Fouling and cleaning of reverse osmosis membranes in the dairy industry.
Thesis DisciplineChemical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Reverse osmosis is used for the concentration of milk to reduce transportation costs from farms, and also within dairy factories to concentrate dairy liquids. There are many components, including salts, proteins, and fat that could potentially foul membranes, and some of these accumulate at the membrane causing increases in the effective osmotic pressure and/or resistance to flux. A SEPA flat sheet membrane system with a Dow FT30 polyamide thin film composite membrane, simulating spiral wound membranes, was used to measure the flux, rejection, fouling and cleaning of model solutions of milk, whole milk and skimmed milk. The rejection coefficient of sodium chloride was significantly affected by transmembrane pressure (TMP) especially at less than 10 bar (81% at 4 bar compared with 99% at 25 bar) while cross-flow velocity had little or no influence. The rejection coefficient of the membrane dropped significantly below pH 4.0. The flux of salt solutions was influenced by the pH, generally with higher fluxes at pH 7 than at pH 3. The rejection coefficient of many salts (sodium citrate, phosphate, sulphate) was greater than 99% at neutral pH, but sodium nitrate had a rejection of 93% while both sodium acetate and sodium chloride had a rejection of 97%. The nitrate ion rejection was measured to be 34% at pH 3 and for a range of solutes, a good correlation between anion radius and rejection was obtained. Both whole milk and skimmed milk were found to have similar fluxes with a maximum at about 14 - 16 bar TMP. The flux decreased below the maximum at higher pressures. A solution with twice the concentration of simulated milk ultrafiltrate (SMUF) had significant flux reduction with a maximum flux at 16 bar TMP. Lactose and whey protein isolate had little or no influence on flux. Altering the mineral content of skimmed milk either up or down increased the fouling resistance, whereas both increases and decreases in protein content reduced the amount of fouling. It is thought that high mineral contents led to mineral precipitation but low mineral contents cause casein micelle disintegration. Flux measurement, TEM, SEM-EDX, FTIR and FT-Raman analysis consistently showed that casein micelles, minerals (especially calcium phosphate) and lipids in milk played a major role in fouling formation while whey protein and lactose has very little influence on fouling. The fouling was found to be greater at higher TMP. FTIR coupled with flux measurement gave a more comprehensive assessment of membrane fouling and cleaning. The membranes were cleaned using a standard cleaning cycle of 10 minutes each of 0.5% w/w NaOH, 0.8% w/w HNO3, 0.5% w/w NaOH with water rinses before each step. One cleaning cycle was required to restore the pure water flux to its pre-treated flux for runs with milk at 12 bar TMP, while two cycles or more were required for runs at 24 bar TMP. Each cycle of cleaning was found to have an additive effect on membrane permeability with lower chemical concentrations producing less of an effect. It was found that cleaning results from a fouled membrane must be compared to results from the same cleaning of a similar membrane but without fouling. The recovery of water permeance for a fouled membrane increased with increasing cleaning duration, increasing chemical concentration and renewal of cleaning chemical. Still, two cleaning cycles were required to recover the water permeance to the pre-treated permeance. A polyamide reverse osmosis membrane that had been used industrially was cleaned in a similar manner in the flat sheet system. The used membrane could not be restored to a high flux with the standard chemical cleaning cycle. FTIR spectra showed species associated with milk lipids. Cleaning by solvent extraction using a two-phase mixture of water, isopropanol and cyclohexane gave a significant flux and rejection improvement. Analysis of the solvent-extracted material indicated the presence of phospholipids with a relatively higher concentration of sphingomyelin. FTIR results showed a clear correlation between the lipid vibration bands and flux before and after cleaning. Alternative cleaning regimes with chemicals, enzyme and urea were unsuccessful. It is hypothesised that some of the lipids are able to bind to the polyamide surface of the membrane. The fouling of used membrane could not be reproduced with 3 hour reverse osmosis runs with lipid enhanced skimmed milk at 24 bar in the laboratory. Collectively the results show the contribution of the major components of milk in determining flux decline, rejection, and cleanability of polyamide membranes.