Anda di halaman 1dari 16

www.catchword.com=titles=09603085.

htm

09603085/02/$23.50+0.00 # Institution of Chemical Engineers Trans IChemE, Vol 80, Part C, September 2002

INVESTIGATION OF ALKALINE CLEANING-IN-PLACE OF WHEY PROTEIN DEPOSITS USING DYNAMIC GAUGING


T. R. TULADHAR, W. R. PATERSON and D. I. WILSON
Department of Chemical Engineering, University of Cambridge, Cambridge, UK

novel thickness measurement technique (DYNA-PROBE) allows the study of the behaviour of layers of soft material undergoing cleaning in owing liquids, in situ and in real time without contacting the surface. The technique is demonstrated here in a study of the swelling and removal of whey protein lms related to cleaning-in-place of dairy heat exchanger fouling deposits. The experimental protocol also allows the gauge to be calibrated in situ after each test, yielding measurement accuracies of 10 mm. Cleaning experiments also featured simultaneous measurement of the mass of protein removed and the thermal resistance of the foulant layer, via a commercial heat ux sensor. Experiments performed over a range of solution concentrations (0.32.0 wt% NaOH), temperatures (2050 C) and velocities (0.030.30 m s 1; Re 50010,000) indicated the existence of an optimal cleaning solution concentration and several other features reported by previous workers. New information provided by the technique, including the sensitivity to shear of the swollen deposit, yields new evidence permitting integration of several con icting earlier hypotheses on protein cleaning behaviour. Keywords: DYNA-PROBE; cleaning-in-place; non-contact gauging; whey protein; swelling; thermal conductivity. INTRODUCTION Fouling is generally de ned as the accumulation of unwanted materials on the surfaces of processing equipment. Cleaning is the removal of such layers from surfaces, and is often performed in conjunction with disinfection and sterilization. The fouling and cleaning of surfaces in contact with thermally labile food materials remains one of the major processing problems in the food industry. The rapid deposition of both organic and inorganic materials, particularly onto heat transfer surfaces, means that process plant must be cleaned regularly. In dairy plants, fouling: (i) causes an increase in pressure drop through the plant; (ii) lowers the ef ciency of heat transfer; (iii) limits the operational time of the plant; and most importantly (iv) endangers plant and process hygiene by harbouring harmful bacteria. Moreover, foulant layers can undergo signi cant ageing transitions, yielding deposits which are more dif cult to remove1 . Optimization of fouling and cleaning cycles is, therefore, of considerable importance, and requires a good understanding of fouling and cleaning kinetics. In the dairy industry, fouling in UHT and HTST heaters can be rapid, requiring cleaning on a daily basis. A signi cant fraction of time can be spent on cleaning (including rinsing and re-conditioning equipment), so there is a clear need to understand the chemical and physical phenomena involved in the removal process. Ultimately, such knowledge will be used to design a heat transfer system to be readily cleaned, just as the aim of fouling mechanism studies is to reduce the propensity to foul. The ultimate hope is that the two sets of criteriaoptimizing anti-fouling and cleaning performanceare compatible and affordable. 199 Cleaning has received considerable attention in the food sector, partly because of the frequent use of multi-purpose plant which uses the same equipment to process different materials. Food soils are frequently cleaned by converting them to softer, weaker or more amorphous materials which can then be readily removed. Dairy-related applications have been studied extensively2 using either milk soils, or representative feedstocks such as whey protein solutions. Investigations of cleaning, particularly related to cleaning-inplace (CIP) methods, have identi ed various aspects of the cleaning process, but fundamental studies of cleaning mechanisms are complicated by the complex microstructure and time-dependent nature of the deposits involved. An important example is alkaline-based cleaning of proteinaceous deposits, where the foulant, consisting of denatured protein, swells on contact with hydroxide to form a gel or matrix which is then broken down by further reaction and=or shear=attrition. Identi cation of the key parameters involved is complicated by the dif culty in studying such materials, which deform readily on contact with solid probes, and collapse when a signi cant liquid fraction is removed. In-situ measurements are thus preferable, but subject to the challenges of: (i) interference, for contact (or near-contact) devices; or (ii) calibration, for remote sensors (particularly given the heterogeneous nature of the deposit). CIP mechanisms in milk and milk-related systems have been studied in the Cambridge Chemical Engineering Department for nearly 20 years. Recent work has focused on the behaviour of the whey protein fraction in milk, which contributes signi cantly to thermal fouling at temperatures < 90 C characterized as type A deposits3 . Cleaning studies have concentrated on CIP using alkaline solutions, which is

