Research of Fusion Diagnostic Algorithm for Aircraft Engine On-Line Fault Diagnostic System

In this paper, a fusion diagnostic algorithm of the hybrid Kalman filter (HKF) and back-propagation neural network (BPNN) is developed for its application to aircraft engine on-line fault diagnosis. HKF is the process of detecting faults and generating filter residuals, while BPNN is the process of isolating the faults using the HKF residuals. The fusion diagnostic algorithm maintains its effectiveness throughout the engine's whole life by periodically updating the reference health baseline of HKF to the health condition of degraded engines. Based on the fusion algorithm, an on-line fault diagnostic system of a high-bypass commercial aircraft engine is developed and its diagnostic capability is evaluated using the Monte Carlo simulation method. The simulation results show that the fault detection rate of the system is above 98%, the fault isolation rate is above 90%, and the false alarm rate is below 1%.