Progressive Knowledge-Guided Large Language Model Framework for Bearing Fault Diagnosis

The paper introduces a progressive physics-guided multi-scale vibration signal processing framework for bearing fault diagnosis, addressing measurement challenges in real-time fault screening. It utilizes an 81-dimensional measurement descriptor and a fault-adaptive segmentation mechanism, achieving 98.49% diagnostic accuracy while reducing computational costs by 12.6 times compared to traditional methods. This framework enhances interpretability and traceability of features to underlying fault mechanics, making it significant for practitioners in safety-critical industrial applications.