200

TULADHAR et al.

one of the main industrial methods. Previous studies using lumped analysis methods (e.g. chemical assays4 ) prompted the use of local sensors to elucidate mechanisms further. The use of heat ux sensors identi ed patterns of behaviour5;6 which required further information on deposit structure and form, prompting the development of the dynamic gauging technique (DYNA-PROBE) described here. This paper reports on the technique development and its application in cleaning studies. GAUGING PRINCIPLES OF THE DYNA-PROBE The authors have developed a novel non-contact proximity gauging technique for studying soft lms deposited on hard surfaces in situ and in a liquid environment. The technique was developed from the method of pneumatic gauging, differing primarily in: (i) the use of liquid, rather than air, as the gauging uid; (ii) employment in suction mode rather than blowing mode; (iii) measurement of ow rate, rather than pressure. In brief, uid is drawn from the reservoir by suction through the nozzle and is discharged externally via a ow meter. The suction mode has clear advantages for aseptic operation. The principle of the technique is illustrated in Figure 1. A full description of the technique is presented in Tuladhar et al.7 . The gauge consists of a tube ending in a convergent nozzle of inside diameter dt , which can be moved towards the deposit surface. The technique exploits a siphon effect, whereby liquid is drawn from the surrounding reservoir into the nozzle and along the tube to a ow ratemeasuring device. The ow rate is determined by the pressure drop across the nozzle approach (1), the nozzle throat expansion (2) and the connecting tubing (3). Contribution (3) is usually small. When the nozzle is located close to a deposit surface, the nozzle approach resistance dominates, so that the ow rate is closely related to the area at the nozzle entry, and thus the clearance, h. Given the location of the nozzle relative to the underlying substrate, the thickness of the deposit layer is then calculated by difference. The key feature of the technique is that the nozzle does not need to contact the surface to make a measurement, merely to approach it. The sole assumed properties of the layer are that it is locally at, and that it does not deform substantially under the stresses imposed by the ow. It is when the nozzle-surface separation, h, is less than dt =4, that the discharge ow rate, m, is sensitive to h=dt , so

that the location of the lm surface can be determined relative to the nozzle. The nozzles location in space, relative to the clean surface, is recorded by a precision location device (e.g. micrometer) attached to the gauge. The thickness of any deposit lm can then be determined by difference, between the nozzle location co-ordinate and the nozzle-surface separation, h. For a nozzle with dt 1 mm, the dt =4 criterion gives a desirable distance of 250 mm and, for the prototype devices described here, an accuracy of 10 mm. The sensitivity range can be increased by increasing the nozzle dimension: this will increase the size of the area being studied, and may result in large gauging ows, which may not be desirable. There is, therefore, likely to be an optimal gauge con guration for each particular application. The technique was developed in quasi-stagnant conditions, where there was no superimposed convective ow in the bulk uid: this paper reports recent work where the technique has been used to study foulant layers where the bulk liquid is owing. This paper reports further investigations of CIP of such protein layers, under well-characterized ow conditions. The gauging device is mounted in a duct, in combination with a heat ux sensor which simultaneously monitors the thermal resistance of the foulant layer. CLEANING EXPERIMENT The cleaning cell shown in Figure 2 is essentially a at plate heat transfer module, where the plate assembly separates liquids owing in ducts of 15 mm square cross-section. The cell body was constructed from Perspex, which allowed visual inspection of the deposit during cleaning. An 8 mm thick brass plate was xed on the top of the lower block and a micro-foil heat ux sensor (MHFS 20455-1, 13 30 mm, Rhopoint Ltd., Surrey) was sandwiched between the fouled plate and the brass plate. It was decided to sandwich the heat ux sensor between the two plates as this prevented it from contacting the uids, hence prolonging the life of the sensor. The signal from the heat ux sensor was ampli ed using a microvoltmeter (DC MicroVoltmeter, TM8 supplied by Levell Electronics Ltd, UK) and converted to a heat ux, q, using a calibration supplied by the manufacturer. The MHFS also included a T type thermocouple for surface temperature measurement (with quoted accuracy 0:5%). A gauge (nozzle i.d. dt 1 mm and tube i.d. d 4 mm, 45 nozzle angle) was mounted vertically above the centreline of the test plate in the upper block. The nozzle could be moved closer to, or further from, the fouled plate using a micrometer mechanism. The discharge mass ow from the gauging probe was collected in a container placed on an electronic mass balance (Sartorius L2200P , accuracy 0:05 g). The two devices were not mounted in close proximity lest the gauging ow disturb the heat transfer rate detected by the heat ux sensor. The effect of the gauging probe in distorting the heat ux reading is discussed in the results section. The cleaning cell was mounted in a simple re-circulating cleaning loop, which is described in detail elsewhere8 . Ten litres of fresh NaOH solution were used for each experiment. These solutions were prepared by dissolving NaOH pellets (98% purity, BDH Laboratory Suppliers, UK) in RO water using an overhead stirrer (Citenco F.H.P. Motors, UK). The solution was initially preheated by circulation through a
Trans IChemE, Vol 80, Part C, September 2002

