文章基本信息

标题：Multiple Fault Detection Using Multi-rate Probability Principal Component Analysis Models
本地全文：下载
作者：Le Zhou ; Junghui Chen ; Zhiqiang Ge 等
期刊名称：IFAC PapersOnLine
印刷版ISSN：2405-8963
出版年度：2017
卷号：50
期号：1
页码：14752-14757
DOI：10.1016/j.ifacol.2017.08.2581
语种：English
出版社：Elsevier
摘要：AbstractIn most chemical processes, both online measurements and offline laboratory analysis can be obtained at different sampling rates. Usually, the online process variables are frequently sampled while the key quality indicators are analyzed at irregular time, sometimes on hourly and sometimes daily basis. To effectively integrate different classes of measurements in a multi-rate process, a multi-rate probability principal component analysis (MPPCA) model is proposed to utilize the efficiently collected data and to improve the performance on both model prediction and process monitoring. In MPPCA, the model parameters are calibrated by the expectation-maximization algorithm. The proposed method is able to handle irregular samples in measurements and incorporate all the observations for model training. Also, the corresponding statistics based on MPPCA is developed for the fault detection purpose. Finally, a TE benchmark is presented to illustrate the effectiveness of the proposed methods.
关键词：KeywordsFault detectionEM algorithmMulti-rate processprobability principal component analysis