A neural network correlation for molar density and specific heat of water: Predictions at pressures up to 100 MPa

• Published in: Fluid phase equilibria 2020-02, Vol.506, p.112411, Article 112411
• Main Authors: Melo, E.B., Oliveira, E.T., Martins, T.D.
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Mathematical modeling; Thermodynamic properties
• ISSN: 0378-3812; 1879-0224
• DOI: 10.1016/j.fluid.2019.112411

The IAPWS-95 formulation is a set of equations that calculates water properties with high precision. To obtain a certain thermodynamic property, one or more iterative procedures are needed, which demand high computational effort. The aim of this work was to obtain an Artificial Neural Network based equation to predict the IAPWS-95 formulation values of density (ρ) and specific heat (Cp) of water in liquid, vapor, and supercritical phases, including the saturation line. Data at temperatures up to 1275 K, and pressures up to 100 MPa were considered. Different sets of input variables were tested and best results were obtained using: temperature, pressure, and speed of sound (used to differentiate liquid from vapor at the saturation line). The network 3-20-15-2 was 99.70% faster than the IAPWS formulation, and presented an overall mean percentage errors of 0.23% and 0.51% for ρ and Cp, respectively, which were lower than those obtained using known correlations. •Speed of sound can be used to obtain thermodynamic variables with high precision.•One ANN for prediction of density and specific heat of water was obtained.•One ANN was capable to describe liquid, vapor, and supercritical phase behavior.•ANN Results were more accurate than those calculated by equations of state.