PREPROCESSING OF LASER-INDUCED BREAKDOWN SPECTRA OF LOW ALLOY STEELS AND CAST IRONS IN THE ANALYSIS USING PARTIAL LEAST SQUARES REGRESSION

Abstract

Based on a set of LIBS spectra of low-alloy steels, regression models were constructed for the analysis of manganese, chromium, nickel, copper, silicon, vanadium, titanium and aluminum using projection onto latent structures. The spectra were recorded in the spectral range of 288–325 nm with a resolution of ~0.04 nm. The laser plasma was excited in a collinear two-pulse excitation mode at a wavelength of 1064 nm. The efficiency of various methods of spectrum preprocessing (normalization to the base line, localization of the spectral range, addition of non-linear components of the spectrum), which allow improving the accuracy of regression models, was studied. The use of the optimal of the considered preprocessing methods made it possible to improve the standard deviation of the analysis results for the test sample in the range from 1.8 times for vanadium to 6.8 times for silicon.