Citation

Stec AA, Hull TR, Purser JA, Purser DA. Fire Safety J. 2009; 44(1): 62-70.

Copyright

(Copyright © 2009, Elsevier Publishing)

DOI

10.1016/j.firesaf.2008.03.005

PMID

unavailable

Abstract

aSchool of Forensic and Investigative Sciences, University of Central Lancashire, Preston PR1 2HE, UK bHartford Environmental Research, 1 Lowlands, Hatfield AL9 5DY, UK Toxic products are the main cause of fire injuries and deaths, but available methods for measuring or calculating toxic product yields have severe limitations. Full-scale or large-scale experimental re-creations of fire scenarios are sometimes used for the assessment of toxic hazard, but such tests are expensive, while small-scale or even larger-scale tests often provide poor simulations of full-scale conditions. From a testing and engineering calculation perspective there is a need for test methods to provide data-enabling calculations of toxic product yields in defined full-scale scenarios. Full-scale and large-scale tests have demonstrated that toxic product yields are highly dependent upon the combustion conditions. Fire stages and types can be characterised either in terms of CO2/CO ratio, or preferably in terms of equivalence ratio, which provide reasonably good predictive metrics for product yields. The steady-state tube furnace (ISO TS 19700) allows individual fire stages to be replicated and shows a good general agreement with product yield data (measured for CO2, CO, HCN, NOx, total hydrocarbons and smoke particulates) obtained from large-scale ISO room tests for the five materials considered here and expressed as functions of equivalence ratio and CO2/CO ratio. The closest direct agreement between the large- and small-scale data were obtained for pool fires involving PP and nylon 6.6 product yield. For materials burned as wall linings, with varying decomposition conditions at different room locations, and/or when a propane flame is also present, direct comparison with tube-furnace data is more problematic. Nevertheless MDF, MDF-FR and PS show reasonable agreement for CO, CO2, HCN and hydrocarbon yields between the scales. Smoke yields tended to be more variable and may be influenced by the presence of different areas of flaming and non-flaming decomposition. Keywords: Toxicity; Purser furnace; BS 7990; ISO 19700; ISO 9705; Steady-state tube furnace; Polyethylene; Polypropylene; Nylon; MDF A number of experimental procedures have been used for the investigation of combustion products [3], mostly consisting of bench-scale test methods, which have been developed with little or no attempt to replicate the decomposition conditions in full-scale fires, or to compare the product yields with those occurring in full-scale fires. Also, differences between the procedures make comparison of the results difficult and potentially confusing. If a correlation could be established between bench- and large-scale tests conditions and results, this would validate bench-scale data for use in engineering hazard calculations, and not only provide useful information for control of the hazard due to combustion products, but also decrease product development costs considerably and allow faster introduction of advanced, fire-safe materials into the market. The data obtained from large and bench scales provide the opportunity to compare significant information concerning the toxic hazard from fires. This paper compares data obtained using a standard fire toxicity assessment, the controlled equivalence ratio method steady-state tube furnace [4] and [5] with published data obtained using the ISO room corner [6] large-scale fire scenario [7], [8] and [9]. The types and quantities of toxic gases produced by a combustion process depend on a combination of factors including flammability of the materials, the chemical composition and the specific conditions of the fire scenario. The toxic hazard depends upon the fuel mass loss rate and effluent dispersal volume, and the yields of different toxic products. If toxic products are present in sufficient quantities for a sufficient time they can cause incapacitation and death. Asphyxiant gases such as carbon monoxide (CO), and hydrogen cyanide (HCN), (and their interaction with the hyperventilatory effect of carbon dioxide (CO2)) are recognised as the most immediately life-threatening products. In addition, irritant combustion products can inhibit respiration, delay or prevent escape, or contribute to post-exposure injury and death [2]. The aim of this paper is to compare the yields of main toxic products (CO, CO2, HCN and NOx) produced in small-scale tests with those produced in large-scale