Variable Time Delay-Based Granger Causality Approach Integrated with Dynamic Coupling Analysis for Root Cause Diagnosis in Chemical Processes

Abstract

Due to the dynamic characteristics of chemical industry systems, the time delay tends to be variable, which leads to changes in coupling intensity. This is contrary to the assumptions in causal analysis, where the time delay and the coupling are typically assumed to be fixed. In this article, a new causal analysis framework that integrates Granger causality with dynamic coupling analysis based on variable time delay is proposed, which not only fully considers the variable time delay in dynamic processes, but also studies the dynamic change of coupling intensity. First, the moving window is used to explore real-time variations in average mutual information to obtain the variable time delay. Then by further analyzing the normal data, the dynamic coupling relationship caused by continuous changes in time delay is distinguished. On this basis, the extended Granger causality and convergent cross mapping are integrated to relax their assumptions of fixed time delay and coupling. Finally, the direction of fault propagation is guided by the results of causal analysis. The effectiveness of proposed method is demonstrated by chemical industry case studies.