Impact of operating parameters on flux decline in microfiltration and ultrafiltration

Abstract

Microfiltration (MF) and ultrafiltration (UF) processes have been applied in the food industry as means of fat and protein separation, clarification and low temperature sterilisation. One of the major factors limiting the use of MF/UF, in practice, is fouling of the membrane during processing. The phenomenon of fouling is very specific for a given application and is difficult to describe theoretically. Even for a certain solution, fouling significance depend upon the combined characteristics of the membrane, feed stream and operating conditions. In this study, the performance of a microfiltration membrane was compared with that of an ultrafiltration membrane in terms of permeability, limiting flux, and fouling characteristics. Three different food components were trialed, reconstituted skim milk (RSM), baker's yeast suspension, and white vinegar using a plate-and-frame module operated in cross-flow mode. Permeability and pure water flux values were important benchmarks for both fouling and cleaning trials. PWF and permeability for MF were approximately 3 times that of the UF membrane. The average pure water permeability for the MF membrane was about (0.166 ± 5% ), while it was about (0.064 ± 4%) for the UF membrane, this is mainly because the different pore size distribution for the two membranes. Limiting flux, i.e. the pressure independent flux, for RSM (10wt%) was measured experimentally and then compared with the predicted values using four different models which were published previously. These models were internal lift, shear-induced diffusion, Brownian diffusion and surface transport models. The predicated values of the limiting flux for the different feed velocities and temperatures using the surface transport model were very close to the measured values for both membranes involved in this part of the study. The impact of different operational parameters were studied and analysed. The transmembrane pressures examined were between 0.69 x 103 kPa and 2.07 x 103 kPa, which correspond to the pressures likely to be encountered in :MF/UF applications in the dairy industry. The effect of temperature was also studied from ambient temperature (-20°C) up to 55°C. Other parameters such as the cross-flow velocity and liquid feed concentration were also considered in this part of the study. Fouling resistances were also calculated and analysed for the different operational conditions. Flux decline results suggest that the operating pressure, temperature and cross-flow velocity are of great importance to the fouling behaviour. The total amount of backflushing time is a significant economical factor controlling the filtration process. The more time spent on backflushing, the more slowly the flux declines. Different backflushing routines were studied using deioinzed water at different frequencies and duration. SEM images were obtained at different stages of fouling and flushing in order to determine a quantitative method to compare the effectiveness of the various backflushing times on irreversible fouling. However, the flux values, with or without backflushing, are difficult to compare in order to determine the etfectiveness of the flushing technique. The flux behavior during cross-flow filtration was examined experimentally, with different membrane/filtering medium combinations. A large number of data sets on flux-time were collected in the first part of the study, which were used to support a mathematical model of the flux decline during MF/UF processes. The semi-theoretical unsteady-state model for the flux decline in cross-flow microfiltration and ultrafiltration has been developed in this study. The model predicts fouling behavior for a wide range of particle ( or molecular) sizes and foulant concentrations. The model uses only two unknown coefficients, k1 and k2, reflecting both the influences of the cake formation and the 'shear cleaning' of the membrane, to describe flux decline. These two parameters have been found to have small dependence on the operating conditions. Predictions of the model agree reasonably well with the experimental results. The model provides both a fundamental understanding of the key physical phenomena governing flux decline and a rational basis for design purposes. Fouling and concentration polarisation is not evenly distributed over the membrane surface. This can affect the flux, which would therefore vary along the cross-flow pathway. Often only an average flux is measured. A laboratory cross-flow module that can be used to yield this information about the flux distribution and decline along the membrane length has been designed and constructed. The rig was used to study fouling during ultrafiltration and microfiltration processes. Here the extent and rate of fouling was found to vary severely along the membrane length (the cross-flow direction) but all showing the exponential like decay. Flux-time data were also collected for the large membrane module and the results were used to validate the unsteady state model mentioned earlier. Here the spatial distribution of the flux has been taken into account. Different cleaning strategies used in the dairy industry including the effects of sodium hydroxide concentration, temperature, cross-flow velocity and transmembrane pressure on the flux recovery percentages have been established. The applicability of a previous mathematical model has been tested. Small improvements have been made to the model approach.