Mimo Process Identification for Controller-Design - Test Signals, Nominal Model and Error-Bounds

YC ZHU,ACPM BACKX

DOI: https://doi.org/10.1016/s1474-6670(17)50621-7

1992-01-01

Abstract:Advanced MIMO (multi-input multi-output) control design techniques need accurate process models. Process identification is an effective way to do industrial process modelling. In this work, some problems in MIMO process identification are studied, where the purpose of identification is control system design. First a recent result on prediction error method identification will be introduced which is the theoretical basis of the work. Then a procedure of MIMO process identification is proposed which consists basically high order model estimation and model reduction. The method of identification is numerically simple and reliable. A special feature of the method is that it can not only estimate a nominal parametric model of the process, but also supply an upper bound of the model uncertainty (modelling errors) in the frequency domain. Input design can be solved easily in this approach. The identified nominal model and error bound can be readily used for robust control system analysis and design. An industrial application is presented to validate our method.