Applications of system identification and the role of regularization (Q1368724)

The authors consider system identification to be model supported testing as well as test-supporting modelling. The article contains several test results that can be used to improve the a priori given mathematical model concerning its structure and its parameters. If the model structure is assumed to be suitable, system identification is reduced to model parameter estimation. By under modelling and investigating the parameter sensitivities concerning measurements (and measurement errors) the model errors that can be corrected are localized in the model prior to correct them. The authors point out the role of the regularization in the estimation procedures. A numerical example with simulated data, demonstrating the error localization and the effect of the regularization, is presented. Many references are given.