Effect of fibre orientation and thermal exposure on the post‐fire mechanical behaviour of carbon fibre reinforced polymer material

This study investigates the influence of fibre orientation and heat flux on the post‐fire (residual) load‐bearing properties of carbon fibre‐reinforced polymer (CFRP) laminates. As a result, a deep insight into the post‐fire load‐bearing response is gained, which is necessary to fully understand and assess the advantages of CFRP laminates containing different fibre orientations for use in load‐bearing structures. Specimens were produced from three CFRP laminates containing different fibre orientations, exposed to varying heat fluxes up to 40  kW/m2 and then loaded in either tension or three‐point bending at ambient room temperature. The study's results have shown that the post‐fire behaviour of CFRP specimens is sensitive to changes in fibre orientation and heat flux. For example, specimens with an anisotropic fibre orientation in tension had the highest tensile load‐bearing capacity, whereas those with bidirectional and multi‐directional fibre orientations demonstrated lower tensile load‐bearing capacities. In bending, however, specimens containing bidirectional and multidirectional fibre orientations had higher load‐bearing capacities than specimens with an anisotropic fibre orientation. Furthermore, the data also shows that exposure to a heat flux reduces the load‐bearing capacity in both the bending and tensile specimens.


| INTRODUCTION
Carbon fibre reinforced polymer (CFRP) laminates have been used in aircraft structures for more than half a century 1 and are at present the most common fibre-reinforced polymer (FRP) material used in constructing new aircraft. 1In past applications, the use of CFRP laminates in aircraft structures has remained limited to secondary non-load-bearing components such as interior sections, cockpit controls, and auxiliary power units to minimise weight and maximise structural efficiency. 2 In recent years, however, due to modern developments such as lower manufacturing costs, more durable matrix resins and increasingly robust and stiffer carbon fibres (CF), CFRP laminates have been utilised in highly hazardous primary structural locations, such as jet engine components (i.e.][5][6] As an engineered material, the manufacture of CFRP laminates can vary.Most importantly, the ability to customise the fibre orientation of each ply layer means their mechanical properties can be optimised to suit the requirements of the application or load. 7Therefore, fibre reinforcements may take different forms.For example, when CFRP laminates are used adjacent to hazardous locations within aircraft structures, such as the wing rib which surrounds the fuel tank, they often employ a near isotropic fibre orientation, 8 such as a woven quasi-isotropic [0 , 90 ] or cross-plied quasi-isotropic [0 , AE45 , 90 ] fibre orientation to distribute stresses.However, in other sections, not necessarily adjacent to the fuel tank but located close to it, such as wing sections main spar, 2 unidirectional [0 ] carbon fibre's orientated parallel to the loading direction are often used to carry longitudinal tensile or bending loads, meaning that CFRP laminates with different fibre orientations are often used near or adjacent to aircraft fuel tanks which poses a fire risk. CFRP laminates are intrinsically different to metallic alloys due to their combustible matrix resin. 9The combustibility of the matrix resin is a well-known, long-standing problem that can result in a loss of load-bearing capacity after exposure to heat from a fire in the event of an in-flight or post-crash fire. 10,11on exposure to heat from a fire, a CFRP laminate's temperature (nearest the heated surface) will rise.As the duration of exposure increases, a thermal wave will begin to penetrate the thickness.
As the temperature increases, the matrix resin will undergo glass transition when it reaches the glass transition temperature (T g ), resulting in a loss in matrix stiffness.Further heating above the T g will lead to the matrix resin's thermal decomposition (pyrolysis) at the thermal decomposition temperature (T d ).In the event of the correct conditions for combustion, the pyrolysis gases may ignite, and a flame will form on the CFRP laminate.This flame will provide further thermal energy to the solid, increasing the heating rate and fire size, thus resulting in an increased hazard to adjacent materials and mechanically resulting in a loss in strength and load-transferring capacity. 12However, the high threshold temperature for the CF may allow them to carry some load in the loading direction at the T d. 13 As the CFRP laminate continues to be heated to very high temperatures, the char and CF will begin to thermally oxidise. 9Oxidation damage begins at the oxidation temperature (T o ) of the char and CF.The exothermic heterogeneous oxidation of the char can reduce the char layer thickness and release additional energy, which can be transferred into the material, promoting further thermal decomposition resulting in a loss in those mechanical properties dominated by the CF, for example, the tensile strength.After oxidation, only inert ash concentrations will remain. 14Further descriptions of the glass transition, pyrolysis and oxidation processes on CFRP laminates and their influence on mechanical properties can be found elsewhere 9,15 ; however, the characteristic behaviour of CFRP laminates means, for the most part, that when a CFRP laminate has been exposed to fire conditions, even after the fire is quickly extinguished, the mechanical performance may be reduced and the load-bearing ability lower than originally. 16This behaviour can be described based on four conditions of exposure.These are at: 1. Ambient (room) temperature conditions: at temperatures below those required for glass transition of the matrix resin and no change in mechanical performance.
2. Low exposure (LE): temperatures equal to the glass transition temperature (T g ) of the matrix resin.
4. High exposure (HE): the temperatures equal to the oxidation temperature (T o ) of the char and CF.

