Rapid Identification and Classification of Metal Waste by Laser-Induced Breakdown Spectroscopy

Abstract

The rapid identification and classification of metal garbage has been experimentally investigated. By combining laser-induced breakdown spectroscopy (LIBS) and machine learning, metal garbage can be effectively identified through spectral analysis. In this work, a novel method for metal garbage classification was developed, and a LIBS system was self-developed. As an example of metal recycling, five types of metal were adopted. Several characteristic lines of Al, W, Fe, Cu, Sn, Pb, and C were identified. For a more effective classification, principal component analysis was conducted to reduce the dimension of the spectra. Samples after the dimension reduction were classified by using K-nearest neighbors, and five types were obtained, exhibiting a final classification accuracy of 97.18%. Moreover, a mathematical model of the linear formulas between spectrum and concentration was established to achieve quantitative analysis with Fe taken as an example, laying the foundation for more refined classification.