Stable Variable Selection Method and Comparison for Quantitative Analysis of Steels Using Laser-Induced Breakdown Spectroscopy

Abstract

In this work, a stable variable selection method based on variable stability correction (VSC) and modified iterative predictor weighting-partial least squares (mIPW-PLS) is proposed for the quantitative analysis of steel samples by laser-induced breakdown spectroscopy (LIBS). This method takes stability as the variable selection criterion, which has strong adaptability to the quantitative analysis of different sample partitions of the steel data set. To demonstrate the feasibility and effectiveness of the proposed method, we compared the successive projections algorithm (SPA) and uninformative variable elimination (UVE) to detect nine sample partitions of different elements in steel. The experimental results showed that the VSC-mIPW-PLS algorithm could achieve credible quantitative analysis accuracy for nine sample partitions of three elements. The root mean square errors of prediction (RMSEP) were no more than 5.1817 (chromium), 1.9759 (nickel), and 2.5848 (manganese), which proved credible prediction ability. This method has the potential for applications using LIBS spectrometers for industrial field and research experiments.