基于知识图谱的数控机床故障问答系统研究  

Research of question-answer system for CNC machine fault on knowledge graph

作　　者：王乾龙 王丽颖[1] WANG Qianlong;WANG Liying(Information Engineering School,Inner Mongolia University of Science and Technology,Baotou 014010,China)

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014010

出　　处：《内蒙古科技大学学报》2023年第4期372-376,393,共6页Journal of Inner Mongolia University of Science and Technology

摘　　要：当前数控机床故障方面的查询机制尚不完善,机床工人在查找故障原因所在时往往需要花费较多的时间.为改善现状,从内蒙古某机械厂的数控机床故障数据中抽取其三元组,构建数控机床故障知识图谱,在此基础上开发知识图谱问答系统.构建数控机床故障知识图谱问答一般会包括4个步骤:数控机床故障问句实体识别;数控机床故障属性映射;数控机床故障实体链接;答案返回.设计ALBERT+Attention+BiLSTM+CRF模型识别问句中的数控机床故障实体;然后在所创的数控机床故障知识图谱中找到对应实体的三元组进行返回;通过训练属性映射模型,选择得分较高的几个三元组作为预选答案,最后通过实体链接重排序返回答案.在内蒙古某机械厂数控机床故障知识图谱数据集上分析并验证了所提方法的有效性.The current query mechanism for CNC machine faults is not perfect,machine workers often need to spend more time to find the cause of the fault.To improve the current situation,this study extracted its triples from the CNC machine fault data of a machinery factory in Inner Mongolia,constructed the CNC machine fault knowledge graph,and developed the knowledge graph Question-Answer System on this basis.The construction of CNC machine fault knowledge graph generally includes four steps:CNC machine fault question sentence entity detection;CNC machine fault attribute mapping;CNC machine fault entity link and answer return.An ALBERT+Attention+BiLSTM+CRF model was desigmed to identify the fault entity of CNC machine in the question sentence.Then the triplet of the corresponding entity in the created CNC machine fault knowledge graph was found and returned.Through the training attribute mapping model,several triples with high scores was selected as the preselected answers,and finally the answers was recorded through entity links.The effectiveness of the proposed method was analyzed and verified on the data set of CNC machine fault knowledge graph of a machinery factory in Inner Mongolia.

关 键 词：数控机床 知识图谱 实体识别 问答系统 属性映射 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]