 [13] Lefeuvre A, Bourmaud A, Morvan C, Baley C. Elementary flax fibre tensile properties: Correlation between stress-
525 strain behaviour and fibre composition. Ind Crops Prod 2014;52:762-69.
526 [14] Dittenber D, GangRao H. Critical review of recent publications on use of natural composites in infrastructure.
527 Composites Part A 2012;43:1419-1429.
528 [15] Yan LB, Chouw N. Crashworthiness characteristics of flax fibre reinforced epoxy tubes for energy absorption
529 application. Mat Des 2013;51:629-40.
530 [16] Yan LB, Chouw N. Behaviour and analytical modelling of natural flax fibre reinforced polymer tube encased coir
531 fibre reinforced concrete composite column. J Comp Mater 2013;47(17):2133-48.
20/34
532 [17] Mak K, Fam A, McDougall C. Flexural Behavior of Sandwich Panels with Bio-FRP Skins Made of Flax Fibers
533 and Epoxidized Pine-Oil Resin. Journal of Composites for Construction 2015;19(6):04015005.
534 [18] Duigou AL, Deux JM, Davies P, Baley C. PLLA/flax mat/balsa bio-sandwich manufacture and mechanical
535 properties. Appl Compos Mater 2011;18:421-38.
536 [19] Uddin N, Kalyankar RR. Manufacturing and structural feasibility of natural fiber reinforced polymeric structural
537 insulated panels for panelized construction. Int J Polym Sci 2011;96:35-49.
538 [20] Shah DU, Schubel PJ, Clifford MJ. Can flax replace E-glass in structural composites? A small wind turbine blade
539 case study. Composites Part B 2013;52:172–81.
540 [21] Corradi S, Isidori T, Corradi M, Soleri F. Composite boat hulls with bamboo natural fibres. Int J Mater Prod
541 Technol 2009;36(1):73–89.
542 [22] Alvarez-Valencia D, Dagher HJ, Lopez-Anido RA, Davids WG, Gardner DJ, Center AC. Behavior of natural-
543 fiber/thermoplastic sheet piling. ACMA-composites and polycon conference, January 2009, Tampa, Florida, USA;
544 2009.
545 [23] Bambach MR. Compression strength of natural fibre composite plates and sections of flax, jute and hemp. Thin-
546 Walled Structures, 2017, 119:103-113.
547 [24] Bambach MR. Geometric optimisation and compression design of natural fibre composite structural channel
548 sections. Composite Structures, 2018, 185:549-560.
549 [25] ISO 527-4:1997. Plastics – Determination of tensile properties. Part 4: Test conditions for isotropic and orthotropic
550 fibre-reinforced plastic composites
551 [26] Papangelis JP, Hancock GJ. Computer analysis of thin-walled structural members. Computers & Structures, 1995,
552 56(1):157-176.
553 [27] AS/NZS 4600:2005. Australian/New Zealand standard: Cold-Formed Steel Structures, Standards Australia; 2005.
554 [28] ANSYS Mechanical APDL, Advanced Analysis Guide. SAS IP Inc, 17th ed.; 2016.
555 [29] AS 1720.3:2016. Australian Standard for Timber structures, Part 3: Design criteria for timber-framed residential
556 buildings. Standards Australia, 2016.
