- Email: [email protected]
An Earth-based equivalent low stretch apparatus for material flammability assessment in microgravity and extraterrestrial environments
Proceedings of the
Combustion Institute
Proceedings of the Combustion Institute 30 (2005) 2335–2343
www.elsevier.com/locate/proci
An Earth-based equivalent low stretch apparatus for material flammability assessment in microgravity and extraterrestrial environments S.L. Olsona,*, H.D. Beesonb, J.P. Haasb, J.S. Baasb a
NASA Glenn Research Center, Cleveland, OH 44135, USA NASA White Sands Test Facility, Las Cruces, NM, USA
b
Abstract The standard oxygen consumption (cone) calorimeter (described in ASTM E 1354 and NASA STD 6001 Test 2) is modified to provide a bench-scale test environment that simulates the low velocity buoyant or ventilation flow generated by or around a burning surface in a spacecraft or extraterrestrial gravity level. The equivalent low stretch apparatus (ELSA) uses an inverted cone geometry with the sample burning in a ceiling fire (stagnation flow) configuration. For a fixed radiant flux, ignition delay times for characterization material PMMA are shown to decrease by a factor of 3 at low stretch, demonstrating that ignition delay times determined from normal cone tests significantly underestimate the risk in microgravity. The critical heat flux for ignition is found to be lowered at low stretch as the convective cooling is reduced. At the limit of no stretch, any heat flux that exceeds the surface radiative loss at the surface ignition temperature is sufficient for ignition. Regression rates for PMMA increase with heat flux and stretch rate, but regression rates are much more sensitive to heat flux at the low stretch rates, where a modest increase in heat flux of 25 kW/m2 increases the burning rates by an order of magnitude. The global equivalence ratio of these flames is very fuel rich, and the quantity of CO produced in this configuration is significantly higher than standard cone tests. These results demonstrate that the ELSA apparatus allows us to conduct normal gravity experiments that accurately and quantifiably evaluate a materialÕs flammability characteristics in the real-use environment of spacecraft or extraterrestrial gravitational acceleration. These results also demonstrate that current NASA STD 6001 Test 2 (standard cone) is not conservative since it evaluates a materialÕs flammability with a much higher inherent buoyant convective flow. Published by Elsevier Inc. on behalf of The Combustion Institute. Keywords: Cone calorimeter; Gravity; Flammability; PMMA; Ignition delay
1. Introduction NASAÕs current method of material screening determines fire resistance under conditions repre-
*
Corresponding author. Fax: +1 216 977 7065. E-mail address: [email protected] (S.L. Olson).
senting a worst-case for normal gravity flammability—the upward flame propagation test (Test 1 [1]). Its simple pass–fail criteria eliminate materials that burn for more than 15 cm. from a standardized ignition source. In addition, if a material drips burning pieces that ignite a flammable fabric below, it fails. The applicability of Test 1 to fires in microgravity and extraterrestrial environments, however, is uncertain because the
1540-7489/$ - see front matter. Published by Elsevier Inc. on behalf of The Combustion Institute. doi:10.1016/j.proci.2004.08.044
2336
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
relationship between this buoyancy-dominated test and actual extraterrestrial fire hazards is not understood. Flames in microgravity are known to preferentially spread upwind (i.e., opposed flow) [2], not downwind (i.e., concurrent flow) as in the normal gravity upward flammability screening Test 1. At low flow velocities, the concurrent (Test 1 configuration) flame spread was not viable over vertical solid cylinders, while the stagnation point flame at the bottom end of the cylinder (low stretch flame) was viable [3,4]. In addition, the maximum flammability in the opposed flow spread configuration is known to be at lower imposed flows and lower oxygen concentrations than that occurring in normal gravity downward spreading flames [5]. Unlike Test 1, the NASA STD 6001 Test 2 [1] standard oxygen consumption (cone) calorimeter (also described in ASTM E 1354) provides quantitative data on ignition delay times and burning rates of materials. However, it currently lacks any pass–fail criteria. In addition, it too is buoyancy-dominated. The objective of this research was to modify the well-instrumented standard cone configuration to provide a reproducible bench-scale test environment that simulates the buoyant or ventilation flow that would be generated by or around a burning surface in a spacecraft or extraterrestrial gravity level. We will then develop a standard test method with pass– fail criteria for future use in spacecraft materials flammability screening. (For example, dripping of molten material will be an automatic fail.) The equivalent low stretch apparatus (ELSA) uses an inverted cone geometry with the sample burning in a ceiling fire (stagnation flow) configuration. In this configuration, the influence of buoyancy is reduced because the hot gases cannot rise directly from the flame zone. The buoyant stagnation flow around the sample acts in the same direction as a variable forced convective stagnation flow, so that the mixed convective stretch rate (velocity gradient) near the fuel surface can be varied. This apparatus will allow us to conduct normal gravity experiments that accurately and quantitatively evaluate a materialÕs flammability characteristics in the real-use environment of spacecraft or extraterrestrial gravitational acceleration. Theoretical predictions of stagnation flames in mixed convective environments by Foutch and TÕien [6] indicate that it should be possible to understand a materialÕs burning characteristics in the low stretch environment of spacecraft (nonbuoyant, but with some movement induced by fans and crew disturbances) by understanding its burning characteristics in an equivalent Earthbased stretch environment (induced by normal gravity buoyancy and/or forced convection). Experimental results with a buoyant low stretch flame [7] demonstrated the transition from a ro-
bust flame at stretch rates of 10–20 s1 to a quenched flame at very low stretch (1–2 s1) in air. Based on these model predictions and experimental results, ELSA was designed and fabricated. 2. Experiment setup The ELSA apparatus, shown conceptually in Fig. 1, uses a mass-flow-controlled, ambient temperature forced-air flow issuing from a 7.5 cm diameter nozzle into the cone to augment the inherent normal gravity low buoyant stretch. Stretch rates have been varied in a prototype facility from purely buoyant (estimated 4 s1 for the 28 cm sample plate) to 33.5 s1 with forced convection. Cone heat flux has been varied from 10 to 25 kW/m2. These low flux levels and sub-buoyant stretch rates are selected because they are more likely in spacecraft or in extraterrestrial environments, where flames will be weak and heat flux to adjacent materials will be low. Samples are mounted 2.5 cm above the cone, which is the same distance as the normal cone configuration. Calibration of the system has been conducted. The flow field of the system was measured using hot wire anemometry, the flow profile is flat across the nozzle. In addition, smoke flow visualization using smoke rosin vaporized by the heated cone (at relatively low rosin vaporization temperature) produced stagnation flow at a forced stretch rate of unity. The spatial heat flux distribution from the cone in purely buoyant flow and at different stretch rates was measured. The flux distribution did not change with stretch rate, and the inverted cone distribution agreed very well with the normal cone distribution [8], only dropping to 90% of the
Fig. 1. Concept of ELSA apparatus, showing fuel sample suspended above radiant cone heater and oxidizer flow jet. The enclosure reflects the WSTF Controlled Atmosphere Cone Calorimeter facility [1].
