Fusion Recognition of Shearer Coal-Rock Cutting State Based on Improved RBF Neural Network and D-S Evidence Theory

Accurate recognition of coal-rock cutting state is a prerequisite for intelligent operation of shearer, so as to achieve safe and efficient production in coal mines. This paper takes the sound signal, Y-axis and Z-axis vibration signals as analytic objects and proposes a fusion recognition method for shearer coal-rock cutting state via the combination of improved radical basis function neural network (RBFNN) and Dempster-Shafer (D-S) evidence theory. First of all, on the basis of original fruit fly optimization algorithm (FOA), the location updating mechanism of moth-flame optimization (MFO) is used to improve the convergence performance and exploration ability of FOA. Thus, a hybrid optimization algorithm of MFO-FOA is accordingly designed and some simulations are conducted to verify the effectiveness and superiority. Then, the optimal network parameters of RBFNN are found out by using proposed MFO-FOA to realize the excellent generalization ability and predictive performance. Moreover, the collected signals are decomposed by variational mode decomposition, and the envelope entropy and kurtosis are used to extract the features of first three intrinsic mode function components. The feature vectors obtained from three-type sensor data are utilized to construct the RBFNN classifiers. Besides, the D-S evidence theory with evidence correlation coefficient is introduced to fuse the preliminary identification results of three RBFNN classifiers. Finally, a self-designed experimental platform for shearer cutting coal-rock is built and some experiments are provided. The experimental results based on measured data demonstrate that the proposed method can effectively identify the coal-rock cutting state with higher accuracy.