Quantitative Analysis of Steels using PLS with Three Data Reduction Methods based on LIBS

Three methods, selecting characteristic lines of elements contained in the samples manually, selecting intensive spectral partitions manually and the whole spectra, were used to reduce dimensions of spectra of 27 steel samples acquired by Laser-Induced Breakdown Spectroscopy. The PLS models were built based on the data after dimension reduction to quantify the Mn concentration of samples. The results show that, PLS model built based on selecting intensive spectral partitions can achieve the best result with the least complexity and the highest generalization ability. Selecting intensive partitions is a promising solution to reduce dimensions for original spectra.