1. Introduction
Fire spread during initial stage influence the safety of egress from fire room. If the interior of a room is lined with combustible materials, fire may spread over lining surface. As a result, the room will be untenable at early stage. Thus, it is important to predict the fire spread over lining materials during initial stage of fire.
Flame spread models have been proposed by many authors mostly to simulate the ISO 9705[1] room corner test. Wickström et al.[2] predicted the results of the room corner test using the ignition time and heat release rate history measured by cone calorimeter tests. Quintiere[3] calculated the burning area geometrically from the upward, lateral, and downward flame spread front and burnout front positions. The heat release rate was estimated by cone calorimeter data. Room temperature was calculated by algebraic equations. Karlsson[4] proposed a mathematical model focusing on the spread of flames on the ceiling. Wade et al.[5] proposed a model combining Karlsson model and a two-layer zone model. Lattimer et al.[6] proposed a method dividing the material surface into square elements and calculated the temperature rise of each element.
The upward flame spread rate was calculated in the similar ways in these models. Lateral and downward flame spread rates were not considered explicitly in the models of Wickström et al., Karlsson and Wade et al. In the Quintiere model, the rates of lateral spread from initial ignition area and downward spread from ceiling were calculated using the flame spread parameters obtained by the LIFT test. When both ceiling and wall linings are combustible, as in the room corner tests, the effect of upward flame spread is dominant. Previous works have been successful in such situations.
All the models assumed that the wall and ceiling materials are thermally thick where the semi-infinite body approximation holds during preheating period. The effect of lining thickness was not explicitly taken into account. However, if the material is thin, the back surface temperature of lining material rises quickly due to heat conduction. As a result, ignition time is affected by the thermal inertia of backing material. In addition, thin lining material is easy to burn through. Then combustion on the back side surface will occur.
The combustion properties of lining materials are obtained by bench scale measurements. Ignitability and heat release parameters can be determined by cone calorimeter tests[7]. The thermal inertia is determined from a series of tests under various heating intensities. In the tests at each heating intensity, the time-history of heat release rate is measured. The heat release rate should be properly modeled depending on heating intensity. The effect of backing material was measured and modelled as effective inertia[8,9].
The lateral flame spread rate is measured by LIFT apparatus determined by ASTM E 1321[10]. As an alternative, Merryweather et al. proposed a measurement method of flame spread rate using a cone calorimeter apparatus and compared the results to the LIFT tests[11]. The specimens were placed at an angle to the heater, and the specimens were wood materials. Similar to Merryweather’s method, Tarumoto et al. carried out flame spread measurements using a cone calorimeter test apparatus, called "reduced-size LIFT (rLIFT) test"[12]. They placed a specimen in parallel to the heater and conducted flame spread tests on flexible polyurethane foam. The authors conducted rLIFT and LIFT tests on wood materials, textile wallpaper, and resin boards, and investigated the correlation between them[13]. The effect of backing material on flame spread rate was investigated, but no explicit effect was found on flame spread rate[14].
The purpose of this study is to investigate the effects of lining thickness on burning and fire spread behavior in a room. Cone calorimeter tests and rLIFT tests were conducted to obtain combustion and flame spread properties. The cone calorimeter tests were conducted in vertical orientation simulating combustion on wall surface. Intermediate- scale room tests were conducted using plywood of varying thickness. The ceiling was made with non-combustible material and plywood were installed on the walls in order to focus on lateral and downward flame spread behavior over wall surface. To predict the burning behavior in intermediate-scale room tests, flame spread calculations were carried out using the combustion properties derived from cone calorimeter tests and rLIFT tests. The Quintiere’s model was used for fire spread calculation because it directly uses the parameters measured by bench scale tests. The calculated results were compared with the intermediate-scale room tests.
2. Specimens
The specimens were lauan plywood with thickness of 5.5, 9, 12 and 30 mm. The specimen with 30 mm thickness was assumed to be thermally thick. All the specimens for the cone calorimeter tests, the rLIFT tests and the intermediate-scale room tests were cut from the same lot of plywood for each thickness. The average of density and moisture content (by oven dry at 105 ℃) measured by cut samples taken from them are shown in Table 1. The thermal penetration time was calculated by
where l is the thickness of specimen [m], a = k/pc is the thermal diffusivity [m2/s], k is the thermal conductivity [W/(mK)], p is the density of specimen [kg/m3], is the specific heat of the specimen [kJ/(kgK)][15]. It was assumed that the thermal conductivity is 0.1 × 10-3 kW/(mK) and specific heat is 1.25 kJ/(kgK).
