Nonlinear continuous dynamic system identification by automatic localized modeling

This paper describes an automated localized modeling method to identify con tinuous nonlinear dynamic systems from their operating data. Using a method similar to finite element method's automatic mesh generation, the input sp ace is partitioned into overlapped regions that are small enough that a loc al model, such as a simple neural network , can approximate the data well i n each legion. Subsequently, adjacent regions are inspected to sec if they call be represented well by a single local model to minimize the number of regions and local,models needed to approximate system. A nonlinear oscillat or is used to test the proposed method, and the method was able to generate models that can simulate the system well.