Discussion

The particle distribution functions were reconstructed for particles with radii of 0.5, 1, 1.5, and 2 pm. Figure 5.20 shows the examples of fitting for particles of these radii with deviations of about 0.001—0.01 pm. The mode reconstruction is accurate enough, even with fitting discrepancy and the level of signal-to-noise (S/N) ratio of 30 on average over the spectrum (minimal S/N ratio in some cases achieved 0.5).

Fitting to model spectra obtained by the Monte Carlo technique

Figure 5.20 Fitting to model spectra obtained by the Monte Carlo technique: (a) 1 pm, (b) 1.5 pm, (c) 0.5 pm, and (d) 2 pm.

Experimental spectra and fitting by genetic algorithm

Figure 5.21 Experimental spectra and fitting by genetic algorithm.

For the experimental spectrum with particle diameter of 1 pm and the profile as in Figure 5.21a, the reconstruction yielded the mode with the radius 0.498 pm. For Figure 5.21b, if we take the spectrum averaged over the set of realizations, then two modes with the radii of 0.27 and 0.6 pm were reconstructed.

Thus, particular attention should be paid to the issue of obtaining the initial spectrum. Recommendations include the obtaining of the spatially resolved spectrum before the cloud or the use of two receivers in the bistatic scheme, when one receiver is directed to the aerosol cloud, while another is directed to the space before the cloud. Analysis of possibilities of reconstruction for aerosol distributions with wider half-width and nearly lognormal forms requires further investigations in this field.

Thus, the distribution functions have been reconstructed from experimental data in accord with the narrow particle size distribution of the artificial aerosol formation based on ethylene glycol, and the possibilities of reconstruction of narrow distributions with a mode in the range 0.5—2 pm with the genetic algorithm have been studied.

 
Source
< Prev   CONTENTS   Source   Next >