Proceedings of the Combustion Institute, Volume 29, 2002/pp. 289 296 GLOWING AND FLAMING AUTOIGNITION OF WOOD N. BOONMEE and J. G. QUINTIERE Department of Fire Protection Engineering University of Maryland College Park, MD 20742, USA A study of the autoignition of wood by a radiant cone heater was conducted. Insulated redwood samples were exposed vertically to incident heat flux ranging from 10 to 70 kw/m 2. IR thermography and normal video recording were used to view the sample surface. The surface temperature and mass loss were continuously recorded. Glowing and flaming autoignition were defined and examined. The times to glowing and flaming autoignition were measured and compared with the times to flaming piloted ignition. The study found that for incident heat fluxes less than 40 kw/m 2, in some cases, the sample surface started to glow (glowing ignition) before a visible flame (flaming ignition) was eventually seen. However, for incident heat fluxes greater than 40 kw/m 2, flaming ignition occurred very quickly (within 30 s). The measured ignition time, ignition temperature, and surface temperature history were compared with theoretical values. The mass flux (pyrolysis rate) was assumed to follow an Arrhenius reaction rate. The activation energy and the pre-exponential factor were determined from a best curve fit of the experimental data. Introduction The ultimate goal of fire research is to reduce fire hazards. Since ignition is the initiation of fire, it is very important to understand the ignition process. For ignition to occur, a number of physical and chemical events must take place. The material must first be heated by means of radiation, convection, or conduction. The rate of temperature rise depends on the thermal properties of the material. Once the material reaches its pyrolysis temperature, it starts to decompose and produce a pyrolysis gas. The pyrolysis gas then travels away from the surface and mixes with the surrounding fresh air to create a combustible mixture. When the concentration of combustible mixture is suitable (i.e., above the lower flammable limit) and the temperature is high enough, ignition occurs. The ignition of combustible gases can be initiated in two fashions: (1) piloted ignition, in which the ignition is initiated from a local energy source; and (2) autoignition, in which the ignition is initiated without an external heat source. This investigation focuses on the autoignition of wood for both glowing and flaming ignition modes. Description of Experiment Redwood cubic samples were 40 by 40 mm of exposed surface area with 40 mm thickness. The wood grain orientation was aligned either parallel (heating parallel to grain) or perpendicular (heating perpendicular to grain) to the incident heat flux. The samples were insulated on the back and side to promote one-dimensional heat transfer and mounted vertically in the sample holder as shown in Fig. 1. Ref. [1] provides additional details regarding the experimental setup. Before the experiment started, a shutter was placed in front of a sample. To begin the test, the shutter was taken away manually, providing a uniform exposed incident heat flux on the sample surface; then the data acquisition system began to record the data. At high incident heat flux (q 40 kw/m 2 i ), the test was terminated when a sustained visible flame occurred; however, at low incident heat flux (q i 40 kw/m 2 ), the sample might not show a visible flame. In this case, the experiment was continued for up to 40 min until the sample was consumed or flaming ignition was eventually seen. A sample mass was continuously recorded by a load cell, and the surface temperature was continuously recorded by an IR camera. The IR camera was calibrated with Fig. 1. Experiment apparatus setup. 289
290 FIRE Ignition and Flame Spread thermocouples in order to correct for the spectral properties of the camera. Mass Loss and Mass Flux (Pyrolysis Rate) Fig. 2. A typical mass loss and mass flux time history (incident heat flux 40 kw/m 2, heating parallel to the grain): (a) mass flux increased like non-charring material, (b) mass flux decreased due to char blocking the pyrolysis products, (c) back effect, (d) mass flux dropped due to depletion of the sample. A typical mass loss and mass flux (pyrolysis rate) time history is illustrated in Fig. 2. At first, when a thin char layer forms on the sample surface, the mass flux increases with time until it reaches a maximum value. Once the char layer becomes thicker, it blocks