DRAFT IMECE PRELIMINARY STUDY OF FIRE SPREAD IN CITIES AND FORESTS, USING PMMA SPECIMEN AS A FUEL IN CFD SIMULATIONS

Transcription

1 Proceedings of IMECE ASME International Mechanical Engineering Congress and Exposition November 13-19, 2009, Lake Buena Vista, Florida, USA DRAFT IMECE PRELIMINARY STUDY OF FIRE SPREAD IN CITIES AND FORESTS, USING PMMA SPECIMEN AS A FUEL IN CFD SIMULATIONS KOYU SATOH State Key Laboratory of Fire Science University of Science and Technology of China Hefei, Anhui , China kosato@fri.go.jp LIU NAIAN State Key Laboratory of Fire Science University of Science and Technology of China Hefei, Anhui , China liunai@ ustc.edu.cn LIU QIONG State Key Laboratory of Fire Science University of Science and Technology of China Hefei, Anhui , China qliu1004@ mail.ustc.edu.cn K. T. YANG Department of Aerospace and Engineering University of Notre Dame Indiana 46556, USA kyang@nd.edu ABSTRACT It is important to examine the behavior of forest fires and city fires to mitigate the damages by fires. There have been many previous studies on forest fires where the fire spreading patterns were investigated, utilizing artificial satellite pictures of forest fires, together with the use of corresponding weather data. On the other hand, large area city fires are very scarce in the world, particularly in modern cities where high-rise concrete buildings are constructed with sufficient open spaces. Thus, the examples of city fires to be referred are few and detailed investigations of city fires are limited. However, there have still been existing old cities where traditional houses built with flammable material such as wood, only separated with very small open spacing. Fires may freely spread in the city, once a big earthquake happens in those areas and then water supply for the fire brigade is damaged in the worst case along with the effect of strong wind. There are some fundamental differences between the forest fires and city fires, as the fuel may distribute either continuously or discretely. For instance, in forest fires, the dead fallen leaves, dry grasses and trees are distributed continuously on the ground, while the wooden houses in cities are discretely distributed with some separation of open spacing, such as roads and gardens. Therefore, the wooden houses neighboring the burning houses with some separation are heated by radiation and flames to elevate the temperatures, thus causing the ignition, and finally reaching a large city fire. The authors have studied the forest fire spread and are planning to start a laboratory experiment of city fire spreading. In the preliminary investigation, a numerical study is made to correlate with the laboratory experiment of city fire propagation, utilizing the three-dimensional CFD simulations. Based on the detailed experimental analysis, the authors are attempting to modify the three dimensional CFD code to predict the forest fires and city fires more precisely, taking into account the thermal heating and ignition processes. In this study, some fundamental information on the city fire propagation has been obtained, in relation to the open spacing between the houses in the cities and also the wind speed. INTRODUCTION It is important to examine the fire behavior of forest fires and city fires, such as fire spreading patterns, not only to mitigate the damages by fires but also recently to pay more attention on the global climate. There have been many previous studies [1-7] on forest fire spreading rate as supported by the corresponding dynamic satellite pictures, along with the corresponding detailed weather data. The present authors [8, 9] have also examined the forest fire spreading behavior, using a three-dimensional CFD code, along with satellite pictures of forest fires and the corresponding weather data and the GIS data.

