Research On Data-driven Cutting Pattern Recognition Method Of Shearer Sound Signal

With the high-quality development strategy of the coal industry,the demand for intelligent coal mining is increasingly urgent.As one of the key pieces of equipment of mechanized mining face,the improvement of the intelligent level of the shearer is an important guarantee for the intelligence of the whole coal mining system.Due to the complex geological conditions coal mining face and the harsh mining environment,how to realize the cutting pattern recognition in the mining process has become a crucial problem to be solved.Therefore,by studying the cutting pattern recognition technology of the shearer,it is of great significance to ensure the safe and reliable operation of the shearer and improve the intelligence level of the shearer.This topic takes drum shearer as the primary research object.It studies the data-driven cutting pattern recognition of the sound signal data generated by the shearer cutting.The main research contents are as follows.(1)Based on the study of the shearer’s basic structure and working mechanism,this study takes a shearer as the prototype.It combines the similarity theory and deduces the main similarity criteria between the model and the prototype according to the dimensional analysis method.Therefore,the paper designs the overall structure of the experimental device of the simulated shearer.On this basis,it takes pulverized coal,cement,sand,and water as simulated similar materials.It uses the orthogonal test method to study the relationship between the compressive strength of simulated coal samples and the material ratio.(2)The paper studies the mechanism of shearer cutting sound signals and the noise reduction theory.Aiming at the problem that the shearer cutting sound signal is easily polluted by noise under intense background noise,it analyzes the primary noise sources and types of cutting sound signal.Therefore,this study proposes a combined noise reduction method of the cutting sound signal based on improved complete ensemble empirical mode decomposition with adaptive noise(ICEEMDAN)and singular value decomposition(SVD).Based on the simulation signal and cutting experiment,the signal processing results show that the ICEEMDAN-SVD combined noise reduction method can effectively remove the noise components in the noisy signal and retain the main frequency components.Compared with the traditional noise reduction methods,the signal-to-noise ratio of the noise reduction signal obtained by the ICEEMDAN-SVD method is the highest,and the root mean square error is the smallest.Therefore,the signal noise reduction effect is better.(3)The cutting sound signal data of the shearer is nonlinear and non-stationary.Therefore,this study proposes to use the variational mode decomposition(VMD)algorithm for processing.It studies the influence of the number of mode components and penalty factor on signal decomposition in the VMD algorithm.In order to adaptively determine the structural parameters of the VMD algorithm,it proposes a method to optimize VMD parameters based on the sparrow search algorithm(SSA).Combined with the characteristics of the shearer cutting sound signal data,this study takes the local minimum envelope entropy of the signal as the objective function.It uses SSA to adaptively determine the optimal parameter combination of the VMD algorithm.The study combines simulation and experimental signals for performance comparison and analysis.The results show that the proposed SSA-VMD algorithm has good anti-noise ability and can accurately extract the feature information in the signal,which is helpful to the subsequent recognition research.(4)This study proposes a cutting pattern recognition model based on the combination of composite multi-scale permutation entropy(CMPE),improved gray wolf optimization algorithm(IGWO),and support vector machine(SVM).It combines the IMF feature component of the cut sound signal data to study the feature expression ability of CMPE and MPE.SVM is used as a classifier for pattern recognition of shearer cutting by solving nonlinear and small sample data classification.In order to determine the penalty parameters and radial basis radius in the SVM model,it uses the gray wolf optimization algorithm(GWO)for optimization.At the same time,this study introduces a differential evolution algorithm and power function convergence factor to improve the global optimization ability and convergence speed of the GWO algorithm.Experimental signals show that the established cutting recognition model can effectively and accurately distinguish different cutting modes.(5)In order to further improve the accuracy of shearer cutting pattern recognition,the study proposes a cutting pattern recognition model based on a depth confidence network(DBN).It studies and analyzes the influence of structural parameters of the DBN model on its performance.For the parameter selection of the DBN model,this study proposes to optimize the DBN structure model by using the IGWO algorithm and adaptively determining the number of hidden layer nodes and learning rate.It constructs the recognition model based on IGWO-DBN.Combined with the existing cutting sound signal data feature set,the average recognition accuracy of the IGWO-DBN model is 99.33%.The comparison results show that the proposed algorithm can improve the recognition accuracy of the shearer cutting mode.The research results of this paper can provide new reference ideas for accurate recognition of shearer cutting mode based on sound signal data.Figure [94] Table [17] Reference [244]...