A novel soot sizing method based on the optimized multi-output support vector machine

Optical method is widely used for soot sizing due to its advantage of high sensitivity and non-contact, but prior knowledge of refractive index and morphology parameters (MPs) are crucial to retrieve the particle size distribution (PSD) of soot. However, it is difficult to measure the refractive index and MPs on-line and in-situ. In this study, a new method is proposed to address this challenge using light scattering angular spectrum (LSAS) and optimized machine learning. The LSAS is utilized to describe the distribution of scattering light intensity that corresponds to different observation angles, and can simultaneously characterize the PSD, refractive index and MPs. Meanwhile, a compacted and miniaturized prototype sensor was meticulously engineered and tested by diverse types of particle samples, where the Kullback-Leibler divergence (DKL) of PSD is ranged from 0.05 to 0.22. The experiment results indicate that the proposed method can provide a unique ability for high precision measurement of soot PSD, and show significant potential for soot analysis on-line and in-situ measurement.