Predicting the Physical Property of Asphaltene Molecule using Fuzzy System

Abstract

Asphaltenes — the heaviest and most aromatic compounds of crude oil— have a strong tendency to aggregate in solution to form complex colloidal structures. They generally impede producing, transporting and refining of crude oil resources. An indepth knowledge of the physical behavior of asphaltenes is needed to mitigate these effects. Predicting the physical behavior of asphaltene such as density is complex and nonlinear as it is a complex molecular mixture and undergoes a thermodynamic liquid–liquid phase separation from sufficiently paraffinic solutions. Therefore, this problem is qualified to be modeled using fuzzy system. A Modified Learning From Examples (MLFE) algorithm is used to build the rules and membership functions while Recursive Least Squares (RLS) algorithm is adopted to tune the system. Thirty training points developed from molecular modeling technique using Compass force field by Materials Studio software are selected to train the fuzzy system. Molecular dynamics simulation values found from OPLSaa force field by LAMMPS software are used to evaluate the performance of the fuzzy system using six simulation points. It is found that the fuzzy system is capable of predicting the physical behavior of asphaltene molecule. The analysis also shows that the error for the simulation points depends significantly on the selected training points.