(574h) Robust Nonlinear Model Predictive Control of An Augmented-Reality Lab-Scale Plant

Model predictive control (MPC) turns out to become the current standard in latest industrial applications. The achieved improvements arising from replacement of outdated PID-control-schemes with MPC-systems confirm this evolution. However, the common approach by now is to utilize linear models to predict the future process behavior. This method has some well known drawbacks e.g. in case of strong nonlinearity of the process, or a large region of operation [1]. These deficiencies become an issue, because the need for matching higher and stringent quality standards is continuously increasing. Therefore current research focuses on nonlinear model predictive control (NMPC) to overcome limitations of linear MPC.

The objective of this work is the application of Robust NMPC to a lab-scale plant of an industrial continuously stirred tank reactor (CSTR) process. The existing assets are extended with a simulation layer, which results in an augmented reality plan. The chemical reaction (reactor hold-up) is the virtual part, represented by a simulation that contributes to the heat exchange process (reactor jacket and heat exchanger) in the real world. Finally, the NMPC-controller is used to control the temperature in the virtual reactor hold-up using the heat exchange process of the real world. With this setup, it is easy to introduce different levels of uncertainty, disturbances and failures into the system. However, excluding the part of the chemical reaction still leaves enough non-artificial phenomena and nonlinearity in the system. It is known that a heat exchanger process possesses strong dependency of the coolant temperature and flow, as well as on the product temperature and flow.

The design of the augmented-reality plant is illustrated by reviewing the design of the robust NMPC-scheme. For the explicit consideration of uncertainty a stochastic optimization framework is used [2]. The results are discussed concerning robustness against model parameter uncertainty, process uncertainty and overall control performance.

[1] HENSON, M.A.: Nonlinear model predictive control: current status and future directions. Comp. & Chem. Eng., Vol. 23, No. 2, pp 187-202 (1998.

[2] ARELLANO-GARCIA, H.: Chance constrained optimization of process systems under uncertainty. PhD Thesis (2006)