Closed-Loop Identification of Dynamic Models for Multivariable Systems with Applications to Monitoring and Redesign of Controllers

Abstract

Closed-loop identification is usually desirable, but more difficult compared to open-loop identification because of the correlation between the noise and the process inputs. This article presents two new closed-loop identification methods for multivariable systems to solve this problem. Based on one closed-loop test, the complete models, including process and disturbance dynamic models, can be identified simultaneously. The finite impulse responses (FIRs) of the closed-loop systems are first estimated by an efficient recursive algorithm derived from a subspace identification method. Then, the fast Fourier transform (FFT) and inverse FFT (IFFT) techniques are used to construct the frequency responses and FIR models, respectively, of process and disturbance dynamics. These algorithms do not need prior knowledge of the process or assumptions about the model structure. Applications of the proposed method to control system monitoring and controller redesign are thus presented. These applications can be accomplished before the process model has been identified and would be very helpful for improving the control performance. The simulation results demonstrate the effectiveness of the proposed identification method and its application.