Model identification for closed-loop multivariable processes based on min-max critical frequency search

This paper presents an improved method of model parameters identification in frequency-domain for closed-loop multivariable processes. Based on reference input and process output data during the closed-loop sequence step tests, the process frequency-responses are estimated with signal frequency analysis. Using a min-max critical frequency search algorithm, only a least possible number of frequency points are obtained for model fitting. Then the first order plus delay time transfer functions are determined by implementing the linear least-square method. Compared with existing methods, the proposed identification technique has the advantage of less computation burden and is easy for industrial applications. Simulation results show the simplicity and effectiveness of the proposed method.