Applications of artificial intelligence for coal mine gas risk assessment

Abstract

Gas accident is the main accident type that affects coal mine safety production. In an effort to ensure workers’ safety and health, and reduce the probability of productivity decrease, it is essential to make the coal mine risks known and controllable through certain technical means. The objective of this study is to propose an innovative and practical method to assess the risk level of coal mine gas and provide help for the prevention and control of coal mine gas accidents. The research main included three steps. In the first step, collected real data about coal mine gas accidents to establish a attributes list that describing coal mine gas accidents. Due to a large number of attribute characteristics, the data set has structural characteristics of too high dimension, large scale and high complexity. In the second step, t-distributed Stochastic Neighbor Embedding (t-SNE) was performed to deal with complex high dimensional gas accident data. In the final step, Genetic Algorithm (GA) was used to optimize Support Vector Machines (SVM) to predict the outcome of coal mine gas accidents in terms of severity. By comparing the performance of prediction effect, error distribution, time cost, etc., the introduction of t-SNE's evaluation model can accurately predict the 89% of accident outcome, while saving about 60% of time cost. Ultimately, the proposed methodology can be generalized and applied to the assessment of other risk types and assist coal mine professionals to examine future safety problems by collecting large scale safety data.