| Existing studies
Whilst existing studies examining the post-fire mechanical behaviour of CFRP laminates have shown that the reduction in the post-fire mechanical properties can be attributed to thermal degradation and damage caused by the heat from a fire, these studies were restricted to measuring the post-fire behaviour of CFRP containing a single fibre orientation [17][18][19][20] without comparing this behaviour to other commonly used fibre orientations, even though the mechanical properties of CFRP laminates are heavily dependent on the type of fibre orientation. 21,22Furthermore, in studies investigating fibre orientation on post-fire behaviour, only one study, to the author's knowledge, has been carried out in tension, 20 with no study evaluating the influence of fibre orientation on post-fire bending behaviour.On the other hand, in the studies that do compare fibre orientation of CFRP laminates in tension and bending, [23][24][25][26] the focus has been on the mechanical properties or influence of fibre orientation and thermal ageing on the performance of said materials with no focus on post-fire residual behaviour.8][29][30] In addition to these points, most of the referenced literature above characterise the 'fire conditions' using circulating hot air with temperature boundary conditions, limiting the conditions that can be tested and typically requiring thermal equilibrium to be established prior to mechanical testing to ensure a well-characterised temperature in the material.This approach is a limitation, as many structures will not reach thermal equilibrium during a fire.Therefore, the thermal boundary conditions are better defined using heat flux 31,32 to allow the material's mechanical response to be explicitly linked to the thermal environment.Consequently, the influence of different thermal exposures on the post-fire properties of CFRP laminates containing different fibre orientations has not been well established.
Because of the abovementioned issues, this study will investigate the influence of fibre orientation and thermal exposure (heat flux) on the post-fire tensile and three-point bending behaviour in terms of temperature distribution, failure load, stress-strain relationship, failure mode, and failure time.The motivation for this study is due to the paucity of existing knowledge regarding the post-fire behaviour of CFRP laminates created using common fibre orientations utilised in hazardous locations of aircraft structures, all of which have been known to compromise aircraft safety. 9quiring this knowledge is therefore important to confidently design safe fire-resilient aircraft made of CFRP laminates.Furthermore, the loading conditions and thermal environments represent those likely to be experienced due to a fire nearby and adjacent to a leak from a fuel tank during an accidental fire during flight.Hence, experiments are performed on aircraft structural-grade CF and a high-performance epoxy resin (ER) following exposure to a one-sided radiant heating representative of the critical (critical due to the impact these heating conditions have on the load-bearing mechanical properties of CFRP laminates) temperatures based on the four conditions of exposure enumerated above.To clarify, these are (1) ambient (room temperature), ( 2)

| MATERIALS
A total of three CFRP laminates were produced for this study.Each laminate contained a unique fibre orientation and comprised 8 plies bonded with ER and a hardener (curing agent).The CF was supplied by Hexcel Composites GmbH, and the ER and hardener were supplied by Easycomposites Ltd, UK, and nominally applied using a 70:30 resin-to-hardener ratio as per the manufacturer's instructions.A nominal 50:50 fibre-to-resin ratio was pursued whilst manufacturing the CFRP laminates to provide a composite with good tensile and compressive strength and stiffness properties.All laminates were produced identically using a hand lay-up technique, which involved applying a release agent to a glass mould and then stacking the dry fabric plies to the glass mould whilst coating each of the 8 plies with resin layer-by-layer.

| Curing process and specimen preparation
After the hand lay-up of each laminate, they were left to cure at room temperature for 24 h.The average thickness of the laminates (calculated from five measurements taken from each) was 6.37 ± 0.3 mm.After curing at room temperature for 24 h, specimens were prepared using a water-cooled diamond wheel saw and dried at 20 C for 12 h in a gravity convection oven.After drying for 12 h, the specimens were visually inspected for surface defects to check that they were free from external delamination, pinholes, cracks or voids.After the visual inspection, small bonding tabs measuring 50 mm Â 30 mm were carefully attached to each end of the tensile specimens using an adhesive (Sikadur 330).In total, 48 specimens were produced, 16 specimens from each of the S1, S2 and S3 CFRP laminates, of which 8 specimens were designated tensile specimens, and 8 were designated bending specimens.
The eight specimens were chosen as this allowed two specimens at each of the ambient, LE, ME and HE, where one specimen would be embedded with thermocouples and one without (other than at ambient room temperature where neither would have thermocouples embedded).A total of three K-type thermocouples were embedded at the exposed (x ¼ 1 mm) denoted TC1, middle (x ¼ 2:5 mm) denoted TC2 and unexposed surfaces (x ¼ 5 mm) denoted TC3 at the horizontal midspan to generate temperature distribution data.Thermocouples were embedded into a select few specimens, not all of them, as their presence could reduce the mechanical performance.

