PLIF AND IMAGE CHEMILUMINESCENCE
4.11 Image Chemiluminescence Results
4.11.1 Reduced Image Chemiluminescence Data
Chemiluminescence experiments were performed for the seven base drive frequencies.
The reduced image chemiluminescence data is presented in its entirety in appendix H. Figure 4-35 shows a sample chemiluminescence image of the flame under steady-state conditions (no acoustic
Figure 4-35: Chemiluminescence image of experimental burner flame under steady conditions (no acoustic forcing). Collected data appears on the left, and is mirrored across the centerline for a visual representation of entire flame. φ = 0.85, VR = 4.0.
forcing). Again, the horizontal axis represents the radius from the centerline of the burner, and the vertical axis represents the height above the nozzle exit plane. The stagnation body height is at z
= 21 mm, which is shown as the top line of the graph frame. The data has been mirrored across the centerline of the burner to complete the visual appearance. The intensity scale presented has been normalized around the maximum intensity in the image.
It is immediately obvious that the spatial resolution of the image is very poor. In comparison with the previously collected PLIF images, this frame provides very little detail. The vast amount of light collected that did not originate in the “object” plane of the lens has essentially washed away any meaningful features. The flame appears artificially thick, and a large amount of luminescence appears well below the location of the reaction zone.
What cannot be seen in the image due to the color scaling (but is nonetheless present in the data) is the background pedestal that has developed. The background value of the image has risen above the floor established by the ICCD camera. This is caused by the out-of-focus light being distributed all over the focal plane. As a result, any attempt to integrate the intensity to generate global values will require summation over the entire plane.
Drive Combustion Forced Rayleigh Frequency (Hz) Response, H(s) Index, Rf
Mag Phase
55 1053 21.7 979
75 1181 -131.9 -788
100 768 80.1 132
150 134 119.9 -66.6
220 7.72 -150.3 -6.70
265 8.18 -133.5 -5.63
375 3.78 -76.1 0.91
Table 4-6: Global unsteady combustion responses as discerned using the chemiluminescence data for the seven base drive frequencies. Positive phases correspond to the heat release lagging the pressure.
Figure 4-36 shows typical results for the phase-resolved chemiluminescence images. The presented images are for the 55 Hz acoustic drive frequency experiment. Only 12 of the 24 generated images are displayed. The phase angles shown follow the same conventions as described previously.
Following the usual methodology, the intensities of the 24 images for each drive frequency case were fully integrated (as is) and then processed by FFT in order to generate the global combustion response and forced Rayleigh index for the seven acoustic drive frequencies. The results are shown in Table 4-6.
As done in the previous cases, the data points are then plotted on the graphs for the global combustion response function and the global forced Rayleigh index that were generated by bulk chemiluminescence measurements for the relevant test condition in chapter 3. These are shown in Figures 4-37 and 4-38.
Inspecting the combustion response data in Figure 4-37, the agreement between the image chemiluminescence data compared and the baseline bulk chemiluminescence data is good, as might
Figure 4-36: Phase-resolved chemiluminescence images for acoustic forcing at 55 Hz. The horizontal axis is the radius (mm) from the centerline of the burner. The vertical axis is the distance (mm) above the nozzle exit plane. Phase angle θ = 0º corresponds to the rising-edge zero crossing of the local unsteady pressure. Burner operating parameters are: φ = 0.85, VR = 4.0.
Magnitude
Frequency (Hz)
100.0 1000.0
1.0E-2 1.0E-1 1.0E+0 1.0E+1 1.0E+2 1.0E+3
20.0
Phase (deg)
Frequency (Hz)
100.0 1000.0
-2160 -1800 -1440 -1080 -720 -360
20.0
Figure 4-37: Combustion response function - points computed from the image
chemiluminescence data (shown as crosshairs) superimposed on the curves computed from the bulk chemiluminescence experiments. Phase data points were placed on the appropriate branch for closest proximity to the curve. Here, negative phases correspond to I’ lagging p’.
Figure 4-38: Forced Rayleigh index - points computed from the image chemiluminescence data (shown as crosshairs) superimposed on the curves computed from the bulk
chemiluminescence experiments.
Forced Rayleigh Index (Rf)
Frequency (Hz)
100.0 1000.0
-2000.0 -1000.0 0.0 1000.0 2000.0
20.0
be expected. A small shortfall in the magnitude is seen in the four lowest frequency data points.
Nonetheless, these results are noticeably better than those seen in any of the PLIF cases. For the two highest frequency data points, the magnitudes are (at least) in the correct order of magnitude.
