Manufacturing controlled asymmetric wrinkles for experimental studies of the effect of defects

Abstract

The strip placement method was used to manufacture nine carbon fibre reinforced epoxy laminates that were tested in bending to study the influence of asymmetric wrinkles on strength. Due to the accuracy of the process and material flow of the material during consolidation, the achieved wrinkles did not match the intended wrinkles in terms of maximum fibre angles or degree of asymmetry, highlighting the difficulty of creating controlled wrinkles in flat coupon specimens. Nevertheless, the experimental data showed decreasing ultimate bending strength with increasing maximum angle and increasing strength with increasing asymmetry. Therefore, for samples with the same maximum angle, the sample with higher wrinkle asymmetry had a higher ultimate bending failure strength, highlighting the importance of considering wrinkle asymmetry alongside the often-used maximum angle.

Simulations of the as-designed coupons were also run to design the experiment. In terms of the relevance of asymmetry, both the numerical and experimental data are in line with previous numerical studies on asymmetric wrinkles under pure tension and compression. Contrasting the as-designed wrinkle morphologies to the manufactured wrinkles, it was found that overall, the study highlights the complexity and difficulty of creating controlled wrinkles within coupon specimens and provides evidence for the relevance of wrinkle shape asymmetry when characterising and simulating wrinkles across multiple load cases.

Keywords

Composites defects characterisation strength

Introduction

The major challenge in composite manufacturing is mitigating and controlling for lay-up features that reduce the mechanical performance of finished parts. Due to the brittle nature of composite materials, stress concentrations caused by micro- and meso-scale deviations can result in premature and sudden failure of the component. This consideration drives both the design and manufacturing decisions that generally lead to heavier components than the theoretical properties of composite materials, resulting in more expensive and slower manufacturing approaches.

As the primary load-carrying constituent, fibre placement according to the engineering drawings is often the most critical manufacturing step. Breakage of fibres, variations in fibre volume fraction, and deviations in fibre path are possible during lay-up and consolidation, with the exact nature and variations being associated with specific raw material forms and deposition processes. Across the literature, out-of-plane wrinkles are the focus of many studies as they occur commonly and induce through-thickness stresses, which act in the weakest direction for traditional layered composite materials. For this reason, as the available literature shows, there has been significant research to understand wrinkles¹ and their causes,^2–4 their impact on the performance of laminates,^5–7 and how to detect and characterise them.⁸

Wrinkling formation

Wrinkles are out-of-plane deviations in the fibre that generally occur at the ply level, with the entire layer shifting and, depending on the size of the wrinkle, the thickness of the laminate, and the deposition process, can deform subsequent plies. Conversely, in-plane fibre waviness may only occur in a portion of the fibres, and there is limited to no effect on other plies. The driving cause behind wrinkles is compression along the fibre length, combined with insufficient compaction force that would otherwise act to flatten the wrinkle into in-plane waviness. Excess length, where the fibres are consolidated into a shorter path than the initial deposition length, is a common driver alongside thermal contraction and chemical cure shrinkage, which can cause the fibres to experience compression strains. While excess length could be resolved by in-plane waviness, due to the aspect ratio of composite materials (i.e. ply thickness being many orders of magnitude smaller than width), the lowest energy solution is for the fibre bundles to bend and move out-of-plane unless pressed down by a significant external force.

Some research activities, like those of Belnoue et al., have demonstrated that there are cases of composite parts where wrinkles can be successfully simulated and the process redesigned, for instance, through modifying the caul plate, to enable the part to be made without wrinkles.^9,10 However, as this is not necessarily universally possible, the occurrence of wrinkles in existing and future parts means that it is important to be able to investigate the effect of wrinkles and use this data to build models to make predictions of the impact of wrinkles. Recent work has focused either on controlled wrinkles that are made in non-natural manners or on leveraging simulation capabilities to predict ‘natural’ defect formation.

The mechanisms for inducing controlled wrinkles are limited, and generally, they lack the complete morphological complexity of naturally occurring defects.¹ Some approaches, including the work by Potter et al.¹¹ and later Bloom et al.¹² did succeed in generating wrinkles with more complex features, by using a method that approximates ‘natural’ forming mechanisms. Specifically, this approach generates a crease in one ply to induce wrinkling in the plies above. With this approach, it is possible to create morphologies that show amplitude decay and asymmetry, both of which are common in naturally occurring wrinkles. However, the main limitation is the difficulty in controlling the process accurately, which affects the repeatability needed for validation of numerical models. Further, the wrinkles formed vary across the width of the laminate, so that a panel cut up into test samples would produce no two nominally identical coupons.