the release of pyrolysis products, resulting in a decrease of mass flux. The mass flux decreases and remains constant (approximately 8 g/[m 2 s]) for a period of time. When the thermal wave reaches the back of the sample, the flux rate increases again, which is shown in the plot at 700 s. Finally, the entire sample becomes charred and the mass flux decreases again due to depletion of the sample. The sample was almost fully consumed after 40 min. Ignition Time Fig. 3. Surface temperature history (incident heat flux 40 kw/m 2, heating perpendicular to the grain): (a) the sample started glowing, time 22 s, T S 422 C; (b) the sample started flaming, time 1204 s, T S 754 C. Fig. 4. Time to flaming and glowing ignition versus incident heat flux. Time to flaming autoignition: (a) parallel, (b) perpendicular; time to glowing autoignition: (c) parallel, (d) perpendicular; time to flaming piloted ignition: (e) parallel, (f) perpendicular. (The arrows indicate transition from glowing to flaming autoignition). Two definitions of ignition time, (1) time to flaming and (2) time to glowing for autoignition, are defined. Time to flaming is the duration of time after the sample is first exposed to incident heat flux until it ignites to produce a visible flame. We define time to glowing as the time when noticeable surface oxidation occurs on the sample surface. This can be indicated when the surface temperature of the sample is greater than the corresponding surface temperature of the inert insulator (Fig. 3). The glowing definition is based on the assumption that, in general, a material with lower thermal inertia (kqc) will have a higher surface temperature than a sample with higher thermal inertia, when exposed to the same heat flux. However, a high thermal inertia material that undergoes a chemical reaction (surface oxidation) can have a surface temperature greater than an inert material of low thermal inertia. The thermal inertia of the sample (redwood) is greater than the inert insulator [1]; therefore, we might expect the surface temperature of the sample to be lower than the insulator, if the sample is inert. However, Fig. 3 shows that the surface temperature of the sample became greater than the corresponding insulator at 22 s after it was exposed to the incident heat flux. Consequently, we can see that the greater surface temperature of the sample resulting from a surface oxidation and 22 s is the time to glowing. The time to glowing and flaming autoignition was measured and compared with the time to flaming piloted ignition [2] as shown in Fig. 4. At high incident heat flux (q 40 kw/m 2 i ), the flaming autoignition time merges with the flaming piloted ignition time. The sample ignited and a visible flame could be readily noticed (within 30 s) after exposure to
GLOWING AND FLAMING AUTOIGNITION OF WOOD 291 Surface Temperature ( C) Surface Temperature ( C) (a) (b) 700 600 500 400 300 200 100 0 900 800 700 600 500 400 300 200 100 0 Experiment data Eq. (1) The sample flaming ignites. 0 10 20 30 40 50 60 70 Time (s) The sample flaming ignites at 1250s. The sample starts glowing at 23s. Experiment data Eq. (1) 0 500 1000 1500 2000 Time (s) Fig. 5. Measured and theoretical surface temperature history. (a) Heating perpendicular to grain, 50 kw/m 2 ; (b) heating perpendicular to grain, 40 kw/m 2. Fig. 6. Plot of ignition surface temperature against incident heat flux. Flaming autoignition: (a) parallel, (b) perpendicular; flaming piloted ignition: (c) parallel), (d) perpendicular; glowing autoignition: (e) parallel, (f) perpendicular. the incident heat flux. For low incident heat flux (q 40 kw/m 2 i ), the flaming autoignition time prominently diverges from the piloted ignition time. The longer flaming ignition time for autoignition is a result of the absence of an external heat source to promote the ignition process, while for flaming piloted ignition, a small pilot flame or an electrical arc can initiate the process. However, the time to glowing autoignition with low incident heat flux still follows the trend of the time to flaming autoignition for the high heat flux and time to flaming piloted ignition. The study found that for high incident heat flux (q 40 kw/m 2 i ), it was hard to distinguish between time to flaming and time to glowing autoignition. The sample started glowing, and then flaming ignition occurred almost immediately. For low incident heat flux (q 40 kw/m 2 i ), the sample surface was first glowing, and in some cases the glowing surface