Performance Investigation of Artificial Neural Networks in Estimation of Second Virial Coefficient

Artificial neural networks are one of the new mathematical methods which have found many applications in chemical engineering in recent years. This method usually gives acceptable results when there is a lack of certain data.In this paper two different kinds of neural networks, multilayer perceptron (MLP) and radial basis function (RBF) are used to predict and estimate the second virial coefficient. Structural parameters and different methods for improving the training step and generality of these two neural networks have been investigated. Finally the behaviors of these two networks to give quick and accurate final results are compared with each other and the weak and strong points of each are discussed. In this paper we have used the experimental data of over 100 different hydrocarbons.