| Fibre-to-resin volume ratio
The fibre-to-resin ratio of each laminate was determined by running three repeat burn-off tests using small 8 mg sample crucibles for each laminate using standard thermogravimetric analysis (TGA) (TGA/DSC 1, Mettler Toledo ® ) and following ASTM standards, 33 Procedure G.
These burn-off tests were run in an air atmosphere at a heating rate of 2.5 C/min.The fibre-to-resin ratio has an important effect on the mechanical properties of CFRP laminates.In theory, the higher the carbon fibre content, the lower the resin content and the higher the strength and stiffness of the composite.Suppose the composite has a resin content that is too low.In that case, it will result in poor interlaminar adhesion (due to the lack of space for the matrix to fully surround and bond with the fibres), causing the carbon fibres to bond too quickly and reducing the mechanical properties of the composite.
If the composite has a resin content that is too high, it will cause a decrease in the strength and stiffness properties of the composite material, as fibres are the main load-carrying members of the composite.Therefore, the resin content of carbon fibre reinforced composites should be controlled optimally.

| Characterising the critical temperatures
The critical temperatures, T g , T d and T o were determined using dynamic mechanical analysis (DMA) and TGA.The T g was quantified using the Perkin Elmer ® DMA 8000 Dynamic Mechanical Analyser with a built-in displacement sensor and a 500 N load cell.The DMA experiments were carried out in a standard three-point bending orientation in which the temperature inside the testing chamber increased from 25 to 250 C at a heating rate of 10 C/min allowing the glass transition to be identified.The specimens were small rectangular specimens measuring 15 mm Â 10 mm Â 5mm (L Â W Â H), weighing approximately 8 mg produced from each of the CFRP laminates.
Three repeat runs under identical conditions were performed on each CFRP laminate, and a mean temperature was taken as the T g .In this study, the temperature at the tangential onset of the increase in the tan δ definition has been used to identify the T g from the DMA data because it defines the beginning of the loss of load-bearing capacity.
Therefore it would be undesirable to use the material beyond this temperature.
The T d and T o were identified using a Mettler Toledo TGA/DSC1.
Experiments were carried out in an air atmosphere using a flow rate of 50 mL/min at a heating rate of 2.5 C/min from 25 to 900 C. The nominal size of the crucibles was 2 mm 3 .Specimens were prepared by drilling through each CFRP laminate using a 5 mm tungsten carbide bit, grinding the chippings using a mortar and pestle and placing them into ceramic crucibles.Crucible preparation was performed carefully to ensure that the intended fibre-to-resin ratio of each of the CFRP laminates was representative of the bulk material since this factor could affect the data due to the heterogeneity of the CFRP laminates.During TGA, three repeat runs under identical conditions were performed on each CFRP laminate to improve the reliability of the data, and a mean T d and T o were reported.In this study, the T d and T o are considered the temperatures corresponding to the tangential onset of the DTG for the ER (first peak) and the CF (third peak).These two definitions have been chosen because they represent conservative values.

| Characterising the thermal exposure
Thermal exposure of the specimens was performed using a cone calorimeter apparatus (Fire Testing Technology, UK) utilising a 3, 15 and F I G U R E 1 An illustration of the fibre orientations of the CFRP laminates used in this study.

| Post-fire experiments
Post-fire tensile and bending mechanical properties were obtained using an industry-standard UTM (Instron 3380 series) using a loading rate (cross-head speed) of 2 mm/min.Each experiment was carried out under identical conditions (i.e.room temperature, airflow).Images were captured during each experiment using a single ultra-high definitions (UHD) camera (Canon ® EOS 5D Mark IV) capable of recording high-quality videography and capturing images simultaneously.Images were captured at 2-s intervals.The camera was placed at an offset distance measuring 450 mm T A B L E 1 Three-point bending failure identification codes based on ASTM D7264. 36rst

| Tensile experiments
In total, 24 tensile experiments were carried out and loaded in the longitudinal x-direction using wedge action grips attached to bonding tabs.
The undamaged specimens were initially tested until failure in orders S1, S2 and then S3.This procedure was followed by the fire-damaged LE, ME and HE CFRP specimens in orders S1, S2 and then S3.The experiments were stopped when mechanical failure of the specimens, indicated by a drop in load and a runaway displacement, was recorded.

| Three-point bending experiments
A total of 24 bending specimens were loaded on their unexposed (undamaged) surface using a constant vertical compressive load.All specimens were placed using a customised double overhanging threepoint bending specimen holder and loading nose.This customised bending apparatus contained steel roller-type supports with a span distance measuring 200 mm.This span distance remained the same for each of the three-point bending experiments.To prevent specimen slippage whilst in the supports, the length of the specimens was equal to the span length plus 50 mm overhangs at either end.The HD imagery of the fire-damaged areas observable with the naked eye.Failure modes were characterised based on the failure codes described by ASTM for tension 35 and three-point bending. 36This classification used a three-part notation to identify each failure characteristic.Tables 1 and 2 represent the notations used for tension and three-point bending, respectively.

