The above phenomena are all heavily influenced by material properties and the testing conditions. To enable the fire safe use of engineered timber, it is necessary to obtain a better understanding of the relative effects of various material, system, and fire properties on the burning behaviour. Throughout the literature, this is typically characterised by “charring rate”—whilst it is known that upon exposure to heat, timber undergoes an initial “peak” charring rate before reaching a steady-state value [78]. The average charring rates presented in this section are all averaged over the entire fire duration—thus including this initial peak. The influence of this peak diminishes with increasing fire duration.
Material Properties
Density
Charring rate is demonstrated to be strongly dependent on density [4, 5, 11, 13, 15, 17, 20, 46, 64, 67, 69, 79,80,81,82,83,84,85,86,87,88,89,90,91], with various charring models [67, 69, 82] using density as a key variable. In general, it has been found that charring rate under exposure to the standard temperature–time curve [92] (to which the majority of recent work has been undertaken) can vary from around 0.8 mm/min for light, dry softwoods, decreasing to 0.4 to 0.5 mm/min for dense, moist hardwoods [46, 84, 91], with 0.6 mm/min being a reasonable average value [84, 91]. Decreasing density occurs due to an increase in void volume, reducing thermal conductivity and thus producing localised heating and heat accumulation, resulting in increased flame spread rates [82].
Lizhong et al. [85] and Tran and White [64] tested small-scale samples (100 mm × 100 mm and 150 mm × 150 mm surface area respectively) under radiant heat exposure to determine the dependency of charring rate on density and various other factors. Both experimental series found a general trend of decreasing charring rate with increasing density.
Lizhong et al. [85] tested horizontal samples of paulownia (ρ = 260 kg/m3), toon (ρ = 530 kg/m3), and elm (ρ = 590 kg/m3) in a radiation platform to linearly increasing heat fluxes, ranging from 0.07 kW/m2s to 0.425 kW/m2s, and constant heat fluxes of 40 and 60 kW/m2. Charring rates were determined by temperature measurements using K-type thermocouples at 0 mm, 2 mm, 6 mm, and 10 mm from the exposed surface. In both cases charring rate was found to increase with decreasing density, such that \( \beta \propto \rho^{ - 0.76} \) for linearly increasing heat fluxes, and \( \beta \propto \rho^{ - 0.5} \) for constant heat fluxes.
Tran and White [64] tested vertical samples of red oak (ρ = 660 kg/m3), redwood (ρ = 312 kg/m3), southern pine (ρ = 508 kg/m3), and basswood (ρ = 420 kg/m3) in a modified Ohio State University (OSU) apparatus to constant heat fluxes of nominally 15, 25, 35, and 50 kW/m2. Charring rates were determined from mass loss data using Eq. (10):
$$ \beta = \frac{{\dot{m}^{\prime \prime } }}{{\rho_{w} }} $$
(10)
Of the species tested, redwood and southern pine were softwoods, and red oak and basswood were hardwoods. For all heat fluxes, redwood charred around 20% faster than southern pine. Similarly, basswood samples charred approximately 60% faster than red oak samples. Interestingly, the trend across all four samples was inconsistent, with basswood samples charring consistently faster than red oak, despite having a higher density. This is likely due to the impact of other factors such as permeability and species (see Sects. 3.1.3 and 3.1.4, respectively).
Furnace tests [27, 67, 69, 82, 83, 89, 93, 94] have also explored the effects of density on charring rate. Of these, White [83], White and Nordheim [69] and Schaffer [82] tested to the ASTM E 119 standard temperature–time curve [95]; Hugi et al. [27], Njankouo et al. [89], Frangi and Fontana [93], and Cedering [94] tested to the ISO 834 standard temperature–time curve [92]. Schaffer [82] also tested to ‘constant’ furnace temperatures of 538°C, 816°C, and 927°C. In each case, charring rates were determined from the position of the 300°C isotherm, except by Frangi and Fontana [93] who determined average charring rates by measuring the residual sample height after each test. The majority of authors found that charring rates decreased with increasing density, as expected and as seen from experiments under direct radiant heating. Some authors however [27, 93, 94], found no correlation between density and charring rate. In the case of Frangi and Fontana [93] and Cedering [94], this is likely due to the small range of densities (~ 125 kg/m3) over which tests were performed. Hugi et al. [27], however, tested 12 species with densities of 350 to 750 kg/m3 yet no strong correlation was observed—species, in particular permeability, appeared to have a greater effect.
Njankouo et al. [89] tested spruce (ρ = 478 kg/m3), fir (ρ = 421 kg/m3), oak (ρ = 557 kg/m3), azobe (ρ = 1038 kg/m3), afzelia (ρ = 860 kg/m3), balau (ρ = 962 kg/m3), bilinga (ρ = 692 kg/m3), meranti (ρ = 522 kg/m3), merbau (ρ = 779 kg/m3), and wenge (ρ = 923 kg/m3) in a vertical furnace, with samples inserted in a frame with 20 holes measuring 280 × 170 mm and subjected to one-sided heating. Charring rate was found to decrease from around 0.60 mm/min at 500 kg/m3 to around 0.40 mm/min at 1000 kg/m3.
White and Nordheim [69] tested five 230 mm × 510 mm × 63 mm boards of eight different wood species (Engelmass spruce, western red cedar, southern pine, redwood, hard maple, yellow poplar, red oak, and basswood). The southern pine samples consisted of seven lamellae where the others had five. Several charring models were proposed based on the results obtained: the standard \( t = mx_{c} \), (where \( x_{c} \) is the char depth and \( m \) is the reciprocal of the charring rate); \( t = mx_{c} - b \) which takes into account the faster initial charring rate; \( t = dx_{c} + gx_{c}^{2} \); and the preferred model of the form \( \ln t = \ln m + a\ln x_{c} \) or \( t = mx_{c}^{a} \), in which \( a \) was found to be equal to approximately 1.23, and m was species-dependent, given by Eq. (11):
$$ m = 0.162 + 0.000809\rho + 1.07w + 0.0689c - 0.00000655\rho d - 0.0024cd $$
(11)
where \( c \) is a (dimensionless) softwood/hardwood classification (1 for softwoods, − 1 for hardwoods), and \( d \) a factor representing permeability (in mm), discussed in Sect. 3.1.3. An alternative expression for \( m \) is given in Eq. (12):
$$ m = - 0.147 + 0.000564\rho + 1.21w + 0.532f_{C} $$
(12)
where \( f_{C} \) is the char contraction factor, given in Eq. (13):
$$ f_{C} = 0.732 - 0.00423d + 0.203c - 0.00164cd - 0.00027\rho c $$
(13)
Equation (11) gives a coefficient of determination of 0.764, and Eq. (12) 0.755. It can be seen again that charring rate decreases with increasing density. For a parameter set giving a charring rate of 0.60 mm/min with a density of 500 kg/m3, using Eq. (11), this reduces to 0.43 mm/min with a density of 1000 kg/m3, giving similar results to those in [89]. All of these equations however are empirical, and thus have limited predictive value.
Schaffer [82] tested Douglas fir, southern pine, and white oak samples in a vertical furnace. Samples were conditioned at 27°C at 30%, 65%, and 80% relative humidity to vary moisture contents to around 6%, 12%, and 18%. Samples were tested with the grain parallel and normal to the fire-exposed surface. The total sample size was 254 mm × 508 mm × 76 mm, formed of lamellae joined together by a phenol-resorcinol adhesive which was found to ‘resist’ delamination (a phenomenon in which the outermost lamella or part(s) thereof detach from the underlying lamellae) under the conditions tested. From the test data, the expression in Eq. (14) was derived:
$$ m = \left( {a + bw} \right)\rho_{w} + c $$
(14)
where a, b and c are empirical species-dependent constants. This is similar to the expressions from White and Nordheim [69], again suggesting that charring rate decreases with increasing density, which was attributed to a greater mass available to absorb energy. Additionally, crack and fissure size and formation were found to vary with density.
