Improvement of classification accuracy during laser-induced breakdown spectroscopy analysis of non-ferrous similar metals

Abstract

The recycling rate of metal scraps produced on non-industrial sites is very low. In particular, the classification of non-ferrous metals is very important due to their economic effect and high demand. However, the classification of non-ferrous metals is not a good situation at the moment. Therefore, we propose that laser-induced breakdown spectroscopy (LIBS) as an alternative technique to classify them. LIBS technology is considered as a very important technique for metal classification. LIBS technology has several advantages. Multiple elements can be analyzed in real time and there is no sample preparation. And it is an analytical method with little environmental impact. In this study, we investigated the classification of five types of non-ferrous metals (aluminum, stainless steel, copper, zinc and lead) by LIBS. In particular, the emphasis was on the classification of similar metals, not the classification of dissimilar metals. The classification rates for each metal were calculated using algorithms called principal component analysis (PCA), linear discriminant analysis (LDA) and support vector machine (SVM). In addition, we compared the classification results of allogeneic metals according to the algorithm of non-ferrous similar metals to determine the most effective algorithm method. This method was also applied to field samples of copper samples to confirm that the similar metals could be classified. We have found it effective to use a non-linear algorithm when classifying homogeneous metals.