Modeling of equilibrium system of Pz-CO2-H2O using neural networks

In this study, equilibrium system of H2O-Pz-CO2 has been modeled using neural networks. In the modeling two networks RBF and MLP were used. Back propagation algorism was used in the learning of the networks. In the RBF network, nonlinear Gaussian function was used as an activation function whereas sigmoid function was used in MLP network. The networks results were evaluated by experimental data presented in the literature. Mean absolute error for RBF and MLP are 4.21 and 4.78 percent, respectively. The results showed that neural networks are suitable tools for decreasing of calculation time and increasing of accuracy in the equilibrium data predictions.