Experimental and Numerical Studies on the Fire Performance of Thin Sustainable Wood-Based Laminated Veneers

Abstract

Wood and wood-based products are abundantly used, especially in structural applications, due to the impetus for sustainable development. The present work helps highlight the fire performance of plywood, one of the most used wood-based laminated structural components, under three different heat fluxes of 35 kW/m², 50 kW/m², and 65 kW/m². The effects on the various fire reaction properties, namely, time to ignition, heat release rate, peak heat release rate, time to peak heat release rate, time to flameout, total burn time, and mass loss, were observed and reported. The times to ignition (42.2% and 35.4%), peak heat release rate (27.7% and 18.9%), flameout (22.2% and 28.6%), burn time (10.6% and 16.1%), and residual mass (25% and 53.3%) were reduced with the increase in heat flux from 35 kW/m² to 65 kW/m², respectively, whereas the peak heat release (21.7% and 2.4%) and ignition temperature (6.5% and 6.6%) were observed to increase. The vertical burning test (UL-94) illustrated the plywood samples to have a V-1 rating, with self-extinguishing capabilities. A numerical predictive model has also been developed based on the Fire Dynamics Simulator to predict the time to ignition, time to flameout, and heat release rate trend along with the peak heat release rate—it is shown to have good agreement with the experimental results, with an average correlation coefficient of 0.87.

1. Introduction

Wood is a natural material and is known to have desirable material properties, resulting in various uses in both structural and non-structural applications [1]. The abundance of wood and its pre-historic applications have resulted in immense technological advancements, enabling widely engineered applications of wood and wood-based composites [2,3]. Although, for the last few decades, polymeric composites have surpassed wood-based products in demand, primarily due to their advantages of facile engineering, relatively superior properties, and the possibility of being tailored as per requirements. However, the majority of polymeric materials, especially thermosetting resins, release toxic gas fumes under fire, which are often labeled as carcinogens [4]. Furthermore, all the synthetic polymers are derived from petroleum-based products, resulting in non-sustainable practices. In addition, anthropogenic climate change is already having significant impacts on the ecological, technological and sociological environments [5]. Therefore, the demand for renewable and biodegradable material has significantly increased, with wood being one of the pioneers and finding its way back even in several restrictive industries (e.g., aviation) [6]. Among all the natural fibers, wood has the highest annual production around the world, and in order to ensure sustainable forest management, it is important to utilize these natural resources efficiently [7]. Some of the advantages of wood include its availability, great strength-to-weight and stiffness-to-weight ratios, as well as its economical, renewable, biodegradable, toxic-free, and aesthetic properties. However, wood being organic and consisting of cellulose, hemicellulose, and lignin, possesses a potential fire hazard [8]. The flame spread in wood is quite large because of the high average released heat [9,10]. Therefore, it is important to have a deeper understanding of the fire reaction properties of wood, and especially plywood, which has significant applications in household furniture, doors, floors, decorations, and even home walls. Radiata pine (Pinus radiata) is a very common wood type used for making plywood in Pacific Rim countries [1], one of the main reasons for its use in the current study. It must be noted that even though fire performance studies on wood have been carried out for decades, there are still limited studies on fire reaction properties and the development of fire dynamics models based on radiata pine. Thus, the current study is critical because the performance of any wood-based product greatly depends on the species of the wood it is made from.

