Rapid classification of steel via a modified support vector machine algorithm based on portable fiber-optic laser-induced breakdown spectroscopy

Abstract

The littering of steel waste increases annually, and it causes environmental pollution and resource wastage. Rapid classification of steel is a critical step in the process of recycling to conserve resources. The method of principal components analysis (PCA) combined with support vector machine (SVM) was developed to establish a model for rapidly classifying 14 kinds of special steel samples, whose spectra were acquired via a portable fiber-optic laser-induced breakdown spectroscopy (FO-LIBS) system. Fifty-one preselected characteristic lines of trace elements were chosen as input variables because the samples of special steel differed in terms of the concentration of trace elements. Results showed that the recognition accuracy of the PCA-SVM model was gradually improved by the increase in principal components (PCs) and reached 100% when 13 PCs were extracted as input; this accuracy value was significantly higher than the 95% accuracy of the SVM model. This finding suggests that FO-LIBS combined with the PCA-SVM algorithm can achieve rapid classification of steel materials and provide a new approach for online detection in the industrial field.