| Fibre-to-resin volume ratio
The thermogravimetry results revealed fibre-to-resin volume ratios of 52%, 55% and 57% for the S1, S2 and S3 CFRP laminates, respectively.The minor variation in the fibre-to-resin ratio between the CFRP laminates can be attributed to the fact that S2 and S3 CFRP laminates required more resin to completely embed the fibres in the resin, reducing the possibility of matrix voids between ply layers and resulting in a smooth surface finish.The S2 and S3 required more resin because they have carbon fibres oriented in multiple directions and, therefore, small lumps caused by the warp and weft yarns of carbon fibre passing over and under one another, which the S1 CFRP laminate does not have as it is a unidirectional CFRP.and 549 s, respectively.On the other hand, the T o is not recorded; therefore, after ME, the S1 CFRP laminate is first dominated by the effects of the glass transition, followed by the pyrolysis of the ER.These temperature measurements also show a large temperature gradient forming between the exposed and middle of the specimen.

| DMA and TGA data
This behaviour is due to char formation, which prevents migrating and pyrolysing gas species flow, reducing heat transfer and pyrolysis rate.
After HE, the data shows that the T g is recorded at x ¼ 1 mm,

| S3 CFRP laminate
Figure 5 presents the temperature distribution data for the S3 CFRP laminate.At LE, the data shows that the T g is measured at x ¼ 1 mm, x ¼ 2:5 mm and x ¼ 5 mm in 98, 141 and 302 s, respectively.Glass transition is the only physico-chemical process occurring at LE.The data also shows that the temperatures at x ¼ 2:5 mm and x ¼ 5 mm begin to converge; this is most likely due to variability in the thermocouple placements and is unlikely to result from a char layer forming and slowing the heat transfer rate.
At ME, the temperature measurements show that the T g is measured at x ¼ 1 mm, x ¼ 2:5 mm and x ¼ 5 mm in 36, 43 and 270 s, respectively, the T d is measured at x ¼ 1 mm and x ¼ 2:5 mm in 302 and 450 s respectively but not at x ¼ 5 mm, whereas the T o is not recorded.This temperature differential shows that half of the through-thickness remains below the T d but above the T g .This behaviour shows that the specimen is weaker at the exposed and middle surfaces than towards the unexposed surface.
At HE, the temperature measurements show that the T g is measured at x ¼ 1 mm, x ¼ 2:5 mm and x ¼ 5 mm in 22, 37 and 142 s, respectively.On the other hand, the T d is measured at x ¼ 1 mm, x ¼ 2:5 mm and x ¼ 5 mm in 59, 103 and 731 s, respectively.In contrast, the T o is not recorded.Overall the data shows that the temperatures after HE, irrespective of through-thickness location, are generally lower than those in the S1 and S2 CFRP specimens.Lower temperatures result from the cross-plied quasi-isotropic nature of the carbon reinforcement used for the S3 CFRP specimen, which provides more stability, is less likely to delaminate and has a lower resin fraction.Less resin means there is less combustible fraction able to exhibit cracks and voids or pyrolyse, the latter of those issues resulting in the formation of pyrolysis gases that are known to increase inter-ply pressure, which is one of the main contributors leading to interfacial delamination of CFRP laminates exposed to fire. 9After ME, discolouration (yellowing) and char formation can be observed between the exposed surface and mid-thickness.In addition, the char appears to have tiny bubbles propagating from its surface.
These bubbles are produced by an accumulation of migrating pyrolysis gases from the unexposed to the exposed surface that eventually erupts closest to the heat source and causes swelling.Bubbles that appear visible but remain trapped below the surface are due to the residue cooling quickly and hardening, causing the gases to remain encircled and unable to propagate.
After HE, the ER at the exposed and middle surface has experienced pyrolysis and further discolouration (yellowing) (increasing degrees of discolouration are observed with increasing heat flux, which translates to higher degrees of thermal decomposition across the depth of the specimen).This pyrolysis has resulted in cracks and delamination forming at the exposed surface and localised char formation at the unexposed surface.
T A B L E 3 Results by each specimen.

| Summary
To summarise the results from this section, the data shows that, generally, the highest temperatures are recorded in the S1 CFRP specimens and the lowest in the S3 CFRP specimens.Higher temperatures result in further damage to the CFRP laminate resulting in them becoming mechanically weaker.However, the data in this section also shows that when a CFRP laminate is exposed to a fire, not all the material can be considered damaged but rather a series of sections, each containing respective residual mechanical properties and performance levels.For example, the unexposed surface may be considered undamaged, whereas the sections that have experienced glass transition but then rehardened may have resulted in further curing, possibly improving their mechanical properties.In contrast, those sections that have pyrolysed but not experienced oxidation may still be able to support applications requiring tensile strength.On the other hand, those sections that have experienced oxidation temperatures may have lost a portion of the load-carrying ability.
Furthermore, as the temperature recorded at TC3 has also exceeded the T d in Figure 3 at 12 and 40 kW=m 2 and Figures 4 and 5 at 40 kW=m 2 , the validity of the testing procedure will be impacted as damage will have occurred in the form of pyrolysis damage to the epoxy resin, which will have weakened these specimens to a greater degree than the those that have not experienced pyrolysis damage at TC3.The result of this damage means these specimens will likely have reduced failure modes, bending strength and different failure modes than the other specimens.