Figure 1. Proximity sensing technique: assignment of ow stations and dimensions.

CLEANING-IN-PLACE OF WHEY DEPOSITS

201

Figure 2. Schematic of the cleaning cell. All dimension in mm. T Thermocouple:

preheater consisting of copper tubes (13 mm i.d. 12 m long) immersed in a hot water bath. The ow was then momentarily stopped, a fouled plate placed in the cell, and the ow restarted. The NaOH solution was pumped around a closed loop, passing in turn through the preheater, a rotameter, the cleaning cell (where it contacted the deposit) and back to the tank. All connections were 15 mm o.d. reinforced plastic tubing. An overhead stirrer in the tank was used to dissolve any lumps of protein removed from the deposit during cleaning. The temperature of the cleaning solution was maintained within 1 C by recirculation through the preheater. Hot deionized water from a constant temperature bath was pumped around a similar closed loop through the reverse side of the test section (at a mean velocity of 0:96 m s1 ) and back to the hot water bath. The hot water loop provided the temperature driving force DT TH2 O TNaOH 20 K required for the heat ux sensor to operate, which gave an initial temperature difference across the lm of around 5 K. Inlet and outlet uid temperatures for both NaOH and hot water were measured using 1 mm diameter K type thermocouples 1 K inserted directly into the ow at the points shown in Figure 2. Due to the short contact time and small

temperature difference between the hot water and NaOH solution, there was little difference between the inlet and outlet temperatures for each uid. Temperatures, heat ux and the discharge ow rate from the gauge were logged on a dedicated PC. Monitoring Techniques The gauging sensor was calibrated in the presence of NaOH bulk ow. Figure 3 shows that the calibration curve using 0:5 wt% NaOH solution at 20 C was affected by the bulk ow rate. The discharge behaviour obtained can be classi ed into two zones: an incremental zone at low clearance, h, where the gauging ow rate increases with increasing h, followed by an asymptotic zone, where the gauging ow rate is relatively insensitive to h. For a given value of h, m increased with increase in bulk ow rate due to increase in the static pressure in the bulk ow while the siphon head, H, remained xed at 320 mm 1 mm. In the asymptotic region h 0:3 dt , a ten-fold increase in bulk ow velocity (0.03 to 0:3 m s1 ) caused an increase of static gauge pressure from 1070 Pa to 2670 Pa leading to a 50% increase in m (1.00 1:54 g s1 ). Work is underway to extend

Figure 3. Calibration of dynamic gauge (d 4 mm; dt 1 mm; H 320 mm) mounted in the cleaning cell (15 15 mm duct) at various bulk velocities at 20 C. Symbols: 0:03 m=s1 (Re 480); &: 0:15 m=s1 (Re 2380); D : 0:22 m s1 (Re 3570); : 0:30 m s1 (Re 4630).