transitioned to flaming ignition. The transition from glowing to flaming autoignition is indicated by the arrows illustrated in Fig. 4. Surface and Ignition Temperature A typical surface temperature time history is shown in Fig. 5. At high incident heat flux (Fig. 5a, q 40 kw/m 2 i ), the sample ignited almost immediately after it was exposed. The surface temperature rose rapidly until flaming ignition occurred, which was indicated by a jump of the surface temperature. From the temperature time history and observation, the sample did not flash before ignition like it sometimes does in piloted ignition [3,4,9]. At low incident heat flux (Fig. 5b, q i 40 kw/ m 2 ), the surface temperature gradually increased and reached a constant value of approximately 700 C. The experiments found that, in some cases for low incident heat flux, the sample was glowing for up to 20 min before a visible flame could be noticed. It should be noted that the surface temperature plotted in the graph was the average surface temperature over the exposed sample area. Figure 6 shows the plot of the surface temperature at ignition as a function of incident heat flux. The 10% error bar represents the error due to the calibration factor of the IR camera. It is obvious that the autoignition temperature varies with incident heat flux. The autoignition surface temperature increases when the incident heat flux decreases. An increase of ignition surface temperature is due to the formation of char on the sample surface when the incident heat flux is low. Fig. 6 also shows the effect of grain orientation on ignition surface temperature. A higher ignition surface temperature occurs when heating is perpendicular to the grain than when heating is parallel to the grain. This is a result of the dependency of thermal conductivity on grain orientation. The thermal conductivity of wood is greater when the heat flows parallel to grain than when it flows perpendicular to the grain [10]. As a result, more energy would accumulate at the surface when heating is perpendicular to the grain before ignition. The temperatures at which the sample surfaces overtake the insulator surfaces (glowing temperature) are also plotted in Fig. 6. The glowing temperature ranges from 383 up to 487 C, depending
292 FIRE Ignition and Flame Spread TABLE 1 Summary of input parameters for the integral model Apparent Thermal Inertia, kqc (kj 2 m 4 K 2 s 1 ) Critical Heat Flux, q cr (kw/m 2 ) Average Glowing/Flaming Ignition Temperature, T ig ( C) Ignition Mode Parallel Perpendicular Parallel Perpendicular Parallel Perpendicular Glowing autoignition 0.16 0.13 11 11 453 424 Flaming autoignition 0.068 0.076 37 35 578 565 Flaming piloted ignition [2] 0.25 a 0.17 a 3 12 204 375 a Calculated from best fit of the piloted ignition data [2]. on the incident heat flux. However, the surface temperature of the sample when surface oxidation occurred could be as high as 750 C. The plot shows an increasing value of ignition temperature as one goes from piloted ignition to glowing and flaming autoignition. Nevertheless, if we consider the trend lines of the flaming autoignition temperature, as incident heat flux increases, the temperature decreases and seems to approach the same average temperature for flaming piloted ignition. These trends imply that if an incident heat flux is large enough, the flaming autoignition temperature will be the same as the flaming piloted ignition temperature. Integral Model A one-dimensional integral model for ignition including reaction effects described by Spearpoint and Quintiere [2] can be expressed as where 4 kqc TS T0 t (1) 3 (1 b)(2 b) q i b 4 4 h(t S T 0) r(t S T 0) q i At the onset of ignition, the surface temperature is equal to ignition temperature (T ig ). Substituting T S T ig into equation 1 and rearranging for 1/ t ig, the ignition time (t ig ) can be obtained in the form of 1/2 1 ig 0 3 (1 b ig)(2 b ig) q i 1 4 kqc T T t ig (2) where b ig 4 4 h(t ig T 0) r(t ig T 0) q i In order to evaluate the model, a number of input 2 parameters need to be determined. These can be obtained from the experiment. The critical heat flux and apparent thermal inertia are derived from the time to glowing and flaming ignition data. The average glowing and flaming ignition temperature can be estimated from an energy balance at the sample surface. As the incident heat flux is very low (q i r q cr), b ig r 1 and ignition time becomes very large (t ig r ). The energy balance at the surface is then written as 4 4 q h(t T ) r(t