Prediction of Molecular Weight of Petroleum Fluids by Empirical Correlations and Artificial Neuron Networks

The exactitude of petroleum fluid molecular weight correlations affects significantly the precision engineering calculations and can make process design trouble-shooting inaccurate. Some methods in literature to predict are used commercial software simulators. According statements made literature, Lee–Kesler Twu most engineering, other do not exhibit any significant advantages over correlations. In order verify which proposed appropriate for fluids with variation between 70 1685 g/mol, 430 data points boiling point, specific gravity, individual hydrocarbons were extracted from 17 sources. Besides existing two different techniques, nonlinear regression artificial neural network (ANN), employed model samples. It was found that ANN demonstrated best accuracy prediction a relative standard error (RSE) 7.2%, followed by newly developed correlation an RSE 10.9%. available those API (RSE = 12.4%), Goosens 13.9%); Riazi Daubert 15.2%). well known Lee–Kesler, Twu, set points, exhibited RSEs 26.5, 30.3% respectively.