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
centerline reading by the edge of the fuel sample for each case. The cone heat flux is set to a fixed level and actively controlled to that level during each test. The heat flux is verified with a heat flux gauge prior to the exposure of the sample in each test. The gas-phase temperature of the stagnation flow was measured at different flux levels and stretch rates. For the range of flux levels reported here, and for a low stretch rate of 3 s1, the gas temperature did not vary by more than 10% across the sample location for a given flux level. While the gas-phase temperature profiles were not corrected for convection and radiation, the temperature distribution mirrors the radiant distribution reasonably well, and the normalized distribution does not change with stretch even though the quantitative readings do (cooler at higher stretch), indicating any local variations in flow are insignificant. The fuel material tested is a 2.4 cm thick clear PMMA slab, which is cut to the standard cone test sample size of 10 cm · 10 cm exposed surface area but which also has a 2.5 cm lip around the opening to prevent the sample from falling through the opening. The holder is 0.2 cm thick stainless steel, 28 cm wide. The test procedure first establishes the cone heat flux using a heat flux gauge. Then, the low velocity ambient temperature flow is established. Finally, the sample is moved into position to expose it to this environment. The pilot (hot source or spark) was positioned to the side of the sample, very near the surface of the plate to ignite the vapors once a flammable mixture is obtained. Thermocouples on the surface of the sample record the surface temperature; video cameras record the ignition and flame behavior. The air mass flow rate, radiant flux, and cone temperature are recorded as well. A gas-phase pilot is used to ignite the gasphase vapor cloud that is generated by the preheated and pyrolyzing surface. The pilot is positioned very close to the edge of the fuel sample surface, to ignite the vapors as soon as they mix with the surrounding air. Once the vapors reach a flammable concentration, the pilot ignites the vapors, and a premixed flame flashes across the fuel surface. These flashes are observed prior to sustained burning, since the premixed vapor is rapidly consumed and flame extinguishes. A test usually has a number of these flashes prior to sustained burning. These flashes indicate that the pilot is correctly located in the premixing region, and the ignition delay to the sustained burning is truly the minimum ignition delay. Tests without these flashes, where the flame sustains on the first Ôflash,Õ invariably had longer ignition delay times, indicating the placement of the pilot was not in the proper position. Once burning is established, the flame reaches a ‘‘pseudo-steady state’’ in approximately 10 s.
2337
While the gas phase stabilizes quickly, the solid phase does not. Because the solid sample is of intermediate thickness, the temperature of the interior continues to increase throughout the test time, so a true steady state is never obtained. Equivalent stretch rates can be determined as a function of gravity, imposed flow, and geometry. For purely buoyant stagnation flow, the equivalent stretch rate is ab = [(qe q*)/qe] [g/R]1/2 [6,7], where the density difference from the average flame temperature to ambient is used, g is the gravity, and R is the radius of curvature of the sample. For purely forced flow, the equivalent stretch rate is characterized by either af = 2U1/R for a cylinder [6], or af = Ujet/djet for a jet impinging on a planar surface [9]. U1 is the velocity of the ambient stream or the jet, R is the radius of curvature of the cylinder, and djet is the diameter of the jet. For a forced stretch rate of 5 s1, the flow exiting the nozzle is greater than 30 cm/s. A generalized expression for stretch rate which captures mixed convection includes both buoyant and forced stretch is defined [6] as aequivalent ¼ af ð1 þ a2b =a2f Þ1=2 . The system has been designed so that the nozzle diameter, sample size, and distance between the two are all of the same order, so that the stretch rate does not vary significantly across the sample. The contributions of the buoyant stretch on the equivalent stretch rate were evaluated by correlating regression rate data for flat disks of various radii [10] and cylindrical samples with reported stretch rates [7]. The correlation, shown in Fig. 2, allows us to determine the inherent buoyant stretch for ELSA to be 4 s1 by matching the regression rates
Fig. 2. Correlation of flat disk radius with equivalent buoyant stretch [7,10]. Regression rates for cylinders and flat disks were correlated, and the equivalent stretch rate for the 14 cm radius ELSA holder was determined from the linear relation found between radius of the disk and equivalent stretch rate. Equivalent stretch for normal cone geometry [11] was determined for ÔidealÕ burning conditions with flame heat flux only.
2338
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
for cylinders and flat disks. Note that this correlation worked even through the range of radii where Rayleigh–Taylor instabilities were found at larger radii (>8 cm) [10]. On the Moon (0.16 g) or Mars (0.38 g), the same size samples would burn with buoyant stretch rates of 1.4 and 2.2 s1, respectively, in the ELSA test configuration. This correlation also allows us to determine that the normal cone buoyant stretch rate is 33 s1 by extrapolating the correlation fit in Fig. 2 to the ÔidealÕ burning rate (heat flux from flame to surface only) for clear PMMA from [11]. While the normal cone flow field is not well characterized, the flame is stabilized near a surface with a cross-flow (entrained air), and velocity gradients in a laminar boundary layer near the surface are constant. For a 5 cm surface (1/2 sample) with an average velocity gradient near the surface of 33 s1, the average entrained cross-flow would be 20 cm/s, which is reasonable. 3. Results and discussion 3.1. Ignition delay The samples are radiantly ignited with the assistance of a gas-phase pilot to ignite the vapors. Ignition of the sample was recorded on video, and
Fig. 3. PMMA sample vapors igniting beneath an irradiated PMMA sample. Spark igniter is located out of the field of view to the left of the images. Irradiation was 25 kW/m2, with an equivalent stretch rate of 8.9 s1 (Ujet = 60 cm/s). Images are roughly 1 s apart. This sequence followed a series of sporadic local ignitions (flashes) that did not sustain. Flame covers sample by the third frame.
the time for ignition was determined for each test condition. Figure 3 shows a sequence of images during ignition. The ignition kernel resulting in a sustained flame beneath the sample is shown in the first frame. The flame is initially planar, but may develop cellular flow under certain conditions, as described later. The surface temperature rises as soon as the sample is exposed to the cone heater, as shown in Fig. 4. The temperature rise is not a square root dependent as predicted by simple theories of constant flux, because at low flux levels the non-trivial surface radiative loss changes as the temperature rises so the net flux actually decreases with time. Sustained ignition of the sample is preceded by a series of ignition flashes. Sustained ignition occurs when the temperature jumps suddenly to a higher plateau which is the fuel pyrolysis temperature beneath the flame. Ignition delay time, defined as the time between when the sample is exposed to the radiant heater and sustained ignition, was measured as a function of stretch rate at 10 and 25 kW/m2 flux levels, as shown in Fig. 5. For a fixed radiant flux, ignition delay times are shown to decrease with decreasing stretch rate or relative g level (based on buoyant scaling discussed above). Since the air flow is not heated, increasing the air flow (stretch rate) increases the convective cooling of the sample, which increases the time it takes to heat to ignition temperatures. The difference between a normal cone [12] (1 g) ignition delay time and ignition delay times at very low stretch is a factor of 3, demonstrating that ignition delay times determined from normal cone tests significantly overestimate the ignition delay times of materials in microgravity or in low stretch extra-
Fig. 4. Surface temperature heating to ignition with time for WSTF test (10 kW/m2, 7.2 s1, cellular flame, black PMMA). Sustained ignition is preceded by a series of ignition flashes. Also shown on the right axis is mass loss with time from load cell data. A cellular flame developed seconds after ignition.
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
Fig. 5. Ignition delay time plotted as a function of equivalent stretch rate or equivalent g level (relative to Earth). n, 10 kW/m2; h, 25 kW/m2; and s, normal cone data (1 g) at 25 kW/m2 [12].
terrestrial environments. In addition, at the 10 kW/ m2 flux level, the ignition delay time at low stretch approaches the value of the normal cone at 25 kW/m2, indicating the sensitivity of low stretch flames to even weak levels of external flux. 3.2. Critical heat flux The critical heat flux for ignition is an important fire safety measurement for material evaluation. At low stretch, radiation from a small adjacent burning material may be sufficient to ignite a material. This was observed on Mir [4] in the Skorost Facility, where a burning molten ball of fuel at the end of a rod of one material heated an adjacent rod to vaporization temperatures (bending, bubbling). The flame from the first rod then acted as the gas-phase pilot for the premixed cloud developed between the two rods. The premixed flame propagated through the cloud and ignited the heated second rod. For the ELSA apparatus, we can estimate the critical heat flux by extrapolating ignition delay time data at different heat fluxes to the zero stretch limit. The ignition delay time for a semi-infinite solid is related to the heat flux and surface ignition temperature [12] as follows: 2 2 ðT ign T 1 Þ tign ¼ ðkqcÞ : 00 q_ 3 As shown in Fig. 6, an inverse square root dependence on heat flux and ignition delay time is found in ELSA tests of all stretch rates tested. The critical heat flux for ignition is where the linear fit to the data crosses the x axis (infinite ignition delay time). The inset plots the critical heat flux values found for each stretch rate, and a roughly linear dependence is found. The zero stretch heat flux is so low that for all practical pur-
2339
Fig. 6. Critical heat flux derived from extrapolation of t0:5 ign ¼ 0. The critical heat flux at the limit of zero stretch is effectively zero. Data: stretch rates (s1), n 25.3, h 15.5, e 10.8, s 8.9, and , 6.4. Inset data: s, this work; n, Ref. [12].