Trans IChemE, Vol 80, Part C, September 2002

202

TULADHAR et al.

Figure 4. Thermal resistance and thickness of whey protein deposit layers. solid line: regressed linear curve passing through origin.

the modelling of gauging ow under quasi-static conditions, as reported previously7, to conditions featuring bulk ow. Meantime, standard calibration curves of m as a function of h at different NaOH solution ow rates and temperatures were established for all the NaOH concentrations employed here. The effect of NaOH concentration 0:32:0 wt% on m h pro les was found to be negligible. The correlations of m as a function of h obtained from these pro les were used as standards in calculating the clearance distance during cleaning runs, taking measurements at 10 s intervals. The deposit thickness could then be calculated by difference, given the location of the cleansurface obtained at the end of the experiment. During cleaning, the gauging probe was manoeuvred so that the nozzle was held < 200 mm from the surface i.e., in the incremental zone for sensor effectiveness, and > 100 mm from the surface to reduce disruption of the deposit layer by the jet into the nozzle. The accuracy of the thickness measurements was estimated as 10 mm. During cleaning, the deposit thickness and structure change. The fouling resistance of the deposit was calculated using: 1 1 Rd 1 Ut Uc
where
1 DT TH2 O TNaOH Ut q q
2

Here, Ut is an overall heat transfer coef cient at time t, and DT is the temperature difference between the hot and cold uids. Uc is an overall heat transfer coef cient measured when the plate is fully cleaned and is set to the mean value of Ut calculated over a ve minute period after the heat ux, q, reached a steady value towards the end of the experiment. Thermal measurements were logged every second. Samples of cleaning solution were collected every minute and the amount of protein removed was determined using an absorbance techniquea modi ed Bradford Micro-assay with Coomasie Plus Protein Assay Reagent (supplied by Pierce and Warriner, UK) described by Gotham et al.9 . Preparation of Fouled Plates Fouled plates for cleaning studies were generated separately in a co-current heat exchanger fouling cell of similar construction to the cleaning cell, featuring removable

SS 316 plates with a depositing region measuring 150 15 1 mm. Details of the fouling protocol are given in Tuladhar8. In the fouling cell, the SS plate separated 15 litres of 3:5 wt% whey protein concentrate (supplied by Carbery Milk Products, Eire) recirculating at 74 C, and hot deionized water circulating at 90 C through ducts of 15 mm square cross-section. Approximately reproducible levels of evenly distributed foulant were generated in 90 minutes, featuring mass coverages of 300430 g m2 w.b. (50 80 g m2 d.b.) and a layer thickness of 300500 mm. Fouled plates were placed on a wire mesh above water in an enclosed vessel and refrigerated until used in cleaning experiments. Plate weight measurements prior to cleaning indicated that the wet deposit coverage value did not change during storage in the saturated, refrigerated environment. To avoid biological degradation of the deposit during extended storage, cleaning experiments were carried out within 36 hours of preparation of a fouled plate. The initial values of the thermal resistance, Rd ;0 , and deposit thickness, dd ;0 , should be a measure of the quantity of deposit before cleaning. Even though data were available from time zero for both Rd ;0 and dd ;0, data from the rst 1020 seconds were usually discarded to allow for initial adjustment of the micro-voltmeter and mass scale readings. Rd ;0 and dd ;0 were obtained from the thermal resistance and thickness pro les respectively by extrapolating the Rd and dd curves to zero. Figure 4 shows the degree of reproducibility obtained in the fouling stage by plotting Rd ;0 against dd ;0 . These data yield a direct estimate of initial deposit thermal conductivity dd ;0 =Rd ;0 of 0:26 W m1 K, which is consistent with the values reported elsewhere1012 for similar proteinaceous deposits. Inspection indicated that the deposit layers generated were evenly distributed across the plates. TYPICAL CLEANING RESULTS Figure 5(a) shows a typical set of results obtained for cleaning experiments on WPC deposits when treated with 0:3 wt% NaOH solution. Figure 5(b) presents these data sets replotted in order to aid interpretation, in the form of protein fraction remaining, and normalized Rd and dd , de ned thus: R d 3 Normalized Rd d ; Normalized dd d Rd ;0 dd ;0
where Rd ;0 , Rd and dd ;0, dd are the thermal resistance and deposit thickness at time t 0 and t respectively. Trans IChemE, Vol 80, Part C, September 2002

