Fast identification and characterization of residual wastes via laser-induced breakdown spectroscopy and machine learning

Elemental composition and heating value are essential properties of residual wastes (RW) for its energy utilization. This paper proposed a highly efficient approach to distinguish inorganic components and characterize organic compounds in RW via laser-induced breakdown spectroscopy (LIBS) and machine learning (ML) models. LIBS data of various RW samples were collected to train and test the hybrid model, which includes a data pretreatment module, a classification module and a regression module. Impacts of different ML model categories and parameters were investigated and discussed. Under optimal conditions, the accuracy for predicting C content, H content, O content and lower heating value reached 96.70%, 92.21%, 87.11% and 94.28%, respectively. The robustness of this system was validated. The future application of the model and their limitation were also discussed. This method provides innovative technical ideas for the identification and characterization of RW, and has important potential value for the energy treatment and utilization of RW.