Regularized neural network load spectrum prediction model of coal-rock cutting based on entropy weight
-
-
Abstract
The load of cutting coal and rock is an important basis for the development of high-performance mining ma- chinery and intelligent mining. By exploring the variation laws and characteristics of the load spectrum of coal cutting, the theoretical support is provided for the study of high efficiency and highly reliable rock breaking method. In view of the randomness of the coal-rock breaking process by this method,the traditional theoretical load model has single value characteristics,which is difficult to accurately describe the load process of coal rock breaking under different cutting conditions. The theoretical deduction of the pick-to-load amplitude mode,the finite experimental load spectrum and the information entropy theory are proposed to reconstruct the theoretical and experimental cutting load spectrum. The reg- ularized neural network is used to model the reconstructed load spectrum. According to the least square method,the load curve family based on the limited wedge angle is proposed to predict the load spectrum of different wedge angles. Combining the load profiles of different parameters with the wedge angle within 30°-50°,taking the load spectrum and theoretical derivation model of different wedge angles as examples,the load spectrum reconstructed under different wedging angles is compared with that modeled regularized neural network,and the load prediction for different wedging angles is also analyzed. The results suggest that the theoretical deduction model of cutting resistance process response is constructed,and the reconstructed load spectrum under wedge angles within 30°-50° is obtained by the combination of theories and experiments. The characterization of load spectrum amplitude and variation law is achieved,and the regularized neural network modeling method of reconstructed load spectrum is obtained. The load profiles under wedge angles of 30°,33°,50°,and 55°are predicted according to the established load prediction models with different wedge angles. The correlation between the predicted load spectrum with the angles of 30° and 50° and the regularized neural network modeled reconstruction load spectrum is 0. 971 7 and 0. 983 9 respectively,which is highly correlated. The relative errors of the amplitude are 4. 04% and 5. 21% respectively. It is verified that the model can represent the load amplitude and the load history of cutting coal and rock. The model has some ad-vantages which provides a reference for the study of the crushing mechanism.
-
-