3. Vertical Cone Calorimeter Tests
3.1 Experimental methods
As this study aims at the flame spread over vertical wall surfaces, the specimen was heated in vertical position in cone calorimeter. The layout of the apparatus and the location of the ignitor followed the Annex E to ISO 5660-1 [7]. Measurements were repeated with various heating intensity between 50 kW/m2 and the ignition limit. Around the ignition limit, measurements were repeated until the difference between the minimum heating intensity with ignition and the maximum heating intensity without ignition was less than 1 kW/m2. The critical heat flux for ignition was determined by the average of them. In the tests, times to ignition were measured visually and heat release rate were measured by oxygen consumption method[16].
3.2 Results of measurements
The measured heat release rates are shown in Figure 1, with a heating intensity of 50 kW/m2 as examples. The heat release rate increased immediately after ignition on the specimen surface and the first peak appeared, followed by a temporal decrease with charring of surface layer. As the charring proceeded inwards and when it reached the back side surface, the combustion area and heat release rate increased causing a secondary peak. PW5.5 showed a secondary peak at about 2 minutes after ignition, PW9 at about 5 minutes, PW12 at about 6 minutes and PW30 at about 24 minutes following slight increase at around 10 minutes. At other heating intensities, the time until the secondary peak after ignition was about 2 minutes for all PW5.5, 4.5 to 7 minutes for PW9 and 5 to 10 minutes for PW12. As it takes considerable time to secondary peak in cases of lower heating intensities, time to second peak was not measured for PW30.
The relationship between heating intensity and time to ignition is shown in Figure 2. In the range of heating intensities above 25 kW/m2, the results were close for all the thickness. In the range below 20 kW/m2, the times to ignition of PW9 were relatively long, which might be due to scatter in material property. The results of other thickness were closer. The critical heat flux for ignition q ˙ ′ ′ t i g q ˙ ′ ′ q ˙ ′ ′ k p c
The relationship between heating intensity and the average heat release rate for 3 minutes after ignition is shown in Figure 4. The relatively large value of PW5.5 is due to the back surface combustion beginning within 3 minutes after ignition. In PW9, heat release rate increased as the heating intensity decreased below 25 kW/m2, which might be due to scatter of material property as has appeared in the results of times to ignition. PW12 and PW30 showed similar results. The values for △H/L were obtained from the slope of Figure 4.
The list of parameters obtained from the cone calorimeter tests is shown in Table 2. These values will be used to predict the intermediate-scale room flame spread in Section 6.
4. Reduced-Size LIFT (rLIFT) Tests
4.1 Experiemntal methods
The procedure for the rLIFT test was developed by Tarumoto et al.[12]. The authors also discussed the details such as the layout of the apparatus, distribution of heating intensity[13]. The size of the specimen was 150 × 500 mm (effective heating range: 130 × 500 mm) for the rLIFT tests. The layout of the rLIFT test apparatus is shown in Figure 5. The photograph is shown in Figure 6. The distance from the left end of heating area of the specimen is designated by x. The center axis of cone heater was located at x = 50 mm. The ignitor was a Meckel burner fueled by propane gas located at x = 10 mm. The specimen and cone heater were installed at 100 mm apart and parallel to each other. The heating intensity on the center axis of the heater was set at approximately 30 kW/m2. The distribution of the heating intensity is shown in Figure 7. The flame arrival times were visually recorded at every 25 mm on the center line of the specimen with reference to the grid drawn on the specimen and the specimen holder. The tests were conducted twice for each thickness[16,17].
4.2 Results of measurements
The flame arrival time to each observation point is shown in Figure 8. The flame arrival was slightly faster in PW9 and PW30 than in PW5.5 and PW12. The distance to the flame spread limit was longer in PW12 than in the other thicknesses.
