Partial least squares regression as novel tool for gas mixtures analysis in quartz-enhanced photoacoustic spectroscopy

Gas mixtures analysis is a challenging task because of the demand for sensitive and highly selective detection techniques. Partial least squares regression (PLSR) is a statistical method developed as generalization of standard multilinear regression (MLR), widely employed in multivariate analysis for relating two data matrices even with noisy and strongly correlated experimental data. In this work, PLSR is proposed as a novel approach for the analysis of gas mixtures spectra acquired with quartz-enhanced photoacoustic spectroscopy (QEPAS). Results obtained analyzing CO/N2O and CH4/C2H2/N2O gas mixtures are presented. A comparison with standard MLR approach highlights a prediction errors reduction up to 5 times.