High-Pressure Polyethylene Process Monitoring Using PCA Based Bayesian Classification

Using PCA based Bayesian Classification to monitor the real plant with different operated conditions is proposed. Since the process condition s are time-variant, as the PCA subspace cannot explain the data of new events, the PCA should be reperformed. In this work the method of updating Bayesian model is developed. Only the data of new events are trained in the newer subspace. The ability of PCA based Bayesian classification for monitoring different operat ing conditions is demonstrated using the real data from high-pressure polyethylene plant. Copyright © 2004 IFAC