Alternative approaches can be found in the study of Çinar and Ersoy,¹³ who made use of natural compaction effects when forming into a corner to induce waviness in L-brackets. The morphology was controlled to some degree by varying the number of plies formed into the radius at one time. While this method creates repeatable and realistic waviness patterns (for L-brackets or similar geometries), the reliance on a specific forming action limits the morphologies that can be achieved. While an ‘ideal’ method would allow wrinkles morphologies as naturally occurring in any manufacturing process to be recreated, this is not necessarily possible due to the wide variety of deposition processes and the other variations (e.g. fibre waviness, changes in fibre volume fraction) that these might cause in addition to wrinkling, which no singular method may easily recreate.

The most widely used approach to forming controlled wrinkles is to use the strip technique,^14–17 where strips of material are added and removed to induce wrinkling in the adjacent plies. This approach is simple and has been demonstrated by various researchers to be capable of producing wrinkling of various sizes and morphologies by adjusting the strip widths and positions. Moreover, this method appears capable of generating constant and repeatable wrinkling in flat plates that allow the wrinkle to be studied separate from the geometry that would normally drive its occurrence, facilitating evaluation of the wrinkles influence on basic measures of strength.

Impact on laminate strength

While wrinkles are broadly understood to be detrimental to strength, a significant fraction of the available data is simulated. Driven by the costly nature and complexity of generating controllable wrinkles, the studies providing experimental data are limited. Mukhopadhyay et al. did provide data for sinusoidal wrinkles both in compression¹⁶ and tension,¹⁸ showing that strength correlated negatively with maximum wrinkle angle. Further work was done by Ju et al.,¹⁹ though here the strength values were shown to correlate negatively with the aspect ratio of the wrinkle rather than the maximum wrinkle angle. While the two metrics are loosely related for simple morphologies, for more complex wrinkle morphologies, the aspect ratio and maximum angle do not share a direct correlation that would allow the datasets to be combined easily. Others, including Xu et al., showed the influence of wrinkles in a limited number of cases for curved components under flexural load.²⁰

A recent study on wrinkled specimens meant to recreate the effect of tow gaps in automated fibre placement (AFP) showed a clear difference between a laminate with a wrinkle and without, but no clear difference between wrinkles with different severities, under 3-point bending.²¹ At the same time, another study using similar approaches showed a more gradual influence on normalised failure load.²² This variability in results is not easily resolved, since in both cases, results are provided against qualitative ‘severity’ rankings rather than metrics describing the wrinkles. This inconsistency in the characterisation of wrinkles complicates the ability of the literature to come together and allow for reliable correlation of wrinkle metrics to laminate strength knockdowns.

Wrinkle characterisation

In the literature, wrinkles are commonly characterised by looking at the fibre path and extracting three key metrics, shown in Figure 1. In many studies, only the maximum wrinkle angle, $θ$ , is used to describe the wrinkles. The aspect ratio, also referred to as the severity factor, is calculated as the ratio between the maximum amplitude, A, and the wavelength, $λ$ , and is the second most common metric.

Figure 1.

Schematic showing the tracing of wrinkle with the key metrics generally measured, including amplitude, A, wavelength, $λ$ , and maximum wrinkle angle, $θ$ , alongside additional metrics considered in this work: mid-point, M, peak point, P, and offset, $δ$ , used for calculating asymmetry.

However, the literature shows many examples of realistic wrinkles whose morphologies are not readily captured by only these metrics. Wrinkles such as those studied by Weber et al.²³ when evaluating the influence of caul sheets, or those observed by Hallander et al.²⁴ when studying novel forming methods like Hot Drape Forming, or those observed by Sebastian²⁵ studying glass fibre bridge decks, all contain more complex morphologies. A recent literature review on feature and defect characterisation suggested several additional possible characteristics, including asymmetry in the form of a skew factor.¹ However, in order to justify developing further metrics and gathering data on them from existing and new samples, their relevance to understanding and predicting the behaviour of wrinkles needs to first be demonstrated.

