A Test Environment to Evaluate the Integration of Operators in Nonlinear Model-Predictive Control of Chemical Processes

Abstract

Abstract: In industrial process control, basic controllers such as linear PI(D) control and cascade controllers are widely accepted by the plant operators because they are easy to understand due to their clear cause-and-effect relation. When the control and performance objectives cannot be met by simple controllers and manual interventions, advanced process control (APC) solutions, usually MPC controllers are employed. The design of APC solutions requires a considerable understanding of the process and more mathematical background—but also during the operational phase (e.g. error diagnosis). Then the issue of the trust of the operators into such complex control solutions and in their ability to monitor their behavior during plant operation arises. It may happen that APC solutions are not accepted or are fully or partly switched off by the operators after a while, which means a wasted effort and lost opportunities for better plant performance. By providing carefully chosen information about the behavior of the controller and well-designed operator interfaces, the trust of the operators and their ability to monitor advanced controllers can be increased. In this contribution we present an environment to investigate trust in automation experimentally and to explore new opportunities for the user-interaction with APC methods. The test environment consists of a chemical process simulator which is controlled online by a NMPC scheme, an interface editor for the fast development of interface concepts as well as a control room simulation in order to evaluate the synergetic cooperation between the operator and the algorithmic control schemes.