Identification of twin-tanks dynamics using adaptive wavelet differential neural networks

In this paper, the identification problem of a general class of nonlinear dynamic plants is considered using differential neural networks approach. In these networks, the activation functions are described by wavelets with parameters that are tuned adaptively. The wavelet´s parameters and weights of network are adapted during the training process using a gradient based algorithm (gradient descent). The proposed neuro identifier is applied to a twin-tanks plant. The results show this identifier is more effective than traditional identifiers (sigmoid based differential neural networks).