GA-ANN modeling for density prediction of hydrocarbons over a wide range of temperature and pressure

Genetic algorithm (GA) and artificial neural network model (ANN) were successfully developed for density prediction of hydrocarbons. A large number of molecular descriptors were calculated with Dragon software and a subset of calculated descriptors was selected with a genetic algorithm as a feature selection technique. Only 11 descriptors were obtained by GA as the most feasible descriptors, and then they were used as inputs for neural network. These descriptors are: pressure, temperature, molar mass, MATs1e, GATs2v,MLOGP, R3u, R4e, E2p, BEN, HATsm. A total of 4464 data points of density at several temperatures and pressures have been used to train, validate and test the model. These data points were randomly divided into three data sets: training (2480), validation (992) and test set (992). The predictive model was built using the Levenberg-Marquardt artificial neural network (LM-ANN) and its architecture and parameters were optimized using training set. The prediction ability of the model was evaluated using the validation and test sets. The mean square error (MSE) and R2were 20.1084, 0.9989 for the validation set and 12.6649, 0.9990 for the test set, respectively. The obtained results showed the excellent prediction ability of the proposed model in the prediction of density for different hydrocarbons.

Keywords:

artificial neural network (ANN) , genetic algorithm (GA) , hydrocarbons