Research gap

In a previous numerical study, the influence of wrinkle asymmetry on tensile and compressive failure load was assessed.²⁶ In this study, manufacturing of asymmetric wrinkles using the strip technique^14–17 is undertaken to test the feasibility of creating such wrinkles in a controlled manner, and to attempt experimental validation of the relevance of asymmetry on laminate strength. To this end, the specimens are tested under 4-Point Bending (4PB) and an accompanying macroscale model is developed to evaluate the influence of wrinkle asymmetry on ultimate bending moment.

Methodology

Sample design

The samples tested in this work were designed and manufactured on the basis of the tensile failure data from a previous study²⁶ and follow the ASTM D3039²⁷ guidelines for balanced and symmetric configurations. The samples are nominally 4 mm thick, 25 mm wide, and 250 mm long, with a cross-ply lay-up. In total, nine combinations of aspect ratio and maximum angle were selected, as listed in Table 1, to create wrinkle morphologies that shared either aspect ratio or maximum angle but never both, while covering a wide range of maximum angles and degrees of asymmetry. No baseline panel without a wrinkle was produced.

Table 1.

Wrinkle parameters for sample groups (G) both as intended and as achieved in experimental samples.

	Aspect ratio [-]		Skew factor [-]		Maximum angle [°]
Sample	Intended	Achieved	Intended	Achieved	Intended	Achieved
G1	0.015	0.011	0	0.06	2.70	2.04
G2	0.050	0.034	0	0.03	8.73	6.20
G3	0.100	0.055	0	0.04	17.44	10.02
G4	0.200	0.092	0	0.08	32.14	16.71
G5	0.050	0.033	0.5	0.08	17.44	6.14
G6	0.100	0.061	0.5	0.06	32.14	11.12
G7	0.100	0.053	0.7	0.01	46.32	9.53
G8	0.200	0.094	0.4	0.07	46.32	17.01
G9	0.200	0.085	0.7	0.16	64.48	16.22

The maximum angle is computed from the assumed sinusoidal shape and on the basis of the aspect ratio and skew factor. The aspect ratio is the maximum amplitude, A, over the wavelength, $λ$ , and the skew factor is the offset, $δ$ , between the peak, P, and the middle of the wavelength, M, normalised by half the wavelength, see Figure 1 for example metrics.

Sample manufacturing

The samples were manufactured as 300 × 300 mm cross-ply laminates using 32 plies of IM7/8552 unidirectional 134 gsm carbon fibre epoxy prepreg in a symmetric stacking sequence: $[90_{2} / 0_{2}]_{4 s}$ . To induce ‘controlled’ wrinkles, the strip technique^14–17 was used, as shown in Figure 2. For each wrinkle configuration, strips were cut from one of any pair of 90° plies above the first set and moved between the first two 90° plies. This causes a local increase in thickness at the bottom of the laminate, deflecting all above plies upwards, recreating wrinkle behaviour in the 0° plies without altering the overall thickness. In theory, this approach also maintains constant laminate thickness and fibre volume fraction, though flow during consolidation can, in reality, lead to resin-rich areas and locally thinner layers with a net higher fibre volume fraction. Material flow can also cause a reduction in the amplitude and an increase in the wavelength of the wrinkle as the 0° fibres attempt to straighten. Given that the asymmetry would induce higher strain energies in the fibre paths on one side, it can also be understood that any flow of the 90° material would also have a tendency to alleviate this imbalance and hence reduce the degree of asymmetry. As such, it is understood a priori that wrinkles will end up being less asymmetric than initially designed.

Figure 2.

Picture from an example plate where strips of 90° plies are stacked to induce a wrinkle in the first 0° ply and the remaining of the width filling the gap (right).

The laminates were laid up on an aluminium tool plate, and silicone top sheets were used to create a smoother upper surface. Autoclave curing, following the material manufacturer's recommended cure cycle, was carried out, with all laminates on a single large tool plate and under a single vacuum bag to achieve a consistent cure history across all laminates. After curing, the plates were cut into coupons 25 mm wide and 250 mm long, and the plate edges were used to obtain micrographs of the wrinkles.

