Distributed Supervised Fault Detection and Diagnosis for a Non-Gaussian Process

In this paper, a novel method named distributed independent component-principal component regression (distributed ICPCR) is proposed to monitor the large-scale non-Gaussian process. First, multiple sub-blocks are obtained, and the key process variables are selected for the following distribution monitoring. Second, the monitoring model of each sub-block is constructed on the basis of the proposed ICPCR method. In this algorithm, the total latent space which contains both Gaussian and non-Gaussian characteristics is constructed, and the regression model between the total latent variables and quality variables is established for the quality-related monitoring. Afterward, the global monitoring result can be obtained by the Bayesian fusion strategy. Third, a probability-based method is proposed to determine the fault sub-blocks, and the ICPCR-based relative contribution plot is presented to locate the fault variables. Finally, a numerical example and the Tennessee Eastman process are used to demonstrate the effectiveness of the proposed method.