文献类型：期刊文献

英文题名：Dynamic process monitoring using adaptive local outlier factor

作者：Ma, Yuxin[1];Shi, Hongbo[1];Ma, Hehe[1];Wang, Mengling[1]

机构：[1]E China Univ Sci & Technol, Key Lab Adv Control & Optimizat Chem Proc, Minist Educ, Shanghai 200237, Peoples R China

年份：2013

卷号：127

起止页码：89

外文期刊名：CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS

收录：;EI(收录号：20235015204333);WOS:【SCI-EXPANDED(收录号:WOS:000324011300012)】；

基金：This research is supported by the Shanghai Leading Academic Discipline Project (No. B504), National Nature Science Foundation of China (No. 61203059), Shanghai Postdoctoral Sustentation Fund (No. 12R21412600), and the Fundamental Research Funds for the Central Universities (No. WH1214039).

语种：英文

外文关键词：Time-varying; Multimode; Non-Gaussian; Local outlier factor; Moving window; Fault detection

摘要：A numerically efficient moving window local outlier factor (LOF) algorithm is proposed in this paper for monitoring industrial processes with time-varying and multimode characteristics. The key feature of the proposed algorithm can be identified as its underlying capability to handle complex data distributions and incursive operating condition changes including both slow dynamic variations and instant mode shifts. With some updating of the rules developed for accelerating the computation speed, a two-step adaption approach is introduced to keep the monitoring model up-to-date. Then, a switch strategy and an update termination rule are designed to deal with operating mode changes. Due to the utilization of local information, the proposed algorithm has a superior ability both in detecting faulty conditions and fast adapting to new operating modes. Finally, the utility of the proposed method is demonstrated through a numerical example and a non-isothermal continuous stirred tank reactor (CSTR). (C) 2013 Elsevier B.V. All rights reserved.