Studies on providing fire-retardancy to wood and plywood have been carried out for more than 70 years [8]. Processes include coating or impregnation with boric acids [11], borax [12], ammonium chloride [8], and ammonium polyphosphate [13,14] vacuum pressure treatment with diammonium hydrogen phosphate and ammonium sulphate [15], and many others. However, studies specific to plywood made of radiata pine veneers, as mentioned earlier, are limited. It is well established that an increase in heat flux will bring adverse effects to wood structures, especially with combustibility being one of the unfavorable properties of wood in high-end applications [16]. It should also be noted that even though the addition of fire retardants helps in improving fire reaction properties, they have often resulted in decreased load-bearing capacities in wood and plywood products, causing rapid catastrophic failures [17]. The use of radiata pine plywood is extremely common in furniture and other building products, especially in Pacific Rim countries. The increase in catastrophic events in buildings due to fires around the world has resulted in the incorporation of stringent regulations in the building codes of many countries. This has resulted in the need for building and structural materials to meet the mandated compliances on fire safety. Therefore, the research focus has greatly been concentrated on the fire performance of both charring and non-charring materials with the potential to be used in the built environment [1]. Additionally, the drive toward sustainable building practices has resulted in the increased demand of wood and other lignocellulosic materials in the built environment. Thus, understanding the inherent fire reaction properties of a material system is critical to quantify its fire performance before applying fire retardants, an aspect that is limited for radiata pine plywood. One of the advantages of wood, although being a combustible product, is its inherent charring capacity due to the presence of lignin [18,19]. This attribute results in immense potential that can be further engineered for use even in restrictive industries.

Various numerical and analytical predictive models have been developed to represent the fire performance of wood-based charring materials in the last few decades. Lautenberger and Fernandez-Pello [20] developed a model to simulate the oxidative pyrolysis of wood and validated it against 38 mm thick white pines. Yeun et al. [21] also introduced a model which could simulate the three-dimensional cone calorimeter experiment and validated it against wet woods. Hostikka and Matala [22] further developed a numerical predictive model for predicting the heat release rate of 20 mm thick birch wood, and Ding et al. [23] predicted the multi-component combustion of 10 mm thick beech wood. Delichatsios et al. [24] additionally developed a numerical predictive model of 19 mm thick Australian white pine under various heat fluxes. Other interesting models included those developed by Emberley et al. [25] that predicted the fire performance of six different wood species having thicknesses from 47 mm to 150 mm, by Lin et al. [26] that again predicted the fire reactions of different wood species 30 mm thick, and others. However, the simulation of thin (~2 mm) plywood made of radiata pine veneers has not been carried out to understand their performances under various heat irradiances, which is particularly important because the thickness plays a critical role in the fire performance of the structure [27,28]. As detailed by Schartel and Hull [28], a single heat release rate (HRR) peak and a steep rise and fall of the HRR curve during cone calorimeter tests are generally seen in thermally thin samples, aspects that are common with thin plywood specimens.

The current work emphasizes understanding the burning behavior of plywood made from radiata pine veneers and the influence of heat flux on their fire reaction properties. Three different heat fluxes of 35 kW/m², 50 kW/m², and 65 kW/m² were selected to study the performance of the typical plywood sample under a cone calorimeter apparatus, commonly used in industrial applications and building standards for grading primary and secondary elements. A value of 50 kW/m² has been widely used as the standard irradiance value for many cone calorimeter tests, with the ASTM E1354 standard [29] also recommending this value. Therefore, 50 kW/m² was taken as the mid-irradiance, and a lower and higher irradiance were selected for comparison. The cone heater at the University of Auckland had a limit of 65 kW/m² for safe operation, limiting the highest irradiance. To maintain symmetry, 35 kW/m² was selected as the lowest irradiance, although the ASTM standard mentions 25 kW/m² as the lowest value used for different studies [29]. Furthermore, a numerical predictive model was also developed to simulate the fire performance of the specific material. Pyrosim^® 2024.1.0702 software incorporated with the Fire Dynamics Simulator (FDS^®) was used to develop the representative numerical model. Validation of the numerical model based on the experimental findings was carried out. This will help provide a base model for understanding the fire reaction properties of radiata pine wood-based material systems without repetitive testing and requiring a cone calorimeter or similar equipment.

2. Materials and Methods

2.1. Material Used

Three-ply natural composite laminates or plywood made from radiata pine (Pinus radiata D. Don) were used in the study. The plywood samples, which were acquired from a local vendor called Plyman Auckland, were manufactured with the help of 2-pot poly (vinyl) acetate (PVA) and had an overall thickness of 2 mm, with individual veneers being 0.6 mm thick. As detailed by the vendor, the plywood was stacked by applying 250 g/m² of the adhesive on each side of the middle layer and then heat consolidated in a hot press for 5 min at 120 °C and under a pressure of about 1.4 MPa. These are generally used for intricate decorative projects, woodworking, furniture, and decorative panels.

