
Deep Belief Network based Coal Mine Methane Sensor Data Classification
Author(s) -
Xue-Feng Wu,
Zhao Zhen,
Li Wang
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1302/3/032013
Subject(s) - methane , coal mining , deep belief network , support vector machine , artificial intelligence , hazard , computer science , methane gas , coal , feature (linguistics) , wireless sensor network , mining engineering , data mining , deep learning , environmental science , pattern recognition (psychology) , engineering , chemistry , waste management , organic chemistry , computer network , linguistics , philosophy
Gas explosion is the main hazard which affects the safety of coal mine production. One way to solve the problem is to predict the dangerous levels of methane concentration using the sensor time series from hazard monitoring systems. In this paper we proposed our method based on DBN to classify the dangerous levels of methane concentration. A multilayer RBM network is built to reconstruct the methane sensor data and depth characteristics of methane disaster classification are extracted. Then a BP network is used for classification learning. We test our method on a real coal mine methane sensor data, and contrast with two classical methods, SVM and KNN. The experiments results showed that our proposed deep feature learning methods could achieve better performance than the two classical shallow methods.