Microfiltration of milk with backpulsing

Transcription

1 Microfiltration of milk with backpulsing Emma Larsson Tetra Pak Processing Systems AB, Lund, Sweden, February 2 Abstract Microfiltration can be done during milk processing to eliminate bacteria or fractionate the proteins. A problem with membrane filtration is fouling and backpulsing has shown to be an effective technique for reducing fouling. Experiments with backpulsing of milk was investigated in laboratory scale with two ceramic membranes with pore sizes of. and.4 µm. Used process conditions were chosen to get the experiments as similar as possible to the industrial process. Two different circulation flows were used where the backpulse only had a positive impact on the process when the lower circulation flow was used. However, a higher circulation flow without backpulsing proved to be better than a lower circulation flow with backpulsing despite the positive results. The conclusion after these experiments was that backpulse can have a positive impact on a process, but not under the investigated operating conditions. To see if the backpulse affect the quality of the milk the protein retention was investigated. A small increase of protein retention was observed for the. µm membrane in all backpulse experiments. Introduction To get milk that keeps fresh during longer time milk gets processed and one step that can be done during the processing is microfiltration. Tetra Pak is one of the world s leading companies in food processing and packaging solutions. They offer a complete portfolio of milk processing and since it has been shown that backpulsing could reduce fouling it was investigated if backpulsing should be introduced into their system. Backpulsing could improve the process by either increase the running time or increase the flux. An improvement in the process was wanted but to the same quality of the milk and that is why it also was investigated if backpulsing had an impact on the protein retention. Theory Membrane filtration is an alternative to other separation processes like evaporation, sedimentation and centrifugal separation. The problem with membrane process is the fouling that occurs over time, which backpulsing should prevent []. Backpulsing or backflushing is when the permeate flow is reversed by an applied pressure on the permeate side of the membrane (Figure ). The difference between backflushing and backpulsing is the frequency of the pulses where backpulsing has a higher frequency. The reverse flow lifts the foulants off the membrane surface and the foulants are swept away by the cross-flow. When the pressure relationship is switched back the flow is again going from the feed side to the permeate side [2, 3]. The transmembrane pressure, TMP, which is the driving force during microfiltrating, is defined according to Equation. TMP= P f+p r -P 2 p () Where P f is the pressure at the feed inlet, P r is the pressure at the retentate outlet and P p is the pressure at the permeate outlet [4]. In many cases has backpulsing shown to increase the filtration rate, make the

2 Figure. Schematic illustration of regular filtration and backpulsing. time between the cleaning occasions becomes longer and maintain a higher flux [5]. The use of backpulsing results in a loss of filtration time and a lower average TMP compared to when backpulsing is not used. This has shown to be a significant limit in how backpulsing can be applied [6].Backpulsing is especially good with ceramic membranes since these membranes easier can withstand the high pressure that is needed with backpulsing. Ceramic membranes have good permanence to chemicals, ph, solvents, high operating pressure and high temperature [7]. How good a membrane is in separating different particles, i.e. how much is retained by the membrane, is given as the retention of the membrane, R (see Equation 2). R=- c p (2) c b Where C p is the permeate concentration and C b is the bulk concentration. When the retention is one all particles are retained in the retentate and when the retention is zero everything passes through the pores []. When the pulse duration and the pulse amplitude are varied, differences in the permeate protein concentration have been observed [8]. According to Rodgers et al. [6] the retention never increases since backpulsing makes pores that usually are plugged more open. With more open pores, more particles can pass through the membrane and the retention decreases [6]. Materials and Methods A cross-flow laboratory scale microfiltration unit supplied by Tetra Pak was used to carry out the experiments (Figure 2). The circulation loop was used to keep the milk at a temperature of 5 C with a heat exchanger (Alfa Laval, AlfaNova 4-2H, Ronneby, Sweden) coupled to a water bath. The pump (FM-OS/95 centrifugal pump, ABB, Motor 3, IP55, 5Hz) was connected to a 3 l tank and was used to create a pressure and a flow through the circulation loop. Ceramic membranes, (Pall Corporation, Pall Exekia) of two mean pore sizes,. µm and.4 µm, were used during the experiments. The membranes were made of alumina oxide and were fitted into a steel house (T 7, Pall Exekia). They had an area of 5 cm 2 and consisted of one flow channel which was 25 cm long and 7 mm in diameter. To drive the piston in the backpulse equipment (and get the reversed pressure) compressed air was used. Milk was first fed to the tank. At startup, both permeate and retentate bleed off (Q p and Q r ) valves were closed. When the pump had started and the desired pressure and cross-flow was reached, permeate and retentate bleed off flows was adjusted with the valves to the right values (see Table ). To get the experiments as similar to the industrial process as possible the same cross flow velocity, flux (J) and Volume Concentration Factor (VCF) was used. VCF is defined as (Q p + Q r )/ Q r. To reach the same velocity, 5.8 m/s, as in one single 4 mm flow channel in a full scale membrane element the circulation

