Coal–Rock Interface Recognition Based on Permutation Entropy of LMD and Supervised Kohonen Neural Network

Abstract

Owing to the difficulty in coal-rock interface recognition during the process of coal mining, the shearer is damaged at a high frequency. To avoid this problem, a method is proposed for coal-rock interface recognition based on permutation entropy calculated using the local mean decomposition (LMD) method and supervised Kohonen neural network (SKNN) by performing sound signal analysis. The complex and nonstationary sound signal is adaptively decomposed by LMD. Given that the decomposed product function (PF) components contain the main information of the features, permutation entropy (PE) is used to reflect the complexity and irregularity in each PF component and is defined as the input of the SKNN model. Finally, the optimal SKNN model is obtained by training the samples. The experimental results show that the comprehensive recognition rate of a coal-rock interface is up to 89%. A coal-rock interface can be recognized effectively by sound signal analysis.