文献类型：会议论文

英文题名：Fault diagnosis of CTA hydrogenation process based on KPCA

作者：Li, Bo[1]; Li, Zhi[1]; Luo, Na[1]; Zhong, Weimin[1]

机构：[1] East China University of Science and Technology, Key Laboratory of Advanced Control and Optimization for Chemical Processes, Ministry of Education, Shanghai, 200237, China

会议日期：October 18, 2013 - October 21, 2013

会议地点：No. 130, Meilong Road, Shanghai, China

语种：英文

外文关键词：Principal component analysis - Statistics - Hydrogenation - Distributed parameter control systems - Fault detection

摘要：As the large amounts of operate data collected from Distributed Control System (DCS) often contain outliers and these data are more complexity and nonlinearity. They can't be used directly to model, optimization and fault diagnosis. In fault diagnosis, the existence of outliers can destroy the covariance structure of Kernel Principal Component Analysis (KPCA), which cause the model can't really reflect the actual normal condition. In this paper, KPCA method is adopted to establish the normal statistic monitor model from the historical data which can represent the normal industrial operate condition. First, the outlier detection algorithm is used to eliminate outliers among normal work condition. Then the primary statistic model for fault diagnosis of the Squared Prediction Error (SPE) and T2 are established according to the data exclude outliers. The effectiveness of this fault diagnosis is demonstrated by the operate data of industrial Crude Terephthalic Acid (CTA) hydrogenation process, and simulation results show that this method can identify the industrial failure condition.