2.2. Experimental Methodology

ASTM E1354 standard [29] was used to perform the experimental analysis on the PVA-glued plywood samples. Heat release rate (HRR), peak heat release rate (PHRR), and total smoke and heat productions were measured according to the ASTM standard with the aid of a cone calorimeter (FTT Limited, East Grinstead, UK). Three different heat fluxes of 35 kW/m², 50 kW/m², and 65kW/m² were chosen to observe the effect of changing heat irradiances on the plywood sample. The ignition temperatures were measured using the K-type thermocouple placed at the center of the sample. The testing procedure as outlined in the ASTM standard was precisely followed for the set-up and experimentation. UL-94 tests, also known as the vertical burning tests, were also performed to analyze the burning characteristics of the structural material, based on the ASTM D3801 standard [30], primarily for comparison, as the standard is designed mainly for testing plastics. A thermogravimetric analysis (TGA) and a derivative thermogravimetric (DTG) study were also conducted on the plywood samples to achieve the required input parameters for the numerical analysis. Three replicates were used for each test.

2.3. Numerical Model Specification

The effect of heat flux on the fire performance of laminated natural composite structures (plywood) was evaluated by developing a numerical model based on the finite volume method available in the software package FDS^® 6.9.1. This package was used to numerically solve the Reynolds-averaged Navier Stokes (RANS) equation on mass conversion, momentum conversion, and energy conversion, which is designed for high turbulent flows of the fire plumes, smoke spread, and combustion process during fires [31]. The turbulence was modeled based on the standard Smagorinsky form of a large eddy simulation [32]. An accurate second-order finite element method in space and time was further incorporated in the FDS, where an explicit second-order Runge–Kutta scheme was employed for updating the flow variables of temperature and heat flux with respect to time [33]. The governing equations of mass, energy conservation, and momentum were given in spatial derivations, which were written on the rectilinear mesh in accurate second-order finite differences [34].

The center of the assigned cell computed the scalar quantity of density, whereas the surface was assigned with the vector quantity of velocity [35]. A particular material system’s physical and thermal properties form the input parameters for computing the pyrolysis and combustion processes of a single solid lumped-mass system. In this case, the majority of the input parameters were achieved from TGA performed on the material system, as shown in Figure 1. The reference temperature was obtained from the normalized mass fraction, and the maximum decomposition rate and temperature were obtained from the mass derivative or DTG curve. Statistical analysis in the form of ANOVA for one-way analysis of variance was carried out with an alpha level of 0.05 to understand whether the variations between the numerical and experimental results were statistically significant [36].

The bulk of the numerical inputs were calculated based on the findings from TGA shown in

Table 1. Physical and thermal properties of standard plywood used in the simulation.

Properties	Unit	Value	Source
Density	kg/m3	566	Supplier
Emissivity	-	0.9	[31]
Specific Heat Capacity	kJ/kg K	2.58	[37]
Thermal Conductivity	W/m K	0.12	[37]
Heat of Combustion	kJ/kg	1.4233 × 104	Cone calorimeter
Heat of Reaction	kJ/kg	430	[37]
Reference Temperature	°C	374	TGA
Reaction Rate (rP)	s−1	0.002	TGA
Heating Rate (Ṫ)	K/min	5	TGA
Exponential constant (e)	-	2.718	[31]
Universal Gas Constant (R)	-	8.314	[31]
Mass Fraction (Ys) (no. of reactions)	-	1	[31]
Pre-exponential Constant (A)	1/s	1.07 × 106	TGA/DSC
Activation Energy (E)	J/mol	4.67 × 104	TGA/DSC
Pyrolysis Range	°C	284–399	TGA
Auto-ignition Temperature	°C	35 kW/m2: 428	Cone calorimeter (tested without igniter)
		50 kW/m2: 329
		65 kW/m2: 318
Surface Temperature of Burner	°C	35 kW/m2: 750	Cone calorimeter (tested without igniter)
		50 kW/m2: 830
		65 kW/m2: 900

