A simulation study of a neural network based approach for the identification of hybrid systems

In a previous paper we proposed a Neural Network (NN) identification approach for a class of Hybrid Dynamic System. However, although the obtained NNs represent average models that can fairly approximate a given HDS, the formulation of mathematical demonstration and/or conditions, which guarantee that the obtained NNs predect the outputs with a similar precision in all the modes still a very hard task. Thus, other alternatives should be investigated in order to study the validity of the global NNs. In this context, different simulation examples are considered in this paper to analyze the accuracy of the identified NNs according to the modes of the HDS.