Answers are grounded in this paper and cite their sources.
Demonstrates improved fault diagnosis in lifeboat davits, suggesting effective adaptation under low temperatures.
To address fault sample scarcity and extreme domain shifts in lifeboat davit (LD) hydraulic systems under polar low-temperature conditions, this paper proposes a meta-transfer learning paradigm based on small-sample knowledge transfer and domain alignment. The meta-transfer domain-aligned fault diagnosis (MTDAFD) network couples the feature extraction of the patch time series transformer with the inner-loop gradient updates of model-agnostic meta-learning (MAML) for rapid adaptation to unknown target conditions. Furthermore, a specifically designed, plug-and-play multi-kernel rational quadratic maximum mean discrepancy (RQM) module integrates into the outer-loop meta-optimization of MAML to flexibly quantify the distribution discrepancy between the source and target domains. The network uses a joint meta-loss function to optimize model parameters, ensuring task discriminability and cross-domain transferability. Experiments on the LD and road paver (RP) hydraulic system datasets demonstrate the effectiveness and engineering feasibility of the MTDAFD method and the RQM module, achieving high classification accuracy even with only 2–10 labeled samples per health state.
No takes yet. Share an insight, caveat, or question.
Niu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: