ASPAD dynamic simulation and artificial neural network for atenolol adsorption in GGSWAC packed bed column.

Abstract

This study aimed to assess the dynamic simulation models provided by Aspen adsorption (ASPAD) and artificial neural network (ANN) in understanding the adsorption behavior of atenolol (ATN) on gasified Glyricidia sepium woodchips activated carbon (GGSWAC) within fixed bed columns for wastewater treatment. The findings demonstrated that increasing the bed height from 1 to 3 cm extended breakthrough and exhaustion times while enhancing adsorption capacity. Conversely, higher initial ATN concentrations resulted in shorter breakthrough and exhaustion times but increased adsorption capacity. Elevated influent flow rates reduced breakthrough and exhaustion times while maintaining constant adsorption capacity. The ASPAD software demonstrated competence in accurately modeling the crucial exhaustion points. However, there is room for enhancement in forecasting breakthrough times, as it exhibited deviations ranging from 6.52 to 239.53% when compared to the actual experimental data. ANN models in both MATLAB and Python demonstrated precise predictive abilities, with the Python model (R2 = 0.985) outperforming the MATLAB model (R2 = 0.9691). The Python ANN also exhibited superior fitting performance with lower MSE and MAE. The most influential factor was the initial ATN concentration (28.96%), followed by bed height (26.39%), influent flow rate (22.43%), and total effluent time (22.22%). The findings of this study offer an extensive comprehension of breakthrough patterns and enable accurate forecasts of column performance.