557
21/34
NOTATION
22/34
FIGURES
Figure 1: Measured uni-axial tension stress-strain curves (fibre volume fractions of 39% for jute,
and 40% for flax)
a b c d e f
a b
Figure 3: Channel fabrication; a) exemplar fabrication mandrel, b) resulting channel section (B, D,
L are outside dimensions)
23/34
a b c d
Figure 4: Exemplar composite channel specimens: a) 4 layered jute (type Figure 2a); b) 4 layered
flax (type Figure 2d); 4 layered jute (type Figure 2e); 6 layered flax (type Figure 2f)
a b
Figure 5: Experimental four-point bending setup; a) elevation, b) section
Figure 6: Exemplar moment-curvature response (flax 4 layers 2WS 1FS LL) with photographs at
periodic intervals; a) 0kNm, b) 0.88kNm, c) 1.13kNm, d) 1.25kNm, e) 1.44kNm
24/34
Figure 7: Moment-curvature responses of exemplar channels with photographs approaching the
ultimate state; 1) jute 4 layers 1WS SL, 2) jute 4 layers 1WS LL, 3) jute 4 layers 2WS SL, 4) flax 4
layers 2WS LL, 5) flax 4 layers 2WS 1FS SL, 6) flax 4 layers 2WS 1FS LL, 7) flax 6 layers 2WS
1FS SL
25/34
a b
Figure 8: Exemplar failure modes (tension flange is the lower flange); a) 6 layered flax, b) 4 layered
jute
Figure 9: Exemplar effective widths (black areas) at the limit state, and the shift of the major axis
bending neutral axis (towards the tension side) between the full section and the effective section
26/34
a b
Figure 10: ANSYS modelling; a) exemplar finite element model geometries, b) exemplar 3mm shell
mesh used in all models
a b
Figure 11: Finite element (ANSYS) solution for specimen flax 4 layers 2WS LL; a) comparison
with experimental moment-curvature response, b) ANSYS solution points and damage initiation
Figure 12: Exemplar deformation at the ultimate state (true scale); flax 4 layers 2WS SL
27/34
a b c
d e
Figure 13: Exemplar progression of damage (jute 4 layers 2WS 1FS LL); a) 1.385kNm, b)
1.399kNm, c) 1.401kNm, d) 1.380kNm, e) 1.349kNm (red is fully damaged, grey is no damage)
28/34
TABLES
Table 1: Measured material properties of the natural fibre composites (ET is the elastic modulus in
tension, εuT is the failure strain in tension, and fuC, fuT and fuS are the ultimate compression, tension
and in-plane shear stresses, respectively)
29/34
Fibre Fibre Ifull Ieff NA Mtest Mtest Mtest N-M M
Fibre Section layers weight D B L t (x103) ET fuT Mtest fult (x103) shift /MApred /MFEpred1 /MFEpred2 limit defl.
(g/m2) (mm) (mm) (mm) (mm) (mm4) (MPa) (MPa) (kNm) (MPa) (mm4) (mm) (mm)
Jute 1WS, SL 4 1600 110.4 57.4 17.9 3.4 1478 6941 62.1 1.189 44.4 1344 3.4 1.01 0.97 0.97 0.84 13.3
Jute 1WS, LL 4 1600 110.9 58.2 27.2 3.4 1548 6941 62.1 1.229 44.0 1364 4.8 1.06 0.92 0.92 0.78 12.7
Jute 2WS, SL 4 1600 112.5 57.9 16.2 3.4 1546 6941 62.1 1.234 44.9 1439 2.6 0.98 0.98 0.98 0.87 12.7
Jute 2WS, LL 4 1600 112.0 57.8 26.4 3.4 1622 6941 62.1 1.301 44.9 1434 4.6 1.07 0.95 0.95 0.82 12.1
Jute 2WS, 1FS, SL 4 1600 109.9 57.3 16.2 3.4 1625 6941 51.9 1.109 37.5 1492 3.1 1.03 1.02 1.02 0.87 12.1
Jute 2WS, 1FS, LL 4 1600 109.2 57.4 26.8 3.4 1698 6941 51.9 1.043 33.5 1485 4.9 0.99 0.85 0.85 0.83 11.5
Jute 2WS, 1FS, SL 6 2400 115.1 61.3 16.6 5.5 2764 6941 62.1 2.216 46.1 2764 0.0 0.90 1.03 1.03 0.46 7.1
Jute 2WS, 1FS, LL 6 2400 115.0 61.4 25.8 5.5 2899 6941 62.1 2.511 49.8 2899 0.0 0.97 1.07 1.07 0.41 6.8
Flax 1WS, SL 4 1600 109.6 57.3 17.3 3.3 1434 6386 55.1 1.303 49.8 1354 2.2 0.92 1.12 1.01 0.82 14.8
Flax 1WS, LL 4 1600 109.9 57.5 27.7 3.3 1502 6386 55.1 1.351 49.4 1337 4.5 0.98 1.09 0.94 0.77 14.2
Flax 2WS, SL 4 1600 111.9 57.2 17.5 3.3 1501 6386 55.1 1.393 51.9 1429 2.8 0.94 1.16 1.05 0.79 14.2
Flax 2WS, LL 4 1600 111.0 57.7 27.0 3.3 1575 6386 55.1 1.447 51.0 1447 4.3 0.97 1.11 0.97 0.71 13.5
Flax 2WS, 1FS, SL 4 1600 108.8 57.1 15.4 3.3 1577 6386 55.1 1.355 46.7 1471 2.6 0.87 1.14 0.99 0.75 13.5
Flax 2WS, 1FS, LL 4 1600 108.4 57.0 25.2 3.3 1648 6386 55.1 1.501 49.4 1447 4.9 1.00 1.15 1.05 0.70 12.9
