Many processes behave like a linear dynamic model with a nonlinear static map at its input (in which case the process is referred to as a Hammerstein system) or at its output (resp. a Wiener system). It is standard industrial practice to mitigate the effects of this nonlinear element by the incorporation of its approximate inverse map at the output (respectively input) of the controller. This inverse map becomes part of the controller and must be tuned along with the other controller’s parameters (such as kp, ki, kd) of the controller. The use of IFT and VRFT for the automatic and model-free tuning of these nonlinear parameters will be the subject of this dissertation.
The student will follow courses on Linear Systems, Nonlinear Systems, Optimization, Stochastic Processes and System Identification. His/her work will include the implementation of the DD methods for the tuning of the nonlinear map in MatLab/Simulink, and assessment of the performance and of the statistical properties of the methods through the analysis of case studies, both simulated and practical.
