Development of Artificial Neural Network Model for Carbon Dioxide Adsorption Parameter Prediction

Currently, renewable energy such as biomass is getting a lot of attention. Burning biomass produces carbon dioxide, and carbon dioxide also comes from feedstock of petrochemical processes and a by-product from processes. This carbon dioxide is a greenhouse gas that causes global warming. Adsorption process is a popular method to reduce carbon dioxide, but it is difficult to estimate adsorption capacity because determination of adsorption capacity requires specific processes such as volumetric method or using kinetic models. In this study, artificial neural network was used to predict carbon dioxide adsorption capacity to estimate the time when carbon dioxide adsorption reaches equilibrium. The carbon dioxide adsorption data were collected from literatures such as adsorption temperature, gas flow rate, BET (Brunauer, Emmett and Teller) surface area, and adsorption capacity at any time (qt). The curve fitting method was used to estimate adsorption rate constant (k), and adsorption capacity at equilibrium (qe). Rate constant and adsorption capacity at equilibrium were output variables, and they are used for artificial neural network (ANN) model training. The optimal conditions for best performance ANN modeling were 15 neurons in 4 hidden layers, TANSIG transfer function for hidden layer, and PURELIN transfer function for output layer. The obtained k and qe were substituted in the pseudo-first-order equation to calculate the qt values at time t. The qt and time were used to plot adsorption capacity graphs for comparing with the experimental data. Here, the obtained results were consistent.