|Title||Novel Methods for Measuring Heat Flux in Industrial Furnaces|
|Conference||American Flame Research Committee, Salt Lake City, Utah, September 5-7, 2012|
|Creator||Lam, C.S.; Ramadan, O.B.; Hughes, P.M.; Wong, J.; Lycett, R.|
|Abstract||Heat flux measurements were conducted in a pilot-scale furnace using an infrared (IR) camera and a non-water-cooled, total heat flux gauge (THFG). The IR camera detected radiation reflected and emitted by the surfaces being viewed, while the THFG was designed such that the heat incident on the sensing surface flowed through the gauge primarily in one dimension, allowing use of a one-dimensional conduction analysis to estimate the total incident heat flux. In the present tests, the IR camera was aimed at the sensing surface of the THFG and the heat flux results from both measurement methods compared. The heat flux levels estimated using the THFG and IR camera generally agreed to within 5%. Factors that may have contributed to differences between the results included (1) convective heat flux, which would have influenced the response of the THFG but not the IR camera, (2) re-emission of only part of the heat flux incident on the THFG from the gauge back to the IR camera, and (3) attenuation and emission of radiation by the combustion gases (especially CO2) and atmosphere between the THFG and IR camera. With the last factor mitigated by the presence of an appropriate filter in front of the IR camera, the other factors did not appear to significantly affect the measured heat flux levels, based on the overall agreement of the results from the THFG and IR camera.|
|Rights||This material may be protected by copyright. Permission required for use in any form. For further information please contact the American Flame Research Committee.|
Novel Methods for Measuring Heat Flux in Industrial Furnaces C.S. Lam, O.B. Ramadan, P.M. Hughes, J. Wong and R. Lycett CanmetENERGY, Natural Resources Canada, 1 Haanel Drive, Ottawa, Ontario, Canada K1A 1M1 Corresponding author: Cecilia Lam CanmetENERGY, Natural Resources Canada 1 Haanel Drive, Ottawa, Ontario, Canada K1A 1M1 Tel: +1 613-996-2759 Fax: +1 613-992-9335 Email: Cecilia.Lam@NRCan-RNCan.gc.ca ABSTRACT Heat flux measurements were conducted in a pilot-scale furnace using an infrared (IR) camera and a non-water-cooled, total heat flux gauge (THFG). The IR camera detected radiation reflected and emitted by the surfaces being viewed, while the THFG was designed such that the heat incident on the sensing surface flowed through the gauge primarily in one dimension, allowing use of a one-dimensional conduction analysis to estimate the total incident heat flux. In the present tests, the IR camera was aimed at the sensing surface of the THFG and the heat flux results from both measurement methods compared. The heat flux levels estimated using the THFG and IR camera generally agreed to within 5%. Factors that may have contributed to differences between the results included (1) convective heat flux, which would have influenced the response of the THFG but not the IR camera, (2) re-emission of only part of the heat flux incident on the THFG from the gauge back to the IR camera, and (3) attenuation and emission of radiation by the combustion gases (especially CO2) and atmosphere between the THFG and IR camera. With the last factor mitigated by the presence of an appropriate filter in front of the IR camera, the other factors did not appear to significantly affect the measured heat flux levels, based on the overall agreement of the results from the THFG and IR camera. KEYWORDS Heat flux, infrared camera, sensor, measurement, radiation, convection 1. INTRODUCTION Accurate measurement of heat flux is needed to characterize and improve the thermal efficiency of industrial furnaces. Typically, heat flux measurements have been made using water-cooled probes that are inserted directly into the furnace, influencing the thermal conditions inside the furnace [1, 2]. Two, less intrusive methods have been recently proposed as viable alternatives for the water-cooled probes. These are the infrared (IR) camera  and a total heat flux gauge (THFG) that requires no water cooling . The IR camera detects radiation reflected and emitted by the surfaces being viewed, thereby acting as a radiometer. As such, it requires only optical access to the combustion chamber and can be placed entirely outside the furnace. It can also be calibrated using a blackbody radiation source. Meanwhile, the THFG is designed such that the heat incident on the sensing surface flows through the gauge primarily in one dimension, allowing use of a one-dimensional conduction analysis to estimate the incident heat flux. Direct calibration is not necessary for this sensor, which detects both radiation and convection. Although this probe needs to be located inside the furnace, it measures more accurately the level of convection, and consequently total heat transfer, than a water-cooled sensor . This is because the sensing surface of the THFG is much closer in temperature to its immediate surroundings, thus minimizing perturbation to the thermal conditions inside the furnace, and is more representative of the receiving surface of interest, which is typically not water-cooled. This paper presents heat flux measurements made in a pilot-scale furnace by both the IR camera and THFG. The measurements were conducted as part of an ongoing project to study the performance of different burners in a high CO2 environment. The measurement techniques are first described, followed by presentation and comparison of the results. 