White [83] tested various composite timber products—Laminated Strand Lumber (LSL), produced by gluing strands of wood together; Laminated Veneer Lumber (LVL); and Parallel Strand Lumber (PSL), formed from 19 mm strands of timber aligned and then pressed together with adhesives. Results were compared with the models presented in [69], with m calculated as in Eq. (14), predicting charring rate to within 12%.
Lau et al. [67] tested to a constant furnace temperature of 500°C. Whilst they did not explicitly explore the effects of charring rate on density, they observed charring rates around knots to be lower due to higher density.
Overall, it can be concluded that samples with higher density will generally char more slowly due to the greater mass of material to pyrolyse; thus more energy is required to fuel this endothermic process. This is illustrated in Fig. 2, which shows the charring rate and density for all standard temperature–time curve tests discussed. From the tests and experiments discussed above, it can be seen that this only has a significant effect over wide density ranges, and thus is unlikely to be the governing factor in design, where typical softwoods used in construction are unlikely to vary in density by more than about ± 200 kg/m3. Additionally, the density variation within an individual sample will often be sufficiently high (around ± 20 kg/m3 for a beam tested in a typical furnace [96]) that predicting charring rate based on a global density value will be meaningless. Frangi and Fontana [93] and Cedering [94] did not find any dependency on density over a range of around ± 125 kg/m3—indeed based on the results of [89] and the models in [69], over this range, it would be expected that the corresponding variation in charring rate would be just 0.05 mm/min, which is less than the scatter in the results.
In a review on timber charring rates, Friquin [5] also found that charring rates decreased with increasing density. Some authors found charring rates varied strongly with density, whereas some tests found no variation of charring rate with density, perhaps due to variations within the sample or to the small range of densities tested. Samples with lower density typically had lower thermal conductivity, thus resulting in a faster temperature rise at the surface, thus pyrolysing and charring earlier. Conversely, higher density samples had higher through-thickness temperatures, and pyrolysed more quickly due to shallower thermal gradients.
Furthermore, as noted by Friquin [5], mass loss rates increase for denser samples. This can be easily calculated from charring rates via Eq. (10), and the results are shown in Fig. 3. It can be seen that whilst charring rate weakly decreases with density, mass loss rate, and thus overall pyrolysis, weakly increases. Thus, the contribution of flammable vapours from denser samples will be greater.
Moisture Content
The presence of moisture is widely acknowledged as retarding pyrolysis [5, 12, 13, 16, 18, 20, 21, 39, 67, 69, 79, 82, 84, 86, 90, 94, 97, 98] due to a heat sink effect—the greater the moisture content, the greater the energy required to evaporate the water, and thus the less energy available for pyrolysis.
Mikkola [12], Moghtaderi et al. [39], Shen et al. [21], and McAllister et al. [41] tested small-scale samples (100 mm × 100 mm, except McAllister et al. [41] who used 90 mm × 90 mm) under radiant heat exposure to determine the effects of moisture content. All recorded different specific outcomes (e.g. charring rate, time to ignition, mass loss rate), but all unsurprisingly observed retarded pyrolysis with increasing moisture content. Mikkola [12] tested pine (ρ = 560 kg/m3), spruce (ρ = 490 kg/m3), and LVL (ρ = 520 kg/m3) samples with moisture contents varying from 0% to 20% in a cone calorimeter at heat fluxes from 20 kW/m2 to 75 kW/m2. Temperature was recorded at various depths throughout the sample, and charring rate calculated from this, assuming a charring temperature of 360°C, 20% higher than other researchers. For an incident heat flux of 50 kW/m2, charring rate varied from an average of 1.07 mm/min at 0% moisture, decreasing to 0.80 mm/min at 10%, to 0.64 mm/min at 20%. From these results, Mikkola presents the empirical relationship in Eq. (15):
$$ \beta \propto \frac{1}{1 + 2.5w} $$
(15)
where w is moisture content (kg/kg). This assumes a strong dependency on moisture content, with dry wood charring 25% faster than wood with 10% moisture content, and 50% faster than wood with 20% moisture content.
Shen et al. [21] tested vertically oriented birch samples (ρ = 740 kg/m3) with moisture contents of 5%, 15%, and 26% to 40 kW/m2 in a cone calorimeter. Temperature was recorded throughout the tests using K-type thermocouples at 1 mm, 7 mm, and 14 mm below the exposed surface, as well as mass loss rate. Variation in surface temperature was explored, and the time to reach 350°C was recorded—this being approximately 30 s, 50 s, and 60 s for moisture contents of 5%, 15%, and 26%, respectively.
McAllister et al. [41] tested poplar samples in a 90 mm × 250 mm × 600 mm wind tunnel capable of varying airflow velocity from 0.8 m/s to 1.6 m/s. Poplar has a consistent grain, low number of knots, and a high cellulose content. Moisture contents of 0%, 8%, and 18.5% were tested at heat fluxes from 20 kW/m2 to 50 kW/m2. The emitter used in these experiments has shorter emission peaks than in a typical fire, of 1.03 to 1.30 μm compared to 1.97 to 2.28 μm, so samples were darkened to increase absorptivity. For a 50 kW/m2 heat flux imposed on a sample with 18.5% moisture content with an airflow of 1 m/s, critical mass flux was found to be 2.83 g/m2s and 3.12 g/m2s for flashpoint and firepoint, respectively. Critical mass flux was found to increase with moisture content, with the overall range being from 1.31 g/m2s to 2.98 g/m2s. With increased moisture content, water is still being evaporated at ignition, absorbing heat, and thus requiring a greater mass flux.
Moghtaderi et al. [39] tested Pacific maple (ρ = 544 kg/m3), sugar pine (ρ = 430 kg/m3), and radiata pine (ρ = 465 kg/m3) samples in a cone calorimeter to heat fluxes varying from 14 kW/m2 to 65 kW/m2. Average moisture content was 10 to 12%. Radiata pine samples were conditioned to achieve moisture contents of 0%, 15%, 22%, and 30%. Times to ignition increased substantially with increased moisture content. At 20 kW/m2, the ignition time increased from 179 s at 0% moisture to 540 s at 30% moisture, and at 60 kW/m2, increased from 3 s to 11 s. Mass flux at ignition also increased slightly as moisture content increased, and surface temperature at ignition increased with moisture content, due to the increased heating time.
Di Blasi et al. [16] tested 40 mm diameter and length beech cylinders (ρ = 728 kg/m3 to 911 kg/m3) in a radiant furnace with a 60 mm diameter quartz reactor. Mass loss was recorded throughout some tests, whilst in others temperature was recorded by five K-type thermocouples along the radius. Moisture content was varied from 0% to 50%; it was found that increased moisture content significantly increased the time to maximum mass loss rate—from c.1400 s at 0% moisture content, to c.2300 s at 11%, to c.3100 s at 29%, to c.3700 s at 49%.
Bamford’s tests [18], discussed previously in Sect. 2.4, found that for two identical 229 mm square panels opposite each other, ignition times were 9 min and 3 min for moisture contents of 14% and 6% respectively. For single panels heated on both sides, ignition time reduced from 12 min to 10 min to 8 min, with reductions in moisture content from 12% to 6.5% to 1%.
