Comparing PCA-based fault detection methods for dynamic processes with correlated and Non-Gaussian variables
作者：
Highlights：
• The article concerns fault detection optimization in engineering systems.
• An evaluation of different PCA-based methods is carried out.
• Two novel methods are proposed and compared with a third well-described method.
• Maximizing FDR and minimizing FAR is the evaluation process objective function.
• The best method for different contexts is defined by applying the TOPSIS method.
摘要
•The article concerns fault detection optimization in engineering systems.•An evaluation of different PCA-based methods is carried out.•Two novel methods are proposed and compared with a third well-described method.•Maximizing FDR and minimizing FAR is the evaluation process objective function.•The best method for different contexts is defined by applying the TOPSIS method.
论文关键词：Fault Detection Methods,Principal Component Analysis,Moving Window Principal Component Analysis,Referenced Moving Window Principal Component Analysis,TOPSIS
论文评审过程：Received 13 June 2021, Revised 25 June 2022, Accepted 25 June 2022, Available online 5 July 2022, Version of Record 9 July 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.117989
