Abstract
Summary Experimental asphaltene research is a high cost, time consuming, and difficult domain. Smart computational tools can help researchers to overcome some of these limitations ( Kondori et al. 2021 ). Golsefatan and Shahbazi (2020) reported using least squares support vector machines model to investigate asphaltene adsorption on surface of NiO/ZSM-5 and NiO/AlPO-5. Their modelling results are in line with the experimental output, while R2, average absolute relative error, and mean squared error are 0.988, 2.943%, and 4.151, respectively. Ali et al. (2021) used various ML techniques to develop predictive models to accurately forecast asphaltene adsorption on surface of MgO NPs. The GBM model was found superior followed by the BG and the RF. Gene Expression Programming (GEP) is a very capable AI algorithm which can automatically develop computer programs using evolutionary algorithms. Use of the GEP to forecast asphaltene adsorption onto SiO2 NPs has never been reported in the literature. We used the GEP for the first time to develop a correlation for prediction of asphaltene adsorption based on particle size, SBET (m2/g), asphaltene content, n-heptane, CNP, C0, and m/C0 as input parameters. The model predictions were compared with experimental data from the literature showing high accuracy of the developed correlation.