The three ‘natural’ fire experiments presented herein all had identical fuel loads, were subjected to the same sustained structural loads during fire exposure, and were manufactured by the same supplier from the same timber strength class with the same adhesive type and ply configuration. However, differences in opening geometry, and hence ventilation factor, resulted in considerably different rates of fire growth and durations of pseudo-steady burning, this in turn resulted in significantly different in-depth temperatures in these slabs that were sufficient to result in different structural behaviour, ranging from surviving the full burnout (notwithstanding subsequent smouldering) to collapse at the onset of the decay phase of the fire. These differing responses highlight the importance of understanding the fire dynamics in a compartment fire in a CLT structure, as well as the resulting changes in exposure temperatures.
Of the three ‘natural’ fire scenarios, Scenario 1 had a large opening, corresponding to an opening factor of 4.64 m−0.5 and indicating a more oxygen-rich environment (i.e. tending to fuel controlled). This in turn resulted in a comparatively high proportion of the fuel being able to burn within the compartment (as opposed to in an external fire plume [35]), and resulted in a high internal heat release rate (HRR) and thus a rapid temperature rise. This also resulted in a shorter pseudo-steady burning duration, as the available fuel was able to burn more rapidly due to enhanced re-radiation from the internal burning. Scenario 2 had smaller openings, corresponding to an opening factor of 14.2 m−0.5, indicating a lower availability of oxygen and tending to ventilation controlled. This led to less internal burning (HRR), and resulted in a lower rise in temperatures, less re-radiation to the fuel load, and thus a longer fire pseudo-steady burning duration. Scenario 3 had only one small opening, corresponding to an opening factor of 23.2 m−0.5, indicating even lower oxygen availability and clearly ventilation-controlled burning. This again led to slower temperature increase and further increased the pseudo-steady burning duration; this ultimately meant that the timber was subjected to a longer thermal attack than for the other two scenarios (however at similar peak temperatures), which led to a deeper char depth and enhanced thermal penetration—and ultimately to structural failure during the decay phase. This is potentially important because it means that, for CLT (and other forms of mass timber such as glued laminated timber), the duration of a fire is likely to be more critical for its load bearing capacity than the peak temperatures in the enclosure (for the likely ranges of temperatures). This is intuitive when considering the relatively high thermal inertia and charring behaviour of timber, meaning that temperature increases are concentrated at the hot surface layer with comparatively steep in-depth thermal gradients.
A longer heating phase results in the CLT being exposed to an incident heat flux for a longer duration, thus resulting in increased in-depth heating, and subsequent loss of mechanical properties. The earlier failure of Scenario 3, in comparison to the standard furnace tests, which experienced deeper charring depths but longer failure durations, can thus be expected when the fire dynamics are properly considered. Such consideration is effectively prevented by the current widespread approach of applying normative fire resistance ratings to fire-exposed mass timber structural elements.
The data presented herein also provide compelling evidence that it is important to explicitly consider the decay phase of real fires, rather than relying solely on normative fire resistance ratings for CLT structural elements. After the peak gas phase temperature is reached and the movable fuel load is consumed, the hot surfaces within the compartment will re-radiate, and thus the in-depth temperatures will continue to increase. Even after the room temperature drops, a thermal wave will continue to propagate throughout the member, further increasing the in-depth temperatures for a period which will depend on both the maximum gas phase compartment temperatures and the duration of steady burning. Furthermore, the risk of continued, localised smouldering of the timber means that continued in-depth heating may continue for many hours after the fire appears to have gone out, as was observed for Scenario 2 which collapsed after 29 h due to unseen smouldering. Such considerations are also precluded by the use of normative fire resistance ratings to assess the structural fire response of CLT structural elements.
Both Scenario 1 and Scenario 2 survived fire exposure and decay phases without experiencing structural failure. It can be seen from Fig. 10 that both slabs’ load bearing capacity stabilised as the compartment temperatures reduced and the fire burned out.
Scenario 3 experienced structural failure (i.e. collapse) after 108 min and, as can be seen in Fig. 5, the temperatures at this point remained close to 1000°C. It would thus be premature to label this as a case of auto-extinction, which is also confirmed in Fig. 7 where it can be seen that charring is ongoing at the time of failure in this case. It thus appears that if auto-extinction can be achieved and charring halted, as was observed for Scenarios 1 and 2, then failure in the decay and timber cooling phases of a natural fire for a CLT element in simple bending appear to be unlikely due to redistribution of heat within the CLT (notwithstanding the possible occurrence of ongoing smouldering, which should be considered as a separate issue and is beyond the scope of the current discussion). This is due to the fact that, for slabs in positive bending (i.e. sagging) heat will mostly affect the tension side, and both tensile strength and elastic modulus are generally assumed to be much less affected by elevated temperatures than in compression [4, 36].
Clearly, for a larger initial fuel load or a higher applied load, Scenario 2 could have sustained flaming combustion longer and may have reached a stage where the applied bending moment would have been exceeded. This should be calculated on a case by case basis, and the sectional analysis presented herein shows—and partially validates—how this can be done from a structural fire engineering perspective. This requires that ongoing smouldering, which was observed to cause structural failure for Scenario 2 after 29 hours, can be prevented.
The proposed sectional analysis model, which has not previously been experimentally verified [31], appears able to accurately predict the flexural failures (or lack thereof) for the experiments described in this paper, and such an approach can therefore be seen as a suitable analysis tool to investigate the load bearing capacity of CLT in bending in compartment fires with a decay phase. This analysis model correctly predicted that continuous decay of the structural capacity in one way spanning CLT slabs is limited and unlikely to cause collapse if auto-extinction of a fire can be achieved before structural failure and the progression of the char front has halted. In a previous paper dealing with similar topics, Wiesner et al. [31] had postulated that the fire decay phase may be more critical for compression elements and, given the partial validation of the model herein, the potential for structural collapse in timber compression elements in a fire decay phase should indeed be considered pertinent and to warrant further research.
The RCSM from EN 1995-1-2, which is currently used in practice for the design of cross-laminated timber projects, resulted in unconservative fire resistance predictions. However, it must be reiterated that the RCSM method, in its current form, was not developed, and should not be used, for CLT. Previous researchers have already demonstrated that application of the RCSM to CLT is likely to result in unsafe capacity predictions [37,38,39]. It is expected that future versions of the Eurocode will suggest modifications to the charring rates and the zero strength layers to attempt to provide more conservative simplified design solutions for CLT and other engineered timber products.
The results of the experiments described herein highlight the importance of the ventilation conditions in a compartment with exposed timber surfaces, but also demonstrate the potential to engineer compartments at the design stage based on considerations of the expected burning rates. An early involvement and consultation with structural fire engineers could therefore avoid problems that might arise if the structural fire safety is considered only at later design stages.