. The reference temperature obtained for the current material system was 400 °C, which further helped in calculating the effective values of the kinematic parameters in Table 1. The reaction fuel was assumed to have the chemical equation of C₆H₁₂O₆, and the mixing-controlled model for a lumped-mass system was used for the simulation. Successful representation of the fire dynamics experienced by thin plywood specimens with the assumed reaction fuel can also be observed from studies by Chanda et al. [19]. This can be attributed to the very thin glue line of the PVA adhesive and wettability of the veneers. The TGA results primarily helped in defining the solid-phase pyrolysis process through a combination of the Arrhenius and power functions. Furthermore, the non-scattering gray gas model’s radiation transport equation was calculated using the finite volume method [32]. The complete simulation domain is illustrated in Figure 2, where the bottom part was kept open for adequate airflow and the vent was provided at the top with inert sides. The mesh boundary was kept at 400 mm × 400 mm × 400 mm. The volume flow rate of the gas through the vent was set to 0.024 m³/s. Three heat fluxes of 35 kW/m², 50 kW/m², and 65 kW/m² were simulated by changing the temperature of the burner, acquired through a sensitivity analysis with the aid of a solid-state device. A detailed mesh convergence study was performed to attain the required cell size of 8 mm, obtained through a grid-size sensitivity analysis, for simulating the burning behavior in a 400 mm cube-meshed area. Pyrosim^® software incorporated with the FDS^® and Smokeview was used for performing the numerical analysis. The auto-ignition temperatures at different heat irradiances were observed through thermocouples in a separate cone calorimeter test without a spark plug. The critical flame temperature was set to 1427 °C for all the heat fluxes, which is also used for simulating common hydrocarbons. A single reaction system was considered, assuming the normalized mass density to be zero and the mass fraction to be unity [37].

3. Results and Discussion

3.1. Fire Rating (V-Rating)

The comparative burning characteristics of the plywood samples in the vertical position were tested based on ASTM D3801 [30] due to the complexity and limitations of the standard used for structural materials. The two cycles of flame impingements were carried out, and the total burn time was calculated based on the first flame impingement time, second flame impingement time, afterflame time, and the afterglow time after the second flame. The afterflame time after the first flame impingement process was about 30 s, whereas there was self-extinguishment after the second impingement process. However, there was an afterglow of 58.4 s following the second flame impingement. Therefore, the total burn time was about 88.4 s, which being less than 250 s, gave the plywood samples a V-1 rating. It should be noted that the biggest advantage of wood-based composite structures under fire is their non-dripping characteristic, which is quite uncommon in polymer composites. The values were averaged based on five replicates.

3.2. Fire Reaction Properties

3.2.1. Time to Ignition (t_Ig)

The FDS^® has no defined mechanism to simulate the ignition temperature (T_Ig) and the time to ignition (t_Ig), which are majorly dependent on the pyrolysis model data. Therefore, the smoke view was used to predict the results outlined in the current section. The numerical study was observed to under-predict all the t_Ig values for all the heat irradiances, shown in Figure 3a. The difference between the experimental and the simulated t_Ig was observed to be about 3.2 s (9.6%) for the heat flux of 35 kW/m², whereas the variation was about 2 s (10.4%) and 1.6 s (12.9%) for the 50 and 65 kW/m² heat fluxes, respectively. The higher heat irradiance showed a little more discrepancy in the t_Ig values, although the actual variations were reduced, which agrees with the study by Dutta [33]. The reason can be attributed to a higher rate of pyrolysis and flow of combustible gases during the simulation than that during the experiment. The ANOVA analysis on the results further prove that the variation in experimental and numerical t_Ig values are statistically not significant, with a p value of 0.06. The reaction rates in the FDS are calculated using the TGA results, provided during the simulation set-up according to Table 1, and acquired based on the decomposition rate and temperature of the material. Therefore, the variation in the TGA and cone calorimeter testing environments might have a significant effect on the pyrolysis rate and eventually on the final variation in the numerical results. The t_Ig might also be affected by the location of the spark ignitor and the formation of gases due to pyrolysis of the material, which occupies a pyramidal volume [38].

