Evaluation of diffusion and Henry's coefficients of CO2 absorption using Response Surface Methodology and Artificial Neural Network models

Abstract

This study investigates CO2 diffusion and Henry's coefficients in a packed column using ANN and RSM. The experimental setup involved circulating a NaOH solution through the packed column. A quadratic model in RSM analysis was employed for optimizing the parameters including liquid flow rate of 73.34, gas velocity of 1.957, temperature of 7.7, and packing's specific surface of 0.45. Multilayer perceptron (MLP) and radial basis function (RBF) models of ANN achieved regression coefficients (R2) exceeding 0.97, with mean squared error (MSE) below 0.0035. The MLP model, featuring three hidden layers (15, 10, 5 neurons) and 1191 epochs, exhibited an MSE of 0.003021, along with an average absolute relative deviation (AARD) of 0.0003 for DCO2 and 0.00001 for HCO2. Similarly, utilizing the Gaussian function with 50 neurons, the RBF model attained an MSE of 0.003544, with an optimal spread of 2. This study concludes that by effectively predicting DCO2 and HCO2 within specified ranges of operational parameters, efficient CO2 capture can be achieved, thereby contributing to advancements in CO2 capture technology.

Keywords

• CO2 absorption
• ANN
• RSM
• Diffusion coefficient
• Henry's coefficient