Fault diagnosis is crucial for ensuring the navigational safety and reliability of unmanned underwater vehicles (UUVs). However, in complex marine environments, UUV fault samples are typically scarce and often contaminated by severe hydraulic noise, which significantly restricts the performance of existing diagnostic methods. To address these challenges, this paper proposes a few-shot fault diagnosis method based on an Adaptive Enhanced Meta-Transformer. First, operational vibration data from UUV actuators are acquired and preprocessed. Second, a feature enhancement module is constructed using an improved Transformer architecture that incorporates a novel Adaptive Head-Weighted Multi-Head Self-Attention mechanism. This mechanism enables the model to precisely localize key fault segments and enhance directional features, even under noisy backgrounds, effectively mitigating attention dispersion. Subsequently, a meta-optimization strategy is employed to iteratively update model parameters, enabling the model to rapidly adapt to new tasks with limited data. Finally, extensive experiments using real-world operational data from UUV actuators demonstrate that the proposed method outperforms state-of-the-art baselines in terms of accuracy and robustness, particularly in cross-component and noisy scenarios.

Thruster health monitoring is essential for unmanned underwater vehicles (UUVs), where propulsion degradation can reduce manoeuvrability, tracking accuracy, and mission safety. Practical diagnosis remains difficult because labelled vibration samples are scarce and deployed vehicles may encounter fault states absent from the support library. Closed-set few-shot classifiers are unreliable in this setting because every query must be assigned to a known state. SAGE-PD (Spectral Attention-Guided Episodic Prototype Displacement) is a few-shot open-set method for UUV thruster vibration monitoring. It encodes each vibration window through raw-waveform and STFT branches, constructs episode-specific prototypes for known thruster states, and models their relational geometry. Unknown-state evidence is obtained by replacing the predicted prototype with the query embedding and measuring the displacement of the transformed prototype structure. Spectral attention weights this displacement toward thruster-related time–frequency components. On the analysed UUV thruster dataset, SAGE-PD achieved $0.9078\pm 0.0573$ Open OA and $0.9011\pm 0.0407$ AUROC in a 5-shot evaluation, and $0.8745\pm 0.0457$ Open OA and $0.8953\pm 0.0409$ AUROC in a 1-shot evaluation. The results show that SAGE-PD improves both known-state recognition and unknown-state rejection by combining support-structure compatibility with vibration-aware spectral evidence.