Dynamic Multimode Process Modeling and Monitoring Using Adaptive Gaussian Mixture Models

Abstract

For multimode processes, it is inevitable to encounter disturbances, such as equipment aging, catalyst deactivation, sensor drifting, reaction kinetics drifting, or adding new operating modes. The existing monitoring algorithms are established either for coping with multimode feature under time-invariant circumstance or for handling the time-varying problem of processes with single operating mode. The purpose of this article is to develop an effective modeling and monitoring approach for complex processes with both multimode and time-varying properties. We propose a novel adaptive monitoring scheme based on Gaussian Mixture Model (GMM). The new method is able to model different operating modes as well as trace process variations. The effectiveness and efficiency of the new method are validated by a numerical example and the Tennessee Eastman (TE) simulation platform in different scenarios.