Research on Distribution Field Reconstruction Technology Based on Markov Random Field-Kriging Model

Abstract

Gas accident is the most serious form of coal mine accidents, so the analysis of gas distribution in coal mining face is particularly important. Aiming at the shortcoming that the numerical simulation of gas distribution cannot timely analyze the gas distribution of the stope based on the measured gas concentration data, the spatial information statistical models such as Kriging model are introduced to construct the stope gas distribution field. However, the Kriging model often uses all the gas concentration data for the estimation of gas concentration distribution in stope, which causes the heavy increasement of computational complexity. To overcome the problem, the neighborhood structure of Markov random field is proposed to embed into the Kriging model, which effectively reduces the computational complexity and mean square error of the estimation of gas concentration distribution in stope. Finally, an experiment study is carried out to show the effectiveness of the proposed method. Estimating 5 regions with 44 data sets, the program operation time is reduced by 28%. The estimation of single points also performed better than the original method and the mean square error is reduced by 19%.