Coal-gangue recognition for top coal caving face based on electromagnetic detection

Abstract

Identifying the coal-gangue mixing ratio during top coal caving is essential for automating the coal caving process efficiently. In this article, a block impression reconstruction method is proposed to create 3D models of coal-gangue mixtures with varying gangue ratios and morphological distributions, based on real coal-gangue blocks that reflect actual coal-falling conditions. These 3D models are then input into CST software for electromagnetic forward simulation. The relationship between electromagnetic signal propagation characteristics and gangue ratio is analyzed, resulting in the creation of a coal-gangue mixture electromagnetic signal dataset. An Optuna-XGBoost-based model is then designed to identify the coal-gangue mixing ratio and the recognition performance is firstly verified by using the electromagnetic forward simulation data. Finally, to verify the method’s practicality, a microwave detection test bench for top coal caving is set up and some comparative experiments are conducted. The experimental results indicate that the electromagnetic signals of coal and rock with different gangue contents exhibit significant differences, and the proposed coal-gangue identification model has significant advantages in accuracy and overall performance compared to other competing models.