A hierarchical granger causality analysis framework based on information of redundancy for root cause diagnosis of process disturbances

Abstract

The Granger causality (GC) test is a widely utilized method for diagnosing process disturbances’ root cause. However, its effectiveness is limited due to the challenge of handling variable redundancy, leading to potentially inaccurate results. To tackle this problem, this paper introduces a new redundancy detection technique incorporated into the GC test framework. The method introduces a sum-of-redundancy index, enabling the development of a variable-layering approach to identify the sequential impact of process variables during disturbances. Furthermore, a hierarchical analysis framework is established, within which GC tests are applied both within and between layers, effectively revealing the propagation paths of process disturbances. Case studies conducted on a benchmark simulation process demonstrate the enhanced accuracy and feasibility of the proposed framework.