3 Figure 2. Schematic figure of the set-up. flow (Q circ ) was adjusted to 8 l/h. Used settings can be seen in Table. The average flux, J, was kept constant which contributed to an increase in TMP with time due to fouling. When the permeate pressure no longer changed the experiment was finished. By calculating how much the TMP had changed and divide with how long time the experiment took the increase in TMP per time unit was calculated for each experiment. A backpulse device, Pall Decolmateur BF3 (Pall Corporation), was used in the experiments. Two parameters could be changed on the backpulse equipment, the pulse duration (T on ) and the time between two pulses (T off ). The pulse duration was when the piston blocked the flow channel and the time between the pulses was when the piston was withdrawn and the flow channel open. T on was.2 or 2 seconds and T off was 4 or 2 seconds. Four different time combinations were tested. Experiments without backpulsing (reference) were also made for comparison with the backpulse experiments. The reference experiments were done in duplicate to have a secure result to compare the experiments made with backpulsing with. The other experiments were only repeated when something went wrong (like too low temperature). Three pressure sensors, three temperature sensors and one flow sensor (see Figure 2) was coupled to the software (Lab View 29, National Instruments Co) which collected data each tenth second. Experiments were also made with a sampling time of. second to study the pressures during a pulse. The milk samples that were analyzed according to their composition were taken at the end of each experiment. To analyze the filtrated milk according to the protein concentration (Minimjölk with.% fat, Skånemejerier, Malmö) a Milko Scan FT2 (FOSS) was used. After each experiment the unit was cleaned with an alkaline cleaning agent (Divos 24, DiversyLever, Huddinge) and an acid cleaning agent (Nitric acid, Brenntag Nordic AB, Malmö). The alkaline cleaning removes proteins and fat and the acid cleaning removes minerals. Table. Settings for the experiments. Membrane pore size (µm) J (l/m 2 h) Q circ Q p Q r VCF Results and discussion Increase in TMP per time unit It was investigated how much the TMP increased per time unit to see if the running time improved with backpulsing. When the membrane with a pore size of.4 µm and a circulation flow at 8 l/h was used together with backpulsing the membrane fouled quicker than when backpulsing was not used (the reference). The increase in TMP per time unit was therefore higher for the experiments with backpulsing than the reference (Figure 3). The process conditions that were used in this case were chosen to get the process as similar as possible to the industrial process, i.e. cross flow velocity 5.8 m/s (created by a circulation flow of 8 l/h), flux 35 l/m 2 h and VCF 2. When a lower circulation flow was used (4 l/h), the results become different.