poses, any heat flux that just slightly more than offsets the experimental heat losses such as surface radiation is sufficient to ignite the material. This agrees with the model of Rhodes and Quintiere [12] in the limit of no convection: ðhc ¼ 0Þ : h i 1 ðT ign T 1 Þ þ erT 4ign rT 4ign : q_ 00crit;a¼0 ¼ c e Surface temperatures at ignition are lower for lower radiant flux levels, as was also observed by Rhodes and Quintiere [12]. The longer preheat times at lower heating rates allow the thermal layer to penetrate deeper, thus increasing the depth over which fuel vaporization occurs [13]. The critical flammable vapor concentration is reached at lower surface temperatures as this subsurface vaporization contribution increases. In addition, with little or no convection in microgravity, the vapor will accumulate. This could result in a violent ignition of the large cloud of accumulated vapor if a pilot/ignition source is positioned far from the source of the vapor. 3.3. Cellular instabilities Due to the flat sample geometry, at low forced stretch rates, Rayleigh–Taylor instabilities were observed beneath the sample after ignition. These instabilities were not observed for curved samples [7]. The instabilities in these tests were notably worse at higher flux levels, where fuel vaporization rates are quite vigorous. Figure 7 shows the decrease in the cellular instabilities with increasing stretch. The nominal transition from cellular to planar one occurs at 8 s1 for a cone heat flux of 25 kW/m2. The cells, once developed, persist for the remainder of the test.
2340
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
Fig. 7. Rayleigh–Taylor instabilities are observed at stretch rates less than 8 s1 for a cone heat flux of 25 kW/m2. These instabilities are due to the flat samples used in these experiments as well as the augmented heat flux levels. At low stretch rates, the cells are very large, and extend into the cone. Equivalent stretch rates: (A) 4.1 s1, (B) 5.6 s1, (C) 8.9 s1, (D) 10.8 s1, and (E) 20.4 s1.
While the cellular instability is clearly gravity dependent, its development occurs well after ignition. The ignition flame is planar, as shown in Fig. 3. Thus, ignition delay time and ignition temperature data are not affected by this instability. The pseudo-steady-state flame shape is dominated by this instability. However, it is not clear if the average burning rates are affected by the gas-phase flame shape. Data presented in Figs. 4 and 9 are
Fig. 8. PMMA regression rates for three levels of imposed heat flux from radiant cone heater, plotted as a function of equivalent stretch rate or equivalent g level (relative to Earth). Data: e, n, this work; s, Ref. [7]; h, Ref. [16]; and ,, Ref. [9].
Fig. 9. WSTF inverted cone data (10 kW/m2, 7.2 s1, cellular flame, black PMMA) showing oxygen, carbon dioxide, and carbon monoxide concentration percent by volume. Heat release rate is also plotted based upon the controlled flow rate and oxygen consumption rate [14].
from a cellular flame. Cellular flame results are included in Fig. 8 as well, with no change in trends, in agreement with [9]. 3.4. Surface regression After ignition, samples were allowed to burn for a period of time to obtain information about the average burning rates of the material. For all stretch rates, the samples were found to regress uniformly across the exposed area, which indicates that the Rayleigh–Taylor cells do not affect the fuel surface regression locally. It is less clear if the cellular flow field significantly affects the burning rate in a global sense. Due to the enhanced convection, the burning rates could be higher than they would be without the cellular flow. On the other hand, they could be lower because of the larger average standoff distance between the flame and the fuel surface. Average regression rates as a function of stretch rate, shown in Fig. 8, were obtained by weighing the samples before and after the test. Using the stretch scaling, a second x axis of effective gravity level is also shown, assuming the stagnation stretch is purely buoyant. Regression rates were calculated using Vregression = Dg/(qAtburn), where Dg is the mass loss in grams, q = 1.19 g/ cc, A = 100 cm2 of the exposed sample, and tburn is the burn time in seconds of the sample from ignition to extinguishment. Burning rates (g/ cm2 s) can also be determined from these data. There is no obvious change in the regression rate trend as the flame changes from a planar to a cellular structure, in agreement with data from [9]. The ELSA apparatus has been incorporated into the White Sands Test FacilityÕs Controlled
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
Atmosphere Cone Calorimeter, and initial shakedown testing has begun. This unit has a load cell for continuous mass loss measurements. Data from the WSTF load cell, shown in Fig. 4, indicate that mass loss is steady with time after ignition for the period of the test, so tests using average mass loss rates are justified. The noise in the data is due to blower vibration. The sample starts losing mass once the sample surface temperature reaches 200 C. It continues a slow mass loss rate until ignition, where the mass loss rate increases by a factor of approximately 6. The mass loss rate after ignition is linear with time until the nitrogen purge is used to extinguish the sample, indicating a steady burning rate during the test time. Regression rates increase with heat flux and stretch rate or g level, but regression rates are much more sensitive to heat flux at the low stretch rates. A modest increase in heat flux from 0 to 25 kW/m2 increases the burning rates by an order of magnitude at the lower stretch rates. These trends are reasonable for low stretch flames, which have been shown to be very sensitive to the ratio of heat loss to heat generated via combustion [7]. Even at 10 kW/m2, the cone heater offsets the surface radiative loss, which is significant relative to the weak heat generation rates [7]. The flame feedback that was previously used to offset the radiative losses is now used to vaporize much more fuel. But at these low stretch rates, there is limited oxygen available with which to react— for example, normal cone tests run flows of 24 liters/s, whereas the flow rate to achieve a stretch rate of 6 s1 is a meager 0.67 liters/s. 3.5. Combustion products The White Sands Test FacilityÕs Controlled Atmosphere Cone Calorimeter also includes an exhaust system set of diagnostics including oxygen consumption, CO, and CO2. Gas species data are shown in Fig. 9. The oxygen concentration drops rapidly after ignition, and then more gradually as the test progresses. The oxygen concentration drops to 17% by the end of the test. It does not plateau to a steady concentration, indicating that the combustion process is strengthening gradually over time. This may be associated with reduced heat losses as the sample heats up during the test. For tests on the order of many minutes such as this one, the inherently transient finite thickness solid-phase timescales are important. With the known inlet air flow rate providing the forced stretch, we can determine oxygen consumption rates and associated heat release rates (12.97 kJ/g O2 [14]). As shown, the heat release rises to 10 W/cm2, or 100 kW/m2. CO and CO2 concentrations rise as well, with CO reaching a maximum of 0.1% (1000 ppm), and CO2 reaching a maximum of 3.65% by the end of the test. The ratio of CO/CO2 is at a maximum of 0.04 at ignition but the ratio has
2341
an average value closer to 0.03 during the test. The CO concentrations in this configuration of 1000 ppm are nearly a factor of 10 higher than standard cone tests (25 kW/m2, 24 liters/s), which have CO concentrations from 0.012% to 0.015% [15]. Exposure to 1000 ppm of CO for longer than 30 min causes nausea, mental confusion, headache, staggering, and heart palpitations. Because oxygen consumption and mass loss are both recorded, we can estimate the global equivalence ratio of the flame. For this test, the global equivalence ratio was 4.0, very fuel rich, indicating the presence of a significant amount of unburned fuel in the exhaust stream. All of the WSTF tests conducted to date have had similar fuel rich global equivalence ratios. The products of combustion from these fuel rich low stretch flames are more toxic (CO, THC, etc.) than normal cone combustion products. In addition, the incomplete combustion poses a significant risk of buildup of flammable gases to explosive levels in a confined space. 4. Conclusions The equivalent low stretch apparatus (ELSA) uses an inverted cone geometry with the sample burning in a ceiling fire configuration that allows for low stretch stagnation flow. Ignition delay times and regression rate data have been measured in a prototype unit. For a fixed radiant flux, ignition delay times for PMMA are shown to decrease by a factor of 3 at low stretch, demonstrating that ignition delay times determined from normal cone tests significantly underestimate the risk in microgravity. The critical heat flux for ignition decreases roughly linearly with stretch rate, with the zero stretch critical heat flux essentially reaching zero net heat flux. The surface temperature at ignition is also lowered at low radiant flux levels, possibly due to in-depth vaporization contributions to the total vaporization rate. Regression rates for PMMA increase with heat flux and stretch rate, but regression rates are much more sensitive to heat flux at the low stretch rates, where a modest increase in heat flux of 25 kW/m2 increases the burning rates by an order of magnitude. The low stretch flame is very fuel rich (global equivalence ratio 4), and significant amounts of CO are produced (1000 ppm). Although not measured, it is expected that there are other toxic gases having concentrations much higher than those generated in the normal cone configuration. These results demonstrate the ability of ELSA to simulate key features of low stretch materialÕs flammability behavior. These results also demonstrate that current NASA STD 6001 Test 2 (standard cone) is not conservative since it evaluates
2342
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
materials flammability with a much higher inherent buoyant convective flow.
