基于动态正交子空间分析的质量相关微小故障检测  

Quality-related incipient fault detection based on dynamic orthogonal subspace analysis

作　　者：潘滔 熊伟丽[1,2] PAN Tao;XIONG Weili(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China;Key Laboratory of Advanced Process Control for Light Industry(Ministry of Education),Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学物联网工程学院,江苏无锡214122 [2]江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122

出　　处：《高校化学工程学报》2024年第5期770-780,共11页Journal of Chemical Engineering of Chinese Universities

基　　金：国家自然科学基金(61773182,62003300);国家重点研发计划(2018YFC1603705-03)。

摘　　要：为进一步提升正交子空间分析(OSA)对过程微小故障的检测能力,提出一种基于动态OSA的故障检测方法。基于输入时滞值构建反映过程数据和质量相关数据动态特性的增广矩阵,采用OSA将增广矩阵分解成3个正交子空间,实现质量相关成分的有效分离;利用主成分分析对每个子空间进行独立监测,形成相应的得分矩阵;进一步通过Kullback-Leibler散度度量故障发生前后得分向量的概率分布差异,建立概率相关故障检测统计量,实现微小故障的检测。通过数值例子和田纳西伊斯曼(TE)过程的仿真实验验证了所提方法的有效性和检测能力。In order to further improve detection ability of orthogonal subspace analysis(OSA)for process incipient faults,a fault detection method based on dynamic OSA was proposed.An augmented matrix reflecting the dynamic characteristics of process and quality-related data was first constructed based on input delay values,and the augmented matrix was decomposed into three quadrature subspaces using OSA to achieve effective separation of quality-related components.Principal component analysis was then used to monitor each subspace independently,and the corresponding score matrix was formed.Kullback-Leibler divergence was used to measure the difference of probability distribution of score vectors before and after fault occurrence,and probability-related fault detection statistics was established to realize the detection of incipient faults.The effectiveness and detection capability of the proposed method are verified by numerical examples and Tennessee Eastman(TE)process simulation experiments.

关 键 词：故障检测 正交子空间分析 Kullback-Leibler散度 质量相关 微小故障 动态过程 

分 类 号：TP227[自动化与计算机技术—检测技术与自动化装置]