Aero-Engine Gas System Fault Diagnosis Method Based on MAML in Few-shot Sample Conditions

The data-driven fault diagnosis method demands a high quantity of data samples, typically necessitating ample samples and labels for effective model training. However, due to the complex operating environment of aircraft engines, it is difficult to collect fault samples. In addition, unpredictable types of faults may occur during operation. A few-shot sample and first occurrence fault diagnosis method based on model-agnostic meta-learning(MAML) is proposed to address the above issues. On the one hand, utilizing MAML to learn generalized fault classification rules enables the model to quickly adapt to tasks with limited sample sizes. On the other hand, by constructing sample pairs for discrimination recognition, the first occurrence fault can be determined. Finally, experiments were conducted on real aircraft engine operation and maintenance datasets to verify the effectiveness of this method.