A quality‐related distributed process monitoring framework for large‐scale manufacturing processes with multirate sampling measurements

Abstract

Quality‐related process monitoring has become a research hot‐spot in the field of industrial control because it is essential to ensure the process safety and product quality. A great number of data driven quality‐related process monitoring methods have been developed for large‐scale manufacturing processes, and most of them are developed based on homogeneous sampling measurement. Therefore, it is necessary to develop quality‐related monitoring methods for large‐scale processes with multirate sampling measurements. In this paper, a new quality‐related distributed monitoring framework for large‐scale manufacturing processes with multirate sampling measurements is proposed. First, a new subsystem decomposition method for multirate sampling processes combining prior knowledge and mutual information is proposed by introducing mathematic expectations. Second, local monitoring model is designed for each subsystem. Multirate partial least squares regression is adopted for modeling among the process and quality variables. The monitoring metrics of the isolation‐based anomaly detection using nearest‐neighbor ensembles are built in prediction space and process space, respectively. Finally, Bayesian inference is introduced to obtain statistical indicators for the plant‐wide processes. The validity of the proposed framework is verified in Tennessee Eastman process and a real hot strip mill process. The results show that the proposed method has favorable effectiveness and significant performance gains compared with state‐of‐the‐art methods.