2. EXPERIMENTAL SETUP All tests were conducted in the Research Tunnel Furnace at CanmetENERGY. As illustrated in Fig. 1, the test section had an overall length of 2.5 m and was insulated with a ceramic fibre lining to simulate adiabatic surroundings for the flame and minimize thermal effects of the furnace walls. The entrance region near the burner was 0.9 m long and had an inner diameter of 0.9 m, while the remainder of the test section was 1.6 m long with an inner diameter of 0.63 m. A wall of refractory bricks covered with a layer of ceramic fibre insulation delineated the end of the test section. A water-cooled coil was situated in front of the wall to simulate a thermal load in the furnace. In addition, two access ports of 0.66 m length and 0.10 m height were available for inserting instrumentation into the entrance region of the furnace. They were located diametrically opposite each other along a horizontal plane bisecting the furnace (Fig. 1). When no instrumentation was present in a port, the opening was blocked using four refractory bricks, each of which had a width of 0.16 m and a height slightly smaller than that of the port. Figure 1. Schematic of interior of Research Tunnel Furnace. 66 cm 13cm 63 cm 250 cm Thermal load Insulating Refractory material brick wall Flow of combustion gases to stack Burner Access port 54 cm Location of heat flux gauge 90 cm 90 cm The burner was the Near Field Aerodynamics (NFA) burner of the International Flame Research Foundation (IFRF), sketched in Fig. 2 [6, 7]. This burner featured a moveable block swirl generator, which contained a set of alternating fixed and moveable blocks that introduced a swirl component to the flow of oxidizer. This flow of oxidizer passed through an annular region surrounding a gas "gun" located along the burner axis. Fuel flowed through the gun and was injected into the swirling flow through 30 small holes surrounding one end of the gun. The fuel and oxidizer then mixed together in the burner quarl. The axial position of the gas gun was such that the fuel injection holes were aligned with the entrance plane of the burner quarl, as shown in Fig. 2. For the present experiments, the swirl setting was fixed at 70%. Figure 2: Schematic of NFA burner with moveable block swirl generator. Tests were performed using natural gas and two different oxidizers. One oxidizer was air, injected at a level of 3% above stoichiometric, while the other oxidizer was a mixture of oxygen (O2) and carbon dioxide (CO2). The ratio of O2 to CO2 was selected to represent the case in which all the nitrogen in the air was replaced (on a volume basis) by CO2. In all tests, the firing rate was kept constant at 200 kW. Various instrumentation were used to characterize the conditions inside the combustion chamber . These included a suction pyrometer for measuring gas temperature and a water-cooled probe for measuring gas composition. This paper will focus on the THFG and the IR camera, which are described in the following sections. 2.1 Total Heat Flux Gauge The THFG was based on the design of the Directional Flame Thermometer [9, 10]. It contained a 159 mm by 83 mm Inconel 601 sensor plate of 9.5 mm thickness. This size of plate was selected so that the gauge would fit in the space occupied by one of the refractory bricks located in the access port of the furnace. Prior to any testing, the Inconel plate was oxidized at 1000ºC for 24 h in order to obtain stable surface properties. The plate was then instrumented with thermocouples as shown in Fig. 3. A Type S (Platinum vs. Platinum-10% Rhodium) thermocouple was mounted on a spring-loaded mechanism so that it pressed up against one side of the plate at the centre. This thermocouple, made from 24-gauge (0.51 mm) wire beaded to form an exposed junction, provided an estimate of the temperature at the centre of the plate surface. A metal-sheathed, ungrounded, 24-gauge, Type K (Chromel-Alumel) thermocouple was also positioned against the surface of the Inconel plate, as close as possible to the Type S thermocouple, to provide a redundant measurement of temperature at the centre of the plate surface. Another Type K thermocouple was then positioned near the first two thermocouples, at a distance of 13 mm away from the plate surface. A layer of 13 mm thick, ceramic fibre blanket insulation (of the