Abstract: Complex internal structure and high coupling of faults make it difficult for precise fault diagnosis of aero-engine,in which the typical machine learning model and ensemble learning model cannot meet the rising flight safety requirements. In order to solve this problem, a fault diagnosis method of aero-engine based on Stacking ensemble learning is proposed. Firstly, four key parameters of aero-engine gas path are selected and the flight cycle observation window is set basing on fault reports provided by OEM. Then, the training data set is built and normalized preprocessing is conducted. Finally, according to the difference degree and typical model performance, the optimal base models and meta model are selected. A two-layer Stacking ensemble learning model is established to realize the intelligent diagnosis of four typical gas path faults of aeroengine. Experimental results show that compared with the typical model, the current model improves both of the accuracy and recall ratio by 3%~16%, which can be better applied to the aeroengine fault diagnosis.

Key words: ensemble learning, data mining, aero-engine, gas path parameters, fault diagnosis, classification model