4 ,9 TMP/time unit (bar/h),7,5,3.4 μm - 8 l/h.4 μm - 4 l/h. μm - 8 l/h, Reference Ton.2 Toff 2 Ton.2 Toff 4 Ton 2 Toff 2 Ton 2 Toff 4 Figure 3. The increase in TMP per time unit in all experiments. Three of the four experiments with backpulsing had a lower increase in TMP per time unit than the reference, as shown in Figure 3. The reason to why the forth experiment did not follow the other three could be explained by a high actual flux. When using backpulsing the actual flux has to be higher than the average flux since there is no forward permeate flow during the pulse when the permeate flow channel is closed. The actual flux was theoretical calculated (Equation 3) based on the fact that the average permeate flow, which is known, should be the same as the actual permeate flow minus the backpulsed flow. J= Q p T+V bp T-t bp A (3) Where J is the theoretical actual flux, Q p is the average permeate flow (.49 ml/s for the.4 µm membrane and.7 ml/s for the. µm membrane), T is the actual time for a cycle (the time between a pulse start and the next pulse start), V bp is the backpulsed volume (3 ml), t bp is the actual backpulse time and A is the membrane area (5 cm 2 ). As can be seen in Table 2 the diverging experiment had a higher actual flux than the other experiments, and the membrane fouled therefore faster. Table 2. The calculated actual flux in all experiments. The time values are from experiments where the pressure was sampled every tenth second. T on and T off are the settings and T and t bp are the actual times. The experiments presented as dash is the experiments without backpulsing, i.e. the references. Membrane pore size (µm) T on T off Circulation flow t bp T J (l/m 2 h)

5 An explanation to why it worked with a circulation flow of 4 l/h but not 8 l/h can be seen when looking on the pressures during a pulse (Figures 4 and 5). The pressure difference between the feed pressure and the backpulse permeate pressure is not high enough when using the higher circulation flow (approximately.2 bar) which means that the backpulse is not strong enough to have an impact on the process. The pressure difference during backpulsing at the lower circulation flow was higher (.55 bar), and therefore more efficient. It can also be seen in the figures that the pressure is not maintained during the whole pulse when it should be 2 seconds. The permeate pressure is higher than the feed pressure only for a short time and have therefore not the whished effect on the process. What also can be concluded is that the permeate pressure during backpulsing is higher at 4 l/h than at 8 l/h. This is because when the lower circulation flow was used the membrane fouled faster and with more fouling the pressure is maintained betteron the permeate side and get thereby higher than for the higher circulation flow where the fouling is less. After the pulse, when the piston is withdrawn, the permeate pressure decreases which means that TMP increases and an instantaneous high flux appears. This flux is higher than the theoretical calculated actual flux and contributes to a faster fouling. When using a membrane with a pore size of. µm lower average flux and VCF were used (see Table ). The experiments with backpulsing showed a higher increase in TMP per time unit than the reference. The pressure difference between the feed inlet and the permeate outlet (see Figure 6) is higher for some experiments than for the.4 µm membrane, but this is another membrane and therefore not comparable. The theoretical actual flow for this membrane was found to be very large (see Table 2) and seems to be the main reason to why the increase in TMP per time unit is smaller for the reference than the backpulse experiments. In Figure 6 where the pressures during a pulse are presented it can be seen that the permeate pressures are maintained during the whole pulse due to a more dense membrane where the milk cannot flow true the membrane as easily. The permeate decrease that could be seen after each pulse for the larger membrane pore size is eliminated for this membrane pore size to the same reason as the pressure is maintained, i.e. a more compact membrane. Retentate outlet, Pr Feed inlet, Pf Permeate, Pp Ton.2 Toff 2 Permeate, Pp Ton.2 Toff 4 Permeate, Pp Ton 2 Toff 2 Permeate, Pp Ton 2 Toff 4,9.2 bar pressure (bar),7,5,3 time Figure 4. The pressures during a pulse with the membrane with pore size.4 µm, a flux of 35 l/m 2 h and a circulation flow of 8 l/h. The dotted line indicates the pulse length at two seconds and the arrow the pressure difference between the feed inlet and permeate outlet.

