Research On Coal And Rock Cutting State Recognition Technology

Shearer is an important mechanical equipment in coal production.Unmanned and intelligent coal mining can improve coal mining efficiency and reduce the occurrence of underground casualties,and the recognition of coal and rock cutting state is the key to realize the intelligent mining of shearer,Based on the research results of the National Natural Science Foundation Project "Research on power transmission law and structure evolution theory of high efficiency cutting drum for coal and rock with gangue",the research on coal and rock cutting state recognition is completed by simulating the cutting process of shearer spiral drum and using the force characteristics of drum in the cutting process.Through the analysis of the physical and mechanical properties of coal and rock mass,the theoretical support is provided for the parameter setting of the discrete coal wall model.Through the research on the force of the pick and drum when the shearer cuts coal and gangue,the reference is provided for the force analysis of the drum in EDEM simulation.Based on MG2×55/250-BW spiral roller,its three-dimensional model was established based on Pro/e.According to the actual coal bed deposit,establish the dis crete element coal wall model with EDEM,introduce the roller model to simulate the coal rock cutting process,extract the roller force data to provide the original signal for coal rock identification.Wavelet threshold denoising is used to denoise the simulation signal,and time domain feature extraction and wavelet packet feature extraction are used to extract the features of the original force signal of the roller respectively.Comparing the two feature extraction methods,it is found that the waveform factor and variation coefficient extracted by time domain analysis can accurately reflect the fluctuation of the roller force,However,the energy feature decomposed by wavelet packet decomposition has no regularity,so the waveform factor and variation coefficient of the roller force data are selected as the characteristic signal.This paper analyzes the principle of BP neural network,constructs the neural network structure according to the characteristic signal,selects its various parameters,trains the neural network through the training set,and obtains the recognition result through the test set.This paper analyzes the principle of genetic algorithm(GA)and particle swarm optimization(PSO),optimizes BP neural network by using two optimization algorithms,establishes GA-BP and PSO-BP neural network structure,and realizes coal and rock cutting state recognition based on optimized neural network,It is proved that the recognition of coal cutting state based on PSO-BP is more accurate and efficient.This paper has 45 pictures,21 tables and 90 references.