PSO-ANFIS and ANN Modeling of Propane/Propylene Separation using Cu-BTC Adsorbent

Abstract

In this work, an artificial neural network (ANN) model along with a combination of adaptive neuro-fuzzy inference system (ANFIS) and particle swarm optimization (PSO) i.e. (PSO-ANFIS) are proposed for modeling and prediction of the propylene/propane adsorption under various conditions. Using these computational intelligence (CI) approaches, the input parameters such as adsorbent shape (SA), temperature (T), and pressure (P) were related to the output parameter which is propylene or propane adsorption. A thorough comparison between the experimental, artificial neural network and particle swarm optimization-adaptive neuro-fuzzy inference system models was carried out to prove its efficiency in accurate prediction and computation time. The obtained results show that both investigated methods have good agreements in comparison with the experimental data, but the proposed artificial neural network structure is more precise than our proposed PSO-ANFIS structure. Mean absolute error (MAE) for ANN and ANFIS models were 0.111 and 0.421, respectively.