Double Layer Distributed Process Monitoring Based on Hierarchical Multi-Block Decomposition

For the purpose of monitoring large-scale distributed processes, a double layer fault detection method based on hierarchical multi-block decomposition is proposed. The process variables are first divided into multiple blocks using mutual information-based hierarchical decomposition. The Gaussian and non-Gaussian information in each block are divided into two layers using the partial least squares and a non-Gaussian regression method, respectively. The corresponding monitoring statistics constructed for both the Gaussian and non-Gaussian components are then integrated using the support vector data description. The performance of the proposed method is demonstrated by the application studies to a numerical example and the Tennessee Eastman (TE) benchmark process. The results show that the superiority of the proposed method over conventional methods.