T ) (3) cr ig 0 ig 0 Solving equation 3 with h 13.5 W/m 2, the average glowing and flaming ignition temperature is obtained. A summary of input parameters for the integral model is shown in Table 1. A comparison of the theoretical surface temperature (equation 1) based on flaming autoignition parameters with the corresponding data is shown in Fig. 5. For the high incident heat flux cases (q i 40 kw/m 2 ), Fig. 5a shows that the integral model overpredicts the surface temperature. This may be due to the variation of thermal inertia with temperature and the evaporation of absorbed water. It could also be due to the transient effects of an incident heat flux on the sample surface between the time when the shutter was taken away and a steadystate heat flux was reached. For the low incident heat flux cases (q 40 kw/m 2 i ) (Fig. 5b), the char layer appears to result in an underprediction of the surface temperature. The measured surface temperature may also be higher due to char oxidation. The plots of theoretical glowing and flaming autoignition and flaming piloted ignition time (equation 2) compared to the experimental data are shown in Fig. 7 and Fig. 8. For flaming autoignition, good agreement between the model and the data is obtained at high incident heat flux. However, at low incident heat flux, disagreement is obtained on some level. The reason may be that the model does not account for the forming of char, which is significant when the incident heat flux is low. For the glowing
GLOWING AND FLAMING AUTOIGNITION OF WOOD 293 0.60 1-1/ 0.50 ( s 2 ) t ig 0.40 Time to Flaming Auto-Ignition Time to Glowing Auto-Ignition Time to Flaming Piloted Ignition (a) (b) (c) 0.30 0.20 0.10 0.00 0 10 20 30 40 50 60 70 80 90 Incident Heat Flux (kw/m 2 ) ig Fig. 7. Plot of 1/ t against incident heat flux (heating parallel to grain). Equation 2: (a) time to flaming autoignition, (b) time to glowing autoignition, (c) time to flaming piloted ignition [2]. (The arrows indicate transition from glowing to flaming autoignition). 0.60 1-1/ 0.50 ( s 2 ) t ig 0.40 0.30 0.20 0.10 0.00 Time to Flaming Auto-Ignition Time to Glowing Auto-Ignition Time to Flaming Piloted Ignition 0 10 20 30 40 50 60 70 80 90 ig Incident Heat Flux (kw/m 2 ) Fig. 8. Plot of 1/ t against incident heat flux (heating perpendicular to grain). Equation 2: (a) time to flaming autoignition, (b) time to glowing autoignition, (c) time to flaming piloted ignition [2]. (The arrows indicate transition from glowing to flaming autoignition). cases, the plots show that the time to glowing for autoignition follows the time to flaming for piloted ignition. These results may imply that if little char does not form before ignition, the flaming autoignition time (in the form of 1/ t ig ) would follow the flaming piloted ignition time at low incident heat flux. Arrhenius Reaction Rate By assuming that a mass flux (pyrolysis rate) of the sample follows a one-step Arrhenius reaction rate, we can write the mass flux as E/RT ṁ Ae s (4) Applying the natural log to equation 4, we obtain E ln ṁ ln A (5) RT S The plots of ln ṁ against 1/T S are shown in Fig. 9 for heating parallel to the grain and Fig. 10 for heating perpendicular to the grain. By calculating the slopes with a linear regression fit to the data, we obtain the activation energy E range of 7 19 kj/mol (b) (a) Fig. 9. Plot of ln ṁ against 1/T S (heating parallel to the grain). with the average value of 13 kj/mol for heating parallel to the grain and 6 28 kj/mol with the average value of 23 kj/mol for heating perpendicular to the grain. The calculated activation energy from this study is comparable to the values from the study by Fredlund on pine, 23.6 kj/mol, and on spruce, 26.3 kj/mol [5]; however, these calculated activation energies are less than the value from Akita s work quoted by Simms [6], which is 60 kj/mol, and by Parker [7], which is 121 kj/mol. The disagreement may result from the fact that this estimated activation energy is based on the mass loss rate and surface temperature measurement. The degradation in depth occurs over a range of temperatures. Therefore, the result reflects an average over this range. Fredlund s value was determined in an identical way. However, Akita s was based on convective heating and ignition time. For this reason, the estimation would be different. For Parker s, it was measured based on the local value. From the 1/T S -axis intercep, we can also calculate the average pre-exponential A as approximately 134 g/(m 2 s) for