Operator Interaction with Model-based Predictive Controllers in Petrochemical Refining

In this paper, we compare the advanced automation characteristics of Model-based Predictive Controllers (MPC) employed in process control to the more thoroughly studied Flight Management System (FMS) employed in airline cockpits. Our analysis shows that operator interaction with cockpit automation and MPC automation may differ in terms of the level of the control loops at which human control actions are directed. Cockpit automation appears to afford human control actions at outer loops through the specification of a flight path or selection of control modes and flight-path parameters. Inner loops, such as the control of flight surfaces, are controlled by the automation. In contrast, MPC affords operator control at the control variable level because operators are required to manipulate limits on those variables. Outer loops are managed by the automation. These differences in level may explain why we observed refinery operators asking questions of the automation pertaining to control intervention whereas reported studies of pilots’ interaction with cockpit automation do not. This could also explain why we did not observe refinery operators having difficulty with automated mode changes, whereas reports from FMS interaction show that mode discrimination is a difficult and challenging task.