Chemical inhibition of asphaltene deposition is considered a cost-effective way to prevent the harsh consequences of asphaltene instability in the produced crude. Thus, a careful screening of asphaltene inhibitors is crucial for an efficient prevention. However, the characteristics of asphaltenes such as their acid–base properties will influence the selection of an asphaltene inhibitor and the inhibition mechanism. Therefore, improved knowledge on asphaltene acidic and basic fractions is important. In this work, the separation of asphaltenes into acid, base, neutral, and amphoteric fractions was performed. Among the existing techniques to fractionate asphaltenes, the method of Ramljack was adopted and applied on a light oil extracted asphaltene. However, this oil was sampled from one of the wells in the Hassi Messaoud field in Algeria that experienced a recurring deposition of asphaltenes. The results of asphaltene fractionation reveal that the half composition of this heavy part of crude oil is active functions gathered acid and base components. However, the main contribution is reported to the neutral fraction. The characterization results of infrared and elementary analyses show that both active fractions are aromatic and polar. Moreover, the acid fraction contains in its structure carboxylic acids, phenols, sulfoxide groups, and aliphatic chains, while the structure of the base fraction contains amines, sulfoxide groups, and aliphatic chains.

No abstract

The deposition of wax is one of the most potential problems that disturbs the flow assurance during production processes of hydrocarbon fluids. In this study, wax disappearance temperature (WDT) that is recognized as a vital parameter in such circumstances is modeled using advanced machine learning techniques, namely, radial basis function neural network (RBFNN) coupled with genetic algorithm (GA) and artificial bee colony (ABC). Besides, an accurate and user-friendly correlation was established by implementing the group method of data handling. Results revealed the high reliability of the proposed hybrid models and the established correlation. Moreover, RBFNN coupled with ABC (RBFNN-ABC) was found to be the best paradigm with an overall average absolute relative error value of 0.5402% and a total coefficient of determination (R 2) of 0.9706. Furthermore, the performance comparison showed that RBFNN-ABC and the established explicit correlation outperform the prior intelligent and thermodynamic models. Finally, by performing the outlier detection, the quality of the utilized database was assessed, the applicability realm of the best model was delineated, and only one point was found as doubtful.

No abstract