Generalized Dice's coefficient‐based multi‐block principal component analysis with Bayesian inference for plant‐wide process monitoring

摘要: Plant-wide process monitoring is challenging because of the complex relationships among numerous variables in modern industrial processes. The multi-block method an efficient approach applied to plant-wide However, dividing original space into subspaces remains open issue. loading matrix generated by principal component analysis (PCA) describes correlation between and extracted components reveals internal relations within process. Thus, a PCA that constructs (PC) sub-blocks according generalized Dice coefficient proposed. PCs corresponding similar vectors are divided same sub-block. sub-block share variational behavior for certain faults. This improves sensitivity A statistic T2 each produced integrated final probability index based on Bayesian inference. contribution plot also developed identify root cause. superiority proposed demonstrated two case studies: numerical example Tennessee Eastman benchmark. Comparisons with other PCA-based methods provided. Copyright © 2014 John Wiley & Sons, Ltd.