1.National Joint Engineering Laboratory of Internet Applied Technology of Mines，Xuzhou ，China;2.Internet of Things Perception Mine Research Center,China University of Mining and Technology, Xuzhou ，China; 3.School of Information and Control Engineering,China University of Mining and Technology,Xuzhou ，China;4.School of Intelligent Manufacturing, Jiangsu Vocational Institute of Architectural Technology, Xuzhou ，China

Current methods for early warning of gas accidents in coal mines are mostly based on manual prevention and control based on monitoring data.In order to realize the intelligent analysis and early warning of gas accidents and the reuse and sharing of related knowledge, based on the ontology and accident tree theory, it is proposed to build a dynamic gas accident early warning knowledge base and reasoning model to meet the description and analysis and warning of gas accident knowledge under the constraints of coexistence of time and space of various disaster-causing factors.Firstly, an ontology model and inference rule base was constructed to realize tacit knowledge inference from root hazards to basic events.Then the occurrence probability of top event and the corresponding levels can be calculated according to the relation between basic events and top event obtained from the previous inference.Finally，corresponding decision support can be given in accordance with the early warning levels.The experimental results indicate that the proposed knowledge base and early warning inference model for gas accident can effectively calculate the occurrence probability of dynamic gas accidents under the joint temporal and spatial constraints.Valuable exploration was made in the field of gas accident knowledge description and intelligent early warning.