Imaging Measurements of Soot Temperature and Volume Fraction in Flames

Transcription

1 Paper 070DI-09 Topic: Diagnostics 8 th U. S. National Combustion Meeting Organized by the Western States Section of the Combustion Institute and hosted by the University of Utah May 19-, 013 Imaging Measurements of Soot Temperature and Volume Fraction in Flames Haiqing Guo, Jose A. Castillo, and Peter B. Sunderland Department of Fire Protection Engineering, University of Maryland, College Park, MD 074 New diagnostics are presented that use a digital camera to measure full-field soot temperatures and soot volume fractions in axisymmetric flames. The camera was a Nikon D700 with 1 megapixels and 14 bit depth in each color plane, and was modified by removing the infrared and anti-aliasing filters. The flame considered here was an 88 mm long ethylene/air coflowing laminar jet diffusion flame on a round 11.1 mm burner. Soot temperatures were measured using ratio pyrometry at 450, 650, and 900 nm and deconvolution. These had a range of K, spatial and temporal resolutions of 3 µm and 0 ms, and an estimated uncertainty of ±50 K. Soot volume fractions were measured using laser extinction at 63.8 nm and deconvolution. These had a range of ppm, spatial and temporal resolutions of 34 µm and 167 ms, and an estimated uncertainty of ±10%. The diagnostics were calibrated with a blackbody furnace. The present measurements agree with past measurements in this flame using traversing optics and probes, but they avoid the long test times and other complications of such traditional methods. 1. Introduction Accurate measurements of soot temperature and soot concentration in flames are essential for gaining insight into many combustion processes. These measurements can be performed optically and nonintrusively in flames. Many flames of interest are axisymmetric and optically thin, which simplifies the measurements significantly. Several studies have performed soot pyrometry in axisymmetric flames. Sunderland et al. [17,18] used ratio pyrometry with a photomultiplier tube at 600, 700, 750, and 830 nm, but this required traversing the optics across the flame at each height and wavelength. Gulder and coworkers [9,11,16] used ratio pyrometry with a spectrometer and imaged the spectra with a charge coupled device (CCD). Again, traversing the burner horizontally at each height was required. Faeth and co-workers [4,19] used grayscale CCD video cameras to perform ratio pyrometry in microgravity flames, but the cameras had a low bit depth of 8 and low pixel counts. Long and coworkers [,10] used more modern color digital cameras without external filters for three-color ratio pyrometry. Soot volume fraction was also found from the soot emissions. Unfortunately, as shown in Ref. [6], the uncertainties were greater than in narrow-band methods. Soot volume fractions have been measured by many studies in axisymmetric flames using laser extinction and assuming Rayleigh scattering from soot. Santoro et al. [13,14] did so in ethylene/air coflowing diffusion flames. As with the early work in soot pyrometry, single point detectors were used, requiring extensive traversing. Full field soot volume fraction measurements with CCD cameras were reported in [1,4,7,15,19]. Faeth and co-workers [4,19] used a laser diode at 63 nm, but, as in their soot pyrometry work, a camera with a low bit depth and pixel count. Gulder and co-

2 workers [15] used a mercury arc lamp and a more advanced camera. However, arc lamps introduce unsteadiness and collimation difficulties. The use of still digital cameras for combustion diagnostics is increasing [,10,1]. As digital camera technology improves, so too improve the measurements that can be performed. Recent advances in camera technology (including higher bit depth, higher pixel counts, larger sensor arrays, and increased signal/noise ratios) allow nonintrusive full-field measurements in flames with increasing accuracy, speed, and spatial resolution. This study involves the development of full-field diagnostics of soot temperature and soot volume fraction in a steady axisymmetric ethylene/air laminar diffusion flame using a digital singlelens reflex (SLR) camera. The results are compared with past measurements involving single point detectors and thermocouples [13,14].. Experimental The flame considered here is an ethylene/air laminar jet diffusion flame. The burner replicates the coflow burner of [13]. It consists of concentric brass tubes of 11.1 and 101 mm inside diameters. For the coflow, 3 mm glass beads followed by 1.5 mm cell size ceramic honeycomb were used to obtain plug flow. The fuel tube extended 4 mm above the honeycomb. The ethylene and air flow rates were maintained at 4.35 and 856 mg/s (or 3.85 and cm 3 /s at laboratory conditions). Rotameters (calibrated with soap bubble meters) were used to monitor the fuel and air flow rates. The visible flame height was 88 mm, as shown in Fig. 1a. Measurements confirmed that the flame was steady, non-sooting, optically thin, and axisymmetric. A Nikon D700 color SLR camera with a 50 mm f/1.4 AF-D Nikkor lens was used for both soot temperature and soot volume fraction measurements. The camera contains a 36 4 mm complementary metal-oxide-semiconductor (CMOS) sensor with 1 megapixels ( pixels) and 14 bit depth in each of the three color planes. The camera was modified by removing the infrared cut filter, allowing measurements at 900 nm for soot temperatures. The anti-aliasing filter was also removed to improve focus. A long pass filter (Schott WG80) was added to restore matched focusing at the CMOS and the eyepiece. All automatic exposure and image post-processing options were disabled. The aperture was f/1.4 and the ISO was 00. A white balance of direct sunlight was selected. Shutter speed was optimized for each image such that no pixels were saturated in any color plane. The shutter was controlled remotely. Images were initially saved in uncompressed Nikon-specific format. To avoid gamma corrections, the conversion to tif format was performed using Dcraw. The three color planes were flattened to grayscale using arithmetic means. A blackbody furnace (Oriel 6703) was used to calibrate the pyrometer and to confirm linear camera Visible (a) 650 nm (b) 63.8 nm (c) Figure 1. Flame images: (a) color flame image, (b) flattened flame image with 650 nm wavelength filter, and (c) flattened laser plus flame image following subtraction of flattened laser only image.

