A practical hybrid modelling approach for the prediction of potential fouling parameters in ultrafiltration membrane water treatment plant

Abstract

Abstract In this work, a novel approach combining first principle equation of Darcy's law on cake filtration and artificial neural network (ANN) predictive models were utilized to represent the dead-end ultrafiltration (UF) process. Common on-line data available in most industrial-scale UF membrane water treatment plant such as feed water turbidity, filtration time and transmembrane pressure were used as inputs parameters. An UF pilot plant was set up to carry out these experiments. This hybrid modelling approach consisting of cake filtration and ANN models have shown promising results to predict the specific cake resistance and total suspended solids of the feed water with good accuracy. These two filtration parameters are often considered as indicators for membrane fouling propensity. Sensitivity analysis has indicated strong linear correlation between feed water turbidity and specific cake resistance in the UF process. The hybrid model provides an alternative method to estimate these parameters besides the conventional laboratory analysis. This practical modelling approach will be beneficial to industrial-scale UF membrane water treatment plant operations to predict the fouling propensity of the UF process based on commonly available on-line data and simple laboratory analysis.