same size as the plate) was placed between the latter thermocouple and Inconel plate. Thus the temperature on both sides of the insulation layer could be obtained, permitting calculation of the heat flux conducted through the insulation. Three additional layers of insulation were placed on top of the first layer of insulation in order to minimize any radiative and convective effects on the thermocouple measurements. Figure 3. Total heat flux gauge (without blanket insulation). The total heat flux incident on the sensor plate could be estimated using the thermocouple measurements and an assumption of one-dimensional conduction through the gauge. Under steady-state conditions, the energy balance of the sensor plate may be written as Eq. 1, where α is the absorptivity of the sensor plate, Qrad is the radiant heat flux incident on the plate, Qconv is the convective heat flux to the plate, Qinsul is the heat flux conducted through the insulation and Qemit is the radiant flux emitted from the exposed (non-insulated) surface of the plate. Equation 1 can be rewritten as Eq. 2, where kinsul is the temperature-dependent conductivity of the insulation, Tplate is the temperature measured by the thermocouple (Type S or Type K) situated against the plate surface, Tinsul is the temperature measured by the Type K thermocouple located in the insulation, x is the distance separating the above thermocouples (13 mm), ε is the emissivity of the sensor plate (assumed to be the same as α = 0.85 ) and σ is the Stefan-Boltzmann constant. Equation 2 provides an estimate of the total heat flux incident on the sensor plate. Inconel plate Type K thermocouple Type S thermocouple Type K thermocouple Spring-loaded mount rad conv insul emit αQ +Q = Q +Q (1) ⎟ ⎟⎠ ⎞ ⎜ ⎜⎝ ⎛ + − + = 1 4 plate plate insul insul conv rad T x T T Q Q k εσ α α (2) For non-steady-state conditions, the measured temperature-time histories can be reduced to heat flux using one-dimensional inverse heat conduction codes, which are available commercially (e.g. IHCP1D of Beck Engineering Consultants Company, Okemos, MI). A detailed description of this type of analysis is available elsewhere [5, 10, 12] and was not used here. With both methods of data analysis (steady-state and transient), no direct calibration of the heat flux gauge is required. 2.2 Infrared Camera The IR camera used in this investigation was the Merlin MID, a mid-wavelength infrared, high-performance camera manufactured by Indigo Systems of Santa Barbara, CA. Details of this camera were previously reported in Ramadan et al.  and are summarized below. The available spectral range of the camera was 1 to 5.4 μm and the thermal sensitivity (noise equivalent differential temperature) was 0.025ºC. An internal cold filter set the operating spectral range to 3 to 5 μm. The camera contained a focal plane array of 320 × 256 sensors. The frame rate and rate of digital output of the camera were 60 Hz. Parameters that had to be input into the camera software prior to recording any images included the ambient temperature, ambient relative humidity, reflected temperature (i.e. apparent temperature of radiation being reflected by the object surface), temperature and transmittance of external optics (e.g. filters), distance between the object and camera, and object emissivity. The camera was calibrated using a blackbody (BB) source. This procedure was described in Ramadan et al. , with relevant details reported here. The BB source was the MIKRON M330, which contained a cylindrical cavity with an effective emissivity of 0.99. The BB source had a manufacturer-specified accuracy of ±0.25% of reading ±1°C, for temperatures above 600°C. The IR camera was calibrated from 600ºC to 1600ºC by taking images of the BB cavity at 100ºC intervals through that temperature range. Different sets of calibrations were performed for different combinations of external filters placed in front of the IR camera. The filters used in this study included two bandpass filters centred at 3.9 μm, referred to in this paper as CO2,BMV and CO2,F, and a zinc selenide (ZnSe) infrared filter. The IR camera does not measure the temperature of a surface, but instead responds to the thermal radiant energy incident on the detector. Due to the presence of an internal filter in the camera, only the energy emitted by the BB source between 3 and 5 μm reached the detector. (When an external filter was used, this range could be limited even further.) The raw signal output from the camera took the form of digital counts produced by the system's A/D converter . In each calibration image of the BB cavity, the average number of counts over the central portion of the image (corresponding to the inside of the cavity) was determined. This number was then expressed as a function of the radiant flux emitted by the BB source. For the present analysis, equations were generated to correlate the number of counts to the radiant flux emitted over all wavelengths (0 to ∞ μm). Figure 4 shows that across a certain range, a quadratic relationship between the number of counts and the radiant flux could be established. (Above this range, saturation of the camera pixels occurred and the signal became invalid .) For comparison, Fig. 5 shows a plot of the number of counts versus the band-limited radiance reaching the IR detector from the BB cavity, after passing through the CO2,BMV filter. (Note that radiance, with units of W/cm2/sr, is the actual parameter detected by the IR camera.) Although the relationship between counts and radiance would not be expected to be entirely linear (since this would result in a measurement of about 475 counts at zero radiance, according to Fig. 5), a linear fit appears to describe the data reasonably well over the measurement range considered. In the present experiments, radiant flux was the parameter of greatest interest (in order to permit comparison with results from the THFG). It may be noted that because radiant flux expresses thermal energy per unit area, the level of flux generated by a BB source can also be generated by a non-blackbody source with a different temperature and effective emissivity. For the purposes of this paper, object temperature was not of direct interest; therefore, object emissivity (which is used to calculate temperature from radiant flux) was not involved. y = 4E-07x2 + 0.0021x - 0.3472 R2 = 0.9998 0 2 4 6 8 10 12 14 16 0 1000 2000 3000 4000 5000 IR camera response [counts] Radiant flux [W/cm^2] Figure 4. Example plot of radiant flux emitted by BB source (over all wavelengths) versus counts output by IR camera with CO2,BMV external filter (data from saturated responses not shown). 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0 1000 2000 3000 4000 5000 IR camera response [counts] Band-limited radiance [W/cm^2/sr] Figure 5. Plot corresponding to Fig. 4, showing band-limited radiance (through CO2,BMV filter) versus counts output by IR camera (data from saturated responses not shown). 2.3 Test Procedure In all tests, the THFG was placed in the access port of the furnace, flush with the inside wall and centred at a distance of 0.54 m from the entrance plane of the furnace test section (Fig. 1). The IR camera was positioned outside the furnace on the side opposite the port containing the THFG, such that it could "see" the sensing (i.e. front) surface of the THFG through the opposing access port and combustion chamber. The camera was located sufficiently far from the furnace (~1 m) that it would not be damaged by heat from the furnace. During each test, the furnace was first allowed to come to steady operating temperature (as monitored by a thermocouple located inside the combustion chamber). The THFG was then inserted into the access port (requiring removal of one refractory brick) and its thermocouple readings allowed to reach steady state. These temperature measurements were subsequently used to determine the total heat flux according to Eq. 2. Next, the four refractory bricks in the access port on the opposite side of the furnace were removed in order to provide optical access by the IR camera to the sensing surface of the THFG. Images inside the furnace were obtained using the IR camera together with one or more external filters, which were inserted manually in front of the camera. The filters were used to minimize effects of emission from the combustion gases along the measurement path and improve the overall quality of the IR images. Once the IR images were captured, the four refractory bricks were replaced in the access port, the THFG in the opposite port removed and replaced with a refractory brick, and the shutdown procedure for the furnace could commence. 3. RESULTS AND DISCUSSION Figure 6 contains a typical image taken by the IR camera, showing the THFG, surrounding refractory bricks and insulation lining the inside of the furnace. Also indicated is the thermocouple used to monitor the temperature inside the combustion chamber. The same image is shown again in Fig. 7, with the colour scale adjusted to have greater resolution over the region of the THFG. Overall, it can be seen that fewer counts were detected near the top left corner and right edge of the THFG sensor plate, in comparison with the centre of the plate. Thus, those outer areas emitted less radiation (by approximately 1 W/cm2 or 9%) than the central region of the plate and were likely affected by heat loss through the edges of the plate. For the present analysis, an average number of counts was determined over a rectangular area slightly smaller than the sensor plate, so that the regions of lower emitted radiation would have minimal influence. This average was then used to determine the corresponding level of radiant flux, as discussed in Section 2.2. The IR results for all tests are shown in Table 1. An uncertainty in the results of 2-3% was estimated based on the standard deviation in the number of counts over the rectangular sampling area in the IR image. This spatial variation was an order of magnitude greater than any temporal variation between images taken with the same filter during each test. On the other hand, the temporal variation may have been more significant when comparing measurements taken with different filters during the same test. Since several minutes were required to manually change filters and the instrumentation port allowing optical access into the furnace (Fig. 1) remained open during this time, measurements taken with different filters could have been affected by changes in the furnace conditions over this period. Nevertheless, any difference due to such changes did not appear to be greater than the 3% uncertainty mentioned above. 