112 / 2025-04-21 19:58:58

Diagnosis-Aware Active Learning Based on Uncertainty Metrics

fault diagnosis,active learning,data enhancement

The healthy operation of electromechanical system products is closely related to the safety of aircraft, and accurate fault diagnosis is one of the key core technologies in order to ensure their high availability, high mission safety and reliability. However, the low and unbalanced failure rate of aircraft electromechanical products leads to the challenges of scarce samples and the high cost of data annotation for fault diagnosis. To address this question, this study proposes a diagnosis-aware active learning (DAAL) method based on uncertainty metrics, introduces an active learning strategy, trains the ranker to capture the diagnostic uncertainty of unlabelled samples by designing the ranking loss and embedding it in a hidden code, and perceives the diagnostic information by using a task-independent active learning model, so as to filter out the unlabelled data that are both informative and high-impact in terms of diagnostic uncertainty metrics, finally achieving efficient data labelling.