Abstract: For problems of low reliable performance and inadequate utilization of monitored data of safety monitoring system of coal mine, a viewpoint was proposed that system reliability could be improved by prolonging mean time between failures of important equipments and credibility of monitored data could be improved by method of data processing or redundancy. A new alarm method of safety monitoring system of coal mine based on relevance model was proposed for improving credibility of monitored data, whose basic idea is connecting data of a measuring point with data of others or process environment by a related model, and the system would alarm when the model is not tenable. The alarm method could be a complement of existing alarm methods and find irregular usage and invalidity of sensors, while usability and suitability of the relevance models in application must be considered. Relevance models of data of gas sensors were established separately based on correlation coefficient, difference between the maximum and the minimum, and the minimum according to analysis of law of gas outburst by example of gas monitoring of working face of coal mine, and the models were verified by use of practical monitored gas data with a conclusion that all the models are suitable for different coal mines and working faces.