Improving quantitative analysis of spark-induced breakdown spectroscopy: Multivariate calibration of metal particles using machine learning.

We have recently developed a low-cost spark-induced breakdown spectroscopy (SIBS) instrument for analysis of toxic metal aerosol particles that we call TARTA (toxic-metal aerosol real time analyzer). In this work, we applied machine learning methods to improve the quantitative analysis of elemental mass concentrations measured by this instrument. Specifically, we applied least absolute shrinkage and selection operator (LASSO), partial least squares (PLS) regression, principal component regression (PCR), and support vector regression (SVR) to develop multivariate calibration models for 13 metals (e.g., Cr, Cu, Mn, Fe, Zn, Co, Al, K, Be, Hg, Cd, Pb, and Ni), some of which are included on the US EPA hazardous air pollutants (HAPS) list. The calibration performance, adjusted coefficient of determination (R) and normalized root mean square error (RMSE), and limit of detection (LOD) of the proposed models were compared to those of univariate calibration models for each analyte. Our results suggest that machine learning models tend to have better prediction accuracy and lower LODs than conventional univariate calibration, of which the LASSO approach performs the best with R > 0.8 and LODs of 40-170 ng m at a sampling time of 30 min and a flow rate of 15 l min . We then assessed the applicability of the LASSO model for quantifying elemental concentrations in mixtures of these metals, serving as independent validation datasets. Ultimately, the LASSO model developed in this work is a very promising machine learning approach for quantifying mass concentration of metals in aerosol particles using TARTA.