Application of BP Neural Network in the Prediction of Periodic Weighting

Based on the theory of BP neural network, the monitored data of the cyclic end resistance of hydraulic support in 02178 working face of Huopu mine is trained. Through analyzing the errors produced by the different nodes of network hidden layer, the periodic weighting prediction model whose network structure is 4-12-1 is built. After field monitoring, the 5th weighting is predicted. And the results showed that the average periodic weighting step is 9.1 m, and the influence range covers 1.53 m, and the average dynamic load coefficient is 1.28. Evidently, the output values of network are basically consistent with the monitored data. Therefore, these predicted data can provide a theoretical basis for supporting design and safety production of roadway with the same conditions.

View article page

Download PDF