Classification of recyclables using laser-induced breakdown spectroscopy for waste management

With the ever-increasing amount of generated waste governments around the world are looking for, and implementing, ways to minimize waste output and maximize waste recovery. The main difficulties are sorting waste items, identifying the different types of plastics, and the time taken to sort them manually. Bioplastics such as polylactic acid and Novatein thermoplastic protein can be incorporated into the recycling stream to minimize waste. Laser-induced breakdown spectroscopy spectra analyzed by k-nearest neighbor and soft independent modeling by class analogy were investigated as methods that can rapidly identify recyclables. Raw, peak normalized, and total intensity normalized spectra were used to identify which would improve classification. Laser-induced breakdown spectroscopy spectra were generated by single laser shots to different locations on nine samples, glass (brown, green, and clear), tin, aluminum, polylactic acid, Novatein, polyethylene terephthalate, and high-density polyethylene. To prove that the system has the potential to be used on a waste sorting stream an autofocus unit was developed to move the laser-induced breakdown spectroscopy beam into focus on the different sample geometries. Two classification methods were investigated, soft independent modeling by class analogy and the k-nearest neighbors algorithm. k-Nearest neighbors on raw spectra produced the best results. Discrimination between bioplastics and plastics were 100%. Glass samples could not be reliably distinguished from each other. Surface contamination produced three misclassifications from 450 spectra. Similar results were obtained when the spectral range was reduced from 182.26-908.07nm to 313.20-495.12nm.