fires. Where available, smoke production and hydrocarbon data are also compared. This work describes experiments performed with six different building product material types (polypropylene (PP), polyethylene (PE), polyamide 6.6 (PA 6.6), polystyrene (PS), medium-density fireboard (MDF) and fire-resistant medium-density fireboard (MDF-FR) (A and B). For the tube-furnace tests PP and PE were provided in pellet form. The PP provided for ISO 9705 tests is specified in Ref. [7]. Chemically, these two materials are more or less identical and it was considered that for the purposes of this study, they are indistinguishable. PA 6.6, a widely used engineering thermoplastic-containing carbon, hydrogen, nitrogen and oxygen was provided for ISO 9705 tests as specified in Ref. [7]. For tube-furnace runs it was obtained as a commercial grade material from Northern Industrial Plastics, and supplied as 2–4 mm pellets. MDF was a BS476 Part 7 “class 2” non-flame-retarded construction material, supplied as 12 mm thick sheets in 1220 mm×2440 mm panels (density 750 kg/m3). It was tested in sheet form in the large scale, and cut into thin strips (12 mm×1.5 mm) segmented at 40 mm for the tube-furnace runs. Fire-resistant medium-density fireboard was an MDF construction material treated with a fire-retardant containing nitrogen, bromine, and phosphorus. The product was 12 mm BS476 Part7 “class 1” fire-rated medium-density fibreboard supplied as sheet material in 1220 mm×2440 mm panels (density 740 kg/m3). It was tested in sheet form in the large scale, and cut into thin strips (12 mm×1.5 mm) segmented at 40 mm for the tube-furnace runs. PS was studied using two comparable grades. In the large-scale tests, a polystyrene lighting diffuser material was used. The product was supplied as 1220 mm×2420 mm sheet (thickness of 1.5 mm). Commercial grade rigid polystyrene for tube-furnace runs was obtained from Atofina, and supplied as 2–4 mm pellets. All the materials were studied in the bench-scale steady-state tube furnace (Purser Furnace, BS 7990 and ISO TS 19700), presented in Fig. 1, equipped with controlled air supply system, an array of gas analysers, sensors and detectors plus an FTIR spectrometer to determine the evolution of toxic gases 4 (Fig. 2). Test specimens undergo steady-state combustion with respect to the degree of ventilation characterised by the equivalence ratio, φ. [10] Equivalence ratios representing different fire stages are defined by the ratio of the actual fuel/air ratio to the stoichiometric fuel/air ratio: Stoichiometric combustion describes burning a material in its nominal chemical oxygen requirement (φ=1) providing just enough oxygen for full oxidation to CO2 and H2O. In practice, for condensed phase fuels this usually results in incomplete combustion because mixing is never perfect. In well-ventilated fires, combustion is fuel lean with excess air (φ<1), whilst in vitiated fires there is less air than the stoichiometric amount (φ>1) [5] and [8]. Each fire stage has a characteristic temperature and equivalence ratio [8] and [11]. The fire can also be characterised by CO2/CO ratio as an indicator of combustion efficiency. However, this only correlates with equivalence ratio (and degree of ventilation) when there is no gas-phase inhibition e.g. by hydrogen halides. Therefore, this parameter corresponds more closely to combustion efficiency, and this is evident when studying FR materials such as the MDF-FR board. For this study, the materials were tested in either granular or pellet format (PA 6.6, PP, PE, PS) or as short segments (MDF, MDF-FR). The combustion products were diluted to a standard 50 l min−1, and sampled from the mixing and measurement chamber, where other parameters were measured. Data available from the tube furnace include heat of combustion, smoke optical density and of yields of toxic species (CO, HCN, NOx, etc.), and O2 depletion. Further data on total airborne unburned fuel content and total unburned hydrocarbons, were obtained by further oxidation of the diluted fire effluent using a secondary tube furnace, with measurement of CO, CO2 and O2. CO2 and CO were measured by NDIR, and O2 by paramagnetic analyser, NOx by chemiluminescence and HCN by spectrophotometric analysis of bubbler solutions, smoke particulates by optical density and gravimetrically using glass fibre filters. The ISO 9705 room test is a method for evaluating the fire characteristics of wall lining materials by determining the contribution to fire growth using a specified ignition source in the corner of a small room with a single open doorway [6]. The test provides data for different stages