Weights Research Of Detectable Amount In Gas Monitoring Evidence Combination

Gas disaster was the worst coal mine disaster. Imperfect gas monitoring system and the lack of scientific identification gas accidents was one of the most important reasons for the frequent occurrence of gas disaster. Therefore, improving the intelligence and data processing of gas monitoring system, depicting mine safety situation accurately was an urgent problem for gas disaster prevention.Based on multi-sensor information fusion technology, with the detectable amount of information in mine gas monitoring system as the research object and weights research as main line,improving DS evidence theory. Using the improved DS evidence theory combination on high-level information fusion of gas monitoring system, achieving real-time evaluation of Coal mine security status levels. Specific content as follows:1. Aiming at the problem that conflict paradox in traditional DS evidence theory combination, this paper analysis corresponding existing improved methods and attempt to resolve the conflict problem using the weighted average method.2. Evidence distance was one of starting point for weights study, research weights determined method based on the Jousselme distance and Pignistic transform. Defined "relative distance of evidence" to determine the weights, using average evidence to improve evidence expression accuracy to compensate for the deficiencies of traditional methods measure the weight of evidence distributional.3. Research for experts weights based on actual project situation was another starting point of weights study which contrast with the evidence distance methods. Trying explore a method of determining the weight coefficient which suit for the background of gas monitoring system.4. Using the weighted average method and DS theory combination step by step with the information of detectable amount. Analyzing the rationality of evidence combination step by step and effect on synthetic sequence and conflict threshold for the final combination results. Simulation results show that this method has good convergence and meet normalization which less affected by the order of synthesis with high reliability combination results.