Experimental testing

Testing of the coupons was done using 4PB, as seen in Figure 3, following the ASTM D6272 standard²⁸ with a support span of $160 mm$ and a load span to support span ratio of 1:2. This ratio was chosen to maximise the bending moment arm and hence minimise the shear load needed to reach the failure bending moment. This was seen as mitigating the risk of local crushing of the coupon surface during loading. A pre-made 4PB jig with a maximum support span of $250 mm$ , support width of $50 mm$ , and $5 mm$ radius support points was used in a $250 kN$ Instron Universal Hydraulic Testing machine using a $10 kN$ load cell. Loading was carried out at a target cross-head displacement of $10.7 mm / \min$ , as specified by the standard, to achieve a rate of straining of the outer fibres of $0.01 (mm / mm) / \min$ .

Figure 3.

Video gauge frame showing a sample loaded in 4 point bending rig.

During the testing, videos were taken, and the cross-head displacement was tracked along with the applied load measured using the load cell. The load cell data was converted to the effective applied moment using:

\begin{matrix} M_{E} = \frac{1}{8} F L_{s}, \end{matrix}

(1)

where the experimental moment,

M_{E}

, is computed from the applied load, F, and the full support span length,

L_{s}

Numerical simulation

As the initial numerical studies, which identified asymmetry as a relevant factor, only evaluated unidirectional tension and compression behaviour in RVEs (Representative Volume Element), additional macro-scale bending models were built to provide simulation results for direct comparison to the experimental data. These macro-scale models were built on the basis of the same simulation tools developed elsewhere.²⁹ The macroscale model, shown in Figure 4, represents the gauge length of the bending specimen with the wrinkle placed in the centre. The mesh consists of C3D8 elements using a single element per layer thickness and with a $0.001 mm$ thick cohesive zone element between plies. In the central region where the wrinkle sits, the elements are $0.25 mm$ long and $1 mm$ wide, giving an aspect ratio of thickness to length to width of 1:2:4, while outside the central area, the elements are $1 mm$ long and wide, giving aspect ratios of 1:4:4.

Figure 4.

Example mesh for sample G7 with refined mesh in the central area containing wrinkle and coarse mesh in remainder of sample. Control reference points and coupled nodes used for applying rotational velocity highlighted.

For loading, a pure bending rotation is specified as a rotational velocity of $2 rad / s$ on the edge boundary faces using a control reference point that is constrained using kinematic coupling of all degrees of freedom, see Figure 4. The simulation is run as an explicit dynamic with a whole model mass scaling targeting a time increment of 1e-6 for a total time period of 1. This model differs from the previous study²⁶ in that it does not apply the periodic boundary constraints, as it is not an RVE; however, it does utilise the same material model and dynamic analysis to capture the onset and progression of failure.²⁹

Post simulation, the reaction moments at the boundary reference nodes were extracted and used to predict the ultimate failure bending moment. As the simulation indicated a more pronounced reduction in the ultimate failure bending moment and influence of asymmetry when bending was applied in a positive sense (i.e. when the bottom half, containing the wrinkle, is in tension), this was the scenario tested during the experiments. These numerical results are in line with those of the simulations under pure tension and compression, which also showed an increased reduction in ultimate failure strength and influence of the asymmetry under tension loading.²⁶

Results and discussion

Designed versus manufactured wrinkles

In order to assess the degree to which the intended wrinkle morphologies were achieved, the micrographs of both plate edges, as well as secondary NDT (Non-Destructive Testing) data gathered from the samples,^30,31 were used to extract the key measures, reported in Table 1. Reviewing these metrics, it is clear that the targeted morphologies were not fully achieved, with increasing deviation for the more severe wrinkle morphologies. Specifically, it can be seen that across the board, the values of three metrics are well below those targeted. The aspect ratios are lower due to both an increased wavelength and reduced amplitude of the achieved wrinkle, implying a flattening and spreading out of the wrinkles. Finally, the skew factors show both a lower than intended degree of asymmetry for those samples with intended asymmetric wrinkles, G5–G9, but small amounts of asymmetry for those samples intended to be symmetric. The fact that across all samples, the achieved wrinkles deviate from the intended wrinkles, speaks to inherent limitations of attempting to create specific wrinkles. For wrinkles where the intended metrics are small, limits in the accuracy of the placement of the strips, as well as the ability of discrete strips to approximate the continuous idealised wrinkle shape, result in achieved values that deviate from the intended values. On the other side of the spectrum, for wrinkles with larger intended metrics, the consolidation and curing processes provide an opportunity for the 90° material in the strips to flow, resulting in final achieved wrinkles that can be drastically different from the initial morphology, which itself may have had discrepancies similar stemming from strip placement and shape approximation. It is likely that for the current approach, the intended wrinkles were never achievable.

