周波数応答を用いた対角優勢を実現する多変数制御器設計

抄録

A model-based controller design method is widely used in the industrial field. However, in this method, a modeling error deteriorates the control performance. This paper proposes a multivariable controller design method using only frequency response data sets. This method omits the time-consuming model identification and designs a controller directly from raw data sets of the controlled plant. The proposed method provides the parameters of a PID controller using convex optimization subject to stability constrains. The optimization problem is formulated in terms of the diagonal dominance of the system. The designed controller suppresses an interaction and achieves high gain of the system in the case of highly diagonal dominance. The stability of MIMO systems can be evaluated by eigenvalue loci without overestimation of eigenvalue loci. The formulated optimization problem involving nonconvex functions and/or constraints is solved using concave-convex procedure that solves a problem written as a difference between two convex functions. The experimental results show the effectiveness of the proposed method in comparison with the conventional method.