Abstract: The fault diagnosis of the bipropellant propulsion system is investigated with partially observed time Petri nets TPN. The internal environment of the bipropellant propulsion system is complex and some key information can not be obtained through the sensors. A method of building modified state class graphs based on partially observed (TPN) is proposed. The node of the system process corresponds to observable transition and unobservable transition. The time interval of each transition is calibrated, and partially observed TPN is built. And then they are transformed into modified state class graphs based on the different relationships of the synchronization and asynchronous transitions. Traverse all paths that satisfy the fire time of observable transition and fire sequences to determine whether the system is faulty. Finally, the partially observed TPN model of the bipropellant propulsion system is built. Based on the observability of each actuator in the system process, fault diagnosis of unobservable part in the system is performed to verify the effectiveness of the algorithm.