Analysis of Factors Influencing Mining Damage Based on Engineering Detection and Machine Learning

Abstract

The direct results of mining damage are overburden fracture and surface subsidence, which may induce groundwater seepage and surface vegetation degradation. Therefore, it is essential to research the factors and mechanisms influencing mining damage. Based on the geological characteristics of the Xiaobaodang minefield in the Yushen Mine area in China, the engineering detection of fractured zone height (FZH), sampling tests of rock mechanical properties, and field measurements of the surface settlement were carried out. Firstly, the factors influencing the FZH were screened by correlation analysis and partial correlation analysis. Next, a model for predicting the maximum height of the fracture zone with the BP neural network (BPNN) was established and trained with Python. Finally, the FLAC3D numerical simulation experiment was adopted to reveal the variation law of overburden stress during coal mining, and the relationship between stress and overburden fracture was analyzed. The results show the following: When the average mining thickness in the study area is 5.8 m, the maximum height of the fractured zone is 157.46 m, and the maximum surface subsidence is 3715 mm. Further, the mining thickness, mining depth, the compressive strength of overburden, the width of the working face, and the mining velocity are the main factors affecting the maximum height of the fractured zone. Additionally, the goodness of fit of the BPNN model can reach 97.22%, meaning that it can effectively predict the maximum height of the fractured zone caused by coal mining. Finally, the area where the stress changes markedly above the goaf is the area where the fractures develop rapidly. Meanwhile, there is a positive correlation between the surface subsidence and the FZH. The research results obtained provide new ideas for reducing mining damage and will be helpful for the green and sustainable development of the mine.