A distributed fault diagnosis system scheme for launch vehicle was presented based on distributed expert system because of the characteristic of the launch vehicle including many subsystems which coupled with each other. In knowledge acquisition, the scheme used the production rule representation and described the experts- knowledge by the four-table structure composed by measure point table, symptom table, rule table and result table, which has the excellent versatility and transportability. Furthermore, the scheme combined the distributed inference and multi-level inference as the inference strategy. Because the distributed inference analyzes the system faults and each subsystem faults respectively at the same time, it can improve the diagnosis speed effectively. In addition, the multi-level inference can diagnose the faults with the high hazard rating preferentially. Therefore, once the highly hazardous faults of the launch vehicle happen, the fault diagnosis system can find them in time and give the maintenance measure to reduce their harm to the lowest level.