3 response for the soot extinction diagnostic. The furnace had a 5 mm cavity opening, an emissivity ε of 0.99 ± 0.01, and a temperature accuracy of ± 0.1 ºC Furnace emissions were obtained from Planck s law: hc E, (1) 5 exp( hc / kt f ) 1 where c is the speed of light, E λ is spectral radiance, h is Planck s constant, k is the Boltzmann constant, T f is the furnace temperature and is wavelength. Within the interested temperature range, exp( hc / λkt f ) 1, so that -1 is assumed to be negligible. Images of the furnace at temperatures of ºC were recorded using the Nikon camera with each of the band-pass filters attached to the front of the camera lens. These filters (Newport 0BPF10) were 50 mm square, had central wavelengths of 450, 650, and 900 nm, and had full width at half maximum (FWHM) bandwidths of 10 nm. The lens was focused on the furnace opening, which was 4 cm from the CMOS sensor. The lens focus was adjusted slightly for each wavelength to account for chromatic aberrations. The results of these blackbody tests are summarized in Fig.. The abscissa is: I ( ) E, () 0 where E λ comes from Eq. (1), I is the irradiance incident on the CMOS sensor, and η(λ) is the bandpass transmissivity as provided by the manufacturer. These integrations were performed in Matlab. The ordinate of Fig. is GS/t where GS is the average measured grayscale indicated by the camera near the image center and t is the camera shutter time. The symbols in Fig. correspond to different blackbody temperatures and/or shutter times. Each band-pass filter yielded linear correlations in Fig., with coefficients of determination (R ) of or higher. Because the filter central wavelengths were far separated compared to the band-pass FWHM, Equation () was simplified to I = η E λ Δλ with a top-hat assumption. η and Δλ are the equivalent band-pass transmissivity and FWHM under the tophat assumption. Combining Eqs. (1) and () and the correlation found in Fig. for any two filters yields Eq. (3): 1 5 GS/ t 1SL 1 hc 1 exp 5 GS/ t SL1 kt f d, (3) where the subscripts indicate the filters, SL is the slope found in Fig., and furnace emissivity is assumed to be independent of temperature. For each line pair, η 1 Δλ 1 / η Δλ is a calibration constant and does not vary with temperature. Quantity η 1 Δλ 1 / η Δλ was obtained from Eq. (3) at each furnace temperature and was found to have a mean of 0.94, 0.88, and 0.93 for the 450/650, 450/900, and 650/900 pairs, respectively. For each of these the 95% confidence interval was +/- 0.04, 0.03, and Figure. Grayscale/shutter time versus irradiance incident on the sensor for each filter. 3

