Fault diagnosis of mechanical equipment involves data collection, feature extraction, and pattern recognition but is often hindered by the imbalanced nature of industrial data, introducing significant uncertainty and reducing diagnostic reliability. To address these challenges, this study proposes the Uncertainty-Aware Bayesian Meta-Learning Framework (UBMF), which integrates four key modules: data perturbation injection for enhancing feature robustness, cross-task self-supervised feature extraction for improving transferability, uncertainty-based sample filtering for robust out-of-domain generalization, and Bayesian meta-knowledge integration for fine-grained classification. Experimental results on ten open-source datasets under various imbalanced conditions, including cross-task, small-sample, and unseen-sample scenarios, demonstrate the superiority of UBMF, achieving an average improvement of 42.22% across ten Any-way 1-5-shot diagnostic tasks. This integrated framework effectively enhances diagnostic accuracy, generalization, and adaptability, providing a reliable solution for complex industrial fault diagnosis.

Accurately predicting the remaining useful life (RUL) of rotating machinery, such as bearings, is crucial for equipment reliability and minimizing unexpected failures in industrial systems. Despite recent advancements, data-driven deep learning methods face challenges in practical industrial settings due to inconsistent data distributions between training and testing phases, and limited generalization capabilities for long-term RUL predictions. To address these issues, we propose LM4RUL, a framework for RUL prediction based on pre-trained Large language Model (LLM). LM4RUL leverages the generalization and reasoning capabilities of LLM to transfer predictive knowledge from pre-training, effectively overcoming data inconsistencies and enhancing prediction accuracy. This represents a meaningful advancement in the artificial intelligence field, being among the first efforts to successfully apply LLM to RUL prediction tasks without the need for additional manual instruction, thereby extending the boundaries of AI applications beyond natural language processing and into complex industrial scenarios. The framework includes the local scale perception representation component, which captures fine-grained bearing degradation trends by tokenizing vibration data, and hybrid embedding learning, which selectively freezes and fine-tunes parameters to model complex nonlinear degradation.