Evaluation of controller performance—use of models derived by subspace identification

Abstract

AbstractA new approach is presented for the estimation of the controller, process, and disturbance models necessary for the calculation of the relative variance index, which was introduced in an earlier paper (Control Eng. Practice 2000; 8:791–797), for the performance of SISO controllers. It involves the use of dynamically, sufficiently rich segments from the normal operating data and the use of the subspace identification technique to estimate the systems mentioned above. This approach improves the estimation accuracy of the performance index in relation to the method presented previously. The estimated models enable the comparison of the present controller performance with that of optimally tuned PI or IMC controllers. This helps identify the potential benefits of either retuning or redesigning the assessed controller. Copyright © 2002 John Wiley & Sons, Ltd.