Abstract: The fault diagnosis of urban rail transit vehicle compartment door system is a weak field of research at home and abroad,for example,the Petri net model for door system fault diagnosis is established on the basis of historical data. In order to make better use of the real-time status information of relevant equipment,inference constraint of time information is suggested to be added to the Petri net model,forward running logic relationship of the system to be adopted to carry out Petri net modeling. The process of the door system is monitored,then combined with the symbol directed graph（SDG）,the reverse order query of abnormal nodes is performed to obtain the candidate fault source. Fault probability is introduced to undetectable nodes as the basis for judging the fault,and suggestions for on-the-road fault diagnosis and early warning maintenance of the door system are given. The method is verified by actual field test data of the door system. Results show that the method is effective in door system fault diagnosis and possesses characteristics of high efficiency,simplified system and easy system control and integration.