[5] [6]
Acknowledgments
[7]
The authors acknowledge NASA Glenn engineers Chris Gallo for designing the prototype unit and Ray Wade for designing the spark ignition system; summer students Lander Coronado-Garcia and FloJaune Griffin for conducting many of the prototype experiments; and Sarah Smith for her work on the engineering modifications to the WSTF ELSA apparatus.
[8]
[9] [10] [11] [12]
References
[13] [1] D. Mulville, Flammability, odor, offgassing, and compatibility requirements and test procedures for materials in environments that support combustion, NASA-STD-6001, 1998. [2] K.B. McGrattan, T. Kashiwagi, H.R. Baum, S.L. Olson, Combust. Flame 106 (1996) 377–391. [3] Y. Halli, J.S. TÕien, NBS-GCR-86-507 (1986). [4] A.V. Ivanov, Ye.V. Balashov, T.V. Andreeva, A.S. Melkhov, NASA Contract NAS3-97160 final report, Russian Space Agency, Keldysh Research
[14] [15]
[16]
Center, Moscow, 1999, also NASA/CR-1999209405 (1999). S.L. Olson, Combust. Sci. Technol. 76 (1991) 233–249. D.W. Foutch, J.S. TÕien, AIAA J. 25 (7) (1987) 972–976. S.L. Olson, J.S. TÕien, Combust. Flame 121 (2000) 439–452. M.T. Wilson, B.Z. Dlugorgorski, E.M. Kennedy, in: Proceedings of the 7th International Symposium on Fire Safety Science, Boston, USA, 2002, pp. 815–826. J.S. TÕien, S.N. Singhal, D.P. Harrold, J.M. Prahl, Combust. Flame 33 (1978) 55–68. J.P. Vantelon, A. Himdi, F. Gaboraiud, Combust. Sci. Technol. 54 (1987) 145–158. A. Tewarson, R.F. Pion, Combust. Flame 26 (1976) 85–103. B.T. Rhodes, J.G. Quintiere, Fire Safety J. 26 (1996) 221–240. S.L. Olson, J.S. TÕien, Fire Mater. 23 (1999) 227–237. V. Babrauskas, S.J. Grayson (Eds.), Heat Release in Fires. Elsevier Applied Science, London and New York, 1992, p. 215. F.-Y. Hsheih, S.E. Motto, D.B. Hirsch, H.D. Beeson, in: Proceedings of the International Conference on Fire Safety, vol. 18, 1993, pp. 299–325. H. Ohtani, K. Akita, T. Hirano, Combust. Flame 53 (1983) 33–40.
Comments Carlos Fernandez-Pello, University of California, Berkeley, USA. The sample size may affect your buoyancy induced flow and strain rate. Have you looked into this potential problem, and potential effects on the ignition delay? Reply. The ELSA sample plus holder were sized to reduce the inherent buoyant stretch to a reasonable level of 4 s1, based on the scaling analysis discussed in Fig. 2 of the paper. Using velocity gradient similarity, the inherent stretch of 4 s-1 is what would be present in a spacecraft with a flow of 5 cm/s (spacecraft ventilation is 5–20 cm/s) for a 5 cm sample (200 Velcro squares, for example). The sample size itself is the same as the standard cone size (10 cm square). A larger sample would have required a larger cone heater to maintain a uniform heat flux over the sample. For flame standoff distances of less than 6 mm [7], Lstandoff Lsample, so the flame can be considered one-dimensional. Thus sample size should not play a role in the ignition delay. d
Subrata Bhattacharjee, San Diego State University, USA. Could you comment on what fraction of the ignition delay can be attributed to the sensible heating time of the fuel?
Reply. The fraction of the ignition delay that can be attributed to sensible heating can be determined from the surface temperature histories for the tests. Degradation of PMMA starts at 200 C [13], so if we estimate the fraction of the time to heat the surface of the material to the start of degradation using t200 C/tign, we find that for both 10 kW/m2 and 25 kW/m2, the fraction is 15%. There is no apparent dependence of the fraction on stretch rate within the scatter in the data, although both times do increase with increasing stretch. The rate of surface heating slows down as the surface heats up, not only due to endothermic pyrolysis (which increases exponentially with Ts), but also due to re-radiation (increases as T4s ) and convective cooling (increases proportional to Ts–Tair). Thus the majority of the ignition delay time is during active pyrolysis, but before a sustainable flammable mixture is generated. d
Michael Delichatsios, University of Ulster, UK. It seems that your results for ignition delay would agree with my published result that at ignition (piloted) a critical mass flux is needed given by ðmdotLj prime ÞcrC p =hc ¼ combustion (independent of imposed or flame heat flux). In this relation hc is the convective heat transfer coefficient. Accordingly, as the straining rate decreases the convective heat transfer coef-
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343 ficient decreases and the critical mass at ignition decreases so that the material can ignite earlier. Thus, there is no contradiction with normal cone detail. The higher yield rates for CO are due to restrained mixing of air with the pyrolyzing gases in the configuration of your experiment. In this sense there is really not a discrepancy with the ‘‘normal’’ cone results. Reply. As we noted in the paper, the surface temperatures at ignition were actually lower at lower imposed flux levels, possibly due to in-depth contributions to fuel vaporization (where lower flux levels allow deeper penetration of the heat). In most combustion situations, the PMMA polymer surface is unsaturated, and the in-depth monomer formation and transport to the surface is the rate-controlling process for pyrolysis [13]. Given these complexities of the solid phase pyrolysis process which become important at low flux levels, we would be hesitant to conclude that the critical mass flux at ignition is independent of imposed flux.
2343
For a given flux level, the surface temperature at ignition does not vary with stretch rate within the scatter in the data, which would seem to indicate that the critical mass flux at ignition does not vary with stretch. Earlier ignition times at low stretch are attributed to the higher net heating rate as convective cooling is reduced. In comparing ELSA results with normal cone results, the normal cone ignition delay times are not conservative if the spacecraft materials selection criteria is based on a long enough ignition delay for a crew to evacuate before flashover, for example. Spacecraft are confined environments with very weak convective mixing. You are correct that the higher yield rates for CO in ELSA are what one would expect of a weakly ventilated flame. However, if we base our materials screening on a buoyancy-dominated, well-mixed test such as the standard cone calorimeter test, then we may be missing one of the most significant hazards of the material in the real-use environment. In that sense, the standard (i.e., ÔnormalÕ) cone result is not conservative.