The relationship between the effective heating intensity q ˙ ′ ′ V q ˙ ′ ′ q ˙ ′ ′ q ˙ ′ ′
5. Intermediate-Scale Room Tests
5.1 Experimental methods
The test specimen was a cubic room of approximately 1.8 m in size as shown in Figures 10 and 11. The corner of the room, the ceiling, and the hanging wall were mocked up. The opposite sides of the walls were openings. The hanging walls about half the height of the room were placed above the openings. The linings of the two walls were plywood (combustible) with calcium silicate boards backing. The ceiling, floor surface and the hangings wall were made of calcium silicate boards (non-combustible).
The fire source was a propane gas burner installed in the corner. The heat release rate of the fire source was set to 50 kW to let the intermittent flame tip reach the ceiling and the continuous flame reach about half the room height.
The total heat release rate in the room was measured by oxygen consumption method. Vertical temperature distribution in the room was measured with a thermocouple tree. Heat flux onto the wall surface was measured with water-cooled heat flux gauges. The oxygen concentration in the smoke layer was measured by a portable gas analyzer. The measurement positions of temperature, heat flux and oxygen concentration are shown in Figure 11. The flame spread behavior was recorded with video cameras. The tests were conducted once for each thickness[16,18,19].
5.2 Measured results
5.2.1 Blank test without plywood linings
In order to understand the heating conditions of the intermediate-scale room tests described in section 5.1, blank test without plywood wall lining was conducted[18]. Figure 12 shows the vertical room temperature distribution 6 minutes after ignition to gas burner, when the temperatures and heat fluxes had reached almost steady state. In the lower layer, the temperature was about 30 °C. In the upper smoke layer, the temperature ranged from 99 °C on the height of lower edge of the hanging wall to 291 °C near the ceiling. The distribution of heat flux onto the wall surface is shown in Figure 13. The heat flux ranged from 18.0 to 25.4 kW/m2 above the fire source. In the area away from the fire plume, the heat flux depended on the height above floor. Comparing the results by height, the values were higher as approaching the ceiling, ranging from 1.3 to 5.7 kW/m2 in the upper smoke layer. In the lower layer, the heat flux was around 1 kW/m2.
5.2.2 Flame spread tests over plywood linings
The photographs of flame spread tests are shown in Table 4. In case of PW5.5, the plywood surface was ignited at about 2 minutes. At 4 minutes, the flame spread horizontally along the corner between the ceiling and the top of wall. At 6 minutes, the flame size decreased, and the charred plywood dropped off around the corner. At this moment, the plywood began to burn out. Thereafter, the flame spread slowly in a lateral direction, but no more significant spread was observed. The test was stopped at 25 minutes after confirming no more flame spread.
In case of PW9, the fire grew from about 2 minutes and the flame continued to spread horizontally along the corner between ceiling and the top of wall. At 11 minutes, as shown by the circle in Table 4, isolated flames were observed in the area where the entrained air by the fire plume penetrated to the smoke layer. At 15 minutes, the isolated flames grew and gas phase burning was observed at the bottom of the smoke layer. It is considered that high-temperature unburnt gas in the smoke layer was mixed with fresh air at the bottom of the smoke layer and burned. The gas phase burning caused flame spread on the plywood surface horizontally along the bottom of smoke layer. At 17 minutes, the flame spread to the opposite end of the room. Thereafter, the flame spread downward rapidly, leading to flashover at 18 minutes 40 seconds.
In case of PW12, the fire grew from about 2 minutes and the flame continued to spread horizontally along the corner between ceiling and the top of wall. At 11 minutes, isolated flames were observed in the area where the air was entrained by the fire plume, similar to the case of PW9. The gas phase burning was also observed. At 15 minutes, the flame on the plywood surface spread along the bottom of the smoke layer to opposite end of the room. The flame spread downward rapidly leading to flashover at 17 minutes 30 seconds.
In case of PW30, the fire grew from about 2 minutes and the flame spread horizontally along the corner between ceiling and the top of wall at 6 minutes. Then the flame size decreased gradually. The flame spread slowly laterally in the smoke layer until 10 minutes. At 17 minutes, isolated flames were observed in the area where the air was entrained by the fire plume, as in the cases of PW9 and PW12. Then the gas phase burning occurred leading the flame on the plywood surface to spread along the bottom of the smoke layer to opposite end of the room at 20 minutes. The flame spread downward rapidly leading to flashover at 23 minutes.
