Closed-loop multivariable process identification in the frequency domain

In this paper, a new technique for on-line multivariable process identification in the frequency domain based on a simple experimental test is presented. Step changes in the controller set point are performed on the process so that the process frequency-response matrix is estimated with signal decomposition and analysis method in the frequency domain, and accurate transform-function matrix is identified as well. The required response data can be obtained while the processes are still in normal closed-loop operation. Simulation example is given to show accuracy of the identification method for multivariable processes.