Abstract
In generalized minimum variance control (GMVC) and in the pole-placement control derived from GMVC, offset removal techniques against unknown disturbances are classified into two methods: (I) disturbance compensation method, and (II) integration method. Method (I) first estimates the disturbance and then compensates it with its estimate, while method (II) puts an integrator into the forward path of the control loop, where the integrator naturally removes offset. Most of the previous studies recommend method (II) rather than method (I), and method (II) is regarded almost as the standard method for offset removal. In this paper, we propose a new technique of method (I) using the disturbance observer and compares it with method (II) through mathematical analysis and numerical simulations. It is clarified that the method (I) with the proposed technique removes offset caused not only by various unknown disturbances but also by model errors. A numerical simulation with GMVC illustrates that the proposed technique is better than method (II) in output response and in robustness to model errors.
