Partially Coupled Inverse Model Design for Multivariable Disturbance Observer-Based Control Based on Generalized Relative Input Disturbance Gain

Abstract

In this paper, a novel inverse model is designed for multivariable disturbance observer-based (MDOB) control system. A new interaction measure, generalized relative input disturbance gain (GRIDG), is proposed to quantify the disturbance rejection capabilities of different inverse model structures. Consequently, an inverse model with partially coupled structure can be determined for better disturbance rejection performance. Besides, by defining a relative input disturbance gain array (RIDGA), the values of GRIDG under different input-output pairings and structures can be derived in an intuitive manner, which greatly facilitates the structure selection of the inverse model. And then, to overcome the difficulty of implementing the partially coupled inverse model, a V canonical structure is used without introducing model approximations. It also has the advantage of simple calculation and being easy to generalize to high-dimensional systems. Finally, simulation examples of several multivariable industrial processes are given to illustrate the design procedure and demonstrate the effectiveness of the proposed strategy.