Machine learning-coupled multi-elemental fingerprinting for high-accuracy source identification of distillation...

Distillation residues represent a significant category of industrial waste with diverse origins. Illegal disposal of these complex mixtures threatens the environment, yet the absence of universal chemical markers complicates the identification of their sources. This study developed a method integrating multiple metallic elemental characteristics with machine learning to identify enterprises responsible for their production. The strategy utilizes inductively coupled plasma optical emission spectrometry (ICP-OES) to determine the metallic elemental composition of the distillation residues. Principal component analysis (PCA) was applied for data denoising and quality enhancement. After dimensionality reduction, a backpropagation neural network (BPNN) maintained an identification accuracy of 98.18%. Shapley Additive Explanations (SHAP) was employed to interpret the machine learning models and evaluate the importance of feature elements. This study proposed a novel strategy for the rapid and accurate identification of distillation residues sources, thereby demonstrating strong potential for integrating artificial intelligence (AI) into solid waste traceability.