of a fire from ignition to flashover, the test being terminated when flames emerge from the door opening. The combustion products were collected in a hood above the door, into a fan-driven duct system past a measurement point for heat release rate and smoke production. The ISO 9705 test method provides data-enabling calculation of the mass rate of production of combustion products and providing there is no secondary combustion after the plume exits the doorway, these represent the products from the fire inside the room. In order to provide the large-scale data considered here the test method was modified in various ways. The main variations involved varying the ventilation conditions (by altering the doorway vent area), varying the form and location of the fuel and varying ignition source and fuel. The primary purpose of these variations was to examine the relationship between toxic combustion product yields and ventilation conditions (i.e. fire type for flaming fires). Polypropylene and nylon 6.6 were tested in an ISO 9705 room at SP, Sweden [7]. The materials (mass 55–75 kg) were tested as pool fires in square pans placed on a load cell. Different ventilation conditions were achieved by reducing the door height from 2.00 to 0.89, 0.68, 0.56 and 0.45 m, resulting in 44.5%, 34%, 28% and 22.5% of full ventilation. Various fire parameters were measured in the door opening and calorimeter duct. Equivalence ratio (φ) was measured using a “phi meter” [12], CO2, CO, and HCN by FTIR at the door opening, CO2 and CO by NDIR in the duct, and smoke particulates by an optical system in the duct. Paramagnetic analysers were used for O2, chemiluminescence for NOx and a flame ionisation detector (FID) for total unburned hydrocarbons at the opening and duct. The MDF “class 2” material was tested in an ISO 9705 room at BRE, UK [8] and [9]. Different ventilation conditions were achieved by reducing the door width from 800 to 400, 280, and 200 mm, resulting in 50%, 35% and 25% of full ventilation. To obtain large-scale yield data, fire parameters were measured in the upper layer inside the ISO room (0.4 m from ceiling and 1.0 m from the from the door end of the room). Grab samples were taken using Tedlar bags, bubblers and glass fibre filters. Data were also captured in the calorimeter duct in the standard manner. Equivalence ratio (φ) was measured by passing bag samples through a high-temperature furnace to fully oxidise the unburnt components, followed by O2 analysis. The oxygen requirement from this analysis was summed with that calculated for oxidation of the particulate sample (soot) to give the total stoichiometric oxygen demand for the airborne fuel during the sample period. This enabled a simplified calculation of φ according as follows [10]: Thus, φ can be calculated as For each grab sample, two possible φ values were calculated, a minimum φ for which all oxygen in the sample was considered to be part of the oxygen supplied and a maximum φ where it was assumed to have been entrained beyond the combustion zone. Both estimates of φ are quoted in Section 4. Unburned hydrocarbons were measured by comparison of the fully oxidised CO2 content of fire effluent with the “CO2+CO” content of the unoxidised sample. The grab samples were quantified for CO2 and CO by NDIR and O2 by paramagnetic analysis of the bag samples, HCN by spectrophotometric analysis of bubbler solutions, and smoke particulates gravimetrically using glass fibre filters. MDF (FR) and polystyrene (PS) were tested in an ISO 9705 room at BRE, UK9. The materials were tested as wall and ceiling linings with ignition using a standard propane burner at 100 kW output for 10 min followed by 300 kW output for 10 min. Product yield data were derived from the standard calorimetry measurements, making corrections for the burner output for CO2, CO and O2 measured under full-width open door conditions. The fuel mass loss data were derived from the corrected HRR measured in the ISO room calorimeter measurement divided by the heat of combustion. The heat of combustion was measured using the tube furnace under comparable combustion conditions. The product yields were calculated from the calorimeter data as the ratio of CO (or other gas) production rate to the mass loss rate. The door width was reduced to 25% or 12.5% of full-width to achieve vitiated fire conditions (MDF 100% and 25%, polystyrene =100% and 12.5%). There were limitations to this method largely due to the small amount of material burned relative to the propane from the burner, and the unknown effect of vitiation on the yields of different combustion gases from the propane. For the