A novel FFT-based robust multivariable process identification method

In this paper, a general robust identification method is proposed for an interactive linear time-invariant multivariable process. Using the fast Fourier transform (FFT), the process frequency response matrix is first calculated from the recorded process input and output time responses. Then the process step response is constructed using the inverse FFT for each process channel. New linear regression equations are derived from such responses and their various-order integrals. The regression parameters are then estimated without iteration and give a first-order plus dead time model or a general second-order plus dead time model. The proposed method is applicable to various experimental scenarios. Its effectiveness is demonstrated through simulation examples and a real-time test.