Predicting permeate flux and NaCl rejection at nanofiltration process of decolorization columns waste water using artificial neural networks

In this study, artificial neural network (ANN) model was used to predict the average permeate flux and NaCl rejection in nanofiltration process of sugar industry decolorization columns waste water. Nanofiltration process was performed at three temperatures 30, 4 and 50°C, at three pressures of 1, 1.5 and 2 MPa, at three concentration levels 60, 80 and 100 g/l and at two pH levels 8 and 9. In order to predict the permeate flux and NaCl rejection multi-layer perceptron neural network with 4 inputs and 2 outputs were used. The results showed a network with 9 neurons in hidden layer and using a hyperbolic tangent transfer function and the Levenberg–Marquardt (LM) optimization technique and 30%-30%-40% data for training/ testing/ validating process can be well predict the permeate flux (0.98) and NaCl rejection (0.88) in the nanofiltration of decolorization column wastewater. Sensitivity analysis results by optimum ANN showed the pressure was the most sensitive factor for prediction of both flux and rejection by the selected ANN.