Bayesian Network based Diagnostics Technique for Civil Aircraft

Current aircraft onboard fault detection and diagnosis (FDD) systems have experienced high probability of false alarms. For example, No Fault Found (NFF) has been a major issue in maintenance efficiency and effectiveness. NFF caused by FDD problems directly leads a lot of unnecessary maintenance activities. This paper adapts Bayesian networks (BNs) to establish fault diagnosis models for aircraft hydraulic system, and analyses the combination of evidences mapping into the corresponding failure modes and the parameters impacting on diagnostic confidence. The simulation results show that BNs-based model is an efficient tool which not only can be used to diagnose the hydraulic components but also can be used to diagnose the sensor failure; it can decrease NFF occurrence rate efficiently. The results are considered promising for the future aircraft onboard FDD system development.