Process Monitoring and Fault Diagnosis Based on a Regular Vine and Bayesian Network

This paper proposes a process monitoring and fault diagnosis method based on a regular vine (R vine) and Bayesian network. The R vine model structure is determined by searching for the maximum sum of combinations of correlations among variables, which makes the model more robust and able to describe data more flexibly. A double-space strategy based on the R vine is used to detect the process fault, which can improve the ability to detect weak faults. Furthermore, a Bayesian network is built according to the first tree of the R vine model to diagnose the detected fault and find the root cause. The causality between the nodes of the Bayesian network is determined via the Granger test. The effectiveness of the proposed method is verified by numerical examples and industrial examples.