Modellbasierte Prädiktive Regelung skalierbarer thermischer Prozesse

  • Model-Based Predictive Control of Scalable Thermal Processes

Zöller, Daniel; Abel, Dirk (Thesis advisor); Horn, Martin (Thesis advisor)

Aachen (2018)
Dissertation / PhD Thesis

Dissertation, RWTH Aachen University, 2018


Thermal processes are nowadays used in various fields of technology. They all have in common that the temperature is an important process variable, which often has to be controlled with maximum precision. In an industrial environment, these systems are usually controlled using conventional PID control circuits, which are optimized for the specific requirements of the respective system. In this case, however, there is always a compromise between good command response and sufficient interference suppression. In addition, an exact parameterization of the controller is very time-consuming, in particular in the case of slow plants, and leads to a moderately dimensioned control behavior. In addition, an exact parameterization of the controller is often very time-consuming, in particular in the case of slow plants, whereby this is often operated away from its optimum setting. In this thesis, a model-based predictive control concept (MPC) for optimizing the dynamic behavior of thermal processes is presented and then evaluated by implementation on two industrial processes. The presented concept uses a non-linear dynamic process model of the plant and consists of a combination of nonlinear state estimator in the form of an Extended Kalman filter and a linear time-variant MPC. The controller is based on a physical model for simple and intuitive transferability. Due to the high importance of the process model, an introduction to modeling is given at the beginning of the work and the basic physical principles for the description of thermal processes are presented. Thus, a corresponding process model can be developed in a simple and comprehensive manner and can be adapted to the different processes. In order to validate this control concept, the methodology described above is first applied to thermal high-vacuum evaporation. In a first step, the analysis and development of a dynamic process model for temperature behavior is performed. In a second step, parameter identification is performed to provide the model equations. The implementation of the MPC is then carried out on a pilot plant. In order to evaluate the transferability, the methodology is applied to thermal conditioning and environmental simulation, the second process considered here. First of all, the control of the air temperature is performed. For this purpose, the thermal behavior of the system is analyzed and modeled to a sufficiently precise degree. For the testing of the control concept, the control algorithm is implemented directly in the process control hardware. Then, the previously developed process model is expanded by the physical influence of the relative humidity and the control algorithm is adapted. In a last step the algorithm is transferred to a similar system of the same series, in order to examine the scalability. The control concept is evaluated by a direct comparison of the measurements of PID control and the MPR. The aim of this thesis is, in addition to the necessary basics principles of model-based predictive control of thermal processes, to describe the essential development steps from modelling to the implementation at the real plant and to give the necessary knowledge about controller maintenance and further development.