2 (simulation result, t=9000 sec) (satellite picture) 1 km wind U=5m/s Figure 1 Wide area CFD simulation result and satellite picture of Setoda Forest Fire, August 30, 2000, Japan [8] t=248 sec t=409 sec 180 m t=511 sec t=705 sec Figure 2 Local area CFD simulation result of very flammable fuel in grassland fires at the wind speed of 4m/s [9] Figure 1 shows the CFD simulation result [8] of wide area forest fires and the corresponding satellite picture of Setoda Forest Fire, August 30, 2000, Japan, where the weather data and GIS elevation data were used and showed quite similar burning pattern. Figure 2 shows the local area simulation results [9] of very flammable fuel, like dead and dried grassland, at the wind speed of 4m/s in the area of 180 m x 180 m, where a small point heat source was put as an original fire in the area and the thermal heat transfer of radiation, convection and conduction to cause the ignition was also taking into account. On the other hand, large area city fires in the world are rare, particularly in the cities where tall concrete buildings are constructed with sufficient open spaces. Therefore, the database to be examined is insufficient and thus detailed investigations of the city fires are limited. However, there have still been existing old cities where traditional houses (maybe including historically valuable one) built with wood or some biomass, separated with only narrow spacing. In the worst case, a big earthquake attacks those areas and then water supply for the fire brigade is lost, and then fires may freely spread in the city, along with the effect of strong wind. There are some fundamental differences between the forest fires and city fires, for example, if the fuel is distributing continuously or discretely. The dead fallen leaves, grasses and trees are distributed continuously in forest fires, while the wooden houses in city fires are distributed with some separation of open spacing, such as roads and gardens in city fires. Therefore, if the city fires continue to propagate, the wooden houses adjacent to the burning houses with some distances must be heated by radiation and flames to elevate the temperatures causing the ignition, and then to propagate next to next widely. Based on the investigations of forest fire spreading [7-9], the authors are planning to start the laboratory experiment on city fire spreading, using PMMA as a fuel material, instead of wood. This study is made for the preliminary investigation of the laboratory experiment of city fire spreading, based on the three dimensional CFD analysis. Previously, such city fire spreading behaviors have been predicted mainly by two-dimensional fire code, assuming fire developing rate using empirical data, since detailed physical analysis of the fire spread between houses were insufficient. Murosaki et al. [10] concluded that the building materials, the weather conditions, and the regional characteristics could most strongly affect the fire spread. Also Gao et al. [11] studied the fire spread in old town based on cellular automaton method. Liu et al. [12] and Satoh et al. [13] studied the burn-out time of fire array combustion relevant to the city fires. The authors have attempted to utilize the three dimensional CFD code to examine the forest fires and city fires, taking into account the thermal transfer and ignition. However, the CFD simulations of city fires have not fully been verified, as yet, although there are many commercial fire software codes available in the world. This preliminary study is aiming to assist the verification of the simulation code. As the first step of the preliminary study, the FDS code by NIST [14, 15] is used. Approximately 50 cases of calculations were conducted, varying the open spacing area and wind speed conditions to examine the fire spreading pattern, together with the heating rate. NOMENCLATURE d : one length of square fuel, [m] h : height of fuel, [m] L : space distance between two fuels, [m] U : wind speed, [m/s] g: gravitational acceleration, [m/s 2 ] Fr: Froude number, U 2 /g/l

