A Likelihood Ratio-Based Approach to Bleed Valve Event Detection in Gas Turbine Engines

Bleed valves are widely used in gas turbine engines (GTEs) for airflow control to prevent compressor surge so as to improve the overall GTE performance and handling. Bleed valve fault detection is a challenging task due to the inhospitable environment that the bleed valve is located and the limited available sensor signals that can be used for detection. The problem is investigated in this paper and a Kalman filtering-based likelihood ratio approach is proposed for bleed valve event detection where only the pressure line signal and the scheduled bleed valve demand signals that are currently available in a GTE are used for detection. With the proposed approach, two models are developed for tracking the change in pressure signal, one with the scheduled bleed valve demand signals as input and one without. Two Kalman filters are designed based on these two models and the likelihood functions of the pressure observations are then evaluated with the state estimates from these Kalman filters. The bleed event detection is eventually achieved via the likelihood ratio test. The developed method is used for detecting bleed events using real flight data and the results are very promising.