4 To obtain soot temperatures, images of the flame were recorded using the 450, 650, and 900 nm filters. The camera was focused on the flame axis, which was 4 cm from the CMOS sensor. The camera focus was adjusted slightly for each wavelength. Figure 1b shows a representative image of the flame using the 650 nm filter after flattening the image to grayscale. Grayscales were first averaged vertically across 0 pixels (0.46 mm) in the object plane. Fourier transforms were then performed with a cutoff frequency of 0.05 pixel -1 to smooth the grayscales. Because the flame was observed to be axisymmetric, grayscales on both sides of the axis were then averaged at each height. Abel deconvolutions were performed for the 450, 650, and 900 nm images using Matlab to convert the line-of-sight projections to radial distributions [] assuming negligible extinction. This assumption was supported by the observation that the maximum extinction by the flame of the 63.8 nm laser was 5%, For optically thick flames, corrections are required to compensate for the self-absorption effect [5,0,1]. These corrections were tested here, but resulted in temperature differences of less than 10 K. The deconvolved grayscales, normalized by camera shutter, were converted to I using the correlations of Fig.. Similar to the derivation for Eq. (3), combining Eqs. (1) and () for any two filters yields the following expression for soot temperature, T, where soot emissivity is assumed to be proportional to λ -n. hc 1 1 k 1 T, (4) n5 1 1 GS/ t SL1 ln GS/ t 1SL 1 Quantity η 1 Δλ 1 /η Δλ in Eq. (4) was obtained from Eq. (3). Various values of n between 1 and 1.38 were tested [8,10,17,18], and because 1.38 was found to give the best agreement between the soot temperatures obtained from the three line pairs this value was used for the results that follow. The uncertainty in the soot temperature measurements is estimated to be ±50 K, with ±0.1 K precision for relative temperatures. Spatial resolution in the object plane is 3 µm. The longest shutter time used was 0 ms. Therefore, although this flame is steady, the diagnostic can also be applied to unsteady axisymmetric flames that are quasisteady on a time scale of about 0 ms. Temperatures were also measured using a thermocouple in soot-free areas. The thermocouple was an uncoated B-type thermocouple (Pt-30% Rh versus Pt-6% Rh) with a wire diameter of 51 µm and a butt welded junction. Radiation correction was performed as in Ref. [1] assuming a thermocouple emissivity of 0.. Measurements were averaged over 10 s at each location. Uncertainty in the corrected thermocouple measurements is estimated to be ±40 K. Soot concentrations were measured with the laser extinction system depicted in Fig. 3. The light source was a 7 mw He-Ne laser (Melles Griot 5LHR171) operating at 63.8 nm. Motivated by Ref. [1], the beam was expanded using two diffuser sets (Thorlabs DG0-0 and DG0-600), the first stationary and the second mounted to a pneumatic vibrator. The vibrator had an amplitude of.5 mm and frequency of 0 Hz and was required to reduce speckle. The beam was collimated to 100 mm using an off-axis parabolic mirror with angle of 30 and a focal length of 30 cm. After the test section, the beam passed a laser line filter at 63.8 nm with 1 nm FWHM (Andover ANDV1564) and a decollimator with a focal length of 5 cm. A neutral density filter with optical density of was used to allow a shutter time (167 ms) longer than the period of the vibrator. A 4

5 Parabolic Mirror Flame Lens Planar Mirror Diffuser Sets Vibrator Laser Line Filter Laser Decollimator Neutral Density Filter Pinhole Lens Camera Figure 3. Schematic of the laser extinction system. 3.8 mm pin hole was used to provide a 0.5 acceptance angle on the optical axis. The camera lens was focused on the object plane. Soot volume fraction was measured for the entire flame using just two images: the flame image with both flame and laser on, and the reference image with only the laser on. Some past studies [1] have also recorded and subtracted images with the flame on and the laser off, but such images here had negligible grayscales because the laser signal was so much stronger than that of the flame emission. Figure 1c shows the subtraction of these images, followed by image was linearly contrast stretched. Small speckle patterns can be observed as a result of the coherent light source despite the use of the vibrator. Subtraction yielded negligible grayscales in the background area appearing black, confirming the steady laser system. Grayscales were extracted from each image and were averaged vertically across 0 pixel (0.68 mm) in the object plane. Fourier transforms were performed with a cutoff frequency of 0.05 pixel -1 to smooth the grayscales. Grayscales on both sides of the axis were then averaged. The laser extinction images were analyzed assuming Rayleigh scattering by soot with a refractive index of m = i, which yields a dimensionless absorption constant, k λ of 4.9 [3]. For soot primary particles smaller than the Rayleigh limit (00 nm at 63.8 nm), this approximation yields [] f s = λa{ln( I λ 0 / I λ )} / k λ, (5) where f s is the local soot volume fraction, I λ is the irradiance measured from the flame image, and I λ 0 is the irradiance measured from the reference image. The symbol A{} denotes the Abel deconvolution operator. Similar to the temperature measurement, a Matlab code was developed to scan the laser extinction images, post-process and deconvolve the grayscales, and determine the soot volume fraction. The uncertainty in the soot volume fraction measurements is estimated to be ±10%, with ± ppm precision for relative soot volume fractions. Spatial resolution in the object plane was 34 µm. The shutter time was 167 ms. Therefore, although this flame was steady, the diagnostic can be applied to unsteady flames that are quasisteady for 167 ms or longer. 3. Results and Discussion Full-field soot temperatures were obtained in the soot containing region with ratio pyrometry. Temperatures from the three line pairs were averaged. The difference between the average temperature and any of the three pairs was less than 50 K. Noises increased in regions with soot 5