Furnace tests [69, 84, 89, 94] have also been used to explore the effects of moisture content on charring rate. Of these, White and Nordheim [69] tested to the ASTM E 119 standard temperature–time curve [95]; Njankouo et al. [89] and Cedering [94] tested to the ISO 834 standard temperature–time curve [92]; and Collier [84] tested to the AS 1530.4 standard temperature–time curve [99]—either 60 min exposure, 30 min exposure, or 30 min exposure starting at the 31st minute. All test series found that increasing moisture content led to reduced charring rates, although there is disagreement as to the precise effects.
The tests undertaken by Njankouo et al. [89], discussed in Sect. 3.1.1, had moisture contents varying from 9% to 20%. Generally, charring rate was found to decrease linearly from around 0.6 mm/min at 9% moisture content to 0.37 mm/min at 20%, but with three or four outliers, suggesting other factors (such as species or density) may be contributing and preventing a definite conclusion.
Collier [84] reports that moisture content significantly affects charring rate through several different mechanisms, but does not state what these are. Some effects oppose each other (such as the driving of moisture deeper into the sample) but overall, increases in moisture content reduce charring rate. As such, charring rates can increase by up to 25% for completely dry timber, however this is of little importance in design, as timber will reach its equilibrium moisture content (typically assumed to be around 10–12%). Collier’s tests on radiata pine showed charring rates decreasing from 0.82 ± 0.10 mm/min at 0% moisture content, to 0.59 ± 0.01 mm/min at 16% moisture content, to 0.53 mm/min at 20% moisture content.
White and Nordheim’s [69] tested eight species at four different moisture contents each: 6%, 9%, 11%, and 16%. From this, they found that charring rate decreased with increasing moisture content, leading to the expressions in Eqs. (11) and (12). This again gives a moderate dependency on moisture content, with dry wood charring 14% faster than wood with 10% moisture content, and 28% faster than wood with 20% moisture content, significantly less of an influence than suggested by Eq. (15) [12].
Cedering [94] tested three solid timber walls of dimensions 40 mm × 140 mm × 1000 mm formed of 55 to 60 timber studs of Norwegian spruce with varying density and moisture content, to the standard temperature–time curve in a furnace to varying oxygen contents. The samples were orientated vertically. The moisture content of the samples was controlled to give values of 6%, 12%, or 18%. The average charring rates across all oxygen contents decreased from 0.70 mm/min at 6% moisture content, to 0.68 mm/min at 12% moisture content, to 0.64 mm/min at 18% moisture content. This gives an average decrease in charring rate of just 4% over a 6% moisture content increase, compared with the 11% given by Eq. (15) [12], or 7% given by Eq. (11) [69]. This small increase could be partially attributed to the fact that timber samples of varying moisture contents were placed in the furnace simultaneously, resulting in unpredictable moisture flows between and in front of samples, corrupting results.
It can be seen that whilst there is strong agreement that increasing moisture content leads to a reduced charring rate due to the latent heat of evaporation of the moisture, there is generally poor agreement as to how much of an effect this has. This is illustrated in Fig. 4, which shows the charring rate and moisture content for all standard temperature–time curve tests discussed. There is poor agreement between tests, although a weak general downwards trend can be distinguished. This is most likely due to the presence of other variables such as density and species dominating the effects on charring rate.
Friquin [5] reports that moisture content has a significant effect on charring rate, with charring rate decreasing with increasing moisture content due to more energy being needed to evaporate the water. Water vapour also serves to delay ignition time, and delay and dilute combustible gases. Further, Babrauskas [97] notes that charring rate at 20% moisture content can be 8.3% lower than charring rate at 8% moisture content. This is a much smaller difference than the 20% reduction given by Eq. (15) [12], or the 13% reduction from Eq. (11) [69].
As with density, the effects of moisture content are unlikely to be of large practical importance, as moisture content is not a variable that can be controlled—it will be dictated by ambient temperature and relative humidity, which in turn will depend on geographical location [100] and building management practice. The equilibrium moisture content of wood, in %, defined as the ratio of water mass to dry mass, can be found from Eq. (16) [100].
$$ w = \frac{1800}{a}\left( {\frac{bh}{1 - bh} + \frac{{bch + 2b^{2} cdh^{2} }}{{1 + bch + b^{2} cdh^{2} }}} \right) $$
(16)
where
$$ a = 349 + 1.29T + 0.0135T^{2} $$
(16a)
$$ b = 0.805 + 0.000736T - 0.00000273T^{2} $$
(16b)
$$ c = 6.27 - 0.00938T - 0.000303T^{2} $$
(16c)
$$ d = 1.91 + 0.0407T - 0.000293T^{2} $$
(16d)
This gives typical values around 18 to 19% in winter and 13 to 15% in summer in the UK. It should be noted that Eq. (16) does not consider different species of wood, which is an obvious shortcoming, since different wood species will have different chemical and anatomical make-ups, allowing different levels of moisture absorption, however it is relatively easy to measure equilibrium moisture content of timber, and can be determined using standards such as ASTM D 4442 [101]. It should also be noted that this is based on outdoor conditions, and thus should be applied with caution for interior conditions.
Permeability
Permeability affects charring rate [5, 20, 27, 64, 67, 69, 79, 80, 84]. This is largely due to grain direction, as permeability along the grain is around four orders of magnitude higher than that across the grain [5, 84, 102]. The effect of grain direction is discussed further in Sect. 3.2.3. Increased permeability allows an increased flow of volatiles, thus contributing to faster pyrolysis.
Tran and White [64] tested redwood, southern pine, red oak, and basswood under a radiant panel, with permeability measured transverse to the grain through the penetration depth of copper chrome arsenate (CCA). This is applied as parameter d in Eqs. (11) and (13), giving d = 4, 31, 3, and 30 mm for redwood, southern pine, red oak, and basswood, respectively. No obvious trend of charring rate was observed with permeability alone, however when considered alongside density, it may be suggested that increased permeability increases charring rate, however too many variables were examined in these experiments to allow meaningful isolation of any of them.
Hugi et al. [27] and White and Nordheim [69] explored the effects of permeability on charring rate under exposure to ISO 834 and ASTM E 119 respectively. Hugi et al. [27] calculated permeability using a coefficient (detailed below), whereas White and Nordheim [69] used the method set forth by Tran and White [64]. Both sets of authors found increasing charring rate with increasing permeability.
Hugi et al. [27] tested 12 wood species of dimensions 1000 mm × 90 mm × 45 mm (abachi, abura, maple, beech, oak, ash, spruce, koto, meranti dark red, meranti light red, sipo, and fir) at tree ring orientations of 0° (tangential), 45°, and 90° (radial). Charring rate was calculated by cutting samples in two after the test and measuring the residual depth over five locations for each sample. This was compared with permeability; the coefficient of permeability was calculated by Eq. (17):
$$ k = \frac{\omega Vgd}{{RAT_{abs} t}}\ln \frac{P\left( 0 \right)}{P\left( t \right)} $$
(17)
where \( \omega \) is the molar mass of gaseous oxygen (g/mol), \( V \) is the volume of the pressure cylinder (m3), \( A \) and \( d \) are the sample cross-sectional area (m2) and thickness (m) respectively, and \( P \) is the pressure (kPa). The oxygen permeability index was then taken as the negative logarithm of this value. Charring rate was found to have a positive linear relationship with oxygen permeability index, suggesting that charring rate increases with increased permeability.