The photographs of back side surfaces of the plywood after the tests are shown in Figure 14. In case of PW5.5, plywood was burned out around the corner and in the flame spread area. In case of PW9, plywood was largely burned out or charred around the corner and in the smoke layer. The burnt-out area of PW12 was slightly larger than that of PW5.5 and smaller than that of PW9. The back side surface of PW12 in the smoke layer was widely discolored. In case of PW30, plywood did not burn through including the corner area.
The measured room temperatures are shown in Figure 15. The smoke layer temperatures were the average of measuring points No.1-5 shown in Figure 11. The lower layer temperatures were the average of measuring points No.6-9. The following discussion focuses on the smoke layer temperatures. All the specimens were ignited on the plywood surface at 1-2 minutes and the temperature rose to around 330 °C. In case of PW5.5, the temperature rose further at 4 minutes. It is because the combustion on the back side surface temporarily intensified the burning. The temperature reached a peak value at about 420 °C as it began to burn out at 6 minutes, and then dropped gradually approaching to the temperature of the blank test. In case of PW9, the temperature rose at 6 minutes, reached about 400°C, which was considered to be the effect of back surface combustion. After that, the temperature stagnated again at 7 minutes, which was considered to be the effect of burnout. The temperature rose again from 10 minutes, it was accelerated by growing flame shown in Table 4 from 15 minutes and led to flashover. In case of PW12, the temperature also rose from 7 minutes. The rise was slower than in PW9, but it did not stagnate as in PW9, and exceeded the temperature in PW9 at 8 minutes. The temperature continued to rise with acceleration and reached flashover. In case of PW30, the temperature gradually dropped from 7 minutes. The temperature began to rise again from 13 minutes, it gradually accelerated and reached flashover.
The heat release rates from the fire source and plywood are shown in Figure 16. The shapes of heat release rate curves are similar to those of the smoke layer temperatures shown in Figure 15. For PW5.5, the heat release rate reached about 150 kW at 6 minutes and then slowly decreased. In PW9, the heat release rate increased from 6 minutes, reached about 120 kW at 7 minutes, then stagnated. The heat release rate increased again at 15 minutes and flashover occurred. In PW12, the heat release rate began to increase at 7 minutes, slightly later than in PW9, and exceeded the heat release rate in PW9 at 8 minutes 30 seconds. the heat release rate continued to increase and flashover occurred. The heat release rate of PW30 decreased at 5 minutes and was close to that of the blank test results between 10 and 13 minutes. The heat release rate increased again from 13 minutes and flashover occurred.
The oxygen concentration in the smoke layer is shown in Figure 17. After ignition, the oxygen concentration decreased to around 10% in all the tests. Thereafter, the oxygen concentration increased slightly to about 12-13%. This suggests that the combustible gas was difficult to burn in the smoke layer, thus the gas phase burning occurred at the bottom of the smoke layer, as shown in Table 4. In case of PW12, the oxygen concentration decreased rapidly with rapid flame spread after 14 minutes and reached almost 0% at flashover. In case of PW9, the decrease of oxygen concentration with flame spread was slower than that in the case of PW12. This was due to burnout at the initially ignited part. In case of PW30, the oxygen concentration began to decrease with flame spread at 14 minutes. The decrease was slower than that of PW12.
5.2.3 Summary of intermediate-scale room test results
In summary, after the plywood surface was ignited, burning became intense when the back surface burned and slowed down when the plywood began to burn out. Both the combustion on the back side surface and burnout occurred quickly in case of thin plywood. In case of PW5.5, the effect of burnout was dominant, thus the flashover did not occur. In cases of plywood thicker than 9 mm, flame spread and back side burning was dominant, which results in flashover. The time to flashover was shorter when the back surface burned such as PW9 or PW12. Among the tests, the flame spread was fastest in PW12.
6. Prediction of Flame Spread
6.1 Methods for flame spread prediction
To predict flame spread, the Quintiere’s model[3] was applied after partial modifications[20]. Figure 18 shows a schematic of the method. The difference from the original model is that the heat loss from fire room temperature is considered to occur only in the upper part of the room.
