Process Simulation an‎d Parameter Study of an Adsorptive Desulfurization Column for Fuel Oil

10/7/2026 12:00:00 AM

The presence of sulfur compounds such as thiophene, benzothiophene, an‎d dibenzothiophene in liquid fossil fuels, especially fuel oil, causes the emission of sulfur oxides (SOₓ) during combustion, resulting in severe air pollution, acid rain formation, an‎d degradation of catalysts in afterburner units. Therefore, the development an‎d optimization of efficient an‎d economical methods for removing sulfur compounds before combustion is of great importance. Among the various desulfurization methods, the adsorption process is considered one of the prominent options due to its favorable performance at low pressure an‎d temperature, adsorbent recycling capability, an‎d environmental compatibility. In this study, a sulfur adsorption column was dynamically modeled an‎d simulated in the MATLAB software environment. The mathematical model was developed based on mass transfer equations, mass balance in fluid an‎d solid phases, an‎d adsorption reaction kinetics to investigate the effects of parameters such as fuel input concentration, adsorbent particle size, an‎d bed height on sulfur removal efficiency. Simulation results showed that increasing the initial concentration from 120 to 5200 ppm reduced the adsorption efficiency an‎d increased the time to reach equilibrium, while decreasing the adsorbent particle size an‎d increasing the bed height improved the adsorption capacity an‎d delayed the curve break point. In order to optimize the operating conditions, the response surface methodology (RSM) was used using Design-Expert software, an‎d then, in order to more accurately predict the system behavior, artificial neural network (ANN) models including multilayer perceptron (MLP) an‎d radial basis function (RBF) were developed. The comparison of the network performance showed that the Levenberg–Marquardt algorithm in the three-layer architecture (10, 15 an‎d 25 neurons) provided the best results an‎d the high values of the correlation coefficient (R²=0.9981 for RBF an‎d R²=0.9985 for MLP) confirm the high accuracy of the models’ predictions. Overall, this study, by combining mathematical modeling, numerical simulation, statistical optimization an‎d machine learning, provides a comprehensive framework for the analysis an‎d design of the adsorption desulfurization process. The obtained results can be used as a basis for the design an‎d industrial scaling of liquid fuel desulfurization processes.

گوگردزدايي جذبي , شبيه‌سازي فرايند , روش‌شناسي سطح پاسخ , يادگيري ماشين

Adsorption Desulfurization , Process Simulation , Response Surface Methodology , Machine Learning

Karim Maghfour Sarkarabad

Dr. Ahad Ghaemi