Combustion Institute
Proceedings of the Combustion Institute 30 (2005) 2335–2343
www.elsevier.com/locate/proci
An Earth-based equivalent low stretch apparatus for material flammability assessment in microgravity and extraterrestrial environments S.L. Olsona,*, H.D. Beesonb, J.P. Haasb, J.S. Baasb a
NASA Glenn Research Center, Cleveland, OH 44135, USA NASA White Sands Test Facility, Las Cruces, NM, USA
b
Abstract The standard oxygen consumption (cone) calorimeter (described in ASTM E 1354 and NASA STD 6001 Test 2) is modified to provide a bench-scale test environment that simulates the low velocity buoyant or ventilation flow generated by or around a burning surface in a spacecraft or extraterrestrial gravity level. The equivalent low stretch apparatus (ELSA) uses an inverted cone geometry with the sample burning in a ceiling fire (stagnation flow) configuration. For a fixed radiant flux, ignition delay times for characterization material PMMA are shown to decrease by a factor of 3 at low stretch, demonstrating that ignition delay times determined from normal cone tests significantly underestimate the risk in microgravity. The critical heat flux for ignition is found to be lowered at low stretch as the convective cooling is reduced. At the limit of no stretch, any heat flux that exceeds the surface radiative loss at the surface ignition temperature is sufficient for ignition. Regression rates for PMMA increase with heat flux and stretch rate, but regression rates are much more sensitive to heat flux at the low stretch rates, where a modest increase in heat flux of 25 kW/m2 increases the burning rates by an order of magnitude. The global equivalence ratio of these flames is very fuel rich, and the quantity of CO produced in this configuration is significantly higher than standard cone tests. These results demonstrate that the ELSA apparatus allows us to conduct normal gravity experiments that accurately and quantifiably evaluate a materialÕs flammability characteristics in the real-use environment of spacecraft or extraterrestrial gravitational acceleration. These results also demonstrate that current NASA STD 6001 Test 2 (standard cone) is not conservative since it evaluates a materialÕs flammability with a much higher inherent buoyant convective flow. Published by Elsevier Inc. on behalf of The Combustion Institute. Keywords: Cone calorimeter; Gravity; Flammability; PMMA; Ignition delay
1. Introduction NASAÕs current method of material screening determines fire resistance under conditions repre-
*
Corresponding author. Fax: +1 216 977 7065. E-mail address: [email protected] (S.L. Olson).
senting a worst-case for normal gravity flammability—the upward flame propagation test (Test 1 [1]). Its simple pass–fail criteria eliminate materials that burn for more than 15 cm. from a standardized ignition source. In addition, if a material drips burning pieces that ignite a flammable fabric below, it fails. The applicability of Test 1 to fires in microgravity and extraterrestrial environments, however, is uncertain because the
1540-7489/$ - see front matter. Published by Elsevier Inc. on behalf of The Combustion Institute. doi:10.1016/j.proci.2004.08.044
2336
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
relationship between this buoyancy-dominated test and actual extraterrestrial fire hazards is not understood. Flames in microgravity are known to preferentially spread upwind (i.e., opposed flow) [2], not downwind (i.e., concurrent flow) as in the normal gravity upward flammability screening Test 1. At low flow velocities, the concurrent (Test 1 configuration) flame spread was not viable over vertical solid cylinders, while the stagnation point flame at the bottom end of the cylinder (low stretch flame) was viable [3,4]. In addition, the maximum flammability in the opposed flow spread configuration is known to be at lower imposed flows and lower oxygen concentrations than that occurring in normal gravity downward spreading flames [5]. Unlike Test 1, the NASA STD 6001 Test 2 [1] standard oxygen consumption (cone) calorimeter (also described in ASTM E 1354) provides quantitative data on ignition delay times and burning rates of materials. However, it currently lacks any pass–fail criteria. In addition, it too is buoyancy-dominated. The objective of this research was to modify the well-instrumented standard cone configuration to provide a reproducible bench-scale test environment that simulates the buoyant or ventilation flow that would be generated by or around a burning surface in a spacecraft or extraterrestrial gravity level. We will then develop a standard test method with pass– fail criteria for future use in spacecraft materials flammability screening. (For example, dripping of molten material will be an automatic fail.) The equivalent low stretch apparatus (ELSA) uses an inverted cone geometry with the sample burning in a ceiling fire (stagnation flow) configuration. In this configuration, the influence of buoyancy is reduced because the hot gases cannot rise directly from the flame zone. The buoyant stagnation flow around the sample acts in the same direction as a variable forced convective stagnation flow, so that the mixed convective stretch rate (velocity gradient) near the fuel surface can be varied. This apparatus will allow us to conduct normal gravity experiments that accurately and quantitatively evaluate a materialÕs flammability characteristics in the real-use environment of spacecraft or extraterrestrial gravitational acceleration. Theoretical predictions of stagnation flames in mixed convective environments by Foutch and TÕien [6] indicate that it should be possible to understand a materialÕs burning characteristics in the low stretch environment of spacecraft (nonbuoyant, but with some movement induced by fans and crew disturbances) by understanding its burning characteristics in an equivalent Earthbased stretch environment (induced by normal gravity buoyancy and/or forced convection). Experimental results with a buoyant low stretch flame [7] demonstrated the transition from a ro-
bust flame at stretch rates of 10–20 s1 to a quenched flame at very low stretch (1–2 s1) in air. Based on these model predictions and experimental results, ELSA was designed and fabricated. 2. Experiment setup The ELSA apparatus, shown conceptually in Fig. 1, uses a mass-flow-controlled, ambient temperature forced-air flow issuing from a 7.5 cm diameter nozzle into the cone to augment the inherent normal gravity low buoyant stretch. Stretch rates have been varied in a prototype facility from purely buoyant (estimated 4 s1 for the 28 cm sample plate) to 33.5 s1 with forced convection. Cone heat flux has been varied from 10 to 25 kW/m2. These low flux levels and sub-buoyant stretch rates are selected because they are more likely in spacecraft or in extraterrestrial environments, where flames will be weak and heat flux to adjacent materials will be low. Samples are mounted 2.5 cm above the cone, which is the same distance as the normal cone configuration. Calibration of the system has been conducted. The flow field of the system was measured using hot wire anemometry, the flow profile is flat across the nozzle. In addition, smoke flow visualization using smoke rosin vaporized by the heated cone (at relatively low rosin vaporization temperature) produced stagnation flow at a forced stretch rate of unity. The spatial heat flux distribution from the cone in purely buoyant flow and at different stretch rates was measured. The flux distribution did not change with stretch rate, and the inverted cone distribution agreed very well with the normal cone distribution [8], only dropping to 90% of the
Fig. 1. Concept of ELSA apparatus, showing fuel sample suspended above radiant cone heater and oxidizer flow jet. The enclosure reflects the WSTF Controlled Atmosphere Cone Calorimeter facility [1].
