Parameter-Influencing Analysis of Aeroengine Operation Reliability

Abstract

The reliability during aeroengine operation is significant in monitoring its time-varying operational status and providing valuable reference for both operation and design. To study aeroengine time-series operational reliability and parameter feature effects in whole flight stages, a tree model is developed to discuss the operational reliability of aeroengine through residual thrust and feature importance analysis. 24 parameter features from an aeroengine (CFM56) operation are applied to investigate the required thrust and residual thrust of aeroengine through quick access recorder (QAR) flight data. Characterizing aeroengine load case by the cosine similarity of aeroengine feature sets and operational reliability of aeroengine are conducted by combining cosine similarity and residual thrust of aeroengine. Further, XGBoost (extreme gradient boosting) model coupled with the Shapley additive ex-planations (SHAP) algorithm is constructed to enhance model interpretability analysis, in which the importance and influence direction of aeroengine features are analyzed. In this study, features similarity and importance were analyzed using aeroengine QAR data from a single flight. The results showed that (1) the operational reliability of aeroengine greatly varies under different flight stages, (2) the load case of aeroengine undulate greatly under take-off and descent stages, but is stable in the cruise stage, and (3) the feature importance and influence direction are indifferent among residual thrust and three aeroengine parameter sets, showing the complexity of aeroengine operation. Based on the operational and applicable restriction data, the efforts of this study provide a new analysis method for aeroengine to improve the design and operation by analyzing operational reliability.