Investigation of the filtration characteristics of pilot-scale hollow fiber submerged MF system using cake formation model and artificial neural networks model Original Research Article

The submerged hollow fiber MF membrane offers a cost-effective means of removing particles and pathogens from water for drinking water production. Fouling, however, has been regarded as a major obstacle to the economical operation of membrane systems. This study focused on the investigation of the filtration characteristics of pilot-scale submerged hollow fiber membrane system using the cake formation model, based on the critical flux concept, and on the prediction of the performance of membrane filtration using artificial neural networks model (ANNs) from a long-term pilot-scale operation data. A cake formation model was applied to analyze the axial flux variations, the specific cake resistance, and the critical flux in hollow fiber membrane. The artificial neural network was applied to the simultaneous simulation of transmembrane pressure (TMP) variations in a pilot-scale membrane system. The experimental results indicated that the seasonal variations in raw water quality parameters significantly affected `the membrane permeability. The temperature of the feed water ranged from 2.1 to 26.4 °C, and the turbidity ranged from 1.0 to 500 NTU, which led to completely different conditions in summer and winter. The algae concentration ranged from 0 to 47 mg/m3, as did the chlorophyll-a concentration, which also led to completely different conditions in spring. A stepwise increase in the imposed flux showed that the apparent critical flux was about 60 L/m2 h in summer and winter. Moreover, the apparent critical flux was about 40 L/m2h in spring. The cake formation model calculation was also compared with the experimental data to better understand the fouling in the hollow fiber membrane modules, and ANNs was applied to simultaneously predict the long-term membrane performance.