3 NUMERICAL SIMULATION METHDS As mentioned above, the objective of this study is the preliminary investigation of the city fire spreading behavior by some reduced scale model experiments in a laboratory and CFD simulations based on the software FDS4 [14, 15] by NIST. The process of the present research is as follows; (1) preliminary simulation study corresponding to the reduced scale experiments, (2) reduced scale model experiments, and (3) reflecting the preliminary studies upon the large real scale simulations, together with the reduced scale experimental study. In the experimental plan, PMMA is to be used as the fuel corresponding to the houses built in cities, instead of wood due to convenience in experiments, where the reduced scale dimension is 1/100 and a model house is a square of 0.08 m x 0.08 m and 0.06 m in height, since a real size house is assumed uniformly as 8 m x 8 m x 6m in height. Therefore, also in this simulation, the dimensions of reduced scale size model are 1/100 (case A) of the real ones. Additionally, in this study, the simulations of the real size houses (case B) are also conducted and compared with reduced scale ones, taking into account the similarity requirements. In the experiments, fuels are to be placed as 15 x 15 square array of PMMA solids, where the open spacing gap (L) between the houses is varied, which corresponds to the roads and gardens in real cities, as shown in Figure 3. In the simulations, too, similar scheme is employed. A line heater is used to heat the PMMA, adjacent to the 2nd row PMMA, since a point heater was difficult to ignite the adjacent PMMA and to cause the fire spread, differently from the very flammable case shown in Figure 2. After the heat release rate of a line heater was tested between 500 and 1000 kw/m2, most appropriate value employed was 900 kw /m2. The vaporization of the PMMA specimen heated by a line heater placed adjacent to the 2nd row causes ignition at the 2nd row PMMA and causes a flame, then radiation, convection and conduction heat will be transferred to the PMMA locating in the downwind direction. If sufficient energy is given to the downwind PMMA, the fire can propagate to the downwind. The physical constant of the PMMA employed here are as follows; Heat of vaporization, 1620 kj/ kg. Heat of combustion, kj/kg. Thermal conductivity, 0.19 [W/m/K. Specific heat, 1.47 kj/kg/k. Density, 1190 kg/m3. Ignition temperature, 380 C deg. The numerical domain was divided into uniform cubic grids, where one length of a single grid was 0.02 m for the reduced scale model (case A) and 2 m for the real size model (case B). In the fringe area enclosing the 15 x 15 fuel array, 0.3 m or 3 m width free space was added for case A or case B, respectively, as seen in Figure 3. The solid PMMA fuel is constructed by 4 x 4 x 3 grids, therefore (d=0.08 m, h=0.06 m) and (d=8 m, h=6 m) for case A and case B, respectively. In the whole domain, the number of grid in the vertical direction was fixed at 30, then 0.6 m for case A or 6 m for case B. The number of grids in the horizontal direction was varied from (90 x 90) at the spacing gap L=0 m to (150 x 150) at L=0.08 m for case A or 8 m for case B. Uniform and constant wind was given from the left boundary, as shown in Figure 3, which was varied at every 2m/s up to 8 m/s for case A and also up to 20 m/s for the real size model for case B, taking into account the Froude number, Fr=U 2 /gl. The spacing distance (L) between the fuels was varied up to 0.08 m at every 0.02m for case A and up to 8 m at every 2 m for case B. RESULTS OF NUMERICAL SIMULATIONS The numerical calculations needed very long cpu hours from the ignition of the PMMA fuel at the 2nd row heater to reach the fire spread toward the 14th right row end in the downwind direction, for example about 500 hours (about 20 days) by a personal computer with 2.7 GHz, for case A at L=0.02 m and U=4 m/s. The two cases of fire front spreading, given by the isotherms in the central vertical plane, are shown in Figure 4. These figures indicate that the fire spread speed depends on the open space gap and the wind speed. (a) L=0.02 m U=4 m/s (b)l=0.04 m U=2 m/s Width of line heater 0.02m (case A) or 2 m (case B) Figure 3 15 x 15 square arrayed PMMA house model of city fires An arrow indicating uniform wind given from the left boundary and line heater to heat up 2nd row PMMA Figure 4 Horizontal view of simulated fire front spreading over the 15 x15 PMMA array

4 Table 1 shows the simulated results of fire spreading possibilities, which means either spreading continuously or stopped intermediary during the burning or never developped, for reduced model (case A) and real size model (case B), respectively. The symbol R shows that the fire propagated to the right end, even if the fire may only partly burnt within the row, the symbol N shows that the fire did not propagate from original burning row even to the adjacent row, the symbol E shows the self-extinguished fire, gradually terminated during the propagation and never reached the right end row and the symbol C shows that the fire is still propagating, but extremely slow and it was difficult to continue the calculation so long and to