Experimental investigation into the influence of ignition location on flame spread and heat release rates of polyurethane foam slabs

This study presents the results from a set of 11 large‐scale open fire tests performed on flexible polyurethane foam slabs/mattresses. The purpose of the study was to investigate the influence of the ignition location on the fire behaviour of the foam slabs and to generate data on a highly characterised material that could be used for modelling work in the future. A method for obtaining spatially resolved flame spread data for this type of material was presented using a gridded array of 5 × 10 thermocouples placed on the underside the foam slab and from this, flame spread was examined using three different approaches. The heat release rate (HRR) results showed clear shapes forming that were dependent on the ignition location, with two distinct behaviours being observed between the various different ignition locations, this was also observed in the calculated flame spread rate (FSR) data. Results within an individual test, showed the calculated range of FSRs over the geometry of the slab varied between approximately 1 and 8 mm/s depending on the ignition location. The average FSR values between tests varied between 3 and 7 mm/s and the maximum and minimum values were calculated to be approximately 11 and 2 mm/s respectively.


| INTRODUCTION
Understanding what may influence a fire scenario in real (large) scales is an important, though a complex task. In general, this may be done by performing large-scale experiments, unfortunately, these are commonly expensive and time consuming to perform. Due to this restraint, repeat tests, and investigating the influence of minor configuration changes may not be feasible in many cases. However, due to the inherent complexity of fire phenomena "minor" changes are also important to investigate when possible as they may bring new insights and can lead to a deeper understanding of fire behaviour.
Horizontal flame spread, for example, over mattress type products, is a much-studied area within fire science and there are many factors that can influence how a fire spreads horizontally over a material's surface. The point-of-ignition is one such factor that is generally known to have an influence on flame spread and heat release rates (HRRs). 1 Despite this, there are very few studies that attempt to investigate its influence systematically.
Mitler and Tu 1 provided a brief study investigating the dependence of the burning behaviour of upholstered chairs on ignition location. As part of a series of full-scale chair burns at NIST, four identical chairs of one chair configuration (style C) were selected and four different ignition locations were tested: (a) centre chair seat cushion, (b) lower centre of chair front, (c) lower centre of chair side, (d) lower centre of chair back. Heat release, mass loss and smoke measurements were recorded, however, the ignition source was not discussed.
The study concluded by stating that, based on the changes in HRR measurements recorded, ignition location can have a significant effect on the time to reach peak HRR, however, no further analysis is given.
In the book by Krasny et al 2 -fire behaviour of upholstered furniture and mattress, the majority of the described work on ignition is about characterising different ignition sources and their effects. Section 4.6.0 mentions the fact the location will affect the results, however, the research quoted from NIST 1 and the CBUF 3 programme was on chairs.
Robson et al 4  Concluding that the ignition location "impacted considerably nearly all the important fire parameters, including peak HRR, time to peak HRR, smoke temperature, CO concentration and the extinction coefficient." However, this was in a vertical flame spread context. Söderbom et al 7 investigated the effects of changing the size and heat output of two types of propane burner ignition source (varying output from 1.7 to 40 kW), and concluded that there was little effect to the overall outcomes after their ignition criteria was met (a HRR of 50 kW), and that the ignition source only affected the ignition period (time to reach 50 kW criteria). However, they did not change the location of the ignition source in this case other than the change in physical dimensions and output of the burners used.

| AIM
The aim of this study was to systematically investigate the dependence of measurable experimental outcomes, namely the HRR and flame spread, on ignition location, using large-scale fire tests for a specific geometrical slab. It is hoped that this work can then be used in the further development of flame spread modelling in the future.

| MATERIALS
FPUF slabs were chosen as the test specimen as they could represent a semi-realistic scenario, in which the foam slabs act as a simplified proxy for bed mattresses, The FPUF investigated in this study was basic, commercially available, non-fire retarded polyether polyurethane, distributed by Sherlock Foams, UK. This particular foam has been used by Laboratoire national de métrologie et d'essais (LNE) as a reference material 8 and in previous research studies by the authors. 9,10 The choice was also made to provide data for modelling purposes. Basic properties are provided in Table 1 and readers are referred to previous research referenced above for more detailed properties on the tested material.
The slabs dimensions were 1200 × 600 mm with a thickness of 50 mm; this represents the standard-sized mattress dimensions of a child's crib/bed. The foam was tested naked, that is, no textile covers or other materials were used as fuel in the study. Naked foam is used predominately to reduce the complexity of experiments and the data obtained, other work by the author 9 investigated the same foam with fabric combinations at a bench-scale level.