CLEANING-IN-PLACE OF WHEY DEPOSITS

203

Figure 5. Typical cleaning pro le of WPC deposit on SS plate in bulk ow conditions. (a) unscaled data; (b) normalized data. Conditions: 0:3 wt% NaOH; TNaOH 30 C, um 0:30 m s1 (Re 5760). Symbols : Rd ; : dd ; : protein.

Firstly, the Rd curve (see Figure 5(a)) shows the sequence of events previously observed , namely an increase, followed by a plateau stage and then rapid decay, in Rd . The protein assay shows that there is little protein removed before the plateau stage. These results suggest that the initial increase in Rd dd /ld must be due either to swelling (because of migration of solvent molecules or some structural changes), or to reduction in deposit thermal conductivity, ld , or both. Figure 5(b) demonstrates the usefulness of the gauging sensor in explaining the thermal resistance pro le, clearly indicating that the increase in thermal resistance is mainly due to swelling of the deposits. These parallel data sets allow us to divide the cleaning of the protein deposits into three stages: Swelling: where the protein reacts and then swells on contact with hydroxide; (ii) Plateau: where the dd and Rd values are approximately steady and small amounts of protein are removed either by diffusion or erosion; (iii) Decay: where the protein matrix breaks down and the protein is rapidly removed. (i) This tri-partite model and the data that support it are essential for understanding and modelling the cleaning behaviour of such lms. Trans IChemE, Vol 80, Part C, September 2002

The reader should note that heretofore the explanation of the initial increase of Rd as being due to swelling of the deposit has been based on a hypothesis or on visual observation. The gauge measurement shows the hypothesis to be correct and delivers quanti cation. The time to complete cleaning (i.e. mass recovery time), as indicated by the protein assay, tc;P was 1080 s. The Rd pro le tended to zero at ca. 1030 s, illustrating the relative sensitivities of these different measuring techniques. In several other cases, the Rd pro le went to zero (and even went negative) before tc;P (i.e. tc;Rd tc;P ); this behaviour is consistent with the nal stages of cleaning involvingscattered islands of deposit on the surface. Small deposits can enhance the local heat transfer rate (negative Rd ) or have negligible effect on heat transfer, depending on the (height)/(viscous sublayer thickness) ratio. Each data point in the protein pro le represents an integrated measure of protein removed over the entire plate surface area, whereas the MHFS and the gauging pro le monitors local heat transfer and thickness respectively over the corresponding sensor areas. Hence, any slow regions, where cleaning occurs less quickly, will skew the mass pro le towards longer cleaning times. Figure 5(b) also shows a noticeable difference in sensor reading in the decay stage. At t 330 s (point A), the thickness sensor registers a rapid, zeroth-order decline in

204

TULADHAR et al.

foulant thickness, suggesting that the gauging ow was perhaps sucking the deposit off the surface. Visual inspection con rmed that the gauge suction was disturbing the deposit at this point, whereas the material near the heat ux sensor was undisturbed. This behaviour usually occurred after the onset of the decay stage as indicated by the thermal or mass measurements. Thus, the cleaning time (corresponding to zero thickness) registered by the thickness pro le, tc;d is less than that registered by the thermal measurement, tc;Rd . The forces exerted by the gauge have yet to be quanti ed, but related work on pneumatic gauging indicates that they are not negligible13 ; indeed, Vaishnav et al.14 reported the use of a laminar impinging jet to characterize the strength of canine endothelium tissue. This behaviour indicated that the onset of the decay period is related to the breakdown in the gel strength on continued exposure to hydroxide. This hypothesis is also consistent with work by Gillham et al.6 on ow pulsing to enhance cleaning rates for WPC deposits, where the increased shear stresses generated by ow pulsing had greatest effect on the decay stage. They observed the cleaning rate in the decay stage to change from rst to (fast) zeroth order behaviour, attributed to fatigue and gross rupture of the weak deposit. These observations demonstrate the range of interactions between the process operating conditions and foulant chemistry involved in NaOH-protein cleaning. Process plant items feature a range of temperature, shear stress, and mass transfer conditions; the knowledge gained from studies such as this can be used to understand how CIP performance is affected by operating policies, or how to optimize equipment design for CIP installations. The transition from