6 concentration lower than 0.5 ppm. The difference between the average temperature and any of the line pairs exceeded 50 K and the results were therefore discarded to avoid increased uncertainties. Temperatures below 8 mm height were not measured because not enough soot was present. Figure 4 shows the pyrometry results in the soot containing area at representative heights of 10, 0, 50, and 70 mm. Also shown are previous results of Santoro [14] using rapid thermocouple insertion and present thermocouple measurements at 50 mm in the soot free area. The pyrometry results were in reasonable agreement with Santoro s. The present thermocouple measurements at 50 mm in the soot free area supported Santoro s measurements. Imaging measurements of temperature and soot volume fraction allow the generation of twodimensional contour plots of these quantities. Figure 5a shows a contour plot of the ratio pyrometry measured soot temperatures, overlaid on the stretched color flame image. Temperatures were measured between K. Near centerline, temperatures below 40 mm were not obtained due to insufficient soot concentration. From the flame considered here, this temperature diagnostic requires temperatures higher than 1500 K and soot volume fractions above 0.5 ppm. In cooler regions, our work in other flames has demonstrated capability of measuring temperatures as low as 1000 K. Figure 6 shows soot volume fraction results at representative heights of 15 and 50 mm above the burner. At both heights the soot was concentrated in an annulus. Near the centerline, small ringing patterns (measurements that fluctuated with radius) were observed owing to noise accumulation inherent in the Abel deconvolution, as seen at 50 mm height in Fig. 6. The results shown were limited to soot volume fractions above 0. ppm, to avoid increased uncertainties. At both heights, the measured peak soot volume fraction was slightly lower and closer to the centerline than in Santoro s [13], but is within expected levels of variation between the two flames and other experimental uncertainties. Reasonable agreement was observed. Figure 4. Measured temperatures versus radius at heights of 10, 0, 50, and 70 mm. Figure 5. Contour plot of (a) ratio pyrometry measured temperature in K, and (b) soot volume fraction in ppm. The radial axis is stretched. 6

7 Figure 5b shows a contour plot of measured soot volume fractions, overlaid on the stretched color flame image. Soot volume fractions were not measured below 8 mm because not enough soot was present. Soot volume fractions were found between ppm. The maximum soot volume fraction was observed at a height of 40 mm. The peak centerline soot volume fraction was found at a height of 50 mm. Compared with Fig. 5a, the soot temperature peaks are radially outside those of soot concentration. Conclusions A Nikon D700 SLR camera was used to measure soot temperature and soot volume fraction in a flame. The camera had a 36 4 mm, 3 14 bit depth, 1 megapixel CMOS sensor. The infrared cut filter was removed to image infrared light. The flame was an 88 mm high ethylene/air coflowing laminar jet diffusion flame on an 11.1 mm burner. It was steady, soot containing, optically thin, and axisymmetric. Temperature measurement with ratio pyrometry and deconvolution required three images of at most 0 ms Figure 6. Measured soot volume fractions versus radius at heights of 15 and 50 mm. each with a filter change between, while soot volume fractions measurement with laser extinction and deconvolution required 167 ms. Temperatures were measured between 1500 and 1850 K in the soot containing region, with an estimated uncertainty of ±50 K. Soot volume fractions were measured between 0. and 10 ppm, with an estimated uncertainty of ±10%. Spatial resolution was between 3 and 34 µm. Precision was ±0.1 K for temperature and ± ppm for soot volume fraction. The results were compared with past measurements and reasonable agreement was observed. These diagnostics can also been performed in optically thick flames and in unsteady periodic flames with a temporal resolution higher than 0 ms. Acknowledgements This work was supported by the National Science Foundation (NSF) Grant No. CBET The authors acknowledge assistance from D. Urban and M. Willnauer. This work has also been supported by the Fulbright program under the Panama Bureau of Educational and Cultural Affairs. References 1. Arana, C.P., Pontoni, M., Sen, S., and Puri, I.K., Field measurements of soot volume fractions in laminar partially premixed coflow ethylene/air flames, Combust. Flame 138, (004).. Connelly, B.B., Kaiser, S.A., Smooke, M.D., and Long, M.B., Two-dimensional soot pyrometry with a color digital camera, Joint meeting of the U.S. sections of the Combustion Institute, Philadelphia, PA, USA, March Dalzell, W.H., Williams, G.C., and Hottel, H.C., A light-scattering method for concentration measurements, Combust. Flame 14, (1970). 4. Diez, F.J., Aalburg, C., Sunderland, P.B., Urban, D.L., Yuan, Z.-G., and Faeth, G.M., Soot properties of laminar jet diffusion flames in microgravity, Combust. Flame 156, (009). 7