6 Retentate outlet, Pr Feed inlet, Pf Permeate, Pp Ton.2 Toff 2 Permeate, Pp Ton.2 Toff 4 Permeate, Pp Ton 2 Toff 2 Permeate, Pp Ton 2 Toff 4 pressure (bar),4,2.55 bar Figure 5. The pressures during a pulse with the membrane with pore size.4 µm, a flux of 35 l/m 2 h and a circulation flow of 4 l/h. The dotted line indicates the pulse length at two seconds and the arrow the pressure difference between the feed inlet and permeate outlet. time Retentate outlet, Pr Feed inlet, Pf Permeate, Pp Ton.2 Toff 2 Permeate, Pp Ton.2 Toff 4 Permeate, Pp Ton 2 Toff 2 Permeate, Pp Ton 2 Toff 4 pressure (bar),6,4, bar time Figure 6. The pressures during a pulse with the. µm membrane, a flux of 2 l/m 2 h and a circulation flow of 8 l/h. The arrow indicates the pressure difference between the feed inlet and permeate outlet. Protein retention An increase in TMP per time unit occurs due to fouling. Fouling means that more pores are plugged and the retention therefore should be high. However, the protein retention of the.4 µm membrane did not follow this rule, as shown in Figure 7. All backpulse experiments at 8 l/h should have larger protein retention than the reference since they had a larger increase in TMP per time unit, but the retention was higher only in one experiment. The protein retention was lower in all backpulse experiments at 4 l/h, even though only one backpulse experiment was expected to have a lower protein retention than the reference. The increase in TMP per time unit was higher in all backpulse experiments with the. µm membrane, as compared with the reference. It was therefore expected that the protein retention should be higher during backpulsing, which also was the case, Figure 7.

7 Retention,9,7,5,3,.4 μm - 8 l/h.4 μm - 4 l/h. μm - 8 l/h Reference Ton.2 Toff 2 Ton.2 Toff 4 Ton 2 Toff 2 Ton 2 Toff 4 Figure 7. Protein retention in all experiments. The performed experiments were few and the protein retention does not differ much between the experiments but according to the performed experiments the backpulse have an impact on the protein retention when a membrane with pore size of. µm was used. Conclusion The aim of this investigation was to study if backpulsing is an effective technique to reduce fouling. With the same process conditions as the industrial process the results show that backpulsing, as applied in these experiments, has no positive impact on the process. By changing the circulation flow the result however become different and it could be seen that backpulsing had a positive impact on the process. The process conditions that are used in the commercial process gave however the lowest increase in TMP per time unit. It could not be concluded that backpulsing affects the protein retention when using a membrane with a pore size of.4 µm. An impact on the protein retention could however be seen when using a membrane pore size of. µm. More experiments should be made to confirm these results since the values do not differ much. References [] Ann-Sofi Jönsson, Membranprocesser- Grundläggande begrepp, p. 4-5, 7, Institution for Chemical Engineering, Lunds university [2] Vinod T. Kuberkar, Robert H. Davis, 2, Microfiltration of protein-cell mixtures with crossflushing or backflushing, Journal of Membrane Science 83, -4 [3] Chen Ning Koh, Thomas Wintgens, Thomas Melin, Frans Pronk, 28, Microfiltration with silicon nitride microsieves and high frequency backpulsing, Desalination 224, [4] Rishi Sondhi, Ramesh Bhave, 2, Role of backpulsing in fouling minimization in crossflow filtration with ceramic membranes, Journal of Membrane Science 86, 4-52 [5] Charles S. Parnhamn, Robert H. Davis, 996, Protein recovery from bacterial cell debris using crossflow microfiltration with backpulsing, Journal of Membrane Science 8, [6] V.G.J. Rodgers, R.E. Sparks, 99, Reduction of Membrane Fouling in the Ultrafiltration of Binary Protein Mixtures, AIChE Journal, vol 37, No., [7] K Scott, 998, Handbook of Industrial membranes, section 7, p , Elsevier Advanced Technology, ISBN [8] Caroline Wilharm, V.G.J. Rodgers, 996, Significance of duration and amplitude in transmembrane pressure pulsed ultrafiltration of binary protein mixtures, Journal of Membrane Science 2,