Density prediction of aliphatic hydrocarbons using nonlinear group contribution method

In this work, we propose a quantitative structure property relationship (QSPR) approach in order tomodel the density of saturated and unsaturated aliphatic hydrocarbons including linear and branched alkanes,substituted and unsubstituted cycloalkanes and cycloalkenes and linear and branched alkenes up to the hightemperature, high pressure conditions. The group contribution method was used to select the most importantdescriptors of compounds structure. Levenberg -Marquardt artificial neural network (ANN) was used to linkmolecular structures and density data. The data set was randomly divided into three data set: training set (4358point), validation set (643 point) and test set (643 point). After training and optimization of the ANNparameters, the performance of the model was investigated by the test set. The result indicates that this modelcan simulate the relationship between the experimental descriptors and the density of the desired moleculesaccurately.

Keywords:

QSPR , Group Contribution Method , Artificial Neural Network , density , hydrocarbons