White and Nordheim [69], (discussed in Sect. 3.1.1) measured CCA penetration depths of 3 mm for Engelmann spruce, western red cedar, and red oak; 4 mm for redwood; 5 mm for yellow poplar; 30 mm for basswood; 31 mm for southern pine; and 47 mm for hard maple. As shown by Eqs. (11) and (13), this gives a charring rate increasing with increasing permeability.
In general, there is good agreement that charring rate increases with increasing permeability, however Tran and White [64] did not observe any obvious trend. Data from Hugi et al. [27] is shown in Fig. 5 to allow a visual observation of the effects of permeability on charring rate.
Species
Another significant factor affecting pyrolysis rate is the species of the wood [5, 15,16,17, 20, 46, 64, 67, 69, 79,80,81,82, 84, 86, 89, 93, 103]. Whilst the species will affect factors discussed earlier such as density, moisture content, and permeability, there will also be additional factors specific to a species, such as chemical composition [5, 20, 64, 67, 79, 84] (primarily lignin content [5, 64, 67]) and anatomy [5, 69] which influence the rate of pyrolysis.
Di Blasi et al. [103], Shen et al. [21] and Lizhong et al. [104] explored the effects of species on thermal response under radiant heat exposure. The experiments are described in Sects. 2.2 and 3.1.2 All authors observed a species dependence, but the dependency found by Lizhong et al. [104] was very small.
Di Blasi et al. [103] found pyrolysis temperature to be lowest for chestnut (due to its higher proportions of hemicelluloses), then Douglas fir, redwood, pine, and highest for beech. Char yields followed an opposing trend, being highest for chestnut, then redwood, Douglas fir, pine, and lowest for beech. Lower lignin contents were found to yield more carbon dioxide, as expected.
Shen et al. [21] conducted further experiments on Mongolian oak (ρ = 888 kg/m3), white birch (ρ = 740 kg/m3), aspens (ρ = 582 kg/m3), spruce (ρ = 469 kg/m3), white pine (ρ = 360 kg/m3), and Masters larch (ρ = 469 kg/m3) under external heat fluxes of 20 to 70 kW/m2. The surface temperatures rose faster for larch than aspens, and faster for aspens than oak, as shown in Fig. 6. In their model, which correlates well with the experiments, thermal diffusivity was found to be the main factor contributing to differences between species.
Lizhong et al. [104] tested acacia, toon, and paulownia samples under linearly increasing heat fluxes, and found ignition temperatures for paulownia at 0.2986 kW/m2s to be 271°C, toon 302°C, and acacia 319°C—this shows only a small influence of species.
A number of furnace tests [64, 69, 82, 87, 88] have explored the effects of species on charring rate. All authors observed a strong species dependence that cannot be attributed to other species-dependent properties such as density, but unlike previous variables, this can only be described qualitatively.
In White and Nordheim’s tests [69], (discussed in Sect. 3.1.1) when a species-dependant factor was incorporated into Eq. (12) giving Eq. (18), the coefficient of determination increased from 0.755 to 0.906.
$$ m = 0.1526 + 0.000508\rho + 0.1475f_{C} + Z_{1} w $$
(18)
where \( m \) is the “reciprocal char rate” (s/mm1.23). Species-dependent factors varied from 0.50 (Southern pine) to 2.05 (Engelmann spruce), thus there is considerable variation (~ 200%) across species, even once density, moisture content, permeability, and hardwood/softwood classification have been accounted for.
Schaffer’s tests on Douglas fir, southern pine, and white oak [82] (discussed in Sect. 3.1.1) also resulted in an empirical model with species-dependent factors. These are given in Table 2.
Table 2 Species Constants for Eq. (14) (Data from [82], Units Adjusted) Again, it is clear that there is considerable variation between species after accounting for density and moisture content. The variability present in parameter \( b \) may suggest that the relative effects of moisture content vary from one species to another, which may account for some of the variability between authors discussed in Sect. 3.1.2. Similarly, the variation in parameters \( a \) and \( b \) may suggest that the relative effects of density may vary from one species to another. Similarly, Mikkola [12] reports a 100% reduction in oxygen concentration caused a 50% decrease in charring rate for white pine [33], but only 35% for Douglas fir [105].
Tran and White [64] recorded the lignin content of the samples they tested: redwood consisted of 37.1% lignin, southern pine 27.9%, red oak 24.5%, and basswood 19.8%. It appears that increasing lignin content may produce an increased charring rate, however as mentioned previously, too many variables were present to allow this to be confirmed with confidence.
Yang et al. [87, 88] tested Taiwania, Japanese cedar, China fir, Douglas fir, and southern pine glulam samples to the standard temperature–time curve in a furnace. They found that charring rate and total heat released decreased from Taiwania, to Japanese cedar, to China fir, to Douglas fir, to southern pine, which does not correlate with density alone, suggesting an influence of species.
Hall [80] reviewed a number of papers giving burn-through times or charring rates for timber panels for various species. Bryan and Doman [106] found greenheart and teak to have burn-through times of 650 s, iroko 450 to 500 s, jarrah 400 to 450 s, opepe and European oak 350 to 400 s, and mahogany and spruce 250 to 300 s. Whilst details of the test setup are not given, it can be seen that the variation between species can be over two times between the extreme cases of greenheart and teak compared to mahogany and spruce. Barnes [107] tested jarrah, teak, and European oak beams for burn-through time, finding jarrah had a time of 16.5 min, teak 12.5 min, and European oak 9.2 to 11.0 min. Webster and Ashton [108] measured charring rates in a furnace for gurjun, spruce, teak, and European redwood. Average values for gurjun were 0.77 mm/min; for spruce 0.74 mm/min; for teak 0.62 mm/min; and for European redwood 0.81. For the common species, the results between these three tests do not correlate well. Hall’s own tests [80] on 15 different species also found variation across species, but again did not correlate well with the results of [106,107,108].
Várhegyi et al. [109] reviewed previous tests on chestnut, and found significant differences between chestnut and other hardwoods and softwoods. Chestnut has a comparatively high char yield and undergoes significant pyrolysis at relatively low temperatures. These samples had a high inorganic extractive content, but acetone extraction was found to reduce the difference between species.
In 1953 the Timber Development Association published a document [17] which grouped 66 species of timber into six arbitrary categories with decreasing fire resistance. Some of these results are to be expected due to density, with balsa, a very light wood typically around 250 kg/m3 density, on its own in Group 6, and Groups 1 and 2 consisting mostly of hardwoods. Groups 3–5 consist of various softwoods.
Whilst no quantitative conclusions can be drawn from the above collected test data, it can be noted that difference in species may have a marked effect on charring rate. Friquin [5] attributes this to lignin content, with higher lignin contents giving higher char yields. Collier [84] states that chemical composition influences charring rate due to its effects on kinetics and energetics of pyrolysis. It can also be noted that species can have a greater effect than density [15, 69, 82], as is evident from Eq. (18). For example, Hall [80] found teak (ρ = 630 kg/m3) to have a charring rate of 0.43 mm/min, whereas Iroko (ρ = 620 kg/m3) tested in the same conditions and with the same moisture content had a charring rate of 0.63 mm/min—a much greater difference than one would expect to be accounted for by density alone.
System Properties
Sample Orientation
As with any material, the pyrolysis rate of wood is affected by its orientation, due to its effects on fire dynamics and airflow. In vertical orientations, buoyancy will drive convection upwards parallel to the sample, resulting in very different conditions to horizontally orientated samples [71]. Charring rates are thus expected to be greater for vertically orientated samples due to increased radiation from the flame.
