Modeling of CO₂ Absorption with Aqueous Piperazine/Potassium Carbonate Solution Using MLP and RBF Neural Networks

Bahmanzadegan, Fatemeh; Ghaemi, Ahad

doi:10.22034/ijche.2025.488020.1461

Modeling of CO₂ Absorption with Aqueous Piperazine/Potassium Carbonate Solution Using MLP and RBF Neural Networks

Articles in Press

Document Type : Original Article

Authors

Fatemeh Bahmanzadegan ¹

Ahad Ghaemi ²

¹ PhD student of Chemical Engineering, Iran University of Science and Technology

² Iran University of Science and Technology

10.22034/ijche.2025.488020.1461

Abstract

In this article, the effect of operating conditions and aqueous solution on the CO₂ absorption rate was modeled using response surface methodology (RSM) and an artificial neural network (ANN). The network inputs included temperature, pressure, piperazine percentage/potassium carbonate as the solvent, with the mass transfer flux of CO₂ as the output. The MLP (multi-layer perceptron) network was trained with three hidden layers containing 10, 40, and 10 neurons, respectively, using the Levenberg-Marquardt training function. The MLP network with three layers and 60 neurons, trained with the Trainlm learning function, achieved an MSE of 0.0018616 and an R² of 0.99924. The RBF (radial basis function) network also reached an MSE of 0.0004 and an R² of 0.99849 after 200 epochs. Overall, the MLP network showed better results as it achieved high accuracy in less time.

Keywords

Reactive absorption

piperazine

potassium carbonate

carbon dioxide

neural network

Subjects