410 S 3144 S 500 1000 1500 2000 2500 3000 Figure 6. Typical image taken by the IR camera, looking through the access port and combustion chamber at the instrumentation and insulation on the opposite side of the furnace (May 13th test). Legend indicates the number of counts. Total heat flux gauge Refractory brick Refractory brick Thermo- Insulation couple 2152 S 3144 S 2200 2400 2600 2800 3000 Figure 7. Same as Fig. 6, with colour scale adjusted to increase lower limit cutoff. Table 1. Results from Total Heat Flux Gauge and IR Camera Total Heat Flux Gauge IR Camera Date Burner Oxidizer Incident heat flux, Type S (W/cm^2) Incident heat flux, Type K (W/cm^2) External Filter Radiant Flux (W/cm^2) 2010- 05-26 NFA Air 17.8 18.1 CO2,BMV+ZnSe 18.7 2010- 05-5 NFA Air 18.2 17.9 CO2,BMV 14.8 2010- 05-26 NFA O2/CO2 14.5 14.6 CO2,BMV 14.5 CO2,BMV+ZnSe 14.3 2010- 05-13 NFA O2/CO2 14.3 14.6 CO2,BMV 14.4 CO2,BMV+ZnSe 13.8 CO2,F+ZnSe 14.5 In addition to the results from the IR camera, Table 1 shows heat flux results from the THFG, determined using Eq. 2. Two heat flux estimates were made during each test, one based on the Type S thermocouple located against the sensor plate together with the Type K thermocouple located in the insulation and the other based on the two Type K thermocouples (Fig. 3). These will be referred to as the "Type S results" and "Type K results," respectively. The uncertainty in the heat flux values was estimated to be approximately 0.8% for the Type S results and 2.5% for the Type K results, based on combining the standard calibration uncertainties of the thermocouples (0.75% for Type K thermocouples and 0.25% for Type S thermocouples) using the root-sum-square method . Given that the 2.5% uncertainty in the Type K results was equivalent to approximately 0.4 W/cm2, the difference between the Type K and Type S results (up to 0.3 W/cm2) was within the estimated measurement uncertainty and was not considered significant. Also, the results from repeat tests conducted with the same oxidizer on different days largely agreed with each other within the corresponding measurement uncertainty, indicating good repeatability of the data. Table 1 shows that the heat flux results from the THFG and IR camera generally agreed to within 5%. For the tests involving air, the results from the IR camera appeared to be less consistent (repeatable) than those from the THFG; therefore, more data would be required to discern any trend. For the tests involving O2/CO2, the results from both measurement methods were reasonably consistent and in good agreement with each other. Nevertheless, several factors could potentially contribute to differences between the two methods. First, as discussed in Section 2, the THFG responds to both convective and radiant heat flux, while the IR camera responds to only radiant flux. Thus, the THFG would detect higher levels of heat flux than the IR camera as a result of any additional convective contribution, which would occur if the combustion gases surrounding the THFG were hotter than the gauge surface. The overall agreement of the IR and THFG results in Table 1 suggests that convective heat transfer to the THFG was not significant in the present experiments. Second, the radiation emitted and reflected by the THFG to the IR camera would be expected to be slightly lower than the heat flux measured directly by the THFG. This can be seen from Eq. 3, which is a rearrangement of Eq. 1, with α substituted by (1-ρ), where ρ is the reflectivity of the sensor plate. The first two terms on the right side of Eq. 3 represent the "true" total heat flux incident on the sensor plate1 and would be slightly lower than the total flux calculated using Eq. 2 because Qconv is not divided by α, which is typically less than 1. The last term on the right side of Eq. 3 represents an additional heat "loss" from the sensor plate that is conducted through the insulation and is not part of the heat emitted or reflected by the sensor plate. (Not included is any extra heat loss occurring through the edges of the plate, as discussed earlier in relation to Fig. 7.) Due to the above factors, the radiation (radiosity) detected by the IR camera from the THFG would be less than the total heat flux incident