Automated classification of clay suspension using acoustic sensing combined with convolutional neural network

Abstract

The dewatering process of fine-containing mine tailing is challenging because of the low settling velocity of fine suspension. Therefore, a long-term reliable monitoring technique during the dewatering process is required to ensure that the stabilization of fine-containing mine tailings is achieved. This study assessed the potential use of acoustic sensing in monitoring the dewatering process by developing an automated classification of clay suspension. The measured backscattered signals of three clays (kaolinite, illite, and bentonite) with three clay concentrations (0.1, 1, and 5 g/L) were measured before developing a classification model using a convolutional neural network. The high accuracy of the CNN model shown in this study indicates the possibility of using low-cost easy-to-measure acoustic sensing in the classification of fine mineralogy and fine concentration for monitoring the dewatering process. Particularly, the highest accuracy for predicting clay concentration of 5 g/L indicates that the proposed framework can predict the low fine concentration of fine suspension during the dewatering process.