基于密度的离群点检测技术在中厚板轧后冷却中的研究  

Research on density-based outlier detection in cooling after rolling

作　　者：黄卫国[1] 金超 张田[2] HUANG Wei-guo;JIN Chao;ZHANG Tian(Technical Reform Office of Shagang Project Jiangsu Shagang Group Co., Ltd.,Zhangjiagang 215625 ,China;The State Key Laboratory of Rolling and Automation, Northeastern University,Shenyang 110819 ,China)

机构地区：[1]沙钢项目技改办江苏沙钢集团有限公司,江苏张家港215625 [2]东北大学轧制技术及连轧自动化国家重点实验室,辽宁沈阳110819

出　　处：《冶金自动化》2019年第3期7-12,共6页Metallurgical Industry Automation

基　　金：中央高校基本科研业务费资助项目(N170703010)

摘　　要：以沙钢3 500 mm中厚板超快冷控制系统为研究对象,系统介绍了该超快冷控制系统的相关组成部分以及所记录的数据特点。针对数据库中所记录数据价值密度低、数据量巨大、"不清洁"等特点,采用基于密度的离群点检测技术对数据进行处理,在空间中筛选出离群因子较大的"异常点"并剔除,以使数据能够完全、充分地反映实际生产过程,在提高数据精度的同时也为后续的钢板生产、故障分析以及规则挖掘提供了必要的保障。In this paper,the ultra-fast cooling control system of Shagang’s 3 500 mm plate is taken as the research object. The relevant components of the ultra-fast cooling control system and the recorded data characteristics are introduced systematically. Aiming at the characteristics of low value density,large amount of data and "unclean"in the data recorded in the database,the density based outlier detection technology is used to process the data. The "abnormal points"with large outliers are selected and eliminated in the space to fully reflect the actual production process. The accuracy of the data is improved and it also provides the necessary guarantee for subsequent steel plate production,failure analysis and mining rules.

关 键 词：离群点检测 密度 数据 精度 超快冷控制系统 

分 类 号：TG335.52[金属学及工艺—金属压力加工]