The upward flame spread rate is given by Eq. (2). The lateral flame spread rate is given by Eq. (3). The downward flame spread rate is given in the same way as the lateral flame spread rate.
where, Ts [K] is the lining surface temperature and Ts,min [K] is the minimum temperature for flame spread. The flame spread front position yf [m] is given by Eq. (4).
where, Q ˙ ′ Q ˙ ′ ′
where, Q ′ ′
The net heat flux q ˙ ′ ′ q ˙ ′ ′ q ˙ ′ ′
Using the latent heat of pyrolysis [kJ/kg] of the material, the pyrolysis rate of the material is described by q ˙ ′ ′ q ˙ ′ ′ Q ˙ ′ ′ Q ˙ ′ ′ Q ˙ ′ ′ q ˙ ′ ′
The average heat release rate during the first 3 minutes after ignition was used to derive the ratio of heat of combustion to heat of gasification, △H/L. As for other properties such as kpc, Tig, Φ were derived from the cone calorimeter and rLIFT tests.
6.2 Prediction results
Fire spread area, heat release rate, and room temperature were calculated using the equations described in previous section. The conditions such as room size and fire source were the same as the intermediate-scale room tests. The thickness of the plywood was 5.5, 9, 12, and 30 mm. The properties were given by the averages for each thickness based on the results in Table 2 and Table 3 except the total heat release per unit area of plywood Q ′ ′
The calculated results are shown in Figure 19 to Figure 21. As shown in Figure 19, the burning of 5.5 mm plywood spread only upward and downstream under the ceiling. Burnt out area expanded sequentially from the initially ignited area at about 7 minutes. In the cases of 9, 12 and 30 mm thickness, lateral and downward flame spread started at about 9 minutes. Accompanying this, the heat release rate and the room temperature increased rapidly shown in Figure 20 and Figure 21. In case of 9 mm, burnout started at 11 minutes, which corresponds with temporal sharp decrease of the heat release rate and temperature. The results for 12 and 30 mm thick plywood were exactly the same.
The calculation results and the experimental results agreed in view of onset of flashover. The plywood of 5.5 mm thick burned out quickly without flashover, while the plywood of 9 mm or thicker, the fire grew to flashover. However, the calculated times to flashover in 9, 12, 30 mm were shorter than the experiments. Regarding the ordinary order of flashover time, the experimental results were shorter in intermediate thicknesses, whereas the calculated results were shorter for thicker thicknesses and the same for thicknesses above 12 mm. To improve the quantitative accuracy, it would be important to consider the time variation of heat release rate per unit area with lining thickness including the effect of burning on the back surface. The effect of the gas phase burning at the bottom of the smoke layer would be important as well.
7. Conclusions
Small-scale tests (cone calorimeter and reduced-size LIFT tests) and intermediate-scale room tests were conducted using plywood of 5.5, 9, 12, 30 mm thickness as specimens. The properties obtained from the small-scale tests were input into a modified Quintiere’s model to calculate the flame spread under the same conditions as the intermediate-scale room tests. The results of this study are summarized below.
Properties for ignition and heat release were obtained from the cone calorimeter tests, and properties for flame spread were obtained from the reduced-size LIFT (rLIFT) tests. These properties can be used to calculate the fire spread in the room.
Based on the intermediate scale room tests, the effect of lining thickness was discussed. When the plywood burned through to the back surface, the heat release became intense. The heat release decreased according to burnout. Both the combustion on the back surface and the burnout occurred quickly in case of thin plywood. In case of 5.5 mm thick lining, burnout occurred quickly without flashover. In case of 9 mm or thicker specimens, the fire grew to flashover. The flame spread over the plywood of 9 and 12 mm thick was faster than that of 30 mm thick plywood. In case of 9 mm and 12 mm lining, the effect of back side burning was significant. In case of 30 mm lining, back side surface did not burn. The time to flashover of 12 mm thick plywood was the shortest. The smoke layer reached high temperatures and the oxygen concentration was reduced to approximately 10% by volume. Thus, it was not easy for volatile gas to burn in the smoke layer due to lack of oxygen. As a result, the gas phase burning occurred at the bottom of the smoke layer where volatile gas and fresh air were mixed. Soon after that, the downward flame spread occurred and flashover occurred.
The fire spread calculation reproduced the intermediate-scale room tests qualitatively. Thin plywood burned out quickly without flashover. The burning of thicker plywood grew to flashover. However, calculated times to flashover were shorter than the test results. In addition, 12 mm thick plywood led to flashover fastest in the tests, but the calculation did not reproduce this tendency. To improve the accuracy, consideration on time variation of heat release rate per unit area including the effect of burning on the back surface may be necessary.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