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
centerline reading by the edge of the fuel sample for each case. The cone heat flux is set to a fixed level and actively controlled to that level during each test. The heat flux is verified with a heat flux gauge prior to the exposure of the sample in each test. The gas-phase temperature of the stagnation flow was measured at different flux levels and stretch rates. For the range of flux levels reported here, and for a low stretch rate of 3 s1, the gas temperature did not vary by more than 10% across the sample location for a given flux level. While the gas-phase temperature profiles were not corrected for convection and radiation, the temperature distribution mirrors the radiant distribution reasonably well, and the normalized distribution does not change with stretch even though the quantitative readings do (cooler at higher stretch), indicating any local variations in flow are insignificant. The fuel material tested is a 2.4 cm thick clear PMMA slab, which is cut to the standard cone test sample size of 10 cm · 10 cm exposed surface area but which also has a 2.5 cm lip around the opening to prevent the sample from falling through the opening. The holder is 0.2 cm thick stainless steel, 28 cm wide. The test procedure first establishes the cone heat flux using a heat flux gauge. Then, the low velocity ambient temperature flow is established. Finally, the sample is moved into position to expose it to this environment. The pilot (hot source or spark) was positioned to the side of the sample, very near the surface of the plate to ignite the vapors once a flammable mixture is obtained. Thermocouples on the surface of the sample record the surface temperature; video cameras record the ignition and flame behavior. The air mass flow rate, radiant flux, and cone temperature are recorded as well. A gas-phase pilot is used to ignite the gasphase vapor cloud that is generated by the preheated and pyrolyzing surface. The pilot is positioned very close to the edge of the fuel sample surface, to ignite the vapors as soon as they mix with the surrounding air. Once the vapors reach a flammable concentration, the pilot ignites the vapors, and a premixed flame flashes across the fuel surface. These flashes are observed prior to sustained burning, since the premixed vapor is rapidly consumed and flame extinguishes. A test usually has a number of these flashes prior to sustained burning. These flashes indicate that the pilot is correctly located in the premixing region, and the ignition delay to the sustained burning is truly the minimum ignition delay. Tests without these flashes, where the flame sustains on the first Ôflash,Õ invariably had longer ignition delay times, indicating the placement of the pilot was not in the proper position. Once burning is established, the flame reaches a ‘‘pseudo-steady state’’ in approximately 10 s.
2337
While the gas phase stabilizes quickly, the solid phase does not. Because the solid sample is of intermediate thickness, the temperature of the interior continues to increase throughout the test time, so a true steady state is never obtained. Equivalent stretch rates can be determined as a function of gravity, imposed flow, and geometry. For purely buoyant stagnation flow, the equivalent stretch rate is ab = [(qe q*)/qe] [g/R]1/2 [6,7], where the density difference from the average flame temperature to ambient is used, g is the gravity, and R is the radius of curvature of the sample. For purely forced flow, the equivalent stretch rate is characterized by either af = 2U1/R for a cylinder [6], or af = Ujet/djet for a jet impinging on a planar surface [9]. U1 is the velocity of the ambient stream or the jet, R is the radius of curvature of the cylinder, and djet is the diameter of the jet. For a forced stretch rate of 5 s1, the flow exiting the nozzle is greater than 30 cm/s. A generalized expression for stretch rate which captures mixed convection includes both buoyant and forced stretch is defined [6] as aequivalent ¼ af ð1 þ a2b =a2f Þ1=2 . The system has been designed so that the nozzle diameter, sample size, and distance between the two are all of the same order, so that the stretch rate does not vary significantly across the sample. The contributions of the buoyant stretch on the equivalent stretch rate were evaluated by correlating regression rate data for flat disks of various radii [10] and cylindrical samples with reported stretch rates [7]. The correlation, shown in Fig. 2, allows us to determine the inherent buoyant stretch for ELSA to be 4 s1 by matching the regression rates
Fig. 2. Correlation of flat disk radius with equivalent buoyant stretch [7,10]. Regression rates for cylinders and flat disks were correlated, and the equivalent stretch rate for the 14 cm radius ELSA holder was determined from the linear relation found between radius of the disk and equivalent stretch rate. Equivalent stretch for normal cone geometry [11] was determined for ÔidealÕ burning conditions with flame heat flux only.
2338
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
for cylinders and flat disks. Note that this correlation worked even through the range of radii where Rayleigh–Taylor instabilities were found at larger radii (>8 cm) [10]. On the Moon (0.16 g) or Mars (0.38 g), the same size samples would burn with buoyant stretch rates of 1.4 and 2.2 s1, respectively, in the ELSA test configuration. This correlation also allows us to determine that the normal cone buoyant stretch rate is 33 s1 by extrapolating the correlation fit in Fig. 2 to the ÔidealÕ burning rate (heat flux from flame to surface only) for clear PMMA from [11]. While the normal cone flow field is not well characterized, the flame is stabilized near a surface with a cross-flow (entrained air), and velocity gradients in a laminar boundary layer near the surface are constant. For a 5 cm surface (1/2 sample) with an average velocity gradient near the surface of 33 s1, the average entrained cross-flow would be 20 cm/s, which is reasonable. 3. Results and discussion 3.1. Ignition delay The samples are radiantly ignited with the assistance of a gas-phase pilot to ignite the vapors. Ignition of the sample was recorded on video, and
Fig. 3. PMMA sample vapors igniting beneath an irradiated PMMA sample. Spark igniter is located out of the field of view to the left of the images. Irradiation was 25 kW/m2, with an equivalent stretch rate of 8.9 s1 (Ujet = 60 cm/s). Images are roughly 1 s apart. This sequence followed a series of sporadic local ignitions (flashes) that did not sustain. Flame covers sample by the third frame.
the time for ignition was determined for each test condition. Figure 3 shows a sequence of images during ignition. The ignition kernel resulting in a sustained flame beneath the sample is shown in the first frame. The flame is initially planar, but may develop cellular flow under certain conditions, as described later. The surface temperature rises as soon as the sample is exposed to the cone heater, as shown in Fig. 4. The temperature rise is not a square root dependent as predicted by simple theories of constant flux, because at low flux levels the non-trivial surface radiative loss changes as the temperature rises so the net flux actually decreases with time. Sustained ignition of the sample is preceded by a series of ignition flashes. Sustained ignition occurs when the temperature jumps suddenly to a higher plateau which is the fuel pyrolysis temperature beneath the flame. Ignition delay time, defined as the time between when the sample is exposed to the radiant heater and sustained ignition, was measured as a function of stretch rate at 10 and 25 kW/m2 flux levels, as shown in Fig. 5. For a fixed radiant flux, ignition delay times are shown to decrease with decreasing stretch rate or relative g level (based on buoyant scaling discussed above). Since the air flow is not heated, increasing the air flow (stretch rate) increases the convective cooling of the sample, which increases the time it takes to heat to ignition temperatures. The difference between a normal cone [12] (1 g) ignition delay time and ignition delay times at very low stretch is a factor of 3, demonstrating that ignition delay times determined from normal cone tests significantly overestimate the ignition delay times of materials in microgravity or in low stretch extra-
Fig. 4. Surface temperature heating to ignition with time for WSTF test (10 kW/m2, 7.2 s1, cellular flame, black PMMA). Sustained ignition is preceded by a series of ignition flashes. Also shown on the right axis is mass loss with time from load cell data. A cellular flame developed seconds after ignition.
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
Fig. 5. Ignition delay time plotted as a function of equivalent stretch rate or equivalent g level (relative to Earth). n, 10 kW/m2; h, 25 kW/m2; and s, normal cone data (1 g) at 25 kW/m2 [12].
terrestrial environments. In addition, at the 10 kW/ m2 flux level, the ignition delay time at low stretch approaches the value of the normal cone at 25 kW/m2, indicating the sensitivity of low stretch flames to even weak levels of external flux. 3.2. Critical heat flux The critical heat flux for ignition is an important fire safety measurement for material evaluation. At low stretch, radiation from a small adjacent burning material may be sufficient to ignite a material. This was observed on Mir [4] in the Skorost Facility, where a burning molten ball of fuel at the end of a rod of one material heated an adjacent rod to vaporization temperatures (bending, bubbling). The flame from the first rod then acted as the gas-phase pilot for the premixed cloud developed between the two rods. The premixed flame propagated through the cloud and ignited the heated second rod. For the ELSA apparatus, we can estimate the critical heat flux by extrapolating ignition delay time data at different heat fluxes to the zero stretch limit. The ignition delay time for a semi-infinite solid is related to the heat flux and surface ignition temperature [12] as follows: 2 2 ðT ign T 1 Þ tign ¼ ðkqcÞ : 00 q_ 3 As shown in Fig. 6, an inverse square root dependence on heat flux and ignition delay time is found in ELSA tests of all stretch rates tested. The critical heat flux for ignition is where the linear fit to the data crosses the x axis (infinite ignition delay time). The inset plots the critical heat flux values found for each stretch rate, and a roughly linear dependence is found. The zero stretch heat flux is so low that for all practical pur-
2339
Fig. 6. Critical heat flux derived from extrapolation of t0:5 ign ¼ 0. The critical heat flux at the limit of zero stretch is effectively zero. Data: stretch rates (s1), n 25.3, h 15.5, e 10.8, s 8.9, and , 6.4. Inset data: s, this work; n, Ref. [12].