Flax 2WS, 1FS, SL 6 2400 113.5 59.8 15.7 5.1 2512 6386 55.1 2.184 49.3 2512 0.0 0.90 1.14 0.99 0.44 8.5
Flax 2WS, 1FS, LL 6 2400 114.0 60.1 24.2 5.1 2635 6386 55.1 2.489 53.8 2635 0.0 0.98 1.19 1.01 0.38 8.1
Mean: 0.97 1.06 0.99
COV: 0.06 0.10 0.06
Table 2: Natural fibre composite channel section dimensions and test results compared with analytical and finite element predictions (dimensions refer
to Figure 3, MApred refers to analytically predicted values, MFEpred refers to finite element predicted values, f refers to stress, M refers to moment, I refers
to second moment of area, ET is the tension elastic modulus) (WS is web stiffener, SL is shorter length flange edge stiffener, LL is longer length flange
edge stiffener and FS is internal flange stiffener)
30/34
APPENDIX A
This Appendix describes the commercially produced fabrics used to fabricate the composite channel
sections, and summarises relevant manufacturers data for the fabric and the epoxy used. The fabrics
were commercially produced by Composites Evolution (compositesevolution.com), specifically for
fibre-resin composites manufacture: Biotex Flax 400g/m2 and Biotex Jute 400g/m2. Both fabric
products are a 2x2 Twill weave, which is a generic pattern consisting of bi-directional yarns woven
over-over-under-under, where the yarns are perpendicular (i.e. 0º/90º), as demonstrated in Figure
A1 (a). Examples of the commercial fabric are shown in Figure A1 (b, c). Composites Evolution
provides technical data for the Biotex fabrics as summarised in Table A1. The commercial epoxy
resin Kinetix R240 with H126 (fast) hardener was used for all composite fabrications. The epoxy is
a room temperature out-of-autoclave cure epoxy resin that does not require heat input to cure.
Kinetix provides technical data for Kinetix R240 with H126 as summarised in Table A2.
a) b) c)
Figure A1: 2x2 Twill weave fabric: a) generic pattern, b) Biotex Jute 400g/m2 c) Biotex Flax 400g/m2
31/34
APPENDIX B
This Appendix contains an example calculation of the moment capacity using AS/NZS4600 [27]
and the procedures outlined in Section 4. Further examples of effective width calculations for
natural fibre channels to AS/NZS4600 are provided in the supplementary file of [24]. The following
calculation is for a four layered flax channel with one intermediate web stiffener and shorter flange
edge stiffener (Figure B1, Table 2), with relevant notation:
b is the flat width of the flange, excluding the corner radii (r)
dL is the flat width of the flange edge stiffener, excluding the corner radii (r)
Is is the second moment of area of the flange edge stiffener
Ia is the limit value for the second moment of area of the flange edge stiffener
k is the buckling factor of the flange
fcr is the buckling stress of the flange
λ is the slenderness of the flange
ρ is the effectiveness factor of the flange
beff is the effective width of the flange
dLeff is the effective width of the flange edge stiffener
EC is the compression elastic modulus
fuC is the maximum compression stress
fuT is the maximum tension stress
Ieff is the effective section second moment of area
yeff is the effective section neutral axis position
MApred is the analytically predicted moment capacity of the effective section
𝑡 𝑥 𝑑3𝐿
𝐼𝑠 = = 141𝑚𝑚3
12
𝐼𝑎 = 399𝑡4 ( 𝑏
𝑆𝑥𝑡 ) (
‒ 0.328 ≤ 𝑡4
115𝑏
𝑆𝑥𝑡 )
+ 5 = 15715𝑚𝑚4
𝑏
𝑛 = 0.582 ‒ ≥ 0.33 = 0.33
4𝑆 𝑥 𝑡
32/34
𝑛
𝑘= ( )(
𝐼𝑠
𝐼𝑎
4.82 ‒
5𝑑𝐿
𝑏 )
+ 0.43 = 0.97
𝑘𝜋2𝐸 𝑡2
𝑓𝑐𝑟 = = 27𝑀𝑃𝑎
12(1 ‒ 𝜐2)𝑏2
𝑓𝑢𝐶
𝜆= = 1.52
𝑓𝑐𝑟
1.28
𝜌= = 0.68
𝜆1.5
𝑏𝑒𝑓𝑓 = 𝜌 𝑥 𝑏 = 26.5𝑚𝑚
𝐼𝑠
𝑑𝐿𝑒𝑓𝑓 = 𝑑𝐿 = 0.1𝑚𝑚
𝐼𝑎
The web is fully effective. The effective section was entered into the cross-section analysis software
ThinWall, providing the following:
Figure B1: Flat widths of the flange (b) and flange edge stiffener (dL) of a channel with one
intermediate web stiffener
33/34
APPENDIX C
34/34