Objective To address the problem in marine propulsion shafting fault diagnosis where fault types can be identified but faulty equipment is difficult to localize, an intelligent diagnostic method integrating mechanism-based feature modeling and knowledge graph-constrained reasoning is proposed.
Methods A three-layer ShaftAgent diagnostic framework is constructed. The mechanism modeling layer is used to extract equipment-level vibration features and auxiliary system features. The interpretable analysis layer adopts XGBoost for fault classification and proposes an equipment-level SHAP attribution aggregation method to enable automatic fault equipment localization. The knowledge-enhanced reasoning layer is designed to build a hierarchical knowledge graph of “equipment-phenomenon-mechanism-fault” and, together with multi-stage prompt engineering, drives large language models to generate diagnostic reports. A consistency verification mechanism is further introduced to ensure that the outputs comply with physical laws.
Results Experimental results show that ShaftAgent achieves a fault classification accuracy of 96.8%, an equipment localization accuracy of 94.2%, and an expert comprehensive score of 4.70 for diagnostic reports. Ablation experiments verify the effectiveness of each module, while case studies demonstrate the complete process from multi-source vibration signals to actionable diagnostic reports.
Conclusion The results indicate that ShaftAgent can effectively address the insufficient equipment-level localization capability and limited interpretability of traditional methods. They also validate the feasibility of applying large language models to industrial fault diagnosis under knowledge graph constraints, providing a new technical approach for intelligent operation and maintenance of marine shafting systems.
DownLoad: