Evaluation of deep coal and gas outburst based on RS-GA-BP

• Published in: Natural hazards (Dordrecht) Vol. 115; no. 3; pp. 2531 - 2551
• Main Authors: Zhu, Junqi, Zheng, Haotian, Yang, Li, Li, Shanshan, Sun, Liyan, Geng, Jichao
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.02.2023; Springer Nature B.V
• Subjects: Accuracy; Adaptive algorithms; Algorithms; Back propagation; Back propagation networks; Civil Engineering; Coal; Coal gas outbursts; Coal mines; Coal mining; Dimensions; Disasters; Earth and Environmental Science; Earth Sciences; Environmental Management; Evaluation; Genetic algorithms; Geophysics/Geodesy; Geotechnical Engineering & Applied Earth Sciences; Hydrogeology; Identification; Methods; Mines; Natural Hazards; Neural networks; Optimization; Original Paper; Gas outburst; BP neural network; Genetic algorithm; Deep coal mine; Rough set
• ISSN: 0921-030X; 1573-0840
• DOI: 10.1007/s11069-022-05652-w

Owing to the high dimension and nonlinear characteristics of gas outbursts in deep coal mines, an intelligent evaluation method for systematically screening and integrating gas data in deep coal mines is proposed herein to effectively identify coal and gas outbursts in deep mines. A rough set improved using a genetic algorithm is introduced to reduce the dimension of complex data pertaining to deep coal mine gas to determine the main control index of deep coal and gas outbursts. Subsequently, the initial weight and threshold of a back propagation (BP) neural network are optimized by combining the characteristics of parallelism and robustness of the genetic algorithm. An adaptive optimization of BP neural network by genetic algorithm back propagation (GA-BP) model is established to identify gas outburst in deep coal mine reasonably. Compared with the standard BP neural network, data fitting shows that the method can significantly improve the detection accuracy of deep coal and gas outburst while improving the speed of disaster identification, as well as improve the efficiency of disaster identification, thereby increasing the risk identification accuracy of deep coal and gas outburst to 90%. This not only provides a new method for the scientific evaluation of deep coal and gas outburst risk, but also an important reference for the scientific evaluation of other high-dimensional and nonlinear fields.