on the THFG. However, neither Qconv (as previously mentioned) nor Qinsul appears to be significant. For the steady state conditions corresponding to the THFG results in Table 1, Qinsul was between 0.1 and 0.2 W/cm2, on the order of 1% of the total incident flux. emit rad rad conv insul Q + ρQ =Q + Q −Q (3) Finally, the radiation from the THFG, as described by Eq. 3, would be further attenuated by the combustion gases in the furnace and the atmosphere outside the furnace before reaching the IR camera. However, such attenuation would be countered by emission of radiation from the atmosphere and combustion gases themselves. With regard to the latter, it should be noted that the presence of a filter centred at 3.9 μm in front of the IR camera would have greatly reduced the contribution of CO2 absorption/emission to the 1 In reality, the total incident heat flux should be the sum of Qrad and Qconv, rather than the sum of Qrad and Qconv,/α, as in Eq. 2. However, in order to calculate Qrad + Qconv from the expression on the right side of Eq. 2, one would need to measure separately the portion of convection in the total heat flux. Despite this, Eq. 2 can be used to estimate total heat flux without significant error in cases where radiation is expected to be much more important than convection. radiation detected by the camera. Thus, based on the overall agreement of the IR and THFG results, effects of the intervening gases was likely negligible in these tests. 4. CONCLUSIONS The THFG and IR camera were both used to measure heat flux in a pilot-scale furnace. The overall agreement (typically within 5%) of the results from the two methods indicates that both techniques were viable for estimating heat flux. However, several factors potentially caused differences between the heat flux levels measured using the different methods. Unlike the THFG, the IR camera did not respond to any incident convective heat flux. The radiosity (reflected and emitted radiation) detected by the IR camera was not equivalent to the heat flux detected by the THFG because some of the heat flux incident on the sensor plate of the THFG was conducted through the insulation and was not re-emitted by the sensor plate back to the IR camera. Further, the radiosity from the THFG was attenuated by the intervening combustion gases and atmosphere before reaching the IR camera. Although effects on the IR results of CO2 absorption and emission could be minimized by using an appropriate optical filter, this was not the case for the THFG, which responded to radiation from all sources within its field of view. Nevertheless, the close agreement of the heat flux results from both methods suggests that none of the above factors was significant in the present tests. REFERENCES  J. Chedaille and Y. Braud, Industrial Flames, Vol. 1, Measurements in Flames, Edward Arnold, London, 1972.  J.P. Vizioz, Techniques and instruments for flame measurements used regularly at the Experimental Station at IJmuiden, IFRF Doc. No. K20/a/47, 1969.  O.B. Ramadan, P.M. Hughes, C.S. Lam, R.J. Lacelle, J. Wong and R. Lycett, 2011, "Incident Radiant Flux Measurements in Pilot-Scale Burners Operating in a High CO2 Environment," 9th European Conference On Industrial Furnaces and Boilers (INFUB), Estoril, Portugal, 26-29 April 2011.  C.S. Lam , O. Ramadan, J. Wong, R. Lycett and P.M. Hughes, "Heat Flux Measurement in Industrial Furnaces," 2010 Spring Technical Meeting, Combustion Institute - Canadian Section, Ottawa, Ontario, 9-12 May 2010.  C.S. Lam and E.J. Weckman, Fire and Materials, 33 (2009) 303-321.  W. Leuckel, Swirl intensities, swirl types and energy losses of different swirl generating devices, IFRF Doc. No. G 2/a/16, 1966.  H.L. Wu and N. Fricker, An investigation of the behaviour of swirling jet flames in a narrow cylindrical furnace, IFRF Doc. No. K 20/a/61, 1972.  P.M. Hughes, O. Ramadan, C. Lam, P. Gogolek, J. Wong, R. Lacelle and R. Lycett, "Industrial Furnace Performance Under Carbon Capture Operation", 2010 Pacific Rim Symposium, American Flame Research Committee, Maui, Hawaii, 26-29 September, 2010.  N.R. Keltner, "Directional Flame Thermometers - A Tool for Measuring Thermal Exposure in Furnaces and Improving Control," Interflam 2007, Interscience Communications, London, UK, 3-5 September, 2007. CD version only.  N.R. Keltner, J.V. Beck and J.T. Nakos, Journal of ASTM International, 7:2 (2010) Paper ID JAI102280.  N.R. Keltner, Personal communication, Fire Instrumentation Research and Engineering Services, Albuquerque, New Mexico, December 2009.  J.V. Beck, B. Blackwell and C.R. St. Clair, Jr., Inverse Heat Conduction: Ill-Posed Problems, Wiley-Interscience, New York, 1985.  FLIR Systems, The Ultimate Infrared Handbook for R&D Professionals, http://www.flir.com/uploadedFiles/Thermography/00_Landing_pages/Direct_ Sales _Landing_pages/RD-Handbook.pdf, downloaded July 2012.  J.P. Holman, Experimental Methods for Engineers, 7th edition, McGraw-Hill, Boston, 2001.