chemical=transport to chemical=mechanical mechanisms is particularly important. From the Rd pro le in Figure 5(a), the swelling time, tswell;Rd , is taken as the time to reach 90% of the maximum thermal resistance Rd ;max . The plateau time is that at which Rd is 0:9 Rd ;max (tswell;Rd to point X in Figure 5(a)) and the decay time, tdecay;Rd , is the interval in which Rd drops from 90% of Rd ;max to zero (X to tc;Rd ). The time to reach Rd ;max is represented by tmax;Rd . A similar set of parameters, namely tswell;d , tplateau;d , tdecay;d and tmax;d were obtained from the dd pro le (see Figure 5(b)), based on the maximum deposit thickness, dd ;max . The swelling and plateau times obtained by each method agreed reasonably well, while the decay times indicated that the suction imposed by the gauge often increased the apparent local cleaning rate. Two maximum swelling ratios were de ned as: F Rd
Rd ;max ; Rd ;0 Fd
dd ;max dd ;0 4

with FRd and Fd being obtained from the Rd and dd pro les respectively. The time fraction for each stage is de ned as: tswell
tdecay

tswell;Rd ; tc;Rd tdecay;Rd


tc;Rd

tplateau

tplateau;Rd ; tc;Rd
5

where tswell , tplateau and tdecay are swelling, plateau and decay time fraction respectively.

Figure 6. Effect of cooking deposits on cleaning pro les at 30 C at 0:15 m s1 : (a) no steaming; (b) 60 minutes steaming. Symbols : normalized Rd ; : normalized dd ; : protein fraction.

Trans IChemE, Vol 80, Part C, September 2002

CLEANING-IN-PLACE OF WHEY DEPOSITS COOKING EFFECT ON CLEANING RESULTS The bulk of the results reported here are for freshly generated deposits. The following section summarizes a brief investigation of deposit ageing which was simulated by extended exposure of soils to higher temperatures. Ageing is an important effect in industrial applications, where the initial foulant material may be cooked on to surface as a result of extended heating. Prior to cleaning, the fouled plate was steamed in an enclosed vessel for different durations by placing the plate on a wire mesh above water, boiling in an enclosed vessel. Wet deposit coverage of the fouled deposit decreased by less than 10% during the steaming process. Figures 6(a) and (b) show the effect of steaming on cleaning pro les of deposits using 0:5 wt% NaOH solution at a bulk ow rate of 0:15 m s1 at 30 C. Steaming the deposit led to longer cleaning times in terms of all measures (Rd , dd and protein removal). Figure 6(b) also indicates that ageing increases the strength of the deposit, as the mismatch between the Rd and dd pro les, discussed with Figure 5(b) and evident at ca. 400 s in Figure 6(a), is eliminated for cooked deposits. This behaviour was observed for other steaming durations and has important consequences for industrial applications, as deposits in different locations will experience different thermal, and therefore ageing, histories. INTERACTION OF GAUGING TECHNIQUES ON HEAT FLUX READING This section considers whether the siphon gauging ow disturbs the heat transfer rate detected by the heat ux sensor when the two devices are placed in close proximity. Two cleaning results were performed when the gauging nozzle was located directly above the MHFS. Experiment 1 was performed by placing the gauging nozzle 1 mm above the deposit surface with no gauge action (but obstructing the bulk ow), yielding only the Rd pro le. Experiment 2 was performed with active gauging, giving both Rd and dd pro les. Both cleaning experiments were performed using 0:5 wt% NaOH solution at 30 C and a bulk ow velocity of 0:15 m s1 . Although the Rd pro les appeared to vary slightly, the cleaning time obtained for both experiments was roughly the same, giving tc;Rd around 900 s. The Rd pro les also agreed reasonably well with Figure 6(a), where the MHFS and the gauging probe were not mounted in close proximity. The small variation in the Rd pro les and the total cleaning time, tc;Rd , in all these three cases may be due to the fact that the area of a nozzle of 1 mm diameter is signi cantly smaller than the MHFS area 13 30 mm. Hence, the location of the gauging probe and its siphon effect acting on a small deposit surface has a negligible effect on the Rd values obtained from the heat ux area. FACTORS AFFECTING CLEANING OF FRESH DEPOSITS Cleaning is a complex process involving reaction and mass transfer effects. The effects of temperature, concentration and velocity on the cleaning of WPC deposits were studied by carrying out a matrix of cleaning experiments over a range of temperatures 1555 C, concentrations 0:32:0 wt% and bulk ow velocities 0:03 Trans IChemE, Vol 80, Part C, September 2002 0:30 m s1 corresponding to 50010,000.

