Knowledge elicitation and structuring for a real-time expert system for monitoring a butadiene extraction system

Abstract A real-time expert system for monitoring a butadiene purification process it has been developed using ONSPEC ™ data acquisition and data processing capabilities. In this paper, the monitoring activities are focused on the two core columns: an extractive distillation column and a solvent stripping column working together in a highly integrated manner through material recycling and energy recovery. The paper emphasizes the steps followed during knowledge elicitation and organization, as well as the interrelation with the expert during these processes. The knowledge elicitation methodology has been mostly devised in a top-down style using HAZOP (HAZard and OPerability analysis), CCA (Cause and Consequence Analysis) and FMEA (Failure Mode and Effect Analysis) techniques. These methodologies provide an efficient means for motivating the experts to think aloud, from a more abstract to a more detailed level of process state understanding, about operational problems and actions or recommendations to restore control and performance. The knowledge acquired has been chunked into sets of rules which are coordinated by a core chunk using a “token-ring” style. This architecture proved to be successful for working in a real-time environment and from the point of view of the expert system expansion.