Comparing the microscope image of the G9 sample and contrasting it to the morphology used for the simulation, see Figure 5, it is clear to see how the achieved wrinkle deviates from the as-designed morphology. The difference is likely caused by a flow of the 90° material build-up below the first 0° plies, as anticipated and discussed in the ‘Sample manufacturing’ section. The image also highlights a minor thickness variation of the plate across the wrinkle, while the targeted morphologies were constant in thickness. This is a consequence of the use of a silicone top-sheet and could be mitigated by using a metallic caul-plate.

Figure 5.

Micrograph image of plate edge for sample G9. The intended wrinkle and external surfaces are overlaid as dashed white lines.

Impact of asymmetry on performance

The experimental results of the actual wrinkles as well as the simulated results of the as-designed wrinkle morphologies are summarised in Table 2 and shown in Figure 6. For the simulated results, this includes both positive bending, where the wrinkle is put under tension and which was the case for the experimental tests, and negative bending, where the wrinkle is put under compression. For the as-manufactured samples, it is possible to observe that for the less severe wrinkles the failure bending moments level off and are about half that predicted numerically at similar angles. Such a difference may stem from damage initiation triggered by secondary effects, such as stress concentrations and damage at the edges, which have been shown to impact strength values in that order of magnitude.³² It is also worth noting that a plateauing strength at lower wrinkle angles has also been observed in other experimental studies,^16,18,33 and the disagreement may be an issue in simulation input parameters controlling the failure conditions; however, this has not yet been investigated due to the large amount of parameters in the model. With increasing maximum angle, as the wrinkle becomes more dominant as the failure initiator, the experimental data and numerical predictions move closer together. For all data sets, similar maximum wrinkle angles show increased skew factor correlating with an increased ultimate bending failure moment. The simulation data shows similar exponential trends in both loading cases, with a rapid drop for wrinkles with a maximum angle up to 20°, followed by more gradual failure load reduction for ever-increasing maximum angles. In both loading cases, it can also be seen that the scenarios of greater asymmetry do show greater maximum bending moment, being between roughly 10% and 60% larger, depending on the pairs used for comparison.

Figure 6.

(a) Ultimate bending failure moment from experiments (orange), including standard deviation shown as error bars, as well as from simulations under positive bending (blue) and negative bending (red) plotted against maximum wrinkle angle, and (b) zoomed in plot. Skew factor is indicated using the marker size, with larger markers indicating larger skew factor values.

Table 2.

Experimental data for sample groups (G) both as intended and as achieved.

	Intended wrinkle				Achieved wrinkle
Sample	Skew [-]	Maximum angle [°]	+’ve bending moment [MNmm]	-’ve bending moment [MNmm]	Skew [-]	Maximum angle [°]	+’ve bending moment [MNmm] (STD)
G1	0	2.70	0.178	0.163	0.06	2.04	0.080 (0.003)
G2	0	8.73	0.108	0.107	0.03	6.20	0.077 (0.001)
G3	0	17.44	0.067	0.055	0.04	10.02	0.071 (0.001)
G4	0	32.14	0.043	0.051	0.08	16.71	0.046 (0.010)
G5	0.5	17.44	0.081	0.069	0.08	6.14	0.078 (0.010)
G6	0.5	32.14	0.065	0.056	0.06	11.12	0.066 (0.003)
G7	0.7	46.32	0.062	0.059	0.01	9.53	0.069 (0.001)
G8	0.4	46.32	0.039	0.049	0.07	17.01	0.051 (0.004)
G9	0.7	64.48	0.037	0.048	0.16	16.22	0.050 (0.004)

Conclusions

In this study, the numerical finding from a previous study²⁶ were supplemented with further numerical simulations and experimental data to highlight the relevance of wrinkle asymmetry in determining laminate strength. The additional numerical simulations consisted of macro-scale models of the gauge length of the 4-point bending specimens, utilising the same damage initiation and progression models deployed in the previous study. The experiments were run using samples manufactured as plates using the strip placing method,¹⁴ so as to create wrinkles with pairwise similar maximum wrinkle angles or aspect ratios but with different levels of asymmetry. The specimens were tested in 4-point bending so as to put the base of the wrinkle in tension, as this was found in the simulation to cause the most pronounced influence of the wrinkle asymmetry.

