Optimal design of GMV self-tuning controllers

A design algorithm for generalized minimum variance (GMV) self-tuning controllers is presented. The weighting polynomials associated with the control cost function are defined by minimizing a newly proposed index which measures the stochastic variances of system input and output signals and violation of dominant closed-loop pole pair from the desired sampling lines in the Z-plane. To obtain the desired control performance, a corresponding constraint is introduced which restricts the closed-loop poles within a permissible region specified by some qualitative engineering criteria instead of using the conventional pole-assignment method to fix the poles to prespecified locations. A nonlinear programming method is utilized to accomplish the constrained minimization. The validity of the proposed algorithm and effectiveness of the GMV controller based on it are illustrated in some representative examples.<>