Abstract
Nowdays, model based development of systems is hilighed to shorten cycles of developments and reduce costs for them. For the model based development, models of target systems are very important. For the modeling of systems, prior information of targets should be used in models. However even if the detailed properties of the system are known, it is hard to identify the targets using too complicated models. Simplified models are used in these cases, but sometimes they are insufficient in performance. In this talk an identification method with combination of simplified physical model and emperical error model is presented. To enhance the performance of physical model, we newly define an error system whose output is estimation error of physical model. And then steady-error model and transient-error model are identified separately. Finally, by compensating the estimation error, more accurate estimation is obtained. An application to an internal combustion engine will be demonstrated.