Microscope imaging and NDT carried out in other studies^30,31 revealed that the manufactured wrinkles did not match those designed. As-manufactured wrinkles deviated in terms of wrinkle wavelength, amplitude, and skew factor (i.e. degree of asymmetry). Overall, the achieved wrinkles had an increased wavelength and reduced amplitude and skew factor. As a consequence, the maximum angles of the wrinkle morphologies were also significantly lower across all samples. Samples designed to have more severe wrinkles also showed minor variation in thickness across the wrinkle. As such, the strip method approach was not successful in producing these more complex wrinkles in a controlled manner. This is assumed to be due to increased wrinkle severity and asymmetry generating strains in the fibres that, during consolidation, result in material flow leading to reduced asymmetry and wrinkle severity.

While the results do indicate characterisation of wrinkle morphologies should be expanded to include measures of skew, alternative approaches to manufacturing complex wrinkles in a controlled manner are also clearly needed. This may require thinner plies for the strips and/or moulds to create more consistent and accurate morphologies and solid caul plates to create more consistent laminate compaction and homogeneous thickness. The ability to manufacture intentional wrinkles, as well as other defects, is crucial to help address the open question around what further defect characteristics might be relevant for high-fidelity simulations,³⁴ especially within the context of virtual testing and other digitisation developments.¹

Footnotes

ORCID iDs

Vincent K. Maes

James Kratz

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is supported by the UK Engineering and Physical Sciences Research Council (EPSRC) through Programme Grant: ‘Certification of Design: Reshaping the Testing Pyramid (CerTest – ), EP/S017038/1. The funding is gratefully acknowledged.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Maes

Potter

Kratz

. Features and defects characterisation for virtual verification and certification of composites: a review. Compos B Eng 2022; 246: 110282.

Hassan

Othman

Kamaruddin

. A review on the manufacturing defects of complex-shaped laminate in aircraft composite structures. Int J Adv Manuf Technol 2017; 91: 4081–4094.

Thor

Sause

MGR

Hinterhölzl

. Mechanisms of origin and classification of out-of-plane fiber waviness in composite materials—A review. J Compos Sci 2020; 4: 30.

Kulkarni

Mali

Singh

. An overview of the formation of fibre waviness and its effect on the mechanical performance of fibre reinforced polymer composites. Compos Part A Appl Sci Manuf [Internet] 2020; 137: 106013.

Senthil

Arockiarajan

Palaninathan

, et al. Defects in composite structures: its effects and prediction methods - A comprehensive review. Compos Struct [Internet] 2013; 106: 139–149.

Centea

Grunenfelder

Nutt

. A review of out-of-autoclave prepregs - material properties, process phenomena, and manufacturing considerations. Compos Part A Appl Sci Manuf [Internet] 2015; 70: 132–154.

Alves

Cimini Junior

. Fiber waviness and its effect on the mechanical performance of fiber reinforced polymer composites: an enhanced review. Compos Part A Appl Sci Manuf 2021; 149: 106526.

Chen

Jin

. Nondestructive testing and evaluation techniques of defects in fiber-reinforced polymer composites: A review. Front Mater 2022; 9: 1–29.

Belnoue

JPH

Hallett

. A rapid multi-scale design tool for the prediction of wrinkle defect formation in composite components. Mater Des [Internet] 2020; 187: 108388.

10.

Wang

Chen

Tretiak

, et al. Virtual data-driven optimisation for zero defect composites manufacture. Mater Des 2024; 241: 112934.

11.

Potter

Khan

Wisnom

, et al. Variability, fibre waviness and misalignment in the determination of the properties of composite materials and structures. Compos Part A Appl Sci Manuf [Internet] 2008; 39: 1343–1354.

12.

Bloom

Wang

Potter

. Damage progression and defect sensitivity: an experimental study of representative wrinkles in tension. Compos B Eng [Internet] 2013; 45: 449–458.

13.

Çinar

Ersoy

. Effect of fibre wrinkling to the spring-in behaviour of L-shaped composite materials. Compos Part A Appl Sci Manuf 2015; 69: 105–114.

