A Model and Data Dual-Driven Framework of Gas Path Fault Diagnosis for Aero Engines

Gas path fault diagnosis technology has recently evolved from condition monitoring to health prediction and diagnosis, and from single algorithms to fused algorithm approaches. The urgency of resolving the accuracy improvement of gas-path fault diagnosis is paramount, particularly when considering performance degradation and individual variation of aero engines. The hybrid gas path fault diagnosis method has been demonstrated to be an effective solution to this problem. In this work, a synthesis of model-based and data-driven architecture for gas path fault diagnosis is proposed to enhance diagnostic accuracy. The novelty of this model and data dual-driven framework is that the confidence-based weighted fusion strategy dynamically allocates weights according to the model-based and data-driven errors. The results show that the fused framework reduces the mean absolute percentage errors from 9.01% to 0.07%, representing a remarkable accuracy improvement. This model and data dual-driven framework leverages the computational efficiency of the data-driven model while utilizing continuous adaptation through physical model-based correction mechanisms. It enables continuous model refinement despite engine degradation during long-term service.