BP neural network prediction model of floor failure depth in North China coalfield

Abstract

North China coalfields are seriously affected by bottom aquifers. In order to accurately the depth of damage at the working face, this paper combined actual measurement with neural network prediction model in analysis. Firstly, DC method with special electrode cable is employed to observe the bottom plate damage depth of 15091 in the comprehensive mining face of Jiulishan mine; secondly, based on large-scale data, genetic algorithm is applied to optimize BP neural network. The prediction model of bottom plate damage depth is set up by optimizing parameters. The mean square error of the prediction model was 0.011, the average percentage error was 5.983 %, and the prediction error based on prediction set was below 10 %. These results indicate that the model can be used for predicting the bottom slab damage depth. Finally, the prediction model was used to analyze the effect of mining thickness and top cutting pressure relief on the depth of damage of the working face floor. Results show that under stratified mining, the depth of damage of the bottom slab is reduced by 77.84 % under cut top pressure relief than uncut top pressure relief; under integrated mining, the depth of damage of the bottom slab is reduced by 59.17 % under cut top pressure relief than uncut top pressure relief; and the effect of mining thickness on the depth of damage of the bottom slab is positively correlated.

Keywords

• floor failure depth
• direct current method
• bp netural network
• fully-mechanized top-caving mining
• layer mining