Research on fault diagnosis based on multi-signal flow graph and branch-and-bound algorithm

A complete fault diagnosis method based on multi-signal flow graph and branch-and-bound algorithm was proposed to deal with real-time online fault diagnosis problems. A multi-signal flow graph model was built for the object system and a dependency matrix was generated as diagnostic knowledge. Conflict sets was generated by the dependency matrix and the system observation vector, which was essential in transforming the problem of finding the minimal diagnosis set to a problem of integer programming. A new version of branch-and-bound algorithm was utilized to calculate the optimal solution of the diagnosis by branching, computing lower and upper bounds and pruning the conflict sets. In this way, explosion problem caused by enumeration could be avoided. By applying the proposed approach to a fuel system of aircraft, the efficiency of this method was verified. Diagnostic results by comparing with the existing TEAMS-RT algorithm demonstrate that the proposed method has a higher accuracy in locating the faults. Besides, both single-fault and multi-fault diagnostic problems can be covered and the method is capable of large-scale complex system fault diagnosis.