Exploring of CO2 adsorption behavior by Carbazole-based hypercrosslinked polymeric adsorbent using deep learning and response surface methodology

In this research, hypercrosslinked polymeric adsorbents based on carbazole as a monomer were used to carbon dioxide capture. In the synthesis of this adsorbent, first, aluminum (III) chloride was considered as a catalyst, then iron (III) chloride was used as the catalyst. Based on the type of catalyst, two types of adsorbents have been obtained. The adsorbent synthesized using then iron (III) chloride catalyst results in better adsorption capacity at the optimal conditions. Based on two methods of artificial neural network modeling and response surface methodology, the adsorption process was investigated, and at a constant temperature and pressure of 25 °C and 5 bar, the effect of parameters such as adsorption time, cross-linker ratio, synthesis time, and their interactions on amount of adsorbed CO2 were investigated. In the response surface methodology, the central composite design was used, and the correlation coefficient for the response based on the quadratic model was 0.9432 considered. ANN modeling uses two algorithms, including multilayer perceptron and radial basis function, and provides convincing predictions from the experimental data. The maximum correlation coefficient values for MLP with two layers with 15 and 10 neurons and RBF with 80 neurons and spread of 1 were 0.9967 and 0.9957, respectively. The optimal values for the input parameters such as adsorption time, cross-linker ratio, and synthesis time were obtained 810 s, 1.54, and 8.8 h, respectively. The optimum amount of adsorption capacity was obtained at 174.59 mg/g.