Hydrocarbon gas properties of viscosity and density are of great importance for gas engineering calculation. These properties are measured experimentally but if unavaliable, they can be predicted through different correlations. This work is aimed to develop new models for gas viscosity and gas density using genetic programing. Large database of experimental measurements were collected and used to develop and test these models. The database consists of gas composition, temperature, pressure, pseudo reduced properties of pressure and temperature, compressibility factor, and experimentally measured viscosity and density, for different hydrocarbon gases and pure and impure gas mixtures containing up to pentane pluse fractions and small concentrations of non-hydrocarbon components. Liquid like gas and gas mixtures containing large percentages of C6+ and measurements conducted below 32°F and 14.7 psi were discarded and a total of 4445 data points constituting of 1853 data points for pure gases and 2592 points for gas mixtures were used in this study. The data was divided into three data sets (training, testing, and validation). Two genetic models were evolved one for viscosity and the other for density. Viscosity is predicted as a function of gas density, pseudo reduced pressure and pseudo reduced temperature. On the other hand, density is predicted as a function of molecular weight, pseudo reduced pressure and pseudo reduced temperature. The two models efficiencies were tested against some commonly used correlations. The omparison indicates the good performance of the developed models over the conventional correlations with absolute average relative error (AARE) of 5.4 and 11.6% for gas viscosity and density models respectivelly

This content is only available via PDF.
You can access this article if you purchase or spend a download.