引用本文:孔祥玉,罗家宇,杜柏阳,等.基于递推MPLS算法的质量相关故障在线监控技术[J].控制与决策,2020,35(9):2094-2102
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览次   下载 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于递推MPLS算法的质量相关故障在线监控技术
孔祥玉,罗家宇, 杜柏阳, 曹泽豪
(火箭军工程大学导弹工程学院,西安710025)
摘要:
改进潜结构投影(MPLS)算法是一种反映过程变量与质量变量相关关系的多元统计分析方法,已有效应用于稳态过程的故障监控.对于缓时变工业系统,MPLS模型难以描述当前过程,因此,需定时更新模型以加强对当前过程的监控.常用的模型更新方式是数据扩充的方法,然而该方法重复使用历史数据导致建模样本不断积累,模型更新效率非常低下.为提高模型动态更新效率,提出递推改进潜结构投影(RMPLS)算法,采用递推结构动态更新MPLS模型,与数据扩充方法相比避免了样本累积,极大地提高了模型更新效率.最后,在田纳西-伊斯曼过程中比较RMPLS和MPLS的模型更新计算量和质量相关故障检测效果,结果表明RMPLS可有效降低模型更新计算量,并全面提高质量相关故障的监测能力.
关键词:  故障检测  质量相关  偏最小二乘  改进潜结构投影  模型更新
DOI:10.13195/j.kzyjc.2018.1738
分类号:TP273
基金项目:国家自然科学基金项目(61673387,61833016);陕西省自然科学基金项目(2020JM-356).
Quality-related fault online monitoring technology based on recursive MPLS algorithm
KONG Xiang-yu,LUO Jia-yu,DU Bo-yang,CAO Ze-hao
(College of Missile Engineering,Rocket Force University of Engineering,Xián 710025,China)
Abstract:
The modified projection to latent structures (MPLS) algorithm is a multivariate statistical process monitoring method reflecting the relationship between process variable and quality variable, which is effectively applied to fault monitoring of the steady state process. For time-varying industrial systems, the MPLS model is difficult to describe the current process effectively, so it is necessary to update the model to enhance the monitoring of the current process. The commonly used model update method is a data expansion method. However, the repeated use of historical data in this method leads to the accumulation of modeling samples and the inefficiency of model updating. In order to improve the dynamic update efficiency of the model, this paper proposes a recursive-MPLS (RMPLS) algorithm, which uses the recursive structure to dynamically update the MPLS model with new data. Compared with the data expansion method,RMPLS avoids sample accumulation and greatly improves the model update efficiency. Finally, in Tennessee-Eastman process, comparing the model updating calculation and the quality-related fault monitoring effect of the RMPLS and the MPLS, the results show that the RMPLS effectively reduces the model updating calculation, and improves the quality-related fault monitoring capability.
Key words:  fault detection  quality-related  partial least squares  modified latent structure projection  model update

用微信扫一扫

用微信扫一扫