Process Control Decision Inference, Monitoring, and Execution

Abstract

Approaches for inferring process control system decision trees from plant data have been heavily researched and demonstrated, but the utility of applying such decision tree inferences for autonomously monitoring, guiding, and executing control philosophies has been lacking. The authors have implemented an architecture leveraging model-based reasoning and graphical programming that investigates the utility of using control decision tree inferencing for validating, monitoring, and executing control strategies for both simple and complex process control problems. The techniques are potentially useful for assessing process control health, as well as extracting process control knowledge that may exist in daily operations but not recognized or well understood by analysts and management. Plant operators and managers can readily employ such insights for improving, augmenting, and extending process control system behavior, but can also potentially employ the refined and validated decision trees as a supervisory control layer on top of their existing control systems.