Research on deep learning-based fault diagnosis methods for aero engine gas path

As the "heart" of an aircraft, the stability and safety of an aero-engine are crucial for ensuring the reliable performance of the aircraft. The aero-engine gas path system, as a crucial part of the aero-engine, is prone to frequent failures, making it crucial to carry out fault diagnosis research on it. In this paper, an improved stacked sparse denoising auto-encoder (SSDAE) fault diagnosis method is proposed. The computational accuracy and training time of the method are demonstrated using airway data from the JT9D engine, with input data containing ten different faults. The results demonstrate the accuracy of the PSO-SSDAE can achieve 96.4% under single operating conditions and 95.6% under variable operating conditions, respectively, which confirms the capability of the methodology to achieve more effective fault detection.