Combined Approach of Rbf Neural Networks, Genetic Algorithms and Local Search and Its Application to the Identification of a Non-linear Process

The identification of non-linear systems by artificial neural networks has been successfully applied in many applications. In this context, the radial basis function neural network (RBF-NN) is a powerful approach for non-linear system identification. An RBF neural network has an input layer, a hidden layer and an output layer. The neurons in the hidden layer contain Gaussian transfer functions whose outputs decrease exponentially according to the square of the distance to the center of the neuron. In this paper, a combined approach including an RBF-NN neural network with training based on a genetic algorithm (GA) and local search is presented. During the identification procedure, the GA guided by local search aims to optimize the parameters of the RBF-NN and the optimal values are regarded as the initial values of the RBF-NN parameters. The validity and accuracy of system identification using the RBF-NN model are tested by simulations, whose results reveal that it is feasible to establish a good model for a non-linear process of pH neutralization.