205 ow Reynolds numbers of

Effect of Temperature Figure 7 compares the effect of temperature on the cleaning pro les of Rd , dd and protein removal at a constant cleaning solution ow rate of 3 l min1 um 0:22 m s1 . The shapes of the curves are similar to those described above. At all temperatures, both Rd and dd at rst increase, followed by a plateau and then a decay. Most of the protein removal occurs in the decay stage. At higher temperatures, between 35 and 50 C, the plateau stage disappears owing to the earlier onset of the decay stage. Comparisons of the Rd and dd data con rm the importance of swelling, and that decay is occurring due to the erosion of the swollen surface. The effect of the gauging sensor in disrupting the deposit in the decay stage is noticeable at all temperatures. At lower temperatures, the onset of the zeroth-order decline in the thickness pro le seems to occur at the end of the decay stage and moves closer towards the plateau stage with increasing temperature. Hence, the total cleaning time registered by dd is shorter than that registered by the Rd and protein pro les. However, at all temperatures, tswell;d and tplateau;d (obtained from the dd pro le) agree reasonably well with tswell;Rd and tplateau;Rd respectively (obtained from the Rd pro le). The mass recovery time, tc;P , obtained from the protein assay (to the nearest minute) agrees reasonably well with that obtained from thermal recovery pro le tc;Rd . Raising the temperature enhanced cleaning, leading to shorter cleaning times. Figure 7(a) shows the cleaning pro le when a NaOH temperature below room temperature was used. This took signi cantly longer > 50% to clean as compared with room temperature cleaning. This may be due to very slow diffusion and reaction of NaOH into the deposit. It is unlikely to be practical to clean below room temperature, so subsequent cleaning studies feature NaOH temperatures at or above room temperature. Similar in uences of temperature on cleaning pro les were also observed at other concentrations and velocities (Tuladhar8 ). Cleaning Time Swelling, plateau and decay times could be measured from both the thickness and thermal resistance pro les. Due to the effect of the gauging sensor in the decay stage, the total cleaning time was taken as the time taken for total recovery of the original heat transfer rate, tc;Rd , (agreeing to ca. 9095% with the time required for complete mass removal). Swelling, plateau and decay times were also obtained from the Rd pro le (except where otherwise stated). The times taken to clean, swell, plateau and decay have been corrected for variation in the deposit coverage by a simple scaling, viz: tc;Rd tswell;Rd tcs ; tcs;swell ; Pcov Pcov tplateau;Rd tdecay;Rd tcs;plateau ; tcs;decay 6 Pcov Pcov where Pcov is the total protein coverage g m2 obtained from protein assay. Note that the scaled times have units of s m2 g1 .

206

TULADHAR et al.

Figure 7. Effect of temperature on cleaning pro les using 0:5 wt% NaOH at 0:22 m s1 : (a) 16 C; (b) 35 C; (c) 50 C. Symbols : normalized Rd ; : normalized dd ; : protein fraction.

Figure 8 shows the scaled time as a function of temperature for 0:5 wt% NaOH solution at a ow velocity of 0:22 m s1 on each stage of cