Research on intelligent implementation of the beneficiation process of shaking table

Abstract

This study utilized image processing and machine vision techniques to extract the concentrate zoning image features of the shaking table in order to assess the concentrate grade and recovery rate and achieve intelligent ore access. Although the deep learning model can properly extract the coordinate information of the concentrate zoning border line separating points, the lack of obtained zoning image features prevents an accurate prediction of the concentrate grade and recovery rate. Therefore, this study developed deep learning semantic segmentation methods using HALCON to successfully extract multi-dimensional zoning information and determine the mapping relationship between zoning attributes, concentration grade, and recovery rate by combining machine learning models. To achieve intelligent implementation, a system consisting of an AGV walking system, an image recognition system, an industrial automation platform, a data processing framework, and a 5-DOF robot with a communication network was developed to continuously and in real-time optimize the processing indicators and operational parameters of the shaking table. Finally, the beneficiation process of the shaking table was intelligently optimized.