Decentralized plant-wide monitoring based on mutual information-Louvain decomposition and support vector data description diagnosis.

A decentralized fault detection and diagnosis method is proposed to monitor the nonlinear plant-wide processes effectively. It includes two theme activities: mutual information-Louvain based process decomposition and support vector data descriptions (SVDD) based fault diagnosis. Firstly, the plant-wide process is preliminarily map as an undirected graph corresponding to the mechanism knowledge and process structure. Mutual information (MI) is introduced to depict the correlation degree between different nodes (i.e., process variables), and a Louvain algorithm with MI correlation is proposed to fine decompose the process into reasonable sub-blocks. Then, decentralized SVDD based fault detection method is presented for each sub-block, and the corresponding variable contribution rate is derived. Finally, a Bayesian fusion inference is given to evaluate the detection results of all sub-blocks in an integrated manner. The proposed method is verified in the Tennessee-Eastman (TE) process.