A Compressor Stall Warning System for Aeroengines Based on the Continuous Wavelet Transform and a Vision Transformer

Abstract

The aerodynamic stability of a compressor has a crucial impact on the performance of modern aircraft power system. It is necessary to design an accurate and reliable rotating stall warning system to take active control measures to avoid compressor instability as much as possible. This paper proposes a compressor rotating stall warning system that combines continuous wavelet transform (CWT) and a vision transformer (ViT), called a CWT-ViT system. Specifically, the system transforms one-dimensional time-series dynamic pressure signal data into two-dimensional color time–frequency images using CWT, which serves as the input to train the ViT classifier. In response to sensor failure, a model ranking execution strategy was adopted to improve the reliability of the whole system. The feasibility and performance of the proposed system were evaluated in different operating modes and sensor failure conditions using compressor stall experiments. The results showed that the average classification accuracy of the proposed system in stall warning tasks was 97.66%, which was the highest among all methods. In addition, the proposed system can maintain an early warning time of over 160 m seven in the case of sensor faults, which was the best warning performance among all methods.