Visual examination of the phase plot shows good agreement throughout. Looking closely at the rapidly descending portion of the phase trace where the first five data points lie, it appears that there may be a small frequency offset leading to the data points lying slightly to the right of the baseline curve. It is unclear if this is due to a set-point error or calibration issue in a parameter such as the burner flowrate, or if it is a bias generated by something more sinister – a problem in the image chemiluminescence setup itself that was only discovered after the data were reduced.
(This is discussed below.)
Figure 4-38 shows the forced Rayleigh index plot produced from the bulk chemiluminescence data with the newly generated data points (image chemiluminescence) superimposed. Again, the agreement is good, despite the small frequency/phase offset mentioned above. As would be expected, the agreement in this case is superior to all the previously reviewed PLIF data cases.
The spatially resolved combustion response plots generated from the chemiluminescence image data are shown in the next two figures. Figure 4-39 presents the spatially resolved combustion response function. The magnitude and phase images are shown for all seven drive frequencies with the magnitude plots on the left and the phase plots on the right. Examination of these images reveals the disservice done by not having a cleanly defined image plane. The integration of luminescence over the entire depth of the flame, combined with the fact that most of this light is out of focus, produces images containing far less useful information compared to their earlier discussed counterparts. Some features can be resolved at higher drive frequencies (such as the banded behavior and the phase relation between the bands) but the resolution is generally poor.
Figure 4-39: Spatially resolved combustion response function produced from
chemiluminescence data. Magnitude plots appear in the left column while phase plots are shown on the right. Note that positive phase values correspond to I’ lagging p’. The coordinate system origin is along the burner centerline at the nozzle exit plane. Data for the seven base drive frequencies are shown. (See next page.)
Figure 4-39 (cont.): Spatially resolved combustion response function produced from chemiluminescence data. Magnitude plots appear in the left column while phase plots are shown on the right. Data for the seven base drive frequencies are shown. (See previous page.) Note the change in magnitude scale versus the plots for 150 Hz and lower.
Figure 4-40: Spatially resolved forced Rayleigh index produced from chemiluminescence data. Data for the seven base drive frequencies are shown. Plots for the four lower frequencies use the top magnitude legend. Plots for the three higher frequencies use the bottom magnitude legend.
The corresponding forced Rayleigh index plots appear in Figure 4-40 and are seen to be equally obfuscated. In these images, all real detail is lost – washed away by the out-of-focus light.
Large regions of yellow and light blue appear (indicating driving and damping zones) that are manifestations of the integration over depth. The small structures that can be discerned in the PLIF data are not seen at all. Together these conspire to make the data of minimal use.
As alluded to above, an additional problem was discovered with the image chemiluminescence technique – a problem that remained unknown until the experiments were complete and the data were being processed. The issue centers on the collection of the out-of-focus light. It was initially assumed, despite the fact that most of the luminescence from the flame was out of focus, that all the light would be captured on the focal plane of the ICCD camera. This was verified to be true in the bulk chemiluminescence experiments where the Nikon lens was coupled to a photomultiplier tube. However, in the bulk chemiluminescence case, the aperture number of the lens was set to f/4 resulting in a much smaller aperture area. This was possible due to the superior sensitivity and selectivity of the lock-in analyzer. Additionally, the photocathode on the PMT was larger in size than that of the ICCD photo-intensifier. As a consequence of these two differences, indeed, a majority of the luminescence collected by the lens was deposited on the active area of the PMT photocathode.
For image chemiluminescence, the situation was quite different. To achieve acceptable signal levels it was necessary to open the lens aperture to the full f/1.2 position. This resulted in greater signal, but also distributed the out-of-focus luminescence over a larger area. This was compounded by the smaller area of the intensifier photocathode (versus the PMT) which is only 12 mm x 12 mm. Finally, adding to the consternation was the fact that the flame image was (purposefully) off-center of the optical axis. The optical axis position was held constant through all experiments (PLIF and image chemiluminescence). It was positioned off-center in order to
provide visibility over a larger radial distance (on one side of the flame) while still maintaining the target resolution of 100 um per pixel. The results of these combined issues were the following:
1) Some chemiluminescence signal was lost beyond the edge of the focal plane.
2) The loss of luminescence was not uniformly distributed over the field of view. Out- of-focus light originating near the optical axis was more likely to land on the focal plane than out-of-focus light originating at distances farther away from the optical axis.
3) Since the optical axis was not centered on the flame, the loss of light from the flame is not symmetric about the flame’s centerline. Therefore, cropping the flame’s image along the flame centerline does not necessarily result in a “half-flame” contribution to the fluorescence signal.
Ultimately, these results can lead to errors in the magnitudes and phases used to generate the response functions. It is quite possible that these account (at least in part) for the discrepancies seen previously between the bulk chemiluminescence data and image chemiluminescence data.