This paper presents a bearing transfer diagnosis method based on digital twin and knowledge graph.The target equa-tion is constructed by encapsulating the physical characteristics of the bearing and using digital twin technology,performing di-mensionality reduction based on an reduced-order model,and integrating multi-degree of freedom dynamics to simulate various fault conditions of the bearing.An end-to-end deep learning structure is employed to learn fault features in the digital twin data-set,and then fine-tuned using on-site experimental data through transfer learning,thereby overcoming the issue of limited data in industrial environments.By the knowledge graph,the diagnostic results are associated with fault causes and improvement strategies,the multi-angle analysis ability of the fault diagnosis outcomes is improved.This method realizes real-time state up-dates and performance evolution of the rolling bearing digital twin model,overcomes the drawbacks of insufficient bearing data and limited data processing capabilities in existing industrial environments.
关键词
数字孪生/轴承故障诊断/迁移学习/知识图谱
Key words
digital twin/bearing fault diagnosis/transfer learning/knowledge graph
维普
万方数据