tube-furnace experiments, data on the relationship between product yields and equivalence ratio were obtained for all five materials. For some of the large-scale tests estimates of global equivalence ratio have been made, while for others (where measurements were made only using the calorimeter), the data necessary to calculate equivalence ratio are unavailable. However, the fire condition can also be characterised by its CO2/CO ratio, which arguably provides a better indicator of combustion efficiency, and is available for all experiments. For this reason the product yield data and comparisons have been expressed as a function of CO2/CO ratio. Fig. 3 shows equivalence ratio plotted as a function of CO2/CO ratio for the five materials covered in this study, in order to establish the relationship between equivalence ratio CO2/CO ratio were available. For the PP and nylon 6.6 pool fires in the ISO room, both opening and duct results are plotted. For the MDF wall linings in the ISO room, mean values of φ are plotted, and maximum and minimum values denoted by error bars. The tube-furnace results show that for all five materials, as equivalence ratio decreases there is a corresponding increase in CO2/CO ratio as combustion efficiency increases. At φ values below 1 the rate of increase of CO2/CO ratio becomes very rapid as the yields of CO become negligible and those of CO2 maximal, except for the fire-retarded MDF. Even at high φ values, the minimum CO2/CO ratios obtained are only as low as around 8 for PP and 5 for nylon 6.6 and PS. Somewhat lower ratios down to 3.6 occurred for MDF in the tube furnace and as low as 2 for the large-scale fires. Low CO2/CO ratios were also obtained for the MDF-FR. For the large-scale pool-fire experiments on PP and nylon 6.6, the data show a reasonably good agreement with the tube-furnace results, although there is a considerable scatter in the large-scale data, and (for individual experiments), differences between the results obtained from the room data and the duct data. For the MDF wall lining experiments, there is also a reasonable agreement, but the room-scale results show a small but consistent shift to lower CO2/CO ratios compared with the tube-furnace results. Large-scale equivalence ratio data are unavailable for PS and MDF-FR. The product yields for PP and nylon 6.6 are compared in Fig. 4 and Fig. 5, respectively. For the PP comparison, tube-furnace data are derived from PP and PE. Two parallel sets of yield data were derived from measurements in the door opening and in the duct, in the ISO 9705 experiments. These data points are plotted separately. The plots for CO and CO2 show that PP and PE give very similar results for the tube-furnace work and thus it is not unreasonable to carry out scale comparisons of PP with PE. For CO yields, there is excellent agreement between small- and large-scale data, and for CO2 there is a good agreement up to a CO2/CO ratio of around 25, but then there is a degree of scatter from the ISO room data, particularly for the CO2 yield data between the duct and room measurements. This highlights the diverse nature of uncontrolled fire conditions in the large scale and the difficulty in obtaining meaningful data. The yields of hydrocarbon and soot appear to be generally lower in the large-scale experiments, and do not show a consistent relationship with the combustion conditions. For nylon 6.6, CO and CO2 yields show good agreement between small- and large-scale data, although CO2 again shows some deviation above a CO2/CO ratio of around 25, especially for the ISO room opening data. HCN, NOx, hydrocarbon, and soot gave low yields at large scale, but the large- and small-scale data are in good agreement for similar CO2/CO ratios. For MDF, the product yields from the crib ignition of linings test, in the ISO 9705 room, are compared with tube-furnace data in Fig. 6. The ISO room testing was carried out at BRE, UK, and yield data were derived from grab samples taken from the upper layer of the fire room. For MDF there is good agreement between small- and large-scale data, for both CO and CO2. HCN, hydrocarbon, and soot yields were consistently lower than those observed for PP, PE or nylon 6.6, but nevertheless are in good agreement with small-scale data for similar CO2/CO ratios. The product yields for MDF (FR) and PS are presented in Fig. 7 and Fig. 8. The ISO room test protocol for these two materials was similar (BRE, UK) using propane burner ignition of walls and ceiling. Product yields were derived from duct data with correction for the propane burner, and heats of combustion obtained from the tube-furnace data