A novel method integrating response surface method with artificial neural network to optimize membrane fabrication for wastewater treatment

Abstract

Developing high performance membranes is an urgent need to resolve water pollution problem. Electroless nickel plating (ENP) has been emerging as a promising membrane fabrication technique due to its distinct advantages over conventional ones. However, the involved intricate mechanisms and multi-factors impede optimization of this fabrication technique. In this study, a method integrating response surface method (RSM) with artificial intelligence was proposed to optimize ENP conditions for polyvinylidene fluoride-nickel (PVDF-Ni) membrane fabrication. RSM simulation based on experimental data showed the optimal conditions of 30 °C reaction temperature, 14.50 min reaction duration and 9 mL ammonia amount in this study. With the data of experiments designed by RSM as input, the combined RSM-artificial neural network (ANN) method enabled to predict the flux and Congo red (CR) rejection of the optimal membrane. The membrane under the optimal conditions was then fabricated. It was found that, the predicted flux and rejection of the optimal membrane were rather close to the real ones with less relative error of −3.64% and −0.58%, respectively. Scanning electron microscopy (SEM) characterization showed that the optimal membrane had an ordered microporous structure with favorable pore size, suggesting feasibility of the proposed method. The proposed method based on artificial intelligence in this work paved a new way and had universal significance to optimize membrane fabrication.