Acoustic Emission for In Situ Monitoring of Solid Materials Pre-Weakening by Electric Discharge: A Machine Learning Approach

Pre-weakening of solid materials using electric discharge is a new technique aiming at reducing significantly the costs and energy consumption as compared with the traditional raw materials processing in mining and recycling industries. However, the absence of an effective pre-weakening process monitoring and control prohibits its introduction into the market. The present contribution aims to fill this gap by investigating the feasibility of combining acoustic emission with machine learning for process monitoring. Hence, this paper is a supplement and enrichment of existing studies on in situ and real-time process monitoring and diagnosis associated with failure mechanism problems. Three categories and six subcategories are defined to describe the major pre-weakening scenarios of solid materials. The acoustic signals are collected and labeled according to the visual control of specially prepared transparent samples subjected to discharge exposure. The acoustic signals are decomposed with data adaptive $M$ -band wavelets and the relative energies of the extracted frequency bands are used as features. Principal component analysis is applied to select the most informative features whereas several classifiers are applied to recognize the pre-weakening quality. The classification accuracy of the defined categories ranges between 84–93% demonstrating the applicability of the proposed method for in situ and real-time control of pre-weakening of solid materials using electric discharge.