heating parallel to the grain and 412 g/(m 2 s) for heating perpendicular to the grain. Types of Ignition Based on the experimental results, the autoignition of wood can be categorized into two modes depending on the intensity of incident heat flux: (1) flaming autoignition and (2) glowing autoignition. Flaming autoignition occurred when a sample was exposed to a high incident heat flux. The pyrolysis product emanated from the sample to mix with fresh air and produced a boundary layer of a combustible mixture. Initially, the boundary was laminar; however, it became turbulent when it traveled beyond the sample surface. Once the combustible mixture reached a lower flammable limit and the temperature conditions were sufficient, flaming autoignition occurred. The flame first appeared in the gas phase above the sample surface and then propagated back
294 FIRE Ignition and Flame Spread of glowing autoignition transition to flaming autoignition, the flame first appeared near the glowing surface (Fig. 12). This observation appeared to confirm that the flaming ignition of combustible mixture near the surface occurred due to additional energy from oxidation on the sample surface. Conclusions Fig. 10. Plot of ln ṁ against 1/T S (heating perpendicular to the grain). to the sample surface (Fig. 11). This ignition behavior is consistent with the observations of Simms [8]. Glowing autoignition occurred when the incident heat flux was low. Surface oxidation occurred, which could be indicated by glowing spots on the sample surface. Since the external heat flux supplying energy to the combustible mixture was low, flaming ignition may not have occurred unless sufficient energy was reached. If the additional energy produced by oxidation of the sample surface could bring the combustible mixture temperature near the sample surface to the ignition temperature, then flaming autoignition would occur. We found that, in the case Autoignition of wood has been studied, and two definitions of ignition have been given. Flaming ignition occurs when the incident heat flux is high. A visible flame first appears in the gas phase above the sample surface and then propagates back to the surface. Glowing ignition occurs when the incident heat flux is low. Surface oxidation occurs, which can be indicated from the glowing surface. When the glowing surface develops to produce a visible flame, the flame first appears relatively close to the glowing surface. The surface temperature history calculated from the integral model overpredicts the surface temperature at high heat flux before the char layer forms. When a char layer has formed under long low heat flux, the integral model underpredicts the surface temperature. The integral model can be used to predict time to flaming and time to glowing for autoignition if we Fig. 11. Flaming ignition process (side view, heating parallel to the grain: 68 kw/m 2 ). Fig. 12. Glowing ignition process (side view, heating perpendicular to the grain: 40 kw/m 2 ).
GLOWING AND FLAMING AUTOIGNITION OF WOOD 295 use appropriate values of thermal inertia, critical heat flux, and glowing/flaming ignition temperature. The overall mass loss rate (pyrolysis rate) of the wood sample can be related to the surface temperature as a one-step Arrhenius pyrolysis rate. ig s ignition surface REFERENCES Nomenclature absorptivity ( ) b dimensionless heat flux parameter ( ) c specific heat (J/kg K) h average convective heat transfer coefficient (W/m 2 K) k thermal conductivity (W/m K) m mass (g, kg) ṁ mass flux (g/m 2 s) q heat flux (kw/m 2 ) q density (kg/m 3 ) T temperature ( C, K) t time (s) r Stefan-Boltmann constant (W/m 2 K) A pre-exponential (g/m 2 s) E activation energy (kj/kmol) R universal gas constant 8.314 kj/(kmol K) Subscripts O i initial condition, ambient incident 1. Boonmee, N., Radiant Auto-Ignition of Wood, Master s thesis, University of Maryland at College Park, 2001. 2. Spearpoint, M. J., and Quintiere, J. G., Fire Safety J. 36(4):391 415 (2001). 3. Atreya, A., and Wichman, I. S., Trans. ASME J. Heat Transfer 111:719 725 (1989). 4. Atreya, A., Carpentier, C., and Harkleroad, M., Effect of Sample Orientation on Piloted Ignition and Flame Spread, in Fire Safety Science: Proceedings of the First International Symposium, pp. 97 109. 5. Fredlund, B., A Model for Heat and Mass Transfer in Timber Structures During Fire: A Theoretical, Numerical and Experimental Study, Report LUTVDG/ (TVBB-1003), Institute of Science and Technology, Department of Fire Safety Engineering, Lund University, Sweden, 1988. 