This study aimed to quantify the effect of membrane surface porosity on particulate fouling predicted by the MFI-UF method at constant flux. Firstly, the surface porosity of polyethersulfone UF membranes (5–100 kDa) was determined using ultra-high resolution SEM. Thereafter, the MFI-UF was measured using suspensions of polystyrene particles (75 nm), which were pre-washed to remove surfactant and particle fractions smaller than the pores of MFI-UF membranes, thus ensuring complete retention of particles during MFI-UF measurements. Consequently, the MFI-UF values of washed polystyrene particle suspensions were independent of the pore size and depended only on the surface porosity of MFI-UF membrane. The results showed that the membrane surface porosity decreased with MWCO from 10.5% (100 kDa) to 0.6% (5 kDa), and consequently the MFI-UF increased from 3700 to 8700 s/L2, respectively. This increase in MFI-UF was attributed to the non-uniform distribution of membrane pores, which is exacerbated as surface porosity decreases. Consequently, preliminary correction factors of 0.4–1.0 were proposed for MFI-UF measured with UF membranes in the range 5–100 kDa. Finally, the surface porosity correction was applied to predict particulate fouling in a full-scale RO plant. However, additional research is required to establish correction factors for different types of feed water.@en

This study aims at applying and verifying the MFI-UF method to predict particulate fouling in RO plants. Two full-scale RO plants treating surface water, with average capacity of 800–2000 m3/h, were studied. Firstly, the MFI-UF of RO feed and concentrate was measured using 5–100 kDa membranes at same flux applied in the RO plants (20–26 L/m2.h). Subsequently, the particle disposition factor (Ω) was calculated to simulate particle deposition in RO cross-flow filtration. Finally, particulate fouling rates were predicted based on MFI-UF and Ω, and compared with the actual fouling rates in the plants. For plant A, the results showed that the fouling rates predicted using MFI-UF measured with 100 kDa membrane have the best agreement with the actual fouling (with 3–11 % deviation). For plant B, the fouling rates predicted based on both 10 and 100 kDa membranes agree well with the actual fouling (with 2 % and 15 % deviation, respectively). However, the fouling predicted based on 5 kDa membrane is considerably overestimated for both plants, which is attributed to the effect of the low surface porosity of 5 kDa membrane. More widespread applications of MFI-UF in full-scale RO plants are required to demonstrate the most suitable MFI-UF membranes for fouling prediction.@en