judge either spreading or extinguished eventually. By these tables, the followings can be found, (1) The fuels distributing with no spacing gap causes the fire spread reaching the right row, independently on the wind speed, for both cases A and B. (2) At weak wind less than 0.5 m/s and if the spacing gap exist, it is difficult to cause the fire spread, independently on the spacing gap size, for both cases A and B. (3) At strong wind of 8m/s for case A and 20 m/s for case B, the fire spread can reach the downwind direction end particularly at the smaller spacing gap, but at the larger spacing gap it is difficult to reach the right end. (4) When the Table 1 Fire spreading possibilities (U: wind speed (m/sec), L: spacing gap (m)) (case A) L=0 L=0.02 L=0.04 L=0.06 L=0.08 U=0.5 R N N N N U=1 R C C N N U=2 R R C C C U=3 R R C C C U=4 R R R C C U=6 R R R C C U=8 R R R E E spacing gap is more than 0.08 m for case A and 8 m for case B, the fire spread is extremely slow or self-extinguished and thus the fire never reached the downwind side row end, even in the strong wind. This means that the spacing gap of more than 8 m for the real size city fires may be very effective to protect the fire spread or at least can delay the fire spread extremely, even in the strong wind. Figure 5 shows the burning rate of PMMA for case A at L=0.02 m and case B at L=2 m. It can be found from Figure 5 and Table 1 that in both cases (1) at the wind speed less than 1 m/s and the spacing gap of 0.02 m for case A and (2) at the wind speed less than 2 m/s and the spacing gap of 2 m for case B, the original fires at 2nd row never propagated to the adjacent 3 rd row PMMA, although the PMMA fuel at the 2nd row is continuously burning. The transferred energy must have been insufficient to elevate the temperature at the adjacent PMMA. Particularly, to be pointed out is that the burning rate of the PMMA fuel for case B is not always increasing depending on the wind speed, as seen in Figure 5. For case B, too, the burning rates at the wind speed of 8 and 14 m/s are far lower that that for 4 m/s. However, it is difficult to very it at present and this may be caused by some initial values or maximum values employed in the software in the simulations and probably more analysis is needed, although it is true that the flame spread over the PMMA surface is difficult in the extremely strong wind in the experiments. Case A L=0.02m Burning rate (kg/s) S2 U=8 m/s U=4m/s 0.01 U=2m/s 0.01 U=0.5m/s Time (sec) (case B) L=0 L=2 L=4 L=6 L=8 U=0.5 R N N N N U=2 R E C N N U=4 R R C E E U=6 R R C E E U=8 R R E E E U=10 R R R E E U=12 R R R E E U=14 R R R E E U=20 R R R E E Case B L=2m Burning rate (kg/s)_ L 2 U=0.5m/s U=2m/s U=4m/s U=8m/s U=14m/s Time (sec) Figure 5 Burning rate vs. time for cases A at L=0.02 m and case B at L=2 m, with variation of wind speed

5 L= 0.02 m, t= 1667 sec L=0.04m, t= 1421 sec L=0.06 m, t= 3215 sec L=0.08 m, t= 3132 sec surface is cooled down, even by 2 m/s wind. In some cases, the flame may self-extinguish over the PMMA at much stronger wind. Figure 9 shows the time-variation of radiation and temperature at each fuel center (from 2nd row to 14th row of PMMA specimen) in the center line of the wind direction (Case A, L=0.02 m at U=4 m/s). The time differences causing the rapid radiation and temperature increase between the adjacent PMMA specimens are not same and therefore the average fire spreading speed was calculated approximately in the central row area of the PMMA array, even in the cases of splitting flames as seen in Figure 6. t= 458 sec t= 704 sec Figure 6 Fire spreading profiles for Case A at U=2m/s t= 381 sec t= 929 sec Figure 8 Time variation of fire spreading profiles for Case A at L=0.02 m and U=4m/s. Radiation t= 1429 sec t= 1915 sec Radiation (kw/m2) rd 4th 6th 5th 7th 8t 3rd Time (sec) 1000 Figure 7 Time variation of fire spreading profiles for Case A at L=0.04 m and U=4m/s. Figure 6 shows the fire spreading profiles for case A at U=2m/s, with variation of the spacing gap and Figures 7 and 8 shows the time variation of fire spreading pattern for case A, at L=0.04 m and U=4m/s and at L=0.02 m and U=4m/s, respectively. In many cases, not always at strong wind speed, the ignition of the PMMA is not uniform and then only partly ignited or split into both sides, although the line heater is uniformly heating the adjacent 3rd row PMMA at constant heat release rate. This may be because that the PMMA Temperature Temperature (deg C) rd 5th 4th 6th 7th 8th Time (sec) Figure 9 Time-variation of radiation and temperature at each fuel center in the center line of the wind direction (Case A, L=0.02 m, U=4 m/s) 3rd

