工业机器人预测式健康管理本体半自动化构建  

Semi-automatic Construction of Predictive Health Management Ontology for Industrial Robots

作　　者：柳少峰 肖红[1] 黄子豪 姜文超[1,2] 熊梦 贺忠堂 LIU Shao-feng;XIAO Hong;HUANG Zi-hao;JIANG Wen-chao;XIONG Meng;HE Zhong-tang(School of Computer Science and Technology,Guangdong University of Technology,Guangzhou 510006,China;Chinese Academy of Cloud Computing Industry technology Innovation and Incubation Center,Dongguan 523808,China)

机构地区：[1]广东工业大学计算机学院,广州510006 [2]中国科学院云计算产业技术创新与育成中心,东莞523808

出　　处：《组合机床与自动化加工技术》2022年第1期29-33,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：2019年佛山市核心技术攻关项目(1920001001367);国家自然科学基金-广东省联合基金项目(U2001201);广东省自然科学基金面上项目(2020A1515010890,2018A030313061);国家重点领域研发计划项目(2018YFB1004202);广东省科技计划项目(2019B010139001);广州市科技计划项目(201902020016)。

摘　　要：针对工业机器人预测式健康管理(PHM)案例信息缺乏深度利用、PHM本体人工构建、过程不智能、构建成本高、本体不完备等问题,提出一种工业机器人PHM本体半自动化构建方法。首先,对多源工业机器人PHM语料进行分词、去除停用词等预处理;其次,使用融合词频、文档频率、TF-IDF、C-value等算法进行本体概念综合抽取;接着,基于CSC语义词库和搜索引擎进行概念间上下属关系抽取,同时基于SAO结构进行交叉关系抽取;最后,使用Protégé工具对工业机器人PHM本体进行持久化与可视化。实验平台采用某国产机器人设备,测试数据为机器人PHM记录以及简书博客、维普期刊获取的相关文档共1690篇,与TF-IDF、C-value、LDA主题模型、BRT等算法进行实验对比,该方法在概念抽取阶段准确率提升10%,在概念关系抽取阶段提升3%,实际应用结果表明本方法有效可行。Aiming at the problems of the lack of deep utilization of the case information of predictive health management(PHM),artificial construction of PHM ontology, unintelligent process, high construction cost and incomplete ontology, a semi-automatic construction method of industrial robot PHM ontology is proposed.Firstly, the PHM corpus of multi-source industrial robot is preprocessed by word segmentation and removal of stop words.Secondly, the algorithm of fused word frequency, document frequency, TF-IDF and C-value is used for ontology concept synthesis extraction.Then, the superior subordinate relationship between concepts is extracted based on CSC semantic lexical base and search engine, and the cross relationship is extracted based on SAO structure.Finally, the PHM ontology of industrial robot is persisted and visualized by Protégé tool.The experimental platform uses a domestic robot equipment, and the test data is provided by the enterprise with robot PHM records and related documents obtained from Simple book blog and VIP journal, with a total of 1690 articles.Compared with TF-IDF,C-value, LDA topic model, BRT and other algorithms, the accuracy of this method is improved by 10% in the concept extraction stage, and by 3% in the concept relationship extraction stage.The practical application results show that this method is effective and feasible.

关 键 词：工业机器人 预测式健康管理 本体 概念抽取 关系抽取 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]