| Post-fire experiments
The data relating to the post-fire experiments using the UTM apparatus is shown in Table 3.    stress-strain relationship for each CFRP laminate, whilst Figure 10 shows the ultimate tensile strength and tensile failure load as a function of temperature for each CFRP laminate.The data in Figure 9 shows that all CFRP specimens exhibit linear behaviour until they reach peak load and fail, irrespective of fibre orientation.The S1 specimens reach peak load over the smallest displacement, whereas the S3 specimen reaches peak load over the largest displacement.The cause of this behaviour is that all the fibres are orientated parallel to the load in the S1 CFRP specimens, whereas, in the S2 and S3 CFRP specimens, they are not.Because CF is stiff and does not yield under stress when undamaged, the fibres and resin retain a good interfacial bond that provides optimal load-bearing support meaning they will fracture rather than deform.
After LE, the effect of the T g on the specimens has increased the variability of the maximum displacement and resulted in lower peak failure load values.However, all CFRP specimens still exhibit linear behaviour.The cause of the variability, however, can be attributed to the effects that the T g has on the specimens, which causes a temporary loss of interfacial bond between the fibre and resin as the specimen is heated.Although temporary, this loss of interfacial bond allows the fibres to slip relative to one another, causing them to move out of optimal alignment with the load direction.When the fibres slip relative to one another, this causes them to lose performance resulting in less consistent behaviour and a reduction in load-bearing capacity.
After ME, the load, displacement and data variability all decrease compared to the undamaged and LE specimens.This behaviour is due to the pyrolysis of the ER, which has resulted in these specimens retaining less than 50% and 34% of their undamaged peak load and displacement, respectively.The reason for the reduction in variability, however, is unknown.Furthermore, although the temperatures recorded in the specimens exposed to ME were well below the T o , meaning the fact that such a large reduction is observed in the load and displacement shows that the ER plays an important fundamental role in assisting the CF to resist failure by providing support through the redistribution of the stresses.Therefore, if the ER pyrolyses, it cannot support the fibre to retain the load.
After HE, the specimen's load-displacement relationship behaviour demonstrates more displacement before failure and very low peak loads relative to the undamaged specimens.This behaviour is due degradation of the CF.It is why the largest decrease in load-carrying capacity is observed after HE (due to oxidation and weakening of the fibre reinforcement) and not after LE or ME, where the CF remain undamaged.
On the other hand, Figure 10 shows that all the specimens, irrespective of fibre orientation, show a small decrease in tensile strength and failure load between the undamaged specimens and LE.However, this increases between LE and ME and between the ME and HE specimens.The data also shows that, on average, the S2 CFRP laminate has the lowest ultimate tensile strength, whereas the S1 CFRP laminate has the highest ultimate tensile strength.and physical weakening of the ER.This process can result in unpredictable behaviour and a loss of interfacial bond after the resin rehardens because the fibres have been allowed to move.This results in the fibres being positioned in a sub-optimal direction to carry the load adequately prevent failure.

| Bending experiments
After ME, the specimen's load-displacement and stress-strain relationship appear very different to the undamaged and LE specimens.They are marked by significant displacements, lower peak loads and smaller yield strengths.The data for these specimens also show small and irregular fluctuations in the loading response.These fluctuations are due to the individual ply layers, whose strength and stiffness properties are independent variables failing in isolation rather than more broadly.Therefore, in principle, if a single-ply layer, or even a single fibre tow (a bundle of fibres), fail in isolation due to a given stress, the overall load-carrying capacity decreases but redistributes across the remaining plies (or fibre tows), causing the materials load-bearing capacity to either sustain itself or increase, albeit for a short duration.Because of this unique behaviour, CFRP laminates can continue to support an applied load (even with some loss in strength) after the ER has partially pyrolysed due to the retained strength of the load-bearing CF.
The HE specimens are also characterised by irregular fluctuations due to the fracture and pull out of the individual plies from what remains of the char on the exposed surface.This behaviour is because, after HE, oxidation has attacked the centre of the specimens, meaning when bending occurs, the strength of the specimen is governed solely by damaged CF above the neutral axis at the convex surface.Therefore, like the specimen's behaviour in the tensile experiments, if a ply or fibre tow fails in isolation, the overall load-carrying capacity reduces but redistributes across the remaining plies and different layers, causing it to continue to maintain partial resistance.
On the other hand, Figure 12 shows increased variability between the specimens and generally shows that the S1 CFRP laminate has the highest ultimate bending strength and failure load.In contrast, the S2 Another factor influencing the behaviour of the bending specimens behaviour during the experiments is which surface (damaged or undamaged) is placed under tension and which one is in compression.This factor is important as, unlike in tensile specimens, in bending, one surface of the material will act in tension, and the other surface will act in compression.The heating effect on either of these faces will likely result in drastically different outcomes.For example, heating of the tension surface will mean that the CF oxidation can dominate the failure and can take a long time compared to heating onto the compression surface, which will be dominated by the loss of the mechanical performance of the epoxy resin.This behaviour is because the carbon fibres perform poorly in compression and become more susceptible to micro-buckling as fibre interactions are reduced due to softening of the polymer matrix.This behaviour is due to the carbon fibre reinforcement primarily controlling the material's response in tension and the epoxy resin primarily controlling the material's response during compression.
Therefore, heating the tensile surface exposes the fibre reinforcement to maximum thermal and mechanical stress.For this reason alone, the data obtained for the bending specimens shown are only representative of such scenarios where the fibre reinforcement dominates the exposed surface whilst the load is applied on the unexposed surface.These data show that the validity of the testing approach is affected by the fact that the damaged areas are no longer in composite action.

| Microstructural analysis
In order to obtain further information about CFRP laminates post-fire response, the specimens have been examined by SEM.In addition, the negligible mechanical properties of the char have resulted in fibres breaking free from their original alignment.Finally, after HE, similar behaviour to the S1 CFRP laminate has occurred, and pyrolysis of the ER is complete, exposing the CF to oxygen and leading to complex oxidation reactions that attack and weaken the surface of the fibre reinforcement.This surface damage is shown in Figure 22C, where a large, damaged section of a single carbon-fibre filament is shown.Regarding the obtained results, the following conclusions can be made: 1.The results have shown that CFRP's post-fire tensile and threepoint bending behaviour and failure modes are sensitive to fibre orientation and heat flux.
2. The results have also shown that, on average, the tensile and bending properties of the CFRP specimens decreased the most between LE and ME.However, the overall largest decrease in tensile and bending properties of the CFRP specimens was found in specimens after HE, as expected.These results show that the ER plays an important fundamental role in assisting the CF to resist failure by providing support by redistributing the stresses.Therefore, if the ER pyrolyses, it cannot support the fibre to retain the load.
3. The results also show that after HE, the post-fire bending properties decrease 61%, 95% and 85% for the S1, S2 and S3 CFRP specimens, respectively, relative to the undamaged CFRP specimens.This behaviour shows that the S1 CFRP specimens retain the most post-fire bending properties, whereas the S2 CFRP specimens retain the least.Furthermore, this shows that during bending, the pyrolysis of the ER has a large impact on the residual properties.This behaviour is because the ER supports and redistributes the stresses across the CF.Hence, when the matrix pyrolyses after ME, a loss of stiffness and interfacial bond strength occurs, resulting in delamination and a general loss of interfacial bond between the residual char that encapsulates the fibres and the fibres themselves.
4. The results also show that after HE, the post-fire tensile properties decrease 93%, 97% and 98% for the S1, S2 and S3 CFRP specimens, respectively, relative to the undamaged CFRP specimens.
This behaviour shows that the S1 CFRP specimens retain the most post-fire bending properties, whereas the S3 CFRP specimens retain the least highlighting the influence of HE oxidation on the CF.
5. The data has also shown that when failure of the undamaged and LE tensile specimens occurs, they exhibit brittle behaviour.After ME and HE, the failure can be characterised as explosive and sudden, irrespective of the fibre orientation.However, a mixture of failure modes was observed in bending depending on the fibre orientation and exposure intensity.
6.The data from this study also shows that the type of failure modes in bending and tensile experiments is sensitive to the in-depth temperature gradient.These in-depth temperatures, in turn, depend on the fibre orientation due to the low through-thickness thermal conductivity of the CF.This behaviour is because the level of exposure directly influences the physico-chemical processes occurring in the specimens, hence the changes to the chemical composition.The results also demonstrate that the validity of the testing approach is also affected by the fact that the damaged areas are no longer in composite action.
Bonding tabs were attached to avoid gripping damage from the universal testing machine (UTM) clamps during the tensile experiments and prevent mechanical failure outside the fire-tested gauge region during the post-fire experiments.All specimen's dimensions were identical, measuring 250 mm Â 30 mm and 5 mm in thickness.An illustration showing the fibre orientations of each CFRP laminate used in this study is shown in Figure 1.

30 kW=m 2
heat fluxes.These heat fluxes correspond to the LE, ME and HE scenarios associated with the known physico-chemical degradation of CFRP laminates.The dimensions of all specimens were identical.Specimens were horizontally orientated perpendicular to the heater.The distance between the cone heater and the exposed surface of the specimens was identical for each experiment and fixed at 25 mm.The 25 mm spacing distance was chosen because it gives the best results concerning the mitigation of 'fire-plume wander' and adequately capturing the combustion products within the confines of the cone calorimeter exhaust hood.34During the experiments, each specimen's 100 mm Â 30 mm midspan section was exposed directly to the external heat fluxes.The lateral specimen surface, unexposed face and ends (nearest the bonding tabs) were insulated to minimise heat losses and promote one-dimensional heat transfer.The duration of the heat fluxes was chosen to allow sufficient damage to the specimens to weaken them to their LE, ME and HE scenarios.The required level of degradation was known when a thermocouple located at the unexposed surface at the centre of the midspan section recorded the critical temperatures.Initially, the solid-phase temperature distribution was measured for one of each specimen at each heat flux under identical conditions (i.e.room temperature, airflow) to determine the time taken for LE, ME and HE to reach the unexposed surface and obtain temperature profiles.After the temperature distribution data had been generated, the cone heater subjected the remaining specimens containing no thermocouples to the required level of damage corresponding to LE, ME and HE.After the cone experiments, those specimens containing thermocouples were carefully trimmed to remove the thermocouples before being mechanically tested.The thermocouples were trimmed to avoid getting caught in the UTM cross-head or pose a tripping hazard.