6 (1) U=2m/s t= 1404 sec t= 3890 sec (3) U=8m/s t= 1683 sec t= 4444 sec t= 5438 sec t= 6620 sec t= 7191 sec t= sec (2) U=4m/s 3851 sec t= 7152 sec (4) U=14m/s t= 1950 sec t= 4070 sec t= sec t= sec t= 7160 sec t= 9961 sec Figure 10 Time-variation of fire spreading profiles for Case B at L=2 m. Figure 10 shows the time variation of fire spreading profiles for case B at L=2m. At the wind speed of 2 m/s, the fire spreading is very slow and the 2nd row fuel has almost burnt out at t=3890 sec. At t=5438 sec the fire reduced very weakly, finally self-extinguished, without reaching the right end row, while at the wind speed more than 4 m/s, the fire spreading reached the right end row. As seen in Figures 6, 7 and 9, in some cases, the fire burning area may split into both

7 sides, affecting the fire spreading rate. In laboratory experiments on forest fires, too, this splitting pattern of the flames can often be observed as shown in Figure 11, which is an example obtained in a forest fire experiment using a wind tunnel (the diameter of the fan=4 m) in a laboratory. This split may be caused by the generation of some vortices due to the cross wind, which obstructs the rapid fire spread. The velocity vectors and the isotherms in the simulated results for case A, at L=0.02 m and U=2.0 m/s, also show the generation of vortices in the splitting and spreading fires, as shown in Figure 12. Case A Fire spreading speed (m/s) L=0m L=0.02m 2m 4m/s Figure 11 Fire spread pattern in laboratory wind tunnel experiment Burning area: 2 m x 3 m in the wind speed of 4 m/s. t=1113sec t=1505sec [Isotherms] [Velocities] Case B 0.08 Fire spreading speed (m/s) L=0.04m L=0.06m Wind speed (m/s) L=0m L=2m L=4m L=6m Wind speed (m/s) Figure 13 Fire spreading speed vs. wind speed, with variation of spacing gap for cases A and B. Figure 12 Simulated results of splitting and spreading fire Case A, L=0.02 m, U=2.0 m/s. Figure 13 shows the fire spreading speed vs. wind speed, with variation of spacing gap for cases A and B. From this figure, following observations can be found, which are consistent with the results that have already been mentioned above; (1) At the continuously distributing fuel, namely without the spacing gap, the fire spread is very fast and highly increasing, depending on the wind speed. (2) Even narrow open spaces can delay the fire spreading, differently from the continuous distribution of fuel. (3) For all cases if the fuel distributes with some open space, the wind effect is minor at the wind speed more than 4 m/s, although the some more simulations are needed, examining the physical properties in the software. (4) If the spacing gap distance is more than 0.08 m for case A and 8 m for case B, which is the double distance of the fuel length, the fire spreading speed is very slow or almost stopped at the original 2nd row fire. If this is verified by the reduced scale experiments and also

8 further simulations, this may provide the effective fire safety plan against the city fires in strong winds, including fires at the earthquakes. However, at this time, it is unknown if the FDS code can predict the city fire spreading behavior accurately, but the results obtained above may remain uncertain. Since this study was done for the preliminary investigation corresponding to the reduced scale experiments, more simulations and experimental investigations are needed. However, the results obtained here are quite reasonable, judging from the experimental results in forest fires made by the authors previously. CONCLUSIONS In this CFD simulation study of city fire spreading behavior, as the preliminary investigations corresponding to the laboratory experiments, the following conclusions can be made; (1) The fuel distributing with no spacing gap easily causes the fire spread, reaching the right row end, although the spreading rate differs depending on the wind speed. (2) At very weak wind less and the fuel distributing with some open spaces, it is difficult to cause the fire spread, independently on the spacing gap size, for both reduced and real size models. (3) At strong wind, the fire spread can reach the downwind direction end for the cases at the smaller spacing gaps, but difficult to reach the right end for the cases at larger spacing gaps. (4) When the spacing gap is more than 0.08 m for reduced scale model and 8 m for real size model, the fire spread is extremely slow or self-extinguished and then never reached the downwind side row end, even in the strong wind. The spacing gap of more than 8 m for the real size city fires may be effective to protect the fire spread or