| Experimental setup
All experiments for this study can be characterised as free burning tests, performed under an open calorimeter at the Danish Institute of Fire and Security Technology (DBI). A total of 11 tests were performed for this specific study. Further tests were also carried out with this setup, however, additional metrics were added that were out of the scope of this study, thus they have been excluded. Before testing, slabs were conditioned for at least 24 hours at 23 C and 50% relative humidity.
A base frame to hold the samples was constructed from timber, on which a 12 mm calcium silicate board (CSB) was placed. Stone wool insulation was pressed in under the CSB between the frame supports to better protect the measurement equipment.
The frame and board were fabricated to be slightly larger than the FPUF slabs and a small raised perimeter edge was added to the base frame so that any potential spillage from the fire tests would be prevented (refer Figure 1). The base frame CSBs and raised edge were covered in aluminium foil before each test so that any residue left could be removed before the next test was performed.

| Measurement equipment
HRRs, mass loss and temperature measurements above and below the slab were performed for most tests. Video recording of each test was also undertaken.
HRR measurements were performed using an open calorimeter setup at DBI. The calorimeter setup is based on the ISO 24473 standard 11 and measures HRR using oxygen consumption calorimetry. 12 The combined expanded relative uncertainty of these measurements is considered to be between 7.1% and 10.6% based on the work from SP. 13 Mass loss measurements were accomplished using a load cell (Sartorius M-177 with a sensitivity of 0.0001 kg) capable of holding the base frame and slabs. The load cell was protected from heat by the base frame and the insulation that was sandwiched in between the base frame supports, which did not touch the load cell.
Temperature measurements were performed using 0.5 mm diameter wire type-K thermocouples. Thermocouple (TC) arrays, set out in a grid of 5 by 10 (total of 50) were placed above and below the FPUF slabs for some tests, others were performed without the TC grid above the slab. The grid was positioned such that, spacing was 120 mm between each TC, and that the TCs on the sides were placed 60 mm in from the edge of the FPUF slabs as illustrated in Figure 1.
Each TC location was identified based on its position in the grid, rows 1 to 5 were named A-E and columns were designated 1 to 10 as shown in Figure 2.
TCs placed below the slabs were done by drilling very small holes (approx. thickness of the TC wire) through the CSB, then pushing the TCs through each hole so that each TC stuck out approximately 3 mm from the surface of the CSB. TCs were then fastened from the backside of the board with aluminium tape and staples. During tests 6 to 8, a TC array was also placed above the slab and positioned in the same grid layout as the TCs below. TCs were suspended and attached 50 mm above the surface of the slabs using five steel wires (corresponding to positions A-E) tensioned between supports on either side of the slab configuration. Temperatures were logged at 2-second intervals using data loggers.
Measuring flame spread rates (FSRs) in this study employed the TC arrays embedded in the tray. This configuration was based on initial testing of the foam slabs in the cone calorimeter, where it was observed that temperature measurements on the underside of the slabs had a distinct temperature rise. In addition, compared to temperature measurement performed above the slabs in the gas phase they were more robust and reliable, due to TC wire damage occurring in the gas-phase and the inherent turbulence within the flames above the slabs surface, making exact positioning of surface spread more difficult to obtain.
Due to the insulative properties of the foam slabs, it was observed that backside TC measurements stayed relatively low right up to the point of foam collapse, upon which a very sharp temperature increase could be observed in the time-dependent temperature readings, exemplified in Figure 3 (C4 or C5 refers to row and column designation). Assuming this was consistent throughout all the experiments, this gave a clear point at which the movement of the collapse could be monitored with relative certainty.
Structural collapse in these foams is a commonly observed phe-

