Comparative study of two neural network approaches for nonlinear identification

Two approaches for identification of nonlinear time-discrete systems by using multilayer neural networks are compared. The first one consists of using a feedforward neural network to simulate the input-output behavior of the unknown system based on a regressive model. The second approach is based on the state variable representation of the system to be identified, and results in an identifier consisting of two feedforward neural nets, one with recurrent feedback. Preliminary results indicate that, while both approaches seem to be effective in identifying general nonlinear systems, the state variable identifier is apparently more useful for control applications, although training is relatively more complex