The agreement between the simulation and experimental t_Ig values were observed to be greater for higher heat fluxes, with minute variations being observed for both 50 and 65 kW/m² fluxes. This might be caused by the inability of the samples to burn uniformly during the experiment under lower heat irradiance. Thus, the rate of burning might be involved in providing the different variations in the t_Ig values. The increase in heat flux resulted in a gradual decrease in t_Ig during both the experimental and simulation analyses. Therefore, it can be concluded that the critical flammable mixture for ignition was reached earlier with the increase in heat irradiance. However, the small variations in the numerical and experimental time to ignition could be easily condoned, showing good agreement between the two. The non-dimensional dependency of the experimental and numerical t_Ig values on the changing heat fluxes, shown in

Table 2. Non-dimensional relationships of the various fire reaction properties, representing their variation due to changing heat fluxes.

Type of Variation	Experimental	Numerical
Time to ignition (tIg)	tIg@35 = 1.73 tIg@50	tIg@35 = 1.74 tIg@50
	tIg@50 = 1.55 tIg@65	tIg@50 = 1.59 tIg@65
	tIg@35 = 2.68 tIg@65	tIg@35 = 2.78 tIg@65
Ignition temperature (TIg)	TIg@35 = 0.95 tIg@50	TIg@35 = 0.93 TIg@50
	TIg@50 = 0.9 TIg@65	TIg@50 = 0.93 TIg@65
	TIg@35 = 0.86 TIg@65	TIg@35 = 0.87 TIg@65
PHRR	PHRR@35 = 0.78 PHRR@50	PHRR@35 = 0.78 PHRR@50
	PHRR@50 = 0.98 PHRR@65	PHRR@50 = 0.98 PHRR@65
	PHRR@35 = 0.76 PHRR@65	PHRR@35 = 0.76 PHRR@65
Time to PHRR (tPHRR)	tPHRR@35 = 1.38 tPHRR@50	tPHRR@35 = 1.11 tPHRR@50
	tPHRR@50 = 1.23 tPHRR@65	tPHRR@50 = 1.24 tPHRR@65
	tPHRR@35 = 1.71 tPHRR@65	tPHRR@35 = 1.38 tPHRR@65
Time to flameout (tf)	tf@35 = 1.28 tf@50	tf@35 = 1.33 tf@50
	tf@50 = 1.27 tf@65	tf@50 = 1.23 tf@65
	tf@35 = 1.64 tf@65	tf@35 = 1.64 tf@65
Burn time (tb)	tb@35 = 1.38 tb@50	tb@35 = 1.11 tb@50
	tb@50 = 1.23 tb@65	tb@50 = 1.19 tb@65
	tb@35 = 1.71 tb@65	tb@35 = 1.33 tb@65

, also had great similarity, with a maximum variation of only 3.6%.

3.2.2. Ignition Temperature (T_Ig)

The T_Ig values were found to increase with the gradual increase in the heat fluxes, as shown in Figure 3b. Moreover, a constant increment was also observed between the samples, both experimentally and numerically. The higher variations (12% and 14%) were found to be for the lower heat irradiances of 35 kW/m² and 50 kW/m², with the lowest variation of about 11% being observed at 65 kW/m². Overall, the variation between the numerical and experimental T_Ig values for all the irradiances, though statistically significant, with a p-value of 0.01, showed a similar observed trend. Furthermore, the numerical simulation can be claimed to have good agreement, although there were slight over-predictions in the T_Ig values for all the fluxes. This could be attributed to the use of thermocouples during the experiments. The thermocouples are known to have less consistency, and slight changes in contact during the burning phase might result in significant variations. Furthermore, the ignition temperature was considered as the temperature at the time to ignition estimated visually from the smoke view. The experimental values were also reported in a similar way, where the temperature recorded by the thermocouple at the ‘visually confirmed’ time to ignition was considered as T_Ig.

