OCR Text |
Show comouter. The computer also controls the dye laser and monchromator scans. A tyDical CARS sDectrum takes aporoximately 30 minutes to acquire, with 600 frequency data points and 30 laser shots averaged at each point. The measurements were performed in coal-seeded methane-air flames stabilized on a stainless steel honeycomb burner. Two sets of stainless steel tubes carry methane and coal through the honeycomb structure, and air flows around the tubes. Fuel and air mix at the upoer surface of the honeycomb. The burner is 100 mm long by 25 mm wide. The flame gases are enclosed by a chimney with quartz windows. The laser beams traverse the 100 mm length of the burner, thereby maximizing any effects, such as absorption, which arise from oropagation through the medium. Coal types and sizes are given in the Figures. RESULTS AND DISCUSSION Measurements of oxygen spectra in coal-seeded flames showed significant spectral interference when high pulse energies were used (26 mJ total pumo and 5 mJ Stokes). At these energies, many random lines were observed in the spectrum in coal-seeded flames, Fig. 1(a), which were not observed in the clean flame spectrum, Fig. 1(b). The appearance of these random lines seemed to be correlated with laser-induced coal particle breakdown, which was observed to occur at random positions along the laser beam oaths. Fig. 1(c) shows an oxygen spectrum obtained in the coal-seeded flame with lower pulse energies (8 mJ pump and 1 mJ Stokes). The occurrence of breakdown was much less frequent, and the spectrum is much less oerturbed. Some random lines are still noticeable, however. A conditional sampling system has been developed to reject those laser shots on which breakdown occurs. The laser beam path is imaged with a demagnification factor of 5 onto a 10 mm horizontal slit. A photomultiolier is placed behind the slit and the photomultipiier signal is amplified and digitized using a gated digitizer. Interference filters are used to reject scattered 532 nm laser light, and time gating of the detection electronics is used to discriminate against particle luminosity. A threshold value for data rejection is set in the data acquisition program. The conditional sampling system has not yet been tested in oxygen CARS measurements, although it is being used in nitrogen CARS measurements. Because the nitrogen CARS signal is approximately 100 times stronger than the oxygen CARS signals, spectral interferences due to coal particle breakdown will be much less significant. Nevertheless, oreliminary results indicate that the use conditional samoling does reduce the occurrence of random lines in the nitrogen spectra. A nitrogen CARS soectrum obtained with high laser pulse energy is shown in Fig. 2. The spectrum shows a few random lines which do not appear in the clean flame, but the number of such lines is reduced relative to the case where conditional sampling is not used. Temoeratures are obtained from nonlinear least squares fitting of theoretical to measured nitrogen CARS spectra, with temperature used as a fitting parameter. The fits of calculated to experimental nitrogen spectra are yery good, as is evident from Fig. 2. The estimated accuracy of the nitrogen CARS temoerature measurements is ±100 K, with most of the uncertainty arising from variation in flame conditions and/or window transmission during the CARS scan. Such systematic errors will be eliminated in the near future when broadband CARS is implemented; broadband CARS spectra can be acquired in a few seconds. |