Removal efficiency optimization of Pb2+ in a nanofiltration process by MLP-ANN and RSM

Using computational intelligence for prediction, modeling, and optimization of chemical process behavior could save costs and time. This study’s main goal was to predict and optimize removal efficiency and permeate flux behavior of Pb 2+ aqueous solution in a nanofiltration process through using response surface methodology (RSM) and multilayer perceptron (MLP) neural network. A regression coefficient R 2 =0.99 was obtained for both removal efficiency and permeate flux in the RSM model. Also, the F-value for the removal efficiency and permeate flux was 394.79 and 1888.85, respectively. Different MLP structures for predicting removal efficiency and permeate flux behavior of lead ion in aqueous solutions were investigated. The best structure was obtained for two hidden layers with nine (tansig transfer function) and three (logsig transfer function) neurons. The values of R=0.9993, R 2 =0.9986, MSE=0.402 and MAE=0.409 for the best structure were obtained. Finally, the the removal efficiency was optimized through RSM based on the experimental data. It was concluded that optimum mode selected for membrane composition of PSF=10.04%, NMP=88.98%, and PAN-CMC-41=0.98% (wt%) 53.17 ppm as lead ion concentration in solution and 30.31 min for filtration time achieved the maximum value of removal efficiency equal to 90.68%.