F
I G U R E 2 Results from the DMA and TGA showing the glass transition (T g ), pyrolysis (T d ) and oxidation (T o ) temperatures.from the lateral surface of the specimen and used to capture the midspan behaviour of the material at the location directly opposite the loading grips/nose.The lighting used during the post-fire experiments was provided by a 200-W LED light (Godox ® SL-200W) at 2000 mm from the specimen's lateral surface.
load and corresponding vertical displacement were recorded continuously by the UTM.The experiments were stopped when mechanical failure of the specimens, indicated by a drop in load and runaway displacement, was recorded.3.3.3 | Mechanical failure modesFailure modes and mechanisms were studied through macroscopic and microscopic observations using a high-resolution flatbed scanner (Epson ® V800) and scanning electron microscopy (SEM) (Jeol ® JSM-6010LA), respectively.SEM was carried out to study the topography of the CFRP laminates and acquire imagery of the specimens before and after the experiments in microscopic detail.At the same time, the high-resolution scanner provided detailed F I G U R E 3 Temperature distribution data for the S1 CFRP laminate.

Figure 2
Figure 2 shows the DMA and TGA results.Analysis of the DMA data shows that the T g of the CFRP laminates is 70 C.On the other hand, the analysis of the TGA data shows that the T d and T o of the CFRP laminates are 230 and 630 C, respectively.Because the method of obtaining these critical temperatures is performed in a kinetically dominated regime (controlled by the rate of the chemical reaction), these temperatures do not depend on the fibre orientation.Hence these temperatures are the same for all CFRP laminates.
¼ 2:5 mm and x ¼ 5 mm in 22, 34 and 105 s, respectively.It also shows that the T d was recorded at x ¼ 1 mm, x ¼ 2:5 mm and x ¼ 5 mm in 51, 102 and 517 s, respectively.The T o , on the other hand, was F I G U R E 6 Post-fire high-resolution images of the S1 CFRP laminate showing damage after different exposure intensities.F I G U R E 5 Temperature distribution data for the S3 CFRP laminate.recorded at x ¼ 1 mm in 218 s.These measurements show that the temperatures are not evenly distributed in the specimen and vary spatially and temporally, resulting in a large temperature gradient.This behaviour is because the ER has quickly heated, raising its temperature to produce pyrolysis products containing gaseous fuel that ignites.When ignition occurs, a char layer forms, impacting the heat and mass transport resulting in additional heat release.Ignition is indicated in the data by a steep temperature increase over a very short time increment.The data also shows that the temperature at x ¼ 1 mm is sufficient to oxidise the char and fibre reinforcement, causing the regression of the char layer resulting in the exposure of the CF, which then oxidises.The regression of the char is also indicated in the TC1 temperature data at HE, which appears to fluctuate when reaching the T o .This behaviour is due to surface regression and char oxidation resulting in TC1 no longer being embedded in the specimen and becoming exposed to the heat source resulting in what looks like gas phase measurements.When this occurs, the recorded temperature fluctuates, becomes unstable and can no longer be considered reliable.This behaviour is reflected in the data for TC1 at x ¼ 1 mm from approximately 450 s until the end of the experiment.Temperatures recorded at x ¼ 2:5 mm and x ¼ 5 mm, on the other hand, are below those required for char and fibre oxidation, resulting in those fibres not being damaged.

Figure 6 Figure 4
Figure 6 displays post-fire images of the frontal and lateral surfaces of the S1 CFRP specimen.After LE, damage appears very limited.After ME, the frontal image shows that the ER has pyrolysed, producing a small char layer and cracks in the residual resin.In

Figure 7
Figure7shows post-fire high-resolution frontal and lateral surface images of the S2 CFRP laminate.After LE, some discolouration (yellowing) about the specimen midspan can be observed; otherwise, physical damage appears limited.After ME, the ER has some discolouration (yellowing) due to the pyrolysis of the ER.The lateral view also shows delamination and the formation of large voids due to a build-up of interlaminar internal pressure and general deformation.This deformation of the S2 CFRP specimens appeared to be the most visible at the centre of the exposed surface around the midspan location, where the heat flux would have been the highest due to the non-uniformity of the cone heater-emitting surface.After HE, the pyrolysis of the ER appears complete at the exposed and midspan locations, which has resulted in delamination of the plies closest to the exposed surface.

F I G U R E 8
Post-fire high-resolution images of the S3 CFRP laminate showing damage after different exposure intensities.

Figure 8
Figure 8 presents post-fire images of frontal and lateral surfaces of the S3 CFRP specimens.After LE, physical damage appears in the form of small bubbles in the resin surface around the midspan location and localised delamination of the surface ply close to the heat source.
The results from the tensile experiments are presented in Figures 9 and 10 whilst the results from the three-point bending experiments are presented in Figures 11 and 12 .

Figures 9 and 13
Figures 9 and 13 show the tensile load-displacement and tensile

F I G U R E 9
Results from the post-fire tensile experiments.F I G U R E 1 0 Ultimate tensile strength and ultimate tensile failure load data.F I G U R E 1 1 Results from the post-fire three-point bending experiments.

Figures 11 and 14
Figures 11 and 14 show the load-displacement and stress-strain relationship for each CFRP laminate, whilst Figure12shows the ultimate

F I G U R E 1 7
General tension failure mechanisms for the CFRP specimens.and S3 CFRP specimens present similar behaviour to one another, albeit with lower ultimate bending strengths and failure loads than the S1 CFRP specimens.

4. 5 |
Post-fire tensile failure modes and mechanismsThis section presents images detailing the tensile failure modes and mechanisms.Failure of the tensile specimens at ambient (undamaged) and LE occurred as a transverse brittle fracture.After ME and HE, on the other hand, failure was attributed to an explosive failure.

Figure 15
Figure15shows a clear chronological series of images showing the propensity for the specimen to fail explosively as heat damage increases.On the other hand, Figure16shows photographs of the tensile failure modes of all the CFRP specimens containing different fibre orientations.

Figure 17 illustrates
Figure 17 illustrates the general tension failure mechanisms for the CFRP specimens at ambient, LE, ME and HE leading up to failure.In addition, an analysis of this kind was performed to provide insight into possible damage that could have contributed to the failure of the specimens.During the tensile experiments, it was observed that brittle (at ambient (undamaged) and LE) and explosive failure (after ME and HE) were the two most common failure modes for the specimens; however, the mechanical failure mechanisms that contributed to failure were different depending on the level of fire damage and fibre orientation of those specimens.

F I G U R E 1 8
Typical failure modes due to three-point bending, showing (A) delamination, (B) more delamination, (C) debonding of the inter-ply layer, (D) matrix and fibre shearing and finally (E) micro-buckling.The tensile failure mechanisms for the S1 CFRP laminate under ambient (undamaged) conditions were due to a rapid build-up of localised strain that occurred disproportionately over a small region at the midpoint leading to all the fibres in this zone fracturing on the same [0 ] plane.This small transverse area was when the nominal stress against nominal strain was maximum.After LE, the failure mechanisms occurred as a brittle failure caused by stress concentrations and the gradual release of energy.This gradual release of energy resulted in an audible loud 'pinging' sound as a fibre tow reached its maximum stress and failed in isolation.When a fibre tow reaches its maximum stress, the stress released is transferred to neighbouring fibres, and consequently, the stress in the fibres nearest to the break magnifies.Transferring the stress overload begins sequential overloading of the other fibre tows, albeit with a shorter duration between each 'ping'.When the stress concentrations increased, the ER responded by cracking.After ME, the S1 CFRP laminate experiences fibre pullout and debonding of the ply layers due to the pyrolysis of the ER.After HE, further delamination damage caused by the internal pressure build-up was observed in addition to the mechanisms observed after ME.Furthermore, at HE, the CF was also damaged due to surface oxidation, with close coupling observed between the completely pyrolysed areas of the ER and the oxidation of the fibres.At ambient (undamaged) conditions and LE, the S2 and S3 CFRP specimens exhibited similar mechanical failure mechanisms to the S1 CFRP laminate.This behaviour can be characterised by a progressive brittle failure caused by stress concentrations and a gradual release of pockets of stored energy.The stored energy culminates in localised tow failure due to stress propagation across the fibre-matrix interface, which results in near-neighbour tow failure.After ME and HE, on the other hand, the failure mechanisms are different from those experienced by the S1 CFRP laminate.For these specimens, failure can be attributed to transverse and splitting matrix cracking caused by a large amount of stored energy released simultaneously; this release of stored energy culminated in explosive delamination damage and failure.

4. 6 |
Figure 20, where a clear chronological series of images detailing the three-point failure modes of the S2 CFRP laminate at ambient, LE, ME and HE.

Figures 21, 22
Figure21shows SEM images detailing the progressive damage to the S1 CFRP specimens after LE, ME and HE.At the LE (Figure21A), the ER appears relatively undamaged other than a few surface flaws due to a loss of temporary interfacial bond and repositioning of some

F I G U R E 2 2
Figure 23A-C shows the progressive damage to the S3 CFRP specimens after LE, ME and HE, respectively.After LE, the ER displays signs of pyrolysis damage and some char formation around the areas of pyrolysis damage.After ME, the images show the widespread damage due to the pyrolysis of the ER and char formation that appears to have spread proportionately across the specimen.Finally, after HE, the pyrolysis of the ER largely occurred, exposing the CF to oxygen and leading to the thermal decomposition of the CF via heterogeneous oxidation reactions.This oxidation damage is shown in the form of crenulations in Figure 23C.