Therefore, the possibility of having slight variations is plausible, although averages of five samples were taken to minimize aberrations. Additionally, the other modeling parameters might also result in variations, including the calculations of pre-exponential function and activation energy, where only a single decomposition reaction system was assumed. Moreover, the uncertainties in the complex gas and solid phases used by the FDS often become difficult to untangle to provide exact quantitative reasoning. The possibility of uncertainties arising from the property values of the materials and the related thermo-physical model in the FDS could also have significant effects on the simulation results. The ignition condition is usually dependent on the material response derived from the surrounding environmental conditions. One such condition is the threshold temperature, which defines the non-Arrhenius pyrolysis functions and ignition criteria. However, the value was considered constant, eliminating its effects on the pyrolysis of the sample, which might result in significant differences. Another important aspect to note is the mixing-controlled combustion model used by the FDS, where the mixing and burning control mechanisms and the lumped-mass system might result in the fuel igniting before the actual time to ignition, resulting in variations. The time to ignition and the subsequent temperature were reported based on visual observations, where the FDS provides enough flexibility for users to assign the required parameters for prediction. However, since the time is based on visual observations, it is prone to human errors [33]. Thus, the ignition condition in the simulation is related to the critical flammable mixture, which is dependent on the user-defined pyrolysis conditions obtained from TGA. The non-dimensional dependencies of the experimental and numerical T_Ig values on the change in heat flux are illustrated in Table 2. An overall variation of only about 2% could be observed between the numerical and experimental relationships, which lies well within the acceptable range.

3.2.3. Heat Release Rate (HRR)

It can be observed from Figure 4 that the numerically achieved HRR curves representing standard PVA-glued plywood are in reasonably good agreement with those of the experimental ones, apart from the later half of the curve due to fire suppression in numerical modeling. All the curves represented the response expected from a thermally thin charring material, as described by Schartel and Hull [28], akin to the plywood used in the current study. The best relation between the numerical and experimental values was identified for a heat flux of 35 kW/m², shown in Figure 4, with little variations observed at higher irradiances. Furthermore, the HRR curves over-predicted the heat release trends under all three heat fluxes. The aspect of char formation can cause this. The FDS has substantial limitations in simulating char formation, hence the residual mass given as input, obtained from TGA, is considered during the simulation as the final char formed. However, in reality, the char forms a barrier and thus reduces the actual HRR. Moreover, the charring process in the FDS is not complex enough to simulate the exact thermo-physical mechanisms of moisture vaporization, contraction, surface recession, cracking, and others. Additionally, the oxygen deficiency during char formation, which also interferes with the transfer of heat between the heat source and the material, could have a critical effect on the results. Therefore, it is extremely common for numerical HRR responses to have a smaller area under the curve, resulting in reduced total heat release. However, the simulation was successful in capturing the effect of changing heat flux on the material system. Another important aspect is the simple chemistry model used in the FDS, which contains only C, H, N, and O as fuel molecules. Consequently, the effects of the aliphatic glue and chorine are not captured, resulting in some minute variations. The shoulders in the numerical HRR trends were also noticeable before the graph reached the peak HRR. This is in accordance with the literature [19] and can be attributed to the assumption that the combustion by-products are 83% wood and 17% char when their stoichiometric equivalents are not considered in the FDS^® model.

In addition, the variation in the environmental conditions during TGA and cone calorimeter tests can again have substantial effects on the validation process. The critical flame temperature (1427 °C) [31], which was set as the default value controlling the flame suppression phenomenon [31], might have resulted in the drastic drops of the HRR curves in the numerical simulations. Another reason behind the drop is the auto-ignition temperature, which was set to prevent spurious burning of the samples. However, the combustion stops drastically when the temperature falls below the auto-ignition temperature, resulting in the HRR curve falling as well. The experimental HRR curves, on the other hand, experienced a gradual drop because of the natural process and no external suppressions in place. The HRR responses in the numerical simulations experienced shoulders in the initial stages of the reaction before reaching the PHRR. This phenomenon can result from the by-products during combustion being segregated only into wood and char, for which the stoichiometric equivalence of each component was not considered. The correlation coefficient helps to observe the linear relationship between two arrays or graphs, which can be simply calculated based on established formulae.

