Modeling Dehydration of Acetone by Use of Polyacrylonitrile Membrane and Polyethylene Glycol with the Help of Artificial Neural Network

The authors of this research analyzed the amount of water-acetone separation in the pervaporation process by means of polyacrylonitrile membranes and polyethylene glycol with the help of artificial neural network. The pervaporation process can be applied for separation of many liquids (separation of acetone from water in this research). Due to azeotropic with water, acetone has purity problems. The amounts achieved from the experimental data were compared to the modeling data, and the results were analyzed. In this research, the effects of such parameters as volumetric flow rate and temperature, as well as feedstuff properties (separation factor and flux) on the dehydration process efficiency were evaluated, and the Multi Layers Perceptron neural network feed forward along with propagation learning algorithm and Levenberg-Marquardt function with 2 inputs and outputs were implemented. Tansig activation algorithm was used for the hidden layer, and purely algorithm was utilized for the output layer. Furthermore, 5 neurons were defined for the hidden layer. After processing the data, 70 percent were allocated for learning, 15% were allocated for validity, and the remaining 15% was allocated for the experience. The achieved results with the aforementioned method had a suitable accuracy. The graphs of the error percentage for the actual values of the separation factor and flux outputs were compared to the achieved values from modeling through polyacrylonitrile membranes and polyethylene glycol for evaluating the efficiency of pervaporation process in the separation of acetone from the water. Finally, the graphs were drawn.