Researchers used machine learning-assisted laser-induced breakdown spectroscopy (LIBS) to classify PP, PET, HDPE, and LDPE under three laser-energy conditions designed to mimic variations in laser irradiance, training the models on standard plastics and testing them on physically separate real-waste specimens. Combining preprocessing with wavelength selection raised the best test accuracy from 0.617 to 1.00, although that result was based on repeated measurements of only four held-out waste specimens, one per polymer class.
