Aggregated asphaltene structural models have been generated through a molecular simulation geometry optimization process, using periodic boundary conditions. This methodology has been validated by first applying it to a pure aromatic system. Initially, a random distribution of 35 molecules was chosen and a geometry optimization process was performed, allowing the cell dimensions to vary without restrictions. The structure factor (S(k)) of an optimized final cell was obtained and compared with experimental results, and the agreement between theoretical and experimental S(k) profiles was satisfactory. This methodology was next used in the analysis of the morphology of 32 asphaltene model molecules and their aromatic cores; asphaltene model molecules were taken from literature. It is remarkable that face-to-face stacking of asphaltene aggregates was observed, as well as π-offset and T-shaped stacking geometries. Finally, the effect of aliphatic chains on the aggregates was also analyzed.

Asphaltene aggregation under vacuum at different temperatures was obtained using classical molecular dynamics (MD) simulations under nonperiodic boundary conditions in a monodisperse system of 96 hypothetical asphaltene molecules. Identical asphaltenes were originally set as an array, where the separation between each other was ∼40 Å. Simulations under the canonical ensemble at NVT conditions, using the Verlet numerical method to solve the motion equations, were conducted. Aggregated systems formed by several asphaltene monomers after 100 ps of classical MD simulations were found. The structure of the solution was analyzed using the radial distribution function. Simulations at four different temperatures (273, 312, 342, and 368 K) were accomplished. Another similar MD simulation at a temperature of 310 K for 300 ps was performed, to validate the stability in the previous systems. After this run, good structures for explaining asphaltene interactions were also observed; these structures were never before proposed. The following effects can be observed from the results: (i) aggregates that have different structures indicate different types of interactions; (ii) decreasing aggregation number values with increasing temperature is consistent with experimental reality; (iii) the average molecular weights obtained for different temperatures agree with the expected range of experimental values; and (iv) the minimum of the potential energy well, in the range of 3.5-4.0 Å, is consistent with the Yen model.

Class imbalance and class overlap are two of the major problems in data mining and machine learning. Several studies have shown that these data complexities may affect the performance or behavior of artificial neural networks. Strategies proposed to face with both challenges have been separately applied. In this work, we introduce a hybrid method for handling both class imbalance and class overlap simultaneously in multi-class learning problems. Experimental results on three remote sensing data show that the combined approach is a promising method.

No abstract

No abstract