Research on coal and gas outburst prediction and sensitivity analysis of influencing factors based on PCA-ICA-BP

Abstract There is a complex nonlinear mapping relationship between the phenomenon of coal and gas outburst and various influencing factors. In order to achieve accurate prediction of the danger of coal and gas outburst, traditional biomimetic algorithms have the disadvantage of easily falling into local optima and the problem of multiple correlations between the influencing factors of coal and gas outburst, This article constructs and uses Matlab to write an Imperial Competition Optimization BP Neural Network (PCA-ICA-BP) algorithm based on principal component analysis to predict the risk of coal and gas outburst. The results showed that after principal component analysis, five principal components were obtained, with a cumulative contribution rate of 89.397%; The average relative errors of PCA-ICA-BP, PCA-BP, ICA-BP, BP, PSO-BP, and GA-BP prediction models are 1.37%, 2.98%, 3.27%, 15%, 7.58%, and 8.65%, respectively. The convergence speed of PCA-ICA-BP model is 4.38 times that of BP model, and the convergence accuracy has been improved by 90.8%. The Sobol index method was used to explore the sensitivity of relevant influencing factors to coal and gas outburst. The sensitivity indicators of coal and gas outburst were sorted based on the main effect and total effect indices. The influencing factors related to gas geology have high sensitivity, and the mine is a gas dominated outburst mine. Sensitivity analysis can provide reference for the deployment of coal and gas outburst sensors in coal mines.