Open Access
Review
Table 10
Analysis of data challenges and mitigation strategies in nuclear pump fault diagnosis.
| Type of data challenge | Specific manifestation | Impact severity | Mitigation strategy |
|---|---|---|---|
| Sample Scarcity | Lack of real fault-labeled samples | Model accuracy drops by 15–25% | VA-WGAN-based sample generation [32] |
| Simulation-to-Reality Gap | Lab data fail to capture on-site noise | Average accuracy degradation of 12.7% | High-fidelity hybrid simulation [28] |
| Multi-source Heterogeneity | Inconsistent data formats/sampling rates | Fusion error up to 18.3% | Federated learning framework [37] |
| Insufficient Validation Depth | Lack of long-term field testing | Model drift rate >5% per year | Digital twin-based validation [45] |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.