Separation of heavy gases from light gases using synthesized PDMS nano-composite membranes: Experimental and neural network modeling

A method for simulation of gas separation with nano-composite membranes was presented using artificial neural network (ANN). In this investigation, a number of nano-composite silica/PDMS (polydimethylsiloxane) membranes were synthesized with different amounts of silica nano-composite loaded into the polymer matrixes. The permeation behavior of pure gases including C3H8, CH4, and H-2 has then been studied as functions of pressure and nano-composite loading. Experimental results were used to develop a black box model to predict gas permeability for the synthesized membranes as a function of operating pressure, nano-composite loading and one of the physical properties of the gases used. A comparison was made between three different groups of data including pressure as the first effective parameter, nano-composite loading as the second, and one out of the four variables, molecular diameter, molecular weight, boiling point, and critical temperature, as the third. It was concluded that feed pressure, nano-composite loading, and boiling point temperature in the feed neurons provides the best combination and leads to the least error in prediction of the gas permeation flux values. The results showed that, increasing pressure increases permeability of the condensable gas, C3H8, whereas permeabilities of the lighter gases, H-2 and CH4, decrease with increasing pressure. In addition, nano-composite loading decreases permeability of the non-condensable gases, H-2 and CH4, while that of C3H8 increases up to 2% of nano-composite loading. Ultimately, it was concluded that ANN method can be successfully used for prediction of gas separation properties of nano-composite membranes after proper network training in this case resulting in predictions of less than 4.0842 RMSE. (C) 2011 Elsevier B.V. All rights reserved.