An ETF-based disturbance observer-based control for multivariable processes with time delays

Abstract

As an effective technique for rejecting unmeasurable disturbances, the disturbance observer-based (DOB) control has two key issues to be addressed when applied to multivariable processes with time delays, i.e., the design of the nominal inverse model and the analytical synthesis of the Q(s) filter. Focusing on the above issues, this paper presents a novel equivalent transfer function-based (ETF-based) DOB control for multivariable processes with time delays. First, considering the potential relation between the ETF matrix and the inverse model, the design of the inverse model can be simplified by calculating the ETF matrix. The method for determining the parameters of the ETF matrix is modified to improve the rationality of the inverse model. Then, a realizable inverse model is achieved by compensating the modified ETF matrix. Unlike the multivariable DOB (MDOB) control using simple diagonal inverse models, the proposed strategy presents a full-element inverse model with full consideration of the process model information. Since the calculation of the modified ETF matrix only involves several simple matrix operations, the design of the inverse model is streamlined. Based on the modified inverse model, the Q(s) filter is synthesized analytically by minimizing the performance index, i.e., the 2-norm of disturbance responses. The adjustable parameters of the Q(s) filter are determined with due consideration of control performance and robustness. Simulation results demonstrate that compared to other MDOB control strategies, the proposed strategy provides the best results in terms of the integral absolute error (IAE) values of the disturbance responses, the total variations (TV) of the inputs, and the robustness indices.