$$r={\displaystyle \frac{\sum \left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)}{\sqrt{\sum {\left(x_i-\overline{x}\right)}^2\sum {\left(y_i-\overline{y}\right)}^2}}},$$

(1)

where r is the correlation coefficient, $x_i$ is the set of experimental values, $\overline{x}$ is the mean of the experimental values, $y_i$ is the set of numerical values, and $\overline{y}$ is the mean of the numerical values [39]. The correlation coefficient between the respective HRR curves for the three heat fluxes was calculated to be 0.89, 0.83, and 0.88 for the responses under 35 kW/m², 50 kW/m², and 65 kW/m² irradiances, respectively. The excellent correlation between the numerical and experimental responses further validates the numerical model predicted.

3.2.4. Peak Heat Release Rate (PHRR)

The experimental and numerical PHRR values were again very similar for the 35 kW/m² heat flux (experimental~419.8 kW/m² and numerical~426.2 kW/m²) along with the time to PHRR (t_PHRR) (experimental~51.2 s and numerical~48.8 s), as shown in Figure 5a,b. The PHRR values for the heat flux of 50 kW/m² also showed a small variation of ~8.67 kW/m² from the numerical results, which might be caused by the greater amount of mass loss rate and higher combustion in the simulation at higher temperatures and heat fluxes compared to that of the experiment. A similar variation of ~10 kW/m² was observed for the heat flux of 65 kW/m², as seen in Figure 5a. A higher numerical prediction of the PHRR value can be related to the boundary conditions assigned during the simulation. An insulating condition was provided to the back surface of the sample during the simulation to represent the aluminum foil used during the experiments for minimizing heat loss from the bottom part of the sample and to prevent dripping, if applicable. The maximum deviation in the numerical and experimental PHRR values was observed to be ~1.7%, which was very much within the acceptable range. An ANOVA analysis of the PHRR values showed that the variation between numerical and experimental values was not statistically significant, with a p value of 0.6, further bringing confidence in the developed model. The non-dimensional dependency of the PHRR on the heat flux for plywood samples was again calculated for both numerical and experimental values, shown in Table 2, where excellent agreements with negligible variations were observed.

Similar to earlier observations, the time to PHRR (t_PHRR) showed a good agreement for the lower heat flux of 35 kW/m², with a variation of only ~3%, shown in Figure 5b. ANOVA analysis further proved that the variation is not statistically significant, with a p value of 0.5. However, the variation increased to about 20% for the higher heat fluxes, primarily again due to the possible increase in the flow of combustibles. These variations also became statistically significant with p values lower than the assigned alpha level of 0.05. Furthermore, the ability of the char to experimentally subdue the PHRR and the t_PHRR was similarly not simulated due to the limitations of the FDS, resulting in significant variations, especially at higher irradiances. The non-dimensional dependency also had a similar observation, with the minimum being ~1% and the maximum being about 19%, shown in Table 2. However, the variation is still acceptable due to the ability of the model to identify the differences due to the changing heat fluxes and achieving a similar trend of decreasing t_PHRR with the rise in irradiances. It must be noted that the fire reaction properties are greatly affected by the material system, variations in thickness, and the adhesive used. Changing the adhesive to an epoxy-based one will influence the PHRR to increase to 571.6 ± 29 kW/m² compared to 536.3 ± 20 kW/m² for plywood fabricated using a PVA adhesive [19], along with variations in increased t_Ig and t_PHRR. Furthermore, solid wood with a 25 mm thickness shows a completely different effect, with PHRR values of only 185.3 kW/m² [18]. All these variations can be primarily attributed to the ability of char formation, with thicker lignocellulosic materials having the capability of forming higher amount of char.

