Design of a Combined Mixing Rule for the Prediction of Vapor−Liquid Equilibria Using Neural Networks

Abstract

This paper addresses the design of an intelligent mixing rule formed by the combination of the Wong and Sandler (W−S) mixing rule and the Huron and Vidal of order 1 (MHV1) mixing rule using the basic principles of neural networks. The basic idea is the use of a perceptron neural network with a least mean squares learning rule in the prediction of vapor−liquid equilibria in isothermal processes. The results obtained were as good as or better than the existing models of similar nature (the maximum deviation was always less than 0.8%). Furthermore, the results can be used to predict systems where no experimental data are available. A sensitivity analysis has been carried out to clarify the effect of the new mixing rule in comparison with the W−S mixing rule and MHVI mixing rule.