A hybrid artificial neural network and genetic algorithm for predicting viscosity of Iranian crude oils

Viscosity is an important measure of fluid resistance to shear stress; therefore, efficient estimation of oil viscosity in various operating conditions is very important. Several variables, such as oil API gravity (API), pressure (P), saturation pressure (Pb), reservoir temperature (Tf), are employed in the estimation of crude oil viscosity. A hybrid group method of data handling (GMDH) artificial neural network, optimized with genetic algorithm (GA), was herein proposed to obtain efficient polynomial correlation to estimate oil viscosity. This correlation was compared with 5 correlations presented in the previous research using the large set of Iranian oil data. Also, sensitivity analysis of the obtained correlation was carried out to study the influence of input parameters on the correlation output. A comprehensive computational and statistical result was provided to evaluate the performance of the proposed methods. Results showed that these models were very good approximations for estimating the viscosity of Iranian crude oils.