Accurate modeling of equilibrium adsorption of liquid phase sulfur compounds using two soft computing approaches

This report introduces the application of two advanced intelligent models, an adaptively trained neuro-fuzzy inference logic in a hybrid configuration (Hybrid-ANFIS) and multilayer perceptron neural network (MLP-NN) to accurately determine the equilibrium sulfur adsorption in the liquid phase of hydrocarbon/ sulfur compound solution. Models were meticulously developed using a dataset of 107 empirical observations of seven types of sulfur compounds. These models incorporate the influence of input parameters, including initial sulfur level, adsorbent weight, molecular weights of the solvent and solute, densities of the solvent and solute, adsorbent particle diameter, temperature, and the Si/Al ratio of the adsorbent. Notably, the equilibrium sulfur adsorption amount was considered as the sole output variable. To evaluate the performance and precision of the implemented models, graphical representations and quantitative analyses were employed. Moreover, an assessment between the results of implemented models of the existing study and outcomes of previous reports were conducted. The results indicate that both developed models provide precise predictions. However, the Hybrid-ANFIS model demonstrates a strong correlation in predicting the adsorption empirical data, with an average absolute relative deviation of 0.36 % and an overall R2 value and 0.9997. In addition, superiority of the Hybrid-ANFIS model in providing the most reliable and accurate predictions of adsorption experimental data among all types of implemented models was concluded. This study sets a new benchmark in adsorption modeling by providing the most accurate and generalizable predictive framework to date.