can delay fire spread speed extremely, even in the strong wind. If this is verified by the reduced scale experiments and also further simulations, this may provide the effective fire safety plan against the city fires in strong winds, including fires at the earthquakes. (5) Without the spacing gap, namely at the continuously distributing fuel, the fire spread is very fast, highly increasing depending on the wind speed, while even narrow open spaces can delay the fire spreading, compared with the continuous distribution of fuel. (6) For all cases if the fuel distribute with some open space, the wind effect is minor at the wind speed more than 4 m/s, although the some more simulations are needed, examining the physical properties in the software. (7) Splitting pattern of the flames is often observed in the simulations, which is similar to the experimental one due to the generation of vortices. This study was done for the preliminary investigation. At this time, it is unknown if the software FDS code used in this simulation study can predict the city fire spreading behavior accurately, therefore the results obtained above may remain uncertain and thus more experimental investigations are needed. However, the results obtained here are quite reasonable, judging from the experimental results in forest fires made by the authors previously. ACKNOWLEDGEMENTS This work was sponsored by National Key Technology R&D Program (2006BAD04B05), National Natural Science Foundation of China under Grants , and R&D Special Fund for Public Welfare Industry (forestry, ). Liu Naian was supported by the Fok Ying Tung Education Foundation. REFERENCES [1] Rothermel, R., 1972, A Mathematical Model for Predicting Fire Spread in Wild-land Fuels, Research Paper INT-115, Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station. [2] Fujioka, F.M., 1985, Estimating Wildland Fire Rate of Spread in a Spatially Non-uniform Environment, Forest Science, 31, 1, pp [3] McDonough, J. M., V. Garzon, Saito, K, 1996, "Time-Dependent Model of Forest Fire Spread in Turbulent Gusting Cross Winds", American Physical Society, Division of Fluid Dynamics Meeting, November 24-26, 1996, abstract #CG.07. [4] Fiorucci P., F. Gaetani, R. Minciardi, 2002, "Dynamic Models for Preventive Management and Real Time Control of Forest Fires", Proceedings of 15th IFAC world congress in automatic control, Barcelona, Spain. [5] Frederic, M., S Xavier and R Lucile, 2006, Fire spread experiment across Mediterranean shrub : Influence of wind on flame front properties, Fire Safety Journal, 41, 3, pp [6] Harzallah1, Y., V. Michel, Q. Liu and G.. Wainer, 2008, "Distributed Simulation and Web Map Mash-Up for Forest Fire Spread", 2008 IEEE Computer Society, Congress on Services, Part 2, pp [7] Satoh, K., L. Z. Yang, K.T. Yang, 2003, Study of Forest Fire Initiation due to Lighted Cigarette; Measurement and Observation of Flaming Probability of Dried Leaves, Proceedings of 2003 Japan Society of Mechanical Engineers, TED-AJ03-551, pp [8] Satoh, K., K. Kuwahara, K.T. Yang, 2004, A Numerical Study of Forest Fire Progression and Fire Suppression by Aerial Fire Fighting, Proceedings of 2004 ASME/HTD, Las Vegas, IMECE [9] Satoh, K., K. Kuwahara, K.T. Yang, 2004, A Numerical Study of Forest Fire Spread, (in Japanese), Proceedings of 2004 Symposium of Japanese Safety Engineering, pp [10] Murosaki, Y., K. Ohnishi and K. Matsumoto, 1984, Research on Spreading Factor of an Urban Propagating Fire, (in Japanese), Journal of the City Planning Institute of Japan, pp [11 Nan Gao, W. Weng, W. Ma, S. Ni, Q. Huang and H. Yuan, 2008, Fire Spread Model for Old Towns Based on Cellular Automaton, Tsinghua Science Technology (ISSN ), 13-5, October 2008, pp [12] Liu, N.A., Q. Liu, Z. Deng, K. Satoh, J. Zhu, 2007, Burn-out

9 Time Data Analysis on Interaction Effects among Multiple Fires in Fire Arrays, Proceedings of the Combustion Institute, 31, (2), pp [13] Satoh, K., N. Liu, J. Zhu, K.T. Yang, 2005, Experiments and Analysis of Interaction among Multiple Fires in Equidistant Fire Arrays, Proceedings of 2005 ASME/HTD, San Francisco, HT [14] McGrattan, K.B., H.R. Baum, R.G. Rehm, A. Hamins, and G.P. Forney., 2000, Fire Dynamics Simulator, Technical Reference Guide. Technical Report NISTIR 6467, National Institute of Standards and Technology, Gaithersburg, Maryland. [15] McGrattan, K.B., B.W. Klein, S. Hostikka, and J.E. Floyd, 2007, "Fire Dynamics Simulator (Version 5), User s Guide", NIST Special Publication , National Institute of Standards and Technology. [16] Johnson, E. A., K. Miyanishi, 2001, "Forest fires: Behavior and Ecological Effects", Academic Press, pp