ABF-FNN: A new fuzzy neural network for predicting coal mine gas concentration hazard

Yimin Sun, Xiaobo Zhang, Zhehao Zhang, Yunyang Wu, Haihao Tang, Haonan Luo

Abstract: Serious accidents in coal mines are often caused by high concentration of gas. Therefore, accurate prediction of mine gas concentration is vital for ensuring the safety of coal mine production. Hence, a new method for predicting the risk of gas content in underground coal mines (ABFFNN, adaptive bandwidth feedback fuzzy neural network) is proposed in this paper. Namely, 1) Fuzzy theory and neural networks are combined in the method, and feedback mechanisms are incorporated to enhance network learning capability and robustness. 2) This method designs a novel approach to calculate adaptive bandwidths and dynamically adapts to changes in univariate and multivariate time series data. Weight training is performed using particle swarm optimization algorithm. 3) The evaluation of the developed prediction model is conducted using two publicly available datasets as well as actual industrial data collected at a mining site. The public datasets used include the Data of Qinggangping coal mine monitoring system in Shaanxi Province and the Box-Jenkins gas stove dataset. It is shown by the results that good prediction results are achieved by the developed prediction model on two public data sets and one private data set. In addition, it can be seen that the gas concentration hazard in underground coal mines can be better predicted by the method proposed in this paper when compared with other models.