Distributed plant-wide process monitoring based on PCA with minimal redundancy maximal relevance

- A new distributed monitoring scheme is developed for plant-wide processes.
- The process can be divided into different subblocks automatically.
- A new block division method is presented by minimal redundancy maximal relevance.

Generally, the plant-wide processes involve abundant measured variables that have complex relationship, and considerable information of variables should be concerned in the process of block division. In this paper, a new distributed monitoring scheme that integrates minimal redundancy maximal relevance (mRMR), Bayesian inference and principal component analysis (PCA) is proposed for plant-wide processes. This method considers not only the relevance between variables, but also their redundancy in block division. Once sub-PCA monitoring models are built in each subblock, the overall results are combined through the Bayesian inference. With a numerical example and the Tennessee Eastman benchmark process, the effectiveness of the proposed method is demonstrated.