Roof pressure data prediction for working face based on back propagation neural network

Abstract

In order to apply the back propagation neural network in artificial intelligence technology to the prediction of roof strata pressure data, 11 main factors affecting the prediction of roof rock pressure data of working faces are taken as input parameters, and roof rock pressure parameters of four working faces are taken as output parameters. The number of hidden layers is determined as 1 layer and the number of hidden layer neural units is 24. On this basis, the prediction model of roof pressure data based on the back propagation neural network is established. Based on the representative roof pressure data for working face of Wangjialing and its surrounding coal mines, the study samples are established, and the measured mine pressure data of 12309 working face in Wangjialing Coal Mine were used as the verification samples to test the accuracy of the prediction. After analysis, the relative errors between the predicted values and the measured values of the first weighting interval, the first weighting strength, the periodic weighting interval and the periodic weighting strength are 0.043 343 653, 0.006 077 606, 0.006 401 138 and 0.020 223 608 8, respectively, which means that the overall relative errors of the mining pressure data are less than 5%, which is in line with the allowable error range of the engineering application. This indicates that the established back propagation neural network model has high accuracy and reliability.

Keywords

• back propagation neural network
• mine pressure
• data prediction
• weighting intensity
• weighting interval