Glass waste analysis and differentiation by laser-induced breakdown spectroscopy associated to support vector machine: The influence of color and particle size

Glasses are widely known for their unique properties, but improper disposal poses several environmental challenges. Laser-Induced Breakdown Spectroscopy (LIBS) has emerged as a promising tool for glass characterization. This study explored the performance of LIBS, associated with machine learning methods and spectral angle mapper, to overcome the matrix effect in glass waste analysis. Using 10-fold cross-validation, an accuracy of 99.04% was achieved in color differentiation, 98.82% in the differentiation of flint glasses from the other glasses, 96.75% in differentiating particle sizes. When particle size and color were analyzed simultaneously, the accuracy remained high at 97.64%. The analytical accuracy was further improved using the spectral angle mapper method, which allowed us to achieve lower standard deviations, particularly for samples of larger particle sizes. These findings highlight the potential of LIBS combined with machine learning to perform a robust analysis of glass waste, contributing to recycling management.