Feedback-linearization using neural process models

This work shows that feedback linearization using neural process models can successfully be applied to nonlinear systems, even to systems with a relative degree higher than one. A highly nonlinear and complex batch polymerization process of relative degree two was used as an example for a simulation case study. The unknown system dynamics was approximated by neural networks, using only measured state variables of one batch for training. Based on the learned model, a feedback-linearization was designed which significantly improved the control performance compared to conventional control