poses, any heat flux that just slightly more than offsets the experimental heat losses such as surface radiation is sufficient to ignite the material. This agrees with the model of Rhodes and Quintiere [12] in the limit of no convection: ðhc ¼ 0Þ : h i 1 ðT ign T 1 Þ þ erT 4ign rT 4ign : q_ 00crit;a¼0 ¼ c e Surface temperatures at ignition are lower for lower radiant flux levels, as was also observed by Rhodes and Quintiere [12]. The longer preheat times at lower heating rates allow the thermal layer to penetrate deeper, thus increasing the depth over which fuel vaporization occurs [13]. The critical flammable vapor concentration is reached at lower surface temperatures as this subsurface vaporization contribution increases. In addition, with little or no convection in microgravity, the vapor will accumulate. This could result in a violent ignition of the large cloud of accumulated vapor if a pilot/ignition source is positioned far from the source of the vapor. 3.3. Cellular instabilities Due to the flat sample geometry, at low forced stretch rates, Rayleigh–Taylor instabilities were observed beneath the sample after ignition. These instabilities were not observed for curved samples [7]. The instabilities in these tests were notably worse at higher flux levels, where fuel vaporization rates are quite vigorous. Figure 7 shows the decrease in the cellular instabilities with increasing stretch. The nominal transition from cellular to planar one occurs at 8 s1 for a cone heat flux of 25 kW/m2. The cells, once developed, persist for the remainder of the test.
2340
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
Fig. 7. Rayleigh–Taylor instabilities are observed at stretch rates less than 8 s1 for a cone heat flux of 25 kW/m2. These instabilities are due to the flat samples used in these experiments as well as the augmented heat flux levels. At low stretch rates, the cells are very large, and extend into the cone. Equivalent stretch rates: (A) 4.1 s1, (B) 5.6 s1, (C) 8.9 s1, (D) 10.8 s1, and (E) 20.4 s1.
While the cellular instability is clearly gravity dependent, its development occurs well after ignition. The ignition flame is planar, as shown in Fig. 3. Thus, ignition delay time and ignition temperature data are not affected by this instability. The pseudo-steady-state flame shape is dominated by this instability. However, it is not clear if the average burning rates are affected by the gas-phase flame shape. Data presented in Figs. 4 and 9 are
Fig. 8. PMMA regression rates for three levels of imposed heat flux from radiant cone heater, plotted as a function of equivalent stretch rate or equivalent g level (relative to Earth). Data: e, n, this work; s, Ref. [7]; h, Ref. [16]; and ,, Ref. [9].
Fig. 9. WSTF inverted cone data (10 kW/m2, 7.2 s1, cellular flame, black PMMA) showing oxygen, carbon dioxide, and carbon monoxide concentration percent by volume. Heat release rate is also plotted based upon the controlled flow rate and oxygen consumption rate [14].
from a cellular flame. Cellular flame results are included in Fig. 8 as well, with no change in trends, in agreement with [9]. 3.4. Surface regression After ignition, samples were allowed to burn for a period of time to obtain information about the average burning rates of the material. For all stretch rates, the samples were found to regress uniformly across the exposed area, which indicates that the Rayleigh–Taylor cells do not affect the fuel surface regression locally. It is less clear if the cellular flow field significantly affects the burning rate in a global sense. Due to the enhanced convection, the burning rates could be higher than they would be without the cellular flow. On the other hand, they could be lower because of the larger average standoff distance between the flame and the fuel surface. Average regression rates as a function of stretch rate, shown in Fig. 8, were obtained by weighing the samples before and after the test. Using the stretch scaling, a second x axis of effective gravity level is also shown, assuming the stagnation stretch is purely buoyant. Regression rates were calculated using Vregression = Dg/(qAtburn), where Dg is the mass loss in grams, q = 1.19 g/ cc, A = 100 cm2 of the exposed sample, and tburn is the burn time in seconds of the sample from ignition to extinguishment. Burning rates (g/ cm2 s) can also be determined from these data. There is no obvious change in the regression rate trend as the flame changes from a planar to a cellular structure, in agreement with data from [9]. The ELSA apparatus has been incorporated into the White Sands Test FacilityÕs Controlled
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
Atmosphere Cone Calorimeter, and initial shakedown testing has begun. This unit has a load cell for continuous mass loss measurements. Data from the WSTF load cell, shown in Fig. 4, indicate that mass loss is steady with time after ignition for the period of the test, so tests using average mass loss rates are justified. The noise in the data is due to blower vibration. The sample starts losing mass once the sample surface temperature reaches 200 C. It continues a slow mass loss rate until ignition, where the mass loss rate increases by a factor of approximately 6. The mass loss rate after ignition is linear with time until the nitrogen purge is used to extinguish the sample, indicating a steady burning rate during the test time. Regression rates increase with heat flux and stretch rate or g level, but regression rates are much more sensitive to heat flux at the low stretch rates. A modest increase in heat flux from 0 to 25 kW/m2 increases the burning rates by an order of magnitude at the lower stretch rates. These trends are reasonable for low stretch flames, which have been shown to be very sensitive to the ratio of heat loss to heat generated via combustion [7]. Even at 10 kW/m2, the cone heater offsets the surface radiative loss, which is significant relative to the weak heat generation rates [7]. The flame feedback that was previously used to offset the radiative losses is now used to vaporize much more fuel. But at these low stretch rates, there is limited oxygen available with which to react— for example, normal cone tests run flows of 24 liters/s, whereas the flow rate to achieve a stretch rate of 6 s1 is a meager 0.67 liters/s. 3.5. Combustion products The White Sands Test FacilityÕs Controlled Atmosphere Cone Calorimeter also includes an exhaust system set of diagnostics including oxygen consumption, CO, and CO2. Gas species data are shown in Fig. 9. The oxygen concentration drops rapidly after ignition, and then more gradually as the test progresses. The oxygen concentration drops to 17% by the end of the test. It does not plateau to a steady concentration, indicating that the combustion process is strengthening gradually over time. This may be associated with reduced heat losses as the sample heats up during the test. For tests on the order of many minutes such as this one, the inherently transient finite thickness solid-phase timescales are important. With the known inlet air flow rate providing the forced stretch, we can determine oxygen consumption rates and associated heat release rates (12.97 kJ/g O2 [14]). As shown, the heat release rises to 10 W/cm2, or 100 kW/m2. CO and CO2 concentrations rise as well, with CO reaching a maximum of 0.1% (1000 ppm), and CO2 reaching a maximum of 3.65% by the end of the test. The ratio of CO/CO2 is at a maximum of 0.04 at ignition but the ratio has
2341
an average value closer to 0.03 during the test. The CO concentrations in this configuration of 1000 ppm are nearly a factor of 10 higher than standard cone tests (25 kW/m2, 24 liters/s), which have CO concentrations from 0.012% to 0.015% [15]. Exposure to 1000 ppm of CO for longer than 30 min causes nausea, mental confusion, headache, staggering, and heart palpitations. Because oxygen consumption and mass loss are both recorded, we can estimate the global equivalence ratio of the flame. For this test, the global equivalence ratio was 4.0, very fuel rich, indicating the presence of a significant amount of unburned fuel in the exhaust stream. All of the WSTF tests conducted to date have had similar fuel rich global equivalence ratios. The products of combustion from these fuel rich low stretch flames are more toxic (CO, THC, etc.) than normal cone combustion products. In addition, the incomplete combustion poses a significant risk of buildup of flammable gases to explosive levels in a confined space. 4. Conclusions The equivalent low stretch apparatus (ELSA) uses an inverted cone geometry with the sample burning in a ceiling fire configuration that allows for low stretch stagnation flow. Ignition delay times and regression rate data have been measured in a prototype unit. For a fixed radiant flux, ignition delay times for PMMA are shown to decrease by a factor of 3 at low stretch, demonstrating that ignition delay times determined from normal cone tests significantly underestimate the risk in microgravity. The critical heat flux for ignition decreases roughly linearly with stretch rate, with the zero stretch critical heat flux essentially reaching zero net heat flux. The surface temperature at ignition is also lowered at low radiant flux levels, possibly due to in-depth vaporization contributions to the total vaporization rate. Regression rates for PMMA increase with heat flux and stretch rate, but regression rates are much more sensitive to heat flux at the low stretch rates, where a modest increase in heat flux of 25 kW/m2 increases the burning rates by an order of magnitude. The low stretch flame is very fuel rich (global equivalence ratio 4), and significant amounts of CO are produced (1000 ppm). Although not measured, it is expected that there are other toxic gases having concentrations much higher than those generated in the normal cone configuration. These results demonstrate the ability of ELSA to simulate key features of low stretch materialÕs flammability behavior. These results also demonstrate that current NASA STD 6001 Test 2 (standard cone) is not conservative since it evaluates
2342
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343
materials flammability with a much higher inherent buoyant convective flow.
