Identification and Reduction of Model Structure for Modeling Distributed Parameter Systems

In this paper, the problem of how to determine the model structure for modeling a distributed parameter system is considered. The model structure error of using model M _A to replace model M _B is defined as a weighted sum of their distances in both observation and prediction spaces. The value of so defined structure error can be obtained by solving a max-min problem. In order to select an appropriate complexity of model structure, the concept of extended identifiabilities is further extended to the case that model structure error is involved. A stepwise regression procedure is then presented for simultaneously determining the model structure and model parameters. Numerical examples are given to explain the presented concepts and methodology.