Feedback control is a basic principle that is employed in all natural and in human-made systems to ensure survival and safe and efficient operation. By using sensor information to adapt the degrees of freedom of the system, complex systems can be operated stably despite the influence of varying external factors and for varying and partly unknown internal dynamics. On the other hand, feedback control can lead to oscillations and instability, and the effect of the control law on the behavior of the controlled system is not straightforward which is why the design of feedback controllers has fascinated generations of control scientists since the 19th century.
Starting from the design of feedback controllers for linear and nonlinear plants with the goal to achieve a good dynamic behavior of the controlled system, our focus of research has evolved to controller design for optimal plant performance. The key idea behind our work on process control is that the purpose of process control is to improve the operation of the plant, e.g. to consume less energy and raw materials, to avoid the production of off-spec material, to maximize the throughput, and to minimize the environmental impact. Controlling some variable to nicely track its set-point can contribute to this goal, but may have little effect and can even be counterproductive.
There are several ways of optimizing plant performance by feedback control: choosing a suitable control structure and suitable controlled and manipulated variables, tracking necessary conditions of optimality (as e.g. operating at the limits of the cooling power), stationary optimization of the set-points and implementation by tracking controllers, and direct optimizing control where the tracking criterion is replaced by an economic criterion, and adding the specifications of product purities and equipment limitations as constraints to the optimization problem. In recent years, we have worked on and applied all four approaches, with the main effort being devoted to direct optimizing control formulations and applications to chemical engineering problems.
On the theoretical side, the main theme of our work is how to cope with the inaccuracies of the models that are used in model-based control. Three main lines of research are being currently pursued:
On the applications side, we work on reactive distillation, chromatographic separations, polymerizations in semi-batch and continuously operated reactors, and on intensified processes.
MOBOCON is an ERC Advanced Investigator Grant Project that was awarded to Prof. Sebastian Engell of the dyn group (PI) and Prof. Hans-Georg Bock of IWR Heidelberg in 2012. The goal of this generously funded project is to overcome the obstacles for the industrial application of optimization-based control by developing methods for the enhanced robustness and failure resilience of model-based optimizing control solutions and for the reduction of the modeling effort, which is the major bottleneck in the development of optimizing controllers. Additionally, MOBOCON addresses the interaction with the plant operators, with the goal to achieve a symbiosis of optimization and human capabilities to detect and react to unforeseen situations. The group at IWR contributes advanced robust numerical solution algorithms for dynamic optimization problems as they arise in optimizing control. The results will be demonstrated for a challenging reactive distillation process in a pilot plant at the BCI department.
Engell, S.: Feedback Control for Optimal Process Operation. J. of Process Control, 17(3), 2007, 203–219.