[5] [6]
Acknowledgments
[7]
The authors acknowledge NASA Glenn engineers Chris Gallo for designing the prototype unit and Ray Wade for designing the spark ignition system; summer students Lander Coronado-Garcia and FloJaune Griffin for conducting many of the prototype experiments; and Sarah Smith for her work on the engineering modifications to the WSTF ELSA apparatus.
[8]
[9] [10] [11] [12]
References
[13] [1] D. Mulville, Flammability, odor, offgassing, and compatibility requirements and test procedures for materials in environments that support combustion, NASA-STD-6001, 1998. [2] K.B. McGrattan, T. Kashiwagi, H.R. Baum, S.L. Olson, Combust. Flame 106 (1996) 377–391. [3] Y. Halli, J.S. TÕien, NBS-GCR-86-507 (1986). [4] A.V. Ivanov, Ye.V. Balashov, T.V. Andreeva, A.S. Melkhov, NASA Contract NAS3-97160 final report, Russian Space Agency, Keldysh Research
[14] [15]
[16]
Center, Moscow, 1999, also NASA/CR-1999209405 (1999). S.L. Olson, Combust. Sci. Technol. 76 (1991) 233–249. D.W. Foutch, J.S. TÕien, AIAA J. 25 (7) (1987) 972–976. S.L. Olson, J.S. TÕien, Combust. Flame 121 (2000) 439–452. M.T. Wilson, B.Z. Dlugorgorski, E.M. Kennedy, in: Proceedings of the 7th International Symposium on Fire Safety Science, Boston, USA, 2002, pp. 815–826. J.S. TÕien, S.N. Singhal, D.P. Harrold, J.M. Prahl, Combust. Flame 33 (1978) 55–68. J.P. Vantelon, A. Himdi, F. Gaboraiud, Combust. Sci. Technol. 54 (1987) 145–158. A. Tewarson, R.F. Pion, Combust. Flame 26 (1976) 85–103. B.T. Rhodes, J.G. Quintiere, Fire Safety J. 26 (1996) 221–240. S.L. Olson, J.S. TÕien, Fire Mater. 23 (1999) 227–237. V. Babrauskas, S.J. Grayson (Eds.), Heat Release in Fires. Elsevier Applied Science, London and New York, 1992, p. 215. F.-Y. Hsheih, S.E. Motto, D.B. Hirsch, H.D. Beeson, in: Proceedings of the International Conference on Fire Safety, vol. 18, 1993, pp. 299–325. H. Ohtani, K. Akita, T. Hirano, Combust. Flame 53 (1983) 33–40.
Comments Carlos Fernandez-Pello, University of California, Berkeley, USA. The sample size may affect your buoyancy induced flow and strain rate. Have you looked into this potential problem, and potential effects on the ignition delay? Reply. The ELSA sample plus holder were sized to reduce the inherent buoyant stretch to a reasonable level of 4 s1, based on the scaling analysis discussed in Fig. 2 of the paper. Using velocity gradient similarity, the inherent stretch of 4 s-1 is what would be present in a spacecraft with a flow of 5 cm/s (spacecraft ventilation is 5–20 cm/s) for a 5 cm sample (200 Velcro squares, for example). The sample size itself is the same as the standard cone size (10 cm square). A larger sample would have required a larger cone heater to maintain a uniform heat flux over the sample. For flame standoff distances of less than 6 mm [7], Lstandoff Lsample, so the flame can be considered one-dimensional. Thus sample size should not play a role in the ignition delay. d
Subrata Bhattacharjee, San Diego State University, USA. Could you comment on what fraction of the ignition delay can be attributed to the sensible heating time of the fuel?
Reply. The fraction of the ignition delay that can be attributed to sensible heating can be determined from the surface temperature histories for the tests. Degradation of PMMA starts at 200 C [13], so if we estimate the fraction of the time to heat the surface of the material to the start of degradation using t200 C/tign, we find that for both 10 kW/m2 and 25 kW/m2, the fraction is 15%. There is no apparent dependence of the fraction on stretch rate within the scatter in the data, although both times do increase with increasing stretch. The rate of surface heating slows down as the surface heats up, not only due to endothermic pyrolysis (which increases exponentially with Ts), but also due to re-radiation (increases as T4s ) and convective cooling (increases proportional to Ts–Tair). Thus the majority of the ignition delay time is during active pyrolysis, but before a sustainable flammable mixture is generated. d
Michael Delichatsios, University of Ulster, UK. It seems that your results for ignition delay would agree with my published result that at ignition (piloted) a critical mass flux is needed given by ðmdotLj prime ÞcrC p =hc ¼ combustion (independent of imposed or flame heat flux). In this relation hc is the convective heat transfer coefficient. Accordingly, as the straining rate decreases the convective heat transfer coef-
S.L. Olson et al. / Proceedings of the Combustion Institute 30 (2005) 2335–2343 ficient decreases and the critical mass at ignition decreases so that the material can ignite earlier. Thus, there is no contradiction with normal cone detail. The higher yield rates for CO are due to restrained mixing of air with the pyrolyzing gases in the configuration of your experiment. In this sense there is really not a discrepancy with the ‘‘normal’’ cone results. Reply. As we noted in the paper, the surface temperatures at ignition were actually lower at lower imposed flux levels, possibly due to in-depth contributions to fuel vaporization (where lower flux levels allow deeper penetration of the heat). In most combustion situations, the PMMA polymer surface is unsaturated, and the in-depth monomer formation and transport to the surface is the rate-controlling process for pyrolysis [13]. Given these complexities of the solid phase pyrolysis process which become important at low flux levels, we would be hesitant to conclude that the critical mass flux at ignition is independent of imposed flux.
2343
For a given flux level, the surface temperature at ignition does not vary with stretch rate within the scatter in the data, which would seem to indicate that the critical mass flux at ignition does not vary with stretch. Earlier ignition times at low stretch are attributed to the higher net heating rate as convective cooling is reduced. In comparing ELSA results with normal cone results, the normal cone ignition delay times are not conservative if the spacecraft materials selection criteria is based on a long enough ignition delay for a crew to evacuate before flashover, for example. Spacecraft are confined environments with very weak convective mixing. You are correct that the higher yield rates for CO in ELSA are what one would expect of a weakly ventilated flame. However, if we base our materials screening on a buoyancy-dominated, well-mixed test such as the standard cone calorimeter test, then we may be missing one of the most significant hazards of the material in the real-use environment. In that sense, the standard (i.e., ÔnormalÕ) cone result is not conservative.