基于模型的机床故障案例命名实体抽取方法比较研究
Comparative Study on Model-Based Named Entity Extraction Methods for Machine Tool Fault Cases
尹昱东 1王保建 1李珂嘉 1王紫平 1张小丽2
作者信息
- 1. 西安交通大学 机械工程学院,西安 710049
- 2. 长安大学道路施工技术与装备教育部重点实验室,西安 710064
- 折叠
摘要
机床出现的故障大多有先例,但故障案例分散,不同工厂又不数据共享且没有标准的数据库管理,以至于对于已有的相似故障,工厂仍需要按照未知故障进行停机维修;因此,急需一套标准服务平台能够集合大量故障案例,同时实现更新维护,增添新故障,以供各工厂做故障参考,尽可能降低维修成本以及时间开销;通过将计算机领域较为流行的知识图谱运用到机床故障诊断领域,全面运用机床故障诊断案例知识,构建以故障现象、故障发生原因以及解决方案为核心的机床故障诊断网络,实现快速确认故障发生部位,提供合理的故障解决方案,提高制造业的生产效率;使用爬虫技术获取故障案例数据,采用BIO标注法完成样本标注,分别使用Bilstrm-crf、Vgg16以及Bert模型完成实体抽取任务,并对上述模型准确率从多个角度进行对比,将知识导入Neo4J图数据库并建立针对机床故障的知识图谱,最终实现知识图谱可视化.
Abstract
Most machine tool failures have the characteristics of scattered fault cases,no data sharing in different factories,and no standard database management.Aimed at existing similar failures,it is still necessary to shut down for maintenance in factories by unknown failures.Therefore,it is urgent for a standard service platform to collect a large number of fault cases,update maintenance and add new faults at the same time,and provide a fault reference for each factory and reduce maintenance and time costs as much as possible.This paper applies the popular knowledge graphs in the computer field to the fault diagnosis networks of machine tools,comprehensively uses the case knowledge of machine tool fault diagnosis,and constructs a machine tool fault diagnosis network with fault phenomena,reasons and solutions as a core to quickly identify the fault location,provide reasonable fault solutions,and improve the production efficiency of manufacturing industry.The crawler technology is used to obtain the fault case data,the BIO tagging method is used to complete the sample tagging,the Bilstrm-crf,Vgg16 and Bert models are used to complete the entity extraction task,respectively,and the accuracy of the above models are compared from multiple aspects,the knowledge is imported into the Neo4J diagram database,it builds the knowledge graph for the machine tool fault,and finally realizes the visualization of knowledge graph.
关键词
机床故障/知识图谱/Bert模型/Neo4j图数据库/命名实体抽取Key words
machine tool fault/knowledge graph/Bert model/Neo4J graph database/named entity extraction引用本文复制引用
基金项目
陕西省自然科学基础研究计划(2021M-169)
陕西省自然科学基础研究计划(2023-JC-YB-477)
西安交通大学本科实验实践与创新创业教育教学改革研究专项(2022)(22SJZX10)
出版年
2024
NETL
NSTL
万方数据