2002 Volume 35 Issue 8 Pages 794-805
This article addresses the complete identification of stable, integrating, and unstable processes. Despite the stability of the process, a biased relay experiment can be performed to provide abundant process input-output data in stable operation. Based on a time-weighted integral filter, an effective method is developed to estimate the orders and time delay as well as the model parameters from the relay-induced data in a straightforward fashion. The method utilizes a small amount of experimental data and is robust with respect to measurement noise and model structure mismatch. Furthermore, it can easily be applied as a model reduction technique by estimating an apparent delay with user-specified orders.