8 5. Elder, P., Jerrick, T., and Birkeland, J.W., Determination of the radial profile of absorption and emission coefficients and temperature in cylindrically symmetric sources with self-absorption, Appl. Opt. 4, (1965). 6. Fu, T., Cheng, X., and Yang, Z., Theoretical evaluation of measurement uncertainties of two-color pyrometry applied to optical diagnostics, Appl. Opt. 47, (008). 7. Greenberg, P.S., and Ku, J.C., Soot volume fraction imaging, Appl. Opt. 36, (1997). 8. Iuliis, S.D., Barbini, M., Benecchi, S., Cignoli, F., and Zizak, G., Determination of the soot volume fraction in an ethylene diffusion flame by multiwavelength analysis of soot radiation, Combust. Flame 115, (1998). 9. Joo, H.I., and Gulder, O.L., Soot formation and temperature field structure in co-flow laminar methaneair diffusion flames at pressures from 10 to 60 atm, Proc. Combust. Inst. 3, (009). 10. Kuhn, P.B., Ma, B., Connelly, B.C., Smooke, M.D., and Long, M.B., Soot and thin-filament pyrometry using a color digital camera, Proc. Combust. Inst. 33, (011). 11. Mandatori, P.M., and Gulder, O.L., Soot formation in laminar ethane diffusion flames at pressures from 0. to 3.3 MPa, Proc. Combust. Inst. 33, (011). 1. Maun, J.D., Sunderland, P.B., and Urban, D.L., Thin-filament pyrometry with a digital still camera, Appl. Opt. 46, (007). 13. Santoro, R.J., Semerjian, H.G., and Dobbins, R.A., Soot particle measurements in diffusion flames, Combust. Flame 51, (1983). 14. Santoro, R.J., Yeh, T.T., Horvath, J.J., and Semerjian, H.G., The transport and growth of soot particles in laminar diffusion flames, Combust. Sci. Technol. 53, (1987). 15. Snelling, D.R., Thomson, K.A., Smallwood, G.J., and Gulder, O.L., Two-dimensional imaging of soot volume fraction in laminar diffusion flames, Appl. Opt. 38, (1999). 16. Snelling, D.R., Thomson, K.A., Smallwood, G.J., Gulder, O.L., Weckman, E.J., and Fraser, R.A., Spectrally resolved measurement of flame radiation to determine soot temperature and concentration, AIAA. J. 40, (00). 17. Sunderland, P.B., Koylu, U.O., and Faeth, G.M., Soot formation in weakly buoyant acetylene-fueled laminar jet diffusion flames burning in air, Combust. Flame 100, (1995). 18. Sunderland, P.B., and Faeth, G.M., Soot formation in hydrocarbon/air laminar jet diffusion flames, Combust. Flame 105, (1996). 19. Urban, D.L., Yuan, Z.-G., Sunderland, P.B., Linteris, G.T., Voss, J.E., Lin, K.-C., Dai, Z., Sun, K., and Faeth, G.M., Structure and soot properties of nonbuoyant ethylene/air laminar jet diffusion flames, AIAA. J. 36, (1998). 0. Wang, Y., Ding, P., Mu, Y., A spline approximation of the Abel transformation for use in opticallythick, cylindrically-symmetric plasmas, J. Quant. Spectrosc. Radiat. Transfer 34, (1995). 1. Young, S.J., Iterative Abel inversion of optically thick, cylindrically symmetric radiation sources, J. Quant. Spectrosc. Radiat. Transfer 5, (1981).. Yuan, Z.-G., The filtered Abel transform and its application in combustion diagnostics, Western States Section of the Combustion Institute, Stanford, CA, USA, October