Combining model based and data-driven approaches: First order nonlinear plants control

Abstract

The paper deals with a data-based robust tuning of a controller for nonlinear first order plants combining exact linearization based nonlinearity compensation with a nonlinear disturbance observer (NDOB) based integral action. A two-step experimental control design is carried out with the aim of the fastest possible monotonic setpoint step responses and of a well damped control signal courses. The first stage of an optimal and robust tuning procedure focuses on collecting and processing data about the process by a sequence of step responses. It is used for determining appropriate parameters of the stabilizing P control extended by a fixed offset. This may be extended by a second stage, when the proposed controller is used to generate new data sets for an optimal robust disturbance observer tuning. All procedures are illustrated by an electrical fan speed control example carried out in a Matlab/Simulink environment.