under similar conditions. Although the combustion conditions in the tube-furnace experiments covered a range from well ventilated to vitiated (0.5<φ<1.9) (see Fig. 3), the CO2/CO ratio maximum was 7 and the minimum CO yield 0.1 g/g. For the MDF-FR, CO and CO2 yields were very similar to those obtained in ISO 9705. The smoke yields obtained using the tube furnace under flaming conditions were considerably lower than those in the ISO 9705 experiments, but under non-flaming decomposition conditions in the tube furnace the yields were similar to those at large scale. The large sets of data points confirm the correspondence between the tube-furnace and large-scale test results for the polystyrene samples (Fig. 8). A highly vitiated condition was obtained in one ISO room experiment (where CO2/CO is 2.5 and the CO yield is 0.41 g/g). The smoke yields are somewhat variable at both small- and large-scale. For this paper we have chosen CO2/CO volume ratio as the fire condition variable against which to compare product yields at small and large-scale. This was partly because equivalence ratio data were not available for all large-scale experiments, but also because it was considered that CO2/CO volume ratio provides a useful metric of the combustion efficiency, or fire condition actually achieved during a fire, as opposed to the equivalence ratio, which is defined by the input conditions for the fire plume in terms of fuel–air ratio. An important consideration when comparing the small- and large-scale results is that while the tube-furnace experiments were all carried out the same way (with the exception of some runs at temperatures higher than 650 °C), there were considerable differences between the methods used for the different large-scale test series, despite them all being based upon the ISO9705 room. The main point addressed by this paper is the extent to which the combustion conditions and product yields obtained from the small-scale apparatus are comparable to those obtained at large scale. In order to make such a comparison it is important to conduct the large-scale experiments in such a way as to provide large-scale conditions that are as well defined as possible. This is best represented by the pool-fire experiments carried out at SP, for which the fuel was confined to a set position near the floor of the room. Under these conditions there is a close agreement between the equivalence ratios and CO2/CO ratios obtained at the two scales and between the yields of CO and CO2 for both PP and nylon 6.6, and for HCN, NOx, and particulates for nylon 6.6. However, another important consideration illustrated by these results is the variability obtained at large scale and the difficulty of obtaining accurate data for equivalence ratios and product yields. For the SP experiments there is a considerable scatter in some of the data and considerable differences between the equivalence ratios and yields obtained by measurements from the calorimeter duct compared with those from the room doorway. This is especially marked for CO2 obtained under well-ventilated conditions (low φ), presumably because these data were captured at an early stage soon after the fire was ignited, where in some cases the CO2 yield is too low for the known conditions while in others it is well above the theoretically available maximum. These differences were acknowledged in the SP project report (SP report 1996, p. 45) for the PP experiments mainly as sampling difficulties in the door opening in the first phase (well ventilated) of a test before a proper outflow region was established. The position of the sampling probe was adjusted before the last PP test, as reported. Hydrocarbon and particulate yields were somewhat lower at large scale than small scale for PP and nylon 6.6, but this may be at least partly due to differences in the methods of measurement used for the two types of test. The large-scale tests carried out at BRE on MDF, PS and MDF-FR all involved wall linings. The problem with this is that during a fire, particularly a well-developed fire, the equivalence ratio and combustion conditions vary in different locations (particularly at different heights) in the room. This situation therefore cannot be readily addressed by any single tube-furnace condition. In order to compare the results with small-scale data it would be necessary to establish the equivalence ratios at different locations within the ISO room and relate them to results from a comparable set of tube-furnace runs. The results obtained for MDF show that for a given global equivalence ratio in the large-scale experiment, the yields of CO are