| Ignition locations
This study aimed to investigate how changing the ignition location on a horizontal slab can affect commonly measured test outputs such as HRR and FSRs, five different ignition locations were chosen (refer to Figure 2). Tests were performed in an open calorimeter, and it was assumed that airflow around the slab was relatively symmetrical, thus flame spread behaviour could also be considered symmetrical (ie, behaviour would be similar igniting from either side of the slab when compared to its opposing side). Based on this assumption, the placement of ignition locations was confined to 1 quadrant of the slab, assuming symmetry in both planes as shown in Figure 2. However, it should be noted that no assessment of the symmetry was performed, thus the influence of, for example, drafts or other asymmetrical influences cannot be discounted.
Five locations were chosen within the quadrant, 1 in each of the quadrants 4 corners, and 1 in the centre of the quadrant, with the aim to spread the locations evenly over the space. Repeat tests were conducted on 4 of the 5 locations, with IL1 and IL3 being tested a total of 3 times, as these represented significant differences in spread mechanics from initial tests and a least 3 tests were required to get a basic measure of repeatability. Table 2 summaries the ignition locations for each test scenario. Tests 9, 10 and 11 were performed separately at a later date with differing environmental conditions.

| Test procedure
At the start of each test, aluminium foil was placed over the base tray and secured using aluminium tape. Each underside TC was checked to confirm that it had pierced the foil to reduce disruptions to the measurements from the foil. The FPUF slab was taken from the conditioning room and placed centrally in position on the base frame tray.
When used, the above slab TC's were then checked to confirm the position and the initial mass of the slab was recorded. HRR measurements, mass loss, TC logging systems and the video camera were then initiated in order to gather baseline data. After a 60s baseline recording, the slab was ignited. Ignition of the slab was performed using a small open flame diffusion gas burner, similar to the standard cigarette lighter; 16

| Data analysis
Data from multiple sources were analysed in this study, HRR measurements were calculated using oxygen consumption calorimetry 12 principles, as per ISO 9705. 17 Mass loss measurements from the load cell were converted from the analogue signal, to mass loss using a predefined conversion input in the data logging system. The calculation of FSRs based on the TC measurements was determined in a number of steps: • The TC measurements were filtered/smoothed using a 10-point moving average to dampen any noise that may interfere with the calculations.
• The derivative of the TC curves was then calculated, and the maximum value was then found-due to the previous filtering step, the maximum value was always calculated to be on the initial temperature rise (as observed in Figure 3). This was taken to be the point in time at which fire spread had reached the TC.
• The time taken to reach this point was then determined.
• The distance between each TC position from the TC position closest to the ignition location (TCig) was also calculated for each scenario. Time t = 0 was when TCig reaches the stated maximum derivative, and distance to each point was calculated via Pythagoras's theorem from TCig. Hence, comparison with other spread rate studies may be misleading and should be considered when reading the results from this study. It was also observed that a distinct decrease occurred (although less pronounced than the observed increase) in backside TC temperature when burnout occurred at the TCs location, thus a burnout/extinction rate (ie, when the burning stops in each TC location) may also be calculated in the same manner.

| RESULTS
The set of results of interest for this study are the HRR and FSRs.
Mass loss, total heat release, burn time, smoke production etc. were measured but are not discussed further in this paper.

| Heat release rates
HRR results are presented, based on their ignition location in Figure 4 with replicate tests when performed. HRR results are presented as time-dependent plots and in terms of peak HRR and time-to-peak HRR ( In general, Table 4 shows that a higher correlation coefficient values tend to be observed for the repeat tests and for the tests where similar behaviour in the HRR is observed (eg, c1-2 and Ee1-2-3).
It should also be noted that the cases of the greatest variation in results were for tests 9, 10 and 11 when compared to others performed. This is likely, in part, to be due to these tests being performed at a different time which may have resulted in slight differences in testing conditions.  (Table 5).

| Visualisation of geometric flame spread
Using the TC method outlined in Section 4-data analysis, the geo-

| DISCUSSION
FSRs showed consistency over repeat tests, however, this consistency was reduced for repeat tests 9, 10 and 11 when compared to others fire safety engineers (FSEs) that may need to consider a fire scenario with a similar fuel source. In this case, FSEs may take a fire curve from each class to better cover the potential effects of that type of fuel source rather than picking one curve arbitrarily. However, it should be noted, that these results are only for one type of geometry, and much more remains to be studied, including the influence of "free vector space," that is, "the allowable space for flames to F I G U R E 9 Legend on next page. Using the FSR maps (Figures 10-14 This suggests that a temperature of between 400 C and 500 C may be the maximum temperature any part of the "solid" foam slab can achieve before it is volatilised. Hence, any TC located in the foam, will not achieve a higher temperature unless it "sees" direct flame impingement.  FIRETOOLS project-a collaboration between Lund University and Danish Institute of Fire and Security Technology (DBI).