6. Simms, D. L., Combust. Flame 7:253 261 (1963). 7. Parker, W. J., Prediction of the Heat Release Rate of Wood, in Proceedings of the First International Symposium, 1986, pp. 207 216. 8. Simms, D. L., Combust. Flame 4:293 300 (1960). 9. Tzeng, L. S., Atreya, A., and Wichman, I. S., Combust. Flame 80:94 106 (1990). 10. Vyas, R. J., and Welker, J. R., Fire Flammability 6:355 361 (1975). COMMENTS Toshisuke Hirano, National Research Institute of Fire and Disaster, Japan. Along the heated surface, a gas flow induced by buoyancy was observed. This means that the condition at the bottom edge of the test piece is different from that at the top edge. Did you realize such a phenomenon? Did you perform flow field measurements? If so, what were the results? Author s Reply. We did not perform a flow field study. However, we did observe the boundary layer of the combustible mixture growth as a sample was heated. Initially the boundary layer was laminar; however, it became turbulent when it traveled beyond the sample surface. The figure below shows a vertical surface temperature distribution along the centerline of the sample for various times taken at 40.2 kw/m 2. Slightly higher temperatures are observed at the top and bottom. Possibly oxygen from the surrounding is easier to cause oxidation with the wood surface at the edges. The surface temperature profile becomes more uniform as time advances showing radiation dominates. 45 Top 40 35 30 25 20 15 10 5 0 Position (mm) 0 200 400 600 800 Temperature ( o C) 0 sec 2 sec 10 sec 20 sec 100 sec 150 sec The heat transfer coefficient (h) varies with distance from the bottom to the top of the sample as the boundary layer grows. However, we used an average value of heat transfer coefficient over the surface in our analysis. The average heat transfer coefficient (h) based on constant surface temperature for a vertical flat plate for laminar flow used in this calculation can be expressed as N u (hl/k) 0.59Ra 1/4. For low radiant heating, we expect to see some temperature variations due to convection. In summary, radiant heating, convective cooling and surface oxidation all play a role.
296 FIRE Ignition and Flame Spread Ignition mass flux (g/s-m 2 ) 25 20 15 10 5 Heated Parallel to Grain Trend: Heated parallel to grain y = 0.1342e 0.006x Trend: Heated perpendicular to grain y = 0.0011e 0.012x Heated Perpendicular to Grain High incident heat flux Low incident heat flux Flaming ignition Glowing ignition 0 0 200 400 600 800 1000 1200 Ignition temperature (K) F. A. Williams, UCSD, USA. It is good to see data on differences between heating parallel and perpendicular to the grain. Since this is an obvious effect to investigate, I wonder if there may be some such data in the literature with which you can compare your data. I also wonder if flow of pyrolysis products parallel to the grain may be better to explain your observed differences than different thermal conductivities. Author s Reply. The plot of mass flux at ignition against ignition surface temperature for various incident heat fluxes shows that the ignition mass flux is greater when heating parallel rather than heating perpendicular to the grain. In order for autoignition to occur, the fuel mass concentration (Y F ) and temperature (T ) must both achieve critical values. More mass flux is required at ignition when heating perpendicular than when heating parallel to the grain for the same incident heat flux since the volatiles move more easily along the xylem tubes as would water from the roots. The mass flux varies proportionally to the surface temperature. For a critical fuel concentration at the surface, there is a critical mass flux. From the figure, this gives a higher ignition temperature for perpendicular heating than parallel heating. Another possible explanation for the higher ignition temperature for heating perpendicular to the grain than parallel is a difference in thermal conductivity. The density and heat capacity of wood does not depend on grain configuration. The experimental data show that the autoignition time is of the same order of magnitude for each heating configuration. By equation 2, the ignition temperature varies inversely proportional to the thermal conductivity. The thermal conductivity is lower for heating perpendicular rather than parallel to the grain. Therefore, the ignition surface temperature is higher when heating perpendicular to the grain. Only more complete modeling can sort out these effects.