| 引用本文: | 朱彦,王坚.基于知识图谱推荐的钢铁产线智能故障诊断[J].控制理论与应用,2024,41(9):1548~1558.[点击复制] |
| ZHU Yan,WANG Jian.Intelligent fault diagnosis of steel production line based on knowledge graph recommendation[J].Control Theory and Technology,2024,41(9):1548~1558.[点击复制] |
|
| 基于知识图谱推荐的钢铁产线智能故障诊断 |
| Intelligent fault diagnosis of steel production line based on knowledge graph recommendation |
| 摘要点击 3617 全文点击 62 投稿时间：2023-02-22 修订日期：2024-06-22 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2023.30077 |
| 2024,41(9):1548-1558 |
| 中文关键词 故障诊断 知识图谱 推荐算法 偏好传播 |
| 英文关键词 fault diagnosis knowledge graph recommendation algorithm preference propagation |
| 基金项目 国家自然科学基金项目(72271188),“2030年科技创新”新一代人工智能重大项目(2018AAA0101800)资助 |
|
| 中文摘要 |
| 钢铁生产需要经过原料输送、冶炼、热轧、冷轧等多个环节, 完成这些流程还需要电力、热风等多种原料与资源的支持. 其工艺流程繁复, 设备种类众多, 设备–故障关系复杂, 使得为其进行故障诊断难度也大大增加. 为了实现精确高效的故障诊断, 本文结合上述钢铁产线故障诊断的领域特点, 在此基础上提出一种基于知识图谱推荐的钢铁产线故障诊断模型, 采用类似水波传播的方式在知识图谱中获取待诊断设备的多阶表示并聚合得到其深度表征以进行故障诊断, 并根据钢铁产线故障诊断的特性优化了其中水波集的构建过程以提升最终的诊断效果. 同时引入成对关系向量知识表示学习模型(PairRE)进行联合训练以学习知识图谱中的复杂关系. 最后通过某大型钢铁公司的热轧产线实际生产数据与多个代表模型进行对比试验和案例分析, 验证了所提出方法的科学性和有效性. |
| 英文摘要 |
| The production of steel requires a number of processes such as raw material transportation, smelting, hot rolling and cold rolling, and the completion of these processes also requires the support of various raw materials and resources such as electricity and hot air. The complicated process flow, many types of equipment and complex equipmentfault relationship make it much more difficult to perform fault diagnosis. In order to achieve accurate and efficient fault diagnosis, a fault diagnosis model based on knowledge graph (KG) recommendation is proposed based on the abovementioned characteristics of steel production line fault diagnosis, which uses a water wave propagation-like approach to obtain a multi-order representation of the equipment to be diagnosed in the KG and aggregates its depth representation for fault diagnosis. The construction process of the water wave set is optimized according to the characteristics of steel production line fault diagnosis to improve the final diagnosis effect. A KG embedding via paired relation vectors (PairRE) model is also introduced for joint training to learn the complex relationships in the KG. Finally, the scientific validity and effectiveness of the proposed method is verified by comparing the actual production data of a large steel company’s hot rolling line with several representative models in experiments and case studies. |
|
|
|
|
|