The Fire Performance of Steel-faced Insulation Panels with Stone Wool or Polymer Cores – A Scaling and Heat Transfer Study Based on Full-scale and Scaled Experiments

The heat transfer and fire dynamics of enclosure fires in compartments made from steel-faced sandwich panels were studied experimentally in order to determine the feasibility of down-scaling the enclosure size of the ISO 13784-1 test compartment. The different scales used for the experiments were 1:5, 2:5, 1:2 and 1:1. For all scales, experiments were conducted with two types of steel-faced sandwich panels, one with a core of stone wool (SW) and the other with a core of a polymeric product (Polyisocyanurate (PIR) or Polyurethane (PUR)). The parameters studied were the heat release rate (HRR), the temperatures of the air and smoke layer inside the compartment, the temperatures at and in the walls, the pressure differences and temperatures in the doorway of the compartment and the time to flashover. For the smallest scale that was studied (a 1:5 ratio relative to the ISO 13784-1), the mass loss as a function of time was also measured.
The majority of the experiments was conducted at 1:5 the scale of the ISO 13784-1 test compartment. The 1:5 scaled compartment measured 0.48 m x 0.72 m x 0.48 m internally in width, length and height, respectively, with a 0.40 m tall by 0.16 m wide doorway in the middle of one of the short walls. A gas burner was located at the wall in one of the corners opposite the doorway. Rather than a section of the compartment the scaling of the entire compartment was hypothesised to reflect the full sized compartment test better than the current standard test, thus making it a more robust test method. By scaling the whole geometry of the compartment on top of the energy provided by the gas burner the boundary conditions seen at full scale would be represented better. This is unlike the current open corner configuration where smoke is immediately extracted and removed from the specimen which is a part of the current way of determining the performance of the panels in Europe. The experiments followed the methodology presented in the ISO 13784-1 with a two-step fire scenario with 100 kW and 300 kW for 10 min of duration each but also other variants of fire scenarios were studied.
The two polymeric cored panels studied, PIR and PUR, were class B with respect to contribution to flashover by EN 13501-1 which is the European classification standard for reaction to fire. Despite the PUR showing a lower calorific energy content than the PIR at component level, it performed significantly worse in the scaled compartment. Exposed to the same fire scenario, the compartment of PUR panels lost 85% of its mass in 10 minutes, all while releasing 220 kW at its peak which was more than 30 times the energy output from the burner. As such, the PUR panel did not have characteristic trade of a class B product such as a limited contribution to flashover and was deemed unsafe to use at larger scales. The PIR core contributed about 15 times the energy output from the gas burner and lost about 60% of its initial mass over a period of 10 minutes. The remaining 40% of the mass consisted of both char, various degrees of decomposed foam and virgin material. The mass of the char, compared to the virgin material, was much less and much of the remaining mass is attributed to virgin material being effectively protected by the char. The stone wool panels were studied under more challenging conditions at 1:5 scale where the input of the burner was sufficient to cause external flaming emanating through the doorway. Despite the average temperature of the compartment being more than 600 °C, the measured HRR was simply following the energy output from the burner. At larger scales the smoke production increased, which suggested a decomposition of the core rather than just a decomposition of the steel-face finish. Across all the scales the thermocouples embedded in the walls showed signs of minor exothermic reactions identified as an increase in the temperature 2 cm in-depth after the passing of the thermal wave.
A newly developed one-dimensional heat transfer model with the thermal properties of the SW core lumped together as an effective parameter was used to analyse and compare the experiments across the scales. The model together with the measured temperature of the boundaries in combination with flow measurements quantified the total energy distribution in the compartment during the experiments. The model was able to predict the thermal wave within 5% for the first 20 mm of the in-depth measured temperature and was therefore able to inversely provide the net heat flux transferred through the internal compartment boundaries. The measurements of hot gases were used to calculate and analyse the convective energy flowing out of the compartments and the fraction of the total heat transfer it accounted for. The nature of the heat transfer model allowed for the quantification of the net heat transfer for the experiments with SW panels. The goal to combine the model with the measured data to show that down-scaling the compartment geometrically was done by scaling the size of the fire based on the Froude number with respect to the geometry of the compartment. The calculated steady state heat flux between the 1:2 scale and 2:5 scale experiments were matching making meaning reducing the size by 25% was possible. Furthermore, the temperature in the compartments for the 2:5, 1:2 and a full scale experiment from literature matched for the first two burning periods. The third burning period was much warmer in the full scale than the 2:5 and 1:2 indicating a limit for the scaling with respect to the size of the fire. The successful reduction in the size of the compartment has many advantages such as work safety, product development, classification and cost.
The data herein presents the use of micro scale and macroscale to understand critical temperatures and incident heat flux where the core material pose a potential risk. In the 1:5 scale compartments both the temperature and the rate of temperature change over time were lower than for the other scales and the temperature did not reach steady or even quasi-steady state. The temperature in the 1:2, 2:5 and full scale (1:1) compartments reached a quasi-steady state during the first 10 minutes burning periods. Meaning the 1:2 and 2:5 scale compartments behaved in the same manner as the full scale. The compartments with stone wool had the same ratios of global heat transfer, consisting of 60% - 80% convective losses through the doorway and 20% - 40% conductive losses through the compartment boundaries for all scales. The 60% is just after the increase in the burner intensity where the gradient between the boundaries and the gas phase is greatest. As time passes and the walls heat up and less heat is lost via conduction and more through the doorway as convective energy increasing its proportion of the total energy to 80%. It was not possible for the heat transfer model to account of 100% of the energy dispensation throughout the duration of the experiments. However, taking the uncertainties of the measurements into consideration, the global heat transfer presented the same energy distribution across all four scales towards the end of the experiments, where steady state was approximated. The experiments therefore show that the energy distribution in compartment with the fire scaled using the Froude number is the same when approaching steady- or quasi steady state.
The successful downscaling of the size of the compartment was limited to 1:2 and potentially 2:5, whereas the 1:5 scale compartments were not able to mimic the exact behaviour seen at larger scales. This is based on the compartment temperatures and non-dimensional HRR prior to failures differed too much. This was evident from the fact that the compartment and walls during the first and second burner step stabilised for the 1:5 scale, whereas failure occurred
immediately after the initiation of the second burner step at both 1:2 and 1:1 scale. The heat transfer through the walls of the compartments with SW panels determined that the third burner step where the HRR was additionally doubled compared to the second step was not successfully scaled as the heat flux was much greater in the full scale experiment compared to the other scales. The experiments with the 1:2 sized compartments had matching compartment temperatures and net heat fluxes across the solid boundaries as the full scale experiment. The net heat flux for the SW compartments showed that the first burner step for the 1:2 and 1:1 scale from literature was matching while the second step had matching net heat fluxes in the near and far field from the burner. The HRR development and time to failure for the 1:2 and two 1:1 scale (of which one is from literature) experiments with PIR panels matched very well and all the compartments failed at the initiation of the second burner step. The fact that the results from the experiments with the 1:2 scale compartments matched the results from the full scale experiments provides; 1) data for successfully conducting research on compartment fires at smaller scale, which greatly benefits intuitions with smaller laboratories, 2) classifying and regulatory bodies data arguing for smaller compartments tests, which are easier to handle, and 3) the manufactures with data showing how reduced scale testing can predict large scale failures for research and development at a reduced expense.