Search papers, labs, and topics across Lattice.
This paper introduces a two-stage transfer learning framework leveraging a lightweight GPT-2-style Transformer for cross-domain bearing fault diagnosis, effectively addressing the challenges of dataset heterogeneity and limited labeled data. By employing hierarchical feature extraction and knowledge modulation through pre-trained encoder weights and fault prototype embeddings, the framework enhances generalizability and adaptability across various fault categories. Experimental results show an impressive 92.61% average accuracy with only 10% labeled data, significantly outperforming existing methods by 17.24 percentage points, underscoring its potential for practical applications in predictive maintenance.
Achieving 92.61% accuracy in bearing fault diagnosis with just 10% labeled data reveals a transformative approach to predictive maintenance in industrial settings.
Bearing fault diagnosis faces critical challenges when dataset heterogeneity, operating condition variations, and limited labeled data occur simultaneously in industrial environments. Existing approaches address these issues in isolation and rely on implicit feature alignment, limiting effectiveness under concurrent challenges. This paper proposes a knowledge-guided two-stage transfer learning framework that employs a lightweight GPT-2-style Transformer with causal self-attention for hierarchical feature extraction from vibration signals, establishing explicit pathways where pre-trained encoder weights and fault prototype embeddings serve as knowledge carriers from multi-source pre-training to target adaptation. The framework addresses the dual-shift challenge through multi-source learning for generalizable representations, prototype-based knowledge modulation for target adaptation, and taxonomy-adaptive classification for seamless transfer across heterogeneous fault categories. Experimental validation on four real-world datasets demonstrates 92.61% average accuracy with only 10% labeled target data, outperforming state-of-the-art methods by 17.24 percentage points, establishing a practical pathway toward cost-effective predictive maintenance in Industry 4.0 applications.
