A novel adaptive monitoring framework for detecting the abnormal states of aero-engines with maneuvering flight data

Abnormal state monitoring of aero-engine is crucial for ensuring the safe and reliable operation of aircraft. Currently, the common approach involves analyzing data from onboard sensors. However, this method faces significant challenges during maneuvering flight due to the considerable data disparities across different flight sorties. This paper proposes an adaptive anomaly monitoring framework for aero-engines operating under diverse flight conditions. Firstly, the flight data is processed using the Mean-Standard Deviation Method to achieve automatic scene segmentation under complex flight conditions. Then, a monitoring model combining the Sparrow Search Algorithm (SSA) and Kmedoids is built for adaptive monitoring under each set of flight conditions. On this basis, the indicators are devised to accurately reflect the risk of abnormal situations for state monitoring. Finally, the proposed method is verified by actual flight data and shows higher accuracy and lower false alarm rate (FAR) in monitoring both abrupt and gradual abnormal situations of aero-engines. Additionally, the method has the capability to automatically identify the parameters corresponding to anomalies. This framework not only enables adaptive monitoring and early warning of aero-engine abnormal states in the absence of prior fault information but also supports comprehensive fault tracing and analysis.