Practical comparison of neural networks and conventional identification methodologies

This paper addresses practical comparison between the conventional identification methodology and identification based on computational intelligence (CI). For this purpose, Auto-Regressive Moving Average with eXogenous inputs (ARMAX), Non-Linear ARMAX (NARMAX) and identifications based on Artificial Neural Networks (ANN) are applied to the modelling of a pilot-scale parallel-tube heat exchanger. First and second-order non-linear optimisation methods are used to train the neural networks. Results of the identification methods are presented and compared. It is shown that the use of second-order non-linear optimisation method for training neural networks yields a significant improvement in the convergence rate.