Data-driven Control Loop Diagnosis: Dealing with Temporal Dependency in Bayesian Methods*

AbstractConventional Bayesian methods commonly assume that the evidences are temporally independent. This condition does not hold for most practical engineering problems. With evidence transition information being considered, the temporal domain information can be synthesized within the Bayesian framework to improve the diagnosis performance. A data-driven algorithm is developed to estimate the evidence transition probabilities. The application in a pilot scale process is presented to demonstrate the data dependency handling ability of the proposed approach.