Shields et al. [40] directly explored the effects of sample orientation using a cone calorimeter. McAllister [42] performed further tests on poplar (ρ = 489 kg/m3) in the apparatus described in Sect. 3.1.2 varying the airflow from 0.8 m/s to 1.3 m/s. Whilst the sample orientation was not altered, the airflow will affect the fire dynamics in a similar manner, so this has been included for discussion in this section.
Shields et al. [40] tested 100 mm × 100 mm × 20 mm Sitka spruce samples under heat fluxes of 20 to 70 kW/m2, and found that time to ignition is sufficiently shorter (and occurs at lower heat fluxes) for horizontal samples than for vertical samples. These distinctions are most noticeable at lower (20 to 40 kW/m2) heat fluxes. For example, Sitka spruce at 50 kW/m2 had piloted ignition times of 12 ± 4 s in the horizontal configuration, and 20 ± 5 s in the vertical configuration; and spontaneous ignition times of 27 ± 11 s in the horizontal configuration and 50 ± 10 s in the vertical configuration. Time to ignition is also greater for the underside of a sample than the topside. Spontaneous ignition times were found to be five to thirteen times greater for ceiling samples. No explanation for this is given.
McAllister [42] found that airflow velocity had only a weak influence on critical mass flux required for ignition in the range studied, with ignition delay increasing from 43 s at 0.8 m/s to 53 s at 1.3 m/s for 18% moisture content.
Lau et al. [67] and Yang et al. [87, 88] compared the charring rates over the sides and bottoms of beams tested in furnaces. With the exception of corner rounding, Lau et al. [67] found charring to be uniform over all sides. However Yang et al. [87, 88] found charring to be faster on the bottom face than the sides.
Ashton [46] reports that charring rate is 1.3 × higher for columns than beams, suggesting that a vertical orientation results in faster pyrolysis.
The averages from all standard temperature–time curve tests reviewed herein give contradictory data; samples orientated horizontally give an average charring rate of 0.70 mm/min with a standard deviation of 0.14 mm/min, and samples orientated vertically give an average charring rate of 0.63 mm/min with a standard deviation of 0.11 mm/min. Due to this contradiction, the small number of authors investigating orientation, and the large standard deviations in the meta-analysis, no firm conclusions on the effects of sample orientation can be drawn.
Sample Size
As well as the orientation of the sample, the size will also affect the pyrolysis behaviour. Increase in scale usually means lesser heat losses per unit volume of material, thus making samples easier to ignite [48].
Ritchie et al. [70] tested circular Douglas fir samples of diameters 100 mm and 600 mm in a cone calorimeter and large radiant panel facility respectively. Samples were exposed to an incident heat flux of 25 kW/m2 after having been conditioned at 22°C and 50 to 55% relative humidity for a month, giving an equilibrium moisture content of 9.2 to 10.0% according to Eq. (16). The 100 mm samples showed initial heat release rate peaks of around 180 to 250 kW/m2, whereas the 600 mm samples showed initial peaks of around 290 to 340 kW/m2.
Hill and Comey [57] found that larger samples took much longer to ignite than smaller samples—small pieces around 0.5 g ignited in 15–20 min, whereas larger pieces around 15 g could take an hour or longer—and ignited at lower temperatures, decreasing from 204°C at 0.5 g to 181°C at 3 g to 175°C at 9 g to 158°C at 16 g. This was attributed to greater heat losses through conduction and radiation.
There is generally good agreement that pyrolysis rate increases with sample size, although relatively few authors have directly investigated this effect. Figure 7 shows the charring rate plotted against sample size for standard temperature–time curve tests reviewed herein. It can be seen that for surface areas up to around 0.25 m2, there is a weak logarithmic increase in charring rate with surface area, however, above this value, the trend discontinues, perhaps in part due to larger surface areas not being completely exposed to the fire. Hadden et al. [110] found that the relative effects of sample size (for polyurethane foams) decrease with increasing sample size, which is observed here through the logarithmic relationship.
Friquin [5] points out that the scale of the samples affects the heat and mass transfer, and thus the pyrolysis behaviour. She cites Suuberg et al. [111], who states that larger samples pyrolyse faster. Additionally, she cites Majamaa [112], who found that a 40 mm thick sample required 10 min to achieve a constant charring rate, and a 80 mm sample 30 min, suggesting that sample thickness, as well as surface area, has an effect on pyrolysis behaviour. This can be attributed to the boundary condition at the back face—a thermally thin sample will achieve thermal penetration throughout the entire sample, resulting in an elevated back face temperature and heat losses from the rear of the sample. A thermally thick sample however, will not achieve an elevated back face temperature, and thus no heat will be lost via the rear of the sample.
Grain Direction
The grain direction is also recognised as having an effect on pyrolysis behaviour [4, 5, 79, 84, 91, 113], largely due to the large increase (around 10,000 times) in permeability parallel to the grain [5, 84], due to the alignment of tracheids. As a result of this, small changes in grain angle can result in large changes in moisture and oxygen movement, thus affecting the charring rate [84]. Additionally, thermal conductivity is greater parallel to the grain than perpendicular to it [5]. As such, it is expected that charring rate will be greater parallel to the grain than perpendicular.
White [83] tested various composite timber products to ASTM E 119 [95] in a vertical furnace, where he found that some tests showed slightly faster charring perpendicular to the grain than parallel, but this was attributed to delamination and fissures.
Buchanan and Barber [114] tested steel connections in timber members, and measured charring rates as 0.68 mm/min parallel to laminations and 0.69 mm/min perpendicular to the laminations.
Perhaps surprisingly, when comparing average charring rates between all the furnace tests reviewed herein which state the grain orientation, no significant difference was observed. It is also noteworthy that in design, timber elements will generally not have faces with parallel grains exposed, and thus only charring perpendicular to grain is likely to occur in practice.
Encapsulation
Schmid et al. [115] tested protected and unprotected CLT beams in bending. They found that protective gypsum board delays the onset of charring and charring rate is slower than in an unprotected case, however charring rate increases significantly once the gypsum falls off, similar to the results of [116] and [117].
Hasburgh et al. [118] tested 12 CLT panels under a standard temperature–time curve with a variety of encapsulation systems. Two tests were undertaken with no encapsulation, four with one layer of gypsum plasterboard (two of which used a 19 mm “furring strip”), three with two layers of gypsum plasterboard (one with a 19 mm “furring strip”), one with three layers of gypsum plasterboard, and two with intumescent coatings. Encapsulation times were noted as 26.3–27.7 min for one layer of gypsum (increasing by ~ 10% to 30.4–30.8 min with a “furring strip”), 70.6–70.8 min for two layers (increasing by ~ 10% to 78.6 min with a “furring strip”), and 130.4 min for three layers. Charring rates of the first lamella (calculated from the onset of charring) were found to vary from 0.80 ± 0.04 mm/min for no encapsulation, 0.44 ± 0.01 mm/min for one layer of gypsum plasterboard (0.54 ± 0.02 mm/min with a “furring strip”), 0.61 ± 0.11 mm/min with two layers of gypsum plasterboard, and 0.95 mm/min with three layers. There is no clear pattern to this variation, however determining the total time taken to char through the first lamella (including the time taken for plasterboard falloff) shows a clear increase with number of layers of plasterboard, increasing from 44 min with no encapsulation, to 101 min with one layer, to 131 min with two layers, to 167 min with three layers (average times over multiple tests, including those with a furring strip).
Hakkarainen [117] tested four timber room structures of length 4.5 m, width 3.5 m, and height 2.5 m with a 2.3 m × 1.2 m window. One compartment was unprotected, one was protected with one 12.5 mm layer regular gypsum plasterboard, and the other two were protected with 12.5 mm regular gypsum plasterboard and an additional 15.4 mm layer of gypsum plasterboard treated to improve core cohesion at elevated temperatures. Three of these tests were constructed of heavy laminated timber, and the other (two layers protection) of wood studs with mineral wool. Four wood cribs were used as a fuel load. Temperature was measured within the timber using K-type thermocouples at 0 mm, 6 mm, 18 mm, 30 mm, 42 mm, and 54 mm below the exposed surfaces. Charring was assumed to occur at 300°C. The evolution of charring depth with time for zero, one, and two layers gypsum protection is shown in Fig. 8. It can be seen that the presence of gypsum plasterboard delayed the onset of charring by around 20 min per layer. It was also observed that for the two-layer test, charring rate significantly increased once the protective layer dropped off, however this was not the case for the one-layer test. This was attributed to the long pre-heating time for the two-layer test.
It can be concluded that the presence of a protective layer of gypsum plasterboard reduces the charring rate of the underlying timber, but once said protection falls off, the charring rate is then significantly increased due to the lack of an insulating layer.
Delamination
Delamination is a phenomenon through which the outer lamella (or part thereof) detaches from the second lamella, thus exposing unburned timber directly to the fire. The presence of a char layer also acts as protection for the underlying virgin wood [4, 5, 13, 15, 27, 39, 69, 70, 81, 82, 84, 85, 89, 93, 116, 119,120,121,122]. As a result, the pyrolysis rate in a fire-exposed timber member is initially high whilst no protective layer exists, before decreasing to a lower quasi-constant value once a char layer has formed [4,5,6, 12, 13, 32, 35, 39, 63, 67, 70, 82, 88, 119, 123, 124]. Schaffer suggests that a char depth of approximately 6.4 mm [82] or 12.7 mm [124] is required before the constant rate is reached, or that the time required to reach a steady state is given by Eq. 19, which is typically around 20 min. This corresponds to a steady-state char layer, and thus steady-state heat transfer into the underlying timber.
$$ t_{ss} = \frac{\alpha }{{\beta_{ss}^{2} }} $$
(19)
where \( \alpha \) is the thermal diffusivity (m2/s), and \( \beta_{ss} \) is the steady-state charring rate (m/s). It is also commonly acknowledged that when protection falls off—either gypsum board or a lamella of an engineered timber product—the pyrolysis rate subsequently increases due to the absence of a protective char layer [115,116,117, 125]. The thickness of the lamellae is thus important to the fire behaviour—if the lamellae are sufficiently thick, then CLT will behave more like solid timber [119].
Frangi et al. [116] tested spruce CLT panels of dimensions 1.15 m × 0.95 m × 0.06 m with either two, three, or five layers. Nine specimens were bonded with a polyurethane (PU) adhesive, and two with a melamine urea formaldehyde (MUF) adhesive. Temperature was measured using K-type thermocouples at the glue-lines. An additional 30 mm timber panel was placed on top of the samples, with an additional thermocouple in this boundary. The evolution of char depth with time is shown in Fig. 9 for all eleven tests. The lamella build-up was uniform for 2- and 3-layer tests, and four 10 mm layers and a 20 mm layer for the 5-layer tests. It can be seen that for samples bonded with PU adhesive, the charring rate increased after around 15 min—this is attributed to the layers falling off leaving an uncharred layer exposed—this was observed visually during the tests. For samples bonded with an MUF adhesive, the lamellae were not observed to fall off, and thus the charring rate remained approximately constant, behaving like a block of solid timber. Thicker lamellae generally showed lower overall charring rates due to delamination occurring less frequently.
Frangi et al. [119] then conducted an FE thermal analysis for CLT exposed to the standard temperature–time curve on specimens with either 3 lamellae of 9 mm, or 3 lamellae of 18 mm. It was assumed that the bonds between lamellae failed at 300°C. Furnace tests were then carried out on 27 mm and 54 mm solid timber panels, and 3 × 18 mm and two 3 × 9 mm CLT panels made of spruce (ρ = 420 kg/m2) and moisture content 11%, bonded with a PU adhesive—the 3 × 9 mm panels were tested one on top of the other—essentially acting as a 6-lamella sample without adhesive between the 3rd and 4th lamellae. The 54 mm thick specimens resulted in experimental charring rates of 0.66 mm/min and 0.84 mm/min for the solid timber and CLT panels respectively, and the 27 mm thick specimens 0.62 mm/min and 1.05 mm/min respectively. Average charring rates are not presented for the FE analysis, but are shown by plots of char depth against time. Taking the values at the failure times (38 min for the 54 mm CLT panel and 28 min for the 27 mm panels) gives average charring rates of 0.93 mm/min for the 54 mm panel and 1.18 mm/min for the 27 mm panel—resulting in overpredictions of 10–12%.
Li et al. carried out a full-scale compartment fire experiment in a 2.5 m × 3.5 m × 4.5 m room with fully exposed CLT surfaces formed of three uniform 35 mm lamellae. They measured an average charring rate of 0.85 mm/min in panels where they did not observe delamination, but 1 mm/min in the ceiling where delamination was observed.
Hadden et al. [126] also carried out a series of compartment fire experiments in a 2.7 m × 2.7 m × 2.7 m room, with partially exposed CLT surfaces. They observed sharp increases in burning rate coinciding with the onset of significant delamination.
Friquin [5] relates the protective quality of the char layer to several factors. The char layer will be thinner than the virgin wood, due to a combination of char oxidation, mechanical surface degradation, and char contraction. The char conversion is the fraction of wood that is converted to char. With increasing char conversion, charring rate decreases due to an increased protective layer. White et al. [64, 69, 81] included a char conversion factor in their models (see Sect. 3.1.1), accounted for through Eq. (13); Eq. (12) shows charring rate to decrease as char conversion factor increases.
It is clear from the available research that delamination will cause an increase in the charring rate until a protective char layer is built up again. Whilst the effects of delamination are understood, its causes remain unpredictable. In particular, the failure modes and conditions of different adhesives used in manufacture of engineered timber products are not well understood.
Thermal Exposure
Heating Scenario
The heating scenario (conventionally this is defined as either a temperature–time curve or an incident heat flux) is known to affect the pyrolysis rate [11, 12, 16, 20, 21, 32, 71, 82, 88, 91, 93, 97, 98, 103, 121, 127,128,129].
Some of the earliest work on the ignition properties of wood is by Hill and Comey [57] from 1886, in which they tested pine samples in a metal heating bath, equivalent to a modern oven with different heating rates, with temperatures recorded by mercurial and air thermometers. It was found that if the sample was heated slowly, it had charred before ignition could occur, and thus it was char, not wood igniting, at temperatures around 405 to 417°C, but if heated quickly, would ignite at lower temperatures, around 360 to 372°C.
Hill and Comey [130] then tested small wood samples in constant gas temperatures, and measured the direct and indirect carbon dioxide production. They found carbon dioxide production per gram per hour to increase exponentially from 0.09 mg/gh at 123°C to 6.67 mg/g at 220°C.
The majority of research exploring the effect of heating scenario on charring rate has been undertaken under radiant heating [7, 12, 32, 71, 81, 85, 88, 98, 103, 104, 129]. The majority of researchers observed a clear increase in charring rate with increasing heat flux.
Mikkola [12] (see Sect. 3.1.2), investigated the effect of incident radiant heat flux, and applied a simple energy balance to determine the relationship between heat flux and charring rate, shown in Eq. (20):
$$ \dot{q}_{n}^{\prime \prime } = \dot{m}^{\prime \prime } \left[ {C_{p} \left( {T_{p} - T_{\infty } } \right) + L_{v} } \right] $$
(20)
where \( \dot{q}_{n}^{\prime \prime } \) is the net heat flux, \( C_{p} \) is the specific heat capacity, and \( L_{v} \) is the heat of gasification of wood. Combining this with Eq. (10) yields Eq. (21):
$$ \beta = \frac{{\dot{q}_{n}^{\prime \prime } }}{{\rho \left[ {C_{p} \left( {T_{p} - T_{\infty } } \right) + L_{v} } \right]}} $$
(21)
Application of Eq. (21) gave charring rates of 0.81 mm/min for spruce and 0.70 mm/min for pine, whereas experimental values obtained were 0.80 mm/min for both. As can be seen, charring rate is linearly dependent on net heat flux.
Di Blasi et al. [103] tested cylindrical beech (ρ = 730 kg/m3), chestnut (ρ = 590 kg/m3), Douglas fir (ρ = 515 kg/m3), redwood (ρ = 488 kg/m3), and pine (ρ = 450 kg/m3) samples of height and diameter 40 mm to radiant heat fluxes of 28 kW/m2 to 80 kW/m2. The heat flux was applied in the radial direction. Steady centreline temperature increased from 327°C to 677°C over the range of heat fluxes tested.
Yang et al. [88] tested China fir, Japanese cedar, and Taiwania samples in the cone calorimeter to a constant 50 kW/m2 for 15 min, and 40 kW/m2 for 5 min. The peak heat release rates were 14%, 27%, and 13% greater for 50 kW/m2 than 40 kW/m2 for Japanese cedar, Taiwania, and China fir respectively.
Lizhong et al. [104] found that surface temperatures under linearly increasing heat fluxes continue to rise steadily, whereas constant heat fluxes rise rapidly, before then continuing to rise much slower. For constant heat fluxes, time to ignition is proportional to the inverse of the heat flux squared, as in Eq. (3), but for linearly increasing heat fluxes, \( t_{ig} \propto \gamma^{ - 0.6897} \), where \( \gamma \) is the rate of increase of heat flux, suggesting significant differences in behaviour between different shapes of heating curve.
White and Tran [81] tested redwood (ρ = 309 kg/m3), pine (ρ = 450 kg/m3), red oak (ρ = 682 kg/m3), and basswood (ρ = 408 kg/m3) under constant radiative heat fluxes of 15, 25, 35, and 50 kW/m2. The specimens tested were 64 mm thick, and had a surface area of 150 mm × 150 mm, formed of four lamellae. Specimens were tested vertically, heated by silicon carbide heater element, with a pilot flame above the sample. Temperature was measured with K-type thermocouples at 6 mm, 12 mm, 18 mm, 24 mm, and 36 mm from the surface. Charring was assumed to occur at 300°C. Using the model \( t = mx_{c}^{a} \) developed in [69], \( a \) was found to increase with incident heat flux, suggesting increasing non-linearity with higher incident heat fluxes.
Considering charring rate, the expression in Eq. (22) was developed:
$$ \beta = 0.0121\dot{q}^{\prime \prime } - 0.00139\rho - 0.0199l + 1.527 $$
(22)
where \( l \) is the lignin content. This shows a linear dependence of charring rate on incident heat flux.
Reszka and Torero [98, 129] tested 100 mm × 97 mm × 67 mm redwood pine (ρ = 545 kg/m3) samples vertically in the cone calorimeter at heat fluxes between 10 and 60 kW/m2 perpendicular to the grain. Temperature was measured by K-type thermocouples inserted through the back face at depths every 5 mm up to 40 mm. An aluminium block was placed at the back to provide a well-defined end boundary condition. Increased heat fluxes were found to provide more rapid in-depth heating and mass loss rates than lower heat fluxes. Repeatability was good for the 60 kW/m2 tests, but poorer for the 10 kW/m2 tests. Two types of result were observed; one which underwent thermal runaway at 200°C, and one which showed long plateaus, taking many hours to reach thermal runaway. The theoretical temperature profile below the char line can be given by Eq. (23):
$$ T\left( x \right) = T_{\infty } + \left( {300 - T_{\infty } } \right)\left( {1 - \frac{x}{40}} \right)^{2} $$
(23)
where \( x \) is the distance below the char line.
Butler [71] compared Baltic redwood tested in furnaces for long durations at heat fluxes between 5 kW/m2 and 60 kW/m2 (shown in Fig. 10), and Guatemala cedar tested in a carbon arc for short durations at heat fluxes of 300 to 3300 kW/m2. This gives the relationship in Eq. (24):
$$ \beta = 0.02202\dot{q}^{\prime \prime } $$
(24)
This can be compared with the same plot for constant heat flux tests reviewed herein, shown in Fig. 10. Whilst the range is much smaller, and the scatter thus more visible (largely due to variation in the other parameters discussed previously), when a line of best fit through the origin is plotted, it yields Eq. (25):
$$ \beta = 0.02014\dot{q}^{\prime \prime } $$
(25)
which is very similar to Eq. (24), thus suggesting that charring rate does have a strong linear dependence on incident heat flux. However it should be noted that since the higher heat fluxes were tested for only short durations, it is likely that only the initial charring peak was captured, and not the subsequent quasi-constant charring rate, giving artificially high readings in Fig. 10. Lie [91] also states that charring rate is a linear function of heat flux. Furthermore, White and Dietenberger [20] state that charring rate varies from 0.8mm/min to 2.6 mm/min for external heat fluxes of 18 to 55 kW/m2, giving a slope twice as steep as than in Eqs. (24) and (25), shown in Fig. 10.
Lizhong et al. [104] tested acacia, elm, toon, and paulownia under linearly increasing heat fluxes of 0.20 kW/m2s, 0.30 kW/m2s, 0.42 kW/m2s, and 0.57 kW/m2s. They found that there was a consistent increase in ignition temperature for all species from 0.30 kW/m2s to 0.42 kW/m2s of around 20 to 30°C, but there was no significant difference in ignition temperature between the heating rates of 0.42 kW/m2s and 0.57 kW/m2s, summarised in Fig. 11. They observed that smouldering ignition occurred prior to a transition to flaming. Mass loss rates at time of ignition did not directly correspond to heating rates, but were generally in the range of 20 to 30 g/m2s.
Lizhong et al. [85] tested 100 mm × 100 mm × 20 mm samples of paulownia (ρ = 260 kg/m3), toon (ρ = 530 kg/m3), and elm (ρ = 590 kg/m3) in a radiation platform to linearly increasing heat fluxes, ranging from 0.07 kW/m2s 0.425 kW/m2s. Temperature was measured using type-K thermocouples at 0 mm, 2 mm, 6 mm, and 10 mm below the exposed surface. Charring was assumed to occur at 300°C. The charring rate increased with heating rate from 0.31 mm/min to 0.95 mm/min, giving the empirical relationship in Eq. (26):
$$ \beta = 136\gamma^{0.51} \rho^{ - 0.76} $$
(26)
where \( \gamma \) is the rate of increase of heat flux.
Kashiwagi et al. [32] tested 38 mm cubes of white pine (ρ = 380 kg/m3) in a radiant heating apparatus consisting of a 100 mm × 130 mm graphite plate heated to 1250°C with a 90 mm × 90 mm water-cooled pipe transmitting spatially uniform radiation to the sample at heat fluxes of 25 kW/m2, 40 kW/m2, and 69 kW/m2. Increased heat flux was found to produce a sharper, earlier peak in pyrolysis rate. Quasi-steady surface temperature increased from 500°C at 25 kW/m2 to 700°C at 69 kW/m2. They found that the thermal wave penetrated much further at higher heat fluxes due to a steeper thermal gradient, tending to increase the char depth. Carbon monoxide, carbon dioxide, and hydrocarbon generation was seen to increase with heat flux.
