Industrial Process Monitoring Based on Parallel Global-Local Preserving Projection with Mutual Information

This paper proposes a parallel monitoring method for plant-wide processes by integrating mutual information and Bayesian inference into global-local preserving projections (GLPP)-based multi-block framework. Unlike traditional multivariate statistic process (MSPM) methods, the proposed MI-PGLPP transforms several sub-block monitoringtasks fully taking advantage of distributed First, original datasets are divided group data blocks quantifying variables. The block indexes new generated automatically. Second, each is modeled GLPP method. variable local structure well preserved during whole projection. Third, introduced to generate final statistics probability To illustrate algorithm performance, detailed case study performed on Tennessee Eastman process. Compared with principle component analysis GLPP-based method, provides higher FDRs superior performance monitoring.