3.2.5. Time to Flameout (t_f) and Burn Time (t_b)

The time to flameout (t_f) was observed to reduce in average intervals of ~22.5s with the increasing heat flux, as expected (Figure 6a). This could be caused by the faster rate of combustion and shorter time to ignition, experienced with rising heat fluxes. The time to flameout for an experimental sample was considered to be the point where the HRR curve stabilized after the combustion process, radiating minimal and constant heat. The under-prediction in the time to flameout could be credited to the critical flame temperature [32], denoted as having the default value in the simulation and resulting in the combustion process to end drastically, as shown in Figure 4. Therefore, the samples had earlier t_f values for each irradiance. Additionally, the ANOVA analysis further proved that the variation between the experimental and numerical values was statistically significant, with p-values lower than 0.05. However, the simulation was capable of capturing the reduction in t_f with increasing heat flux, with a similar trend being captured in the model. Additionally, the non-dimensional dependency of t_f on the heat flux had further good agreement between the numerical and experimental results, with a maximum deviation of 3% (Table 2). The values themselves had a maximum deviation of only ~17%, again falling within an acceptable range given the complexities of the thermos-physical properties.

The overall burn time had a similar observation, which was calculated by subtracting the time to ignition from the time to flameout (Figure 6b). A similar reduction in t_b in average intervals of ~7 s was observed with the gradual increase in heat flux, which the numerical model was able to capture. The numerical simulation values had a maximum notable variation of ~19% with variations that were statistically significant as per the ANOVA analysis, although the non-dimensional relationship only deviated by ~5% from the experimental values (Table 2). Therefore, the general trend is definitely captured by the numerical model, with further research being needed on increasing the area under the HRR curve to have a better similarity, primarily at the end of the combustion cycle. Moreover, the inability to model the effects of char might again be a significant issue for determining t_f and t_b and addressing this will definitely provide an exact representative numerical model.

3.2.6. Mass Loss

The increase in the mass loss rate and the subsequent decrease in the residual mass were observed with the rising heat fluxes that caused greater degradation of the char, as shown in Figure 7. The increase in the amount of mass loss can be directly attributed to the higher amount of heat flux, as it is expected for samples to degrade faster and more rapidly under higher irradiances. The residual char weighed ~2 g after the combustion process under the heat flux of 35 kW/m² (Figure 7a), which decreased to ~1.5 g under 50 kW/m² (Figure 7b), and ~0.7 g under 65 kW/m² (Figure 7c). The reduced mass of the char residue denotes that the plywood sample becomes less stable structurally when exposed to higher heat fluxes. Therefore, applying fire-retardant coatings or flame-retardant fillers is desirable for various possible internal and external applications, with the thin structures being possibly extended to be a part of any façade system or conduits when formed and modified with fire-retardant fillers, and as fire-retardant laminates that can be applied as face sheets in multiple layered systems and structures.

4. Conclusions

The current work attempted to comprehend the flammability of wood-based laminated structures or plywood made from radiata pine, which is important due to the spike in demand for natural and renewable structural material systems in recent times. The concluding remarks can be summarized as follows:

• The vertical burn test gave a rating of V-1 to the plywood samples with a self-extinguishing ability, proving that the material system can have substantial applications if fire retardancy can be introduced;
• The plywood samples were observed to experience a decrease in time to ignition by an average of about 39%, flameout by about 21.5%, peak heat release rate by about 24.5%, and burn time by about 13% with the gradual increase in the heat fluxes;
• The increase in heat irradiance also resulted in the samples experiencing a higher ignition temperature that increased roughly on average by ~7.25%, peak heat release by ~12%, and mass loss by ~30% when tested under the cone-calorimeter;
• The numerical model showed satisfactory agreement with the experimental data from all the heat fluxes. The highest correlation coefficient was calculated to be 0.89 for the heat irradiance of 35 kW/m², followed by 0.88 for 65 kW/m², and 0.83 for 50 kW/m²;
• ANOVA analysis further proved that the variations in HRR, PHRR, t_Ig, and t_PHRR-@35 values between the numerical and experimental analyses were statistically not significant, while the other factors showed similar trends observed from experimental values. This helped develop confidence in the numerical model;
• The non-dimensional dependencies of the various fire reaction parameters of the three heat fluxes used were calculated both experimentally and numerically. They had considerably good agreement, except for the time to peak heat release rate. This variation can be attributed to the inability of the FDS to simulate the fire suppression experienced in the real experiment due to char formation, an aspect that needs further research.