14.

Wang

Potter

Hazra

, et al. Experimental fabrication and characterization of out-of-plane fiber waviness in continuous fiber-reinforced composites. J Compos Mater [Internet] 2012; 46: 2041–2053.

15.

Wang

Potter

Etches

. Experimental investigation and characterisation techniques of compressive fatigue failure of composites with fibre waviness at ply drops. Compos Struct 2013; 100: 398–403.

16.

Mukhopadhyay

Jones

Hallett

. Compressive failure of laminates containing an embedded wrinkle; experimental and numerical study. Compos Part A Appl Sci Manuf 2015; 73: 132–142.

17.

Xie

Smith

Mukhopadhyay

, et al. A numerical study on the influence of composite wrinkle defect geometry on compressive strength. Mater Des [Internet] 2018; 140: 7–20.

18.

Mukhopadhyay

Jones

Hallett

. Tensile failure of laminates containing an embedded wrinkle; numerical and experimental study. Compos Part A Appl Sci Manuf [Internet] 2015; 77: 219–228.

19.

Xiao

Wang

, et al. Effect of gaps/overlaps induced waviness on the mechanical properties of automated fiber placement (AFP)-manufactured composite laminate. Mater Res Express [Internet] 2022; 9: 045305 .

20.

Jones

Ali

, et al. Effect of out-of-plane wrinkles in curved multi-directional carbon/epoxy laminates. Compos Sci Technol 2020; 197: 1–25.

21.

Tang

Wang

, et al. Influences of wrinkles induced by tows’ offset in automated fiber placement on the bending properties of composites. Compos Struct 2022; 297: 115964.

22.

Kulkarni

Mali

Singh

. Experimental investigations on flexural performance of polymer composites with non-uniform out-of-plane waviness defect. J Thermoplast Compos Mater 2022; 36: 1–15.

23.

Weber

Englhard

Arent

, et al. An experimental characterization of wrinkling generated during prepreg autoclave manufacturing using caul plates. J Compos Mater 2019; 53: 3757–3773.

24.

Hallander

Akermo

Mattei

, et al. An experimental study of mechanisms behind wrinkle development during forming of composite laminates. Compos Part A Appl Sci Manuf 2013; 50: 54–64.

25.

Sebastian

. Fibre waviness in pultruded bridge deck profiles: geometric characterisation and consequences on ultimate behaviour. Compos B Eng 2018; 146: 270–280.

26.

Maes

Elsaid

BSF

Hallett

, et al. The effect of out-of-plane waviness asymmetry on laminate strength. In: 36th Technical conference of the American society for composites. Texas, USA: Virtual, 2021, pp.670–683.

27.

Test Method for Tensile Properties of Polymer Matrix Composite Materials. ASTM D3039/D3039M-17. Pennsylvania, USA: ASTM International, 2017.

28.

Test Method for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials by Four-Point Bending. ASTM D6272-17. Pennsylvania, USA: ASTM International, 2017.

29.

El Said

Hallett

. Multiscale surrogate modelling of the elastic response of thick composite structures with embedded defects and features. Compos Struct [Internet] 2018; 200: 781–798.

30.

Mussatayev

Fitzgerald

, et al. Directional eddy current probe configuration for in-line detection of out-of-plane wrinkles. Compos B Eng 2024; 268: 111048.

31.

Maes

Woo

, et al. Imaging and reconstruction of asymmetric wrinkles in carbon fibre composites using high-frequency eddy current and full matrix capture-based ultrasound. Compos B Eng 2024; 287: 111873.

32.

Chuaqui

TRC

Sebastian

Sahadevan

, et al. Edge treatment of short beam shear tests for improved assessment of structural strength. Compos Part A Appl Sci Manuf [Internet] 2020; 137: 105991.

33.

Journoud

Bouvet

Castanié

, et al. Experimental and numerical compressive tests on curved laminate structures with embedded wrinkle. Mech Adv MaterStruct [Internet] 2024; 31: 550–570.

34.

Maes

Kratz

. Applicability of historical laminate strength data for predicting knockdowns for wrinkles in complex parts. In: Falzon

McCarthy

(eds) Proceedings of the 2023 International Conference on Composite Materials. Belfast, Northern Ireland: International Committee on Composite Materials (ICCM), 2023, pp.1–16.