somewhat higher than those obtained at the same equivalence ratio in the tube furnace. It is considered that this difference may at least be partly due to the complexity of the fuel/air ratios in the lined ISO room, and the possibility that fuel at high level in the room may be thermally decomposed under non-flaming conditions, which would increase the CO yield compared with the CO2 yield. For the work on MDF-FR and PE the ISO 9705 data are further complicated by the presence of the propane burner flame. Although the CO2 and CO yields from this were measured separately under well-ventilated conditions and subtracted from the test data, the signal from the propane burner was large compared with that from the test fuel and the effect of restricted ventilation on the gas yields from the propane burner are unknown. It is interesting to note that for these experiments the smoke yields obtained from the tube-furnace under flaming combustion conditions were low compared with those from the room corner test, but those obtained from the tube-furnace under non-flaming conditions were similar. It is considered that much of the wall-lining decomposition in response to the propane burner radiation was in fact non-flaming. Another finding from this work is that the tube-furnace provides an excellent method for exploring the relationship between combustion conditions and product yields. The relationships between equivalence ratio and yields of carbon oxides have been reported previously [13] and [14], but this work shows how for several fire gases (notably CO, hydrocarbons and HCN) yields tend to increase with increase in fuel–air ratio from fuel-lean (φ=0.5) to fuel-rich (φ=1.5) conditions, while NOx, which is favoured by more oxygen-rich conditions, shows a decreasing yield with ventilation in the tube furnace. The yields of toxic gases in large-scale fires are highly dependent on the decomposition conditions, in particular the fuel/air equivalence ratio, or the CO2/CO ratio, which provides a measure of combustion efficiency. When product yields in large-scale ISO9705 experiments are compared with those obtained from the tube furnace expressed as a function of φ or CO2/CO ratio, a good level of general agreement has been found for a range of products including CO, CO2, HCN, NOx, hydrocarbons and smoke particulates. It is therefore concluded that the tube-furnace model has the potential to predict toxic product yields from different fire scenarios such as those obtained from the ISO 9705 test room. The BS7990 or ISO TS 19700 steady-state tube furnace, as a bench-scale toxicity assessment, can simulate a range of fire stages or conditions, characterising the fire behaviour of materials under controlled and well-defined laboratory conditions. Studies such as these are relatively inexpensive compared with large-scale tests, taking less time and having higher reproducibility. Further, since each test run represents the burning behaviour for a particular fire stage, the results are more generally applicable than those of a single large-scale test, where the different fire stages are indistinguishable. Large-scale fire testing has played an integral role in advancing the understanding of fire behaviour, both for prediction of scaling effects, and to study the burning of larger objects, however, it also suffers limitations. The complexity of a large-scale test environment, with combustion conditions varying both temporarily and spatially, and the sensitivity to apparently minor details such as the position or mounting of the sample can make results difficult to interpret. The time required to prepare the test safely, and the very high cost required add to the burden. In addition, the large-scale test results are not always appropriate to the full-scale application or system being studied, through a lack of sufficient knowledge about the behaviour and factors leading to the resulting fire. This may result from an inappropriate set-up methodology or inadequate locations and methods for sampling the fire effluent. When analysing data from large-scale tests the reproducibility may be low and the data may not reflect the real situation. The validity of a bench-scale study of fire behaviour is dependent on how it translates to real-scale fire behaviour. In general, real-scale fires (both laboratory tests and unwanted fires) are poorly defined, and exhibit high sensitivity to a number of uncontrolled variables, giving poor reproducibility. Therefore, there is a need for a well defined a large-scale reference fire scenario for assessment of fire smoke toxicity. Its design should address the following problems: