Application of Artificial Intelligent Modeling for Predicting Activated Carbons Properties Used for Methane Storage

Performance characterization and optimization of activated carbons are extensively studied using artificial intelligence modeling. In this study, the effect of several parameters on the preparation of activated carbon by chemical activation is investigated. Various preliminary parameters have been considered. The study has resulted in finding four parameters, which are of higher importance compared to the others. These parameters include chemical agent type, chemical agent to precursor ratio, activation temperature, and activation time. In our previous study, 36 activated carbon (AC) samples were prepared using the aforementioned parameters at various levels. In the present investigation, these experimental results have been used for the modeling. As a novel approach, an adaptive neuro-fuzzy inference system (ANFIS) is also applied to the experimental data presented in this study. ANFIS is established by combining artificial neural network (ANN) with fuzzy inference system. After determining the model parameters, some additional data points are used to validate the models. Finally, the outcomes are compared with the experimental results. The normalized mean square error (NMSE) has been obtained as 0.00327, which is very satisfactory for the model validation. These attempts to simulate the preparation stage of activated carbons would provide a simple and flexible route with various AC preparations. Such an effort is essential to develop the adsorbed natural gas (ANG) technology.