Moghtaderi et al. [39] found that ignition temperature decreased with external heat flux, from 366°C at 20 kW/m2 to 335°C at 60 kW/m2 at 15% moisture content, due to shorter ignition times.
Furthermore, Inghelbrecht [7] found that the mass loss rate (and thus charring rate) tended towards an asymptotic value regardless of heat flux, suggesting that it is only the initial, high charring rate, which can be 4–5 times the steady-state charring rate, (see Sect. 3.2.4) that is dependent on external heat flux. Charring rate under steady-state conditions has minimal dependence on external heat flux.
Schaffer [82] tested Douglas fir, southern pine, and white oak in a furnace at constant exposure temperatures of 538°C, 816°C, 927°C. It was observed that higher temperatures yielded higher charring rates. Additionally, it was found that char formed at gas temperatures of 538°C had a finer texture than that formed above 538°C. It was found for the higher exposure temperatures that two structurally distinct char layers had formed of approximately equal thickness, with the inner layer having greater structural strength. These differences are not defined by the author, and are likely to be a gradient of strength as a function of how complete the pyrolysis is.
Silcock and Shields [128] compared char depth to local fire severity, which they defined in Eq. (27).
$$ LFS = \mathop \int \limits_{0}^{\tau } \dot{q}\left( t \right)^{\prime \prime } dt $$
(27)
where \( \tau \) is the total fire duration, and \( \dot{q}\left( t \right)^{\prime \prime } \) is the incident heat flux. Comparing cone calorimeter and furnace tests with this definition, they found the expression in Eq. (28).
$$ x_{c} = 0.675LFS^{0.645} $$
(28)
Where \( x_{c} \) is char depth, again showing char depth to increase with incident heat flux.
There is good agreement between authors that charring rate increases with increasing heat flux. From the works available, this is an approximately linear function. Because exposures are not consistent between works, it is hard to draw definite conclusions. It is also noteworthy that only Inghelbrecht [7] explores the effect of the initial peak, and observes a much weaker dependence thereafter. This is particularly important to consider for longer (or shorter) fire durations.
Oxygen Concentration
The oxygen concentration in the environment surrounding the exposed timber also influences its pyrolysis and combustion behaviour—a lower oxygen concentration will result in lower char oxidation rates, thus there will be more char to reduce the heat transfer into the underlying timber. This has been found to result in a decreased charring rate with decreasing oxygen concentration [12, 32, 84, 94, 97], with the exception of Butler [71] found that tests in nitrogen and air showed little difference in charring rate, however this is likely due to the large range of heat fluxes tested, which would dominate over variations in oxygen concentration.
Several researchers [12, 32, 33, 72, 97] explored the effects of oxygen concentration on timber pyrolysis under radiative heating, finding that charring rate decreased with decreasing oxygen concentration.
Mikkola [12] found that mass loss rates decrease by approximately 20% for a 50% reduction in oxygen concentration, and 50% with 100% for white pine [33], and 35% for a 100% reduction for Douglas fir [105].
Babrauskas [97] found that in furnace tests, typical oxygen concentration varies from 4% to 10%. At around 8% to 10% O2, charring rate decreases by around 20% from ambient conditions, and a further 33% at 4% O2 (to just over 50% of charring rate at ambient conditions).
Kashiwagi et al. [32] varied oxygen concentration from 0% to 10.5% to 21%. Mass flux was found to increase with increased oxygen concentration. After 3 min exposure to 25 kW/m2, the mass flux in air was approximately four times that in nitrogen.
Delichatsios [72] exposed 100 mm × 100 mm × 4 mm regular and fire retarded plywood to 25, 35 and 50 kW/m2 heat fluxes in a cone calorimeter at oxygen concentrations of 21, 18, and 15%. The samples were insulated at the back face by 46 mm ceramic board. The fire retarded plywood was found to have ignition times increasing very slightly with decreasing oxygen concentration, from 54 s at 21%, 57 s at 18%, and 59 s at 15% under 35 kW/m2, and decreasing with increasing heat flux. The regular plywood samples also showed very little dependence on oxygen concentration.
Ohlemiller et al. [33] tested 38 mm × 38 mm white pine and red oak samples in a radiative heating apparatus at oxygen concentrations of 0%, 10.5%, and ambient, parallel to the grain. They found that at 40 kW/m2, mass flux was about twice that in nitrogen; at 25 kW/m2 it was approximately four times. Different oxygen concentrations start having effects on pyrolysis at surface temperatures around 350°C. Char forms more rapidly in an oxygen-rich environment, which speeds up local temperature rise and pyrolysis reactions. Between 120 and 240 s, the front surface achieved a quasi-steady value of 650°C in air, and 550°C in nitrogen. Furthermore, the pyrolysis zone thickness was found to be just 10 mm in nitrogen, but 15 mm in air. The exothermic char oxidation process adds to the external heat flux, increasing pyrolysis rates. Thermogravimetric analyses at 5°C/min found char to oxidise at 400 to 500°C, but at temperatures in excess of 650°C at 40 kW/m2. In nitrogen, the initial product released is water, followed at surface temperatures around 300 to 400°C by tars which then dilute the water. Carbon monoxide, carbon dioxide and hydrocarbons follow more slowly at the same temperatures. As tars pass through the char layer, they may undergo cracking to smaller hydrocarbons, or polymerisation to a char. Since the hydrocarbon yield does not increase as rapidly as tar production falls, this suggests some tars do repolymerise to char, possibly forming water in the process. The yield of carbon monoxide and carbon dioxide increases dramatically in oxygen, due to the greater availability of oxygen to form these compounds through char oxidation. Greater heat fluxes led to earlier and higher mass loss peaks. Quasi-steady surface temperature also increases with heat flux, from 550°C at 25 kW/m2, to 650°C at 40 kW/m2, to 750°C at 69 kW/m2.
In Cedering’s [94] tests on timber stud walls (see Sect. 3.1.2), oxygen concentration was set at 4, 7, and 10%. It was found that the charring rate increased with oxygen content—with averages of 0.65 mm/min at 4%, 0.68 mm/min at 7%, and 0.70 mm/min at 10%. This was shown to be less important at higher moisture contents.
Overall Summary
The effects of various parameters on the pyrolysis rates of wood have been extensively, if not systematically, studied. This is most easily quantified through “charring rate”. A summary of these parameters and their relative importance is given in Table 3. In the works discussed above, several different criteria have been used to determine charring rates, however this is not critical due to sharp mass loss peaks occurring at well-defined temperatures. Various different temperatures have been used to determine the location of the char front; this is commonly defined as 288°C in North America and 300°C elsewhere, but has been found to be as high as 360°C in some cases. The precise temperature is not critical due to the steep temperature gradient at the charring front [5].
Table 3 Summary of Factors Affecting Charring Rate and Their Significance Whilst there is considerable scatter in the results presented in this paper, it can be clearly seen that the effect of incident heat flux is the most dominant parameter, with an order of magnitude higher influence than the other parameters considered, over the ranges to be expected in normal design. Therefore, it is vital that this influence be properly understood to allow safe, robust structural fire engineering design. It is therefore also crucial to develop a proper, scientific understanding of the fire exposures that are likely to manifest in timber compartments, particularly those with a significant amount of exposed structural timber.