Taguchi Optimization of Operating Conditions of a Microfiltration Alumina Ceramic Membrane and Artificial Neural-Network Modeling

Abstract

A comprehensive study using a novel aluminum oxide microfiltration (MF) membrane and a thorough analysis of the effects of various operating conditions on the treatment of produced water were carried out. To set up the experiments and optimize the process parameters, an L9 orthogonal array of the Taguchi method and the larger-the-better target group analysis method was used. The impacts on filtrate flux and fouling control of operating conditions such as pH, temperature, crossflow velocity (CFV), and transmembrane pressure (TMP) were examined. Optimum operating conditions were determined to be 50°C, 1.8 bar, 1.8 m/s, and a pH of 5 and allowing for a maximum flux of 975 L/h·m2. The microfiltration (MF) membrane showed an oil-rejection rate of 98.25%, and the CFV was considered to be the most significant operating variable contributing to the regulation of the flux. Furthermore, 97% recovery was achieved with a mixture of cleaning solutions combining NaOH and HNO3. Two flux decline models were used to interpret the data including the Hermia and an artificial neural network (ANN). Hermia’s cake-forming process had the average highest correlation with permeate flux decline data for the nine experiments (R2=0.83). Using ANN simulation, the best results were obtained with two hidden layers and 25 neurons in each layer. Its performance in terms of the mean squared error as a percentage of the maximum flux was 0.4%.