Abstract: The CBR(Case Based Reasoning) based online fault diagnosis method of helicopter actuator loop is studied in this paper. As to the shortcomings of the traditional CBR method which just adops the discrete data and the symbol vector to establish a case, a new multi-source and dissimilar information case representation method is brought forward for dynamic systems, which uses the system characteristics set, symptom set and feature set hybrid model as the condition attribute of a case. Then the synthetic similarity measure method for case retrieve is studied. At last, the above method is validated on a helicopter pitching actuator loop, and experiment results show that the CBR method can diagnose the control parameter drift fault of actuator loop. The proposed method can serve as an effective method for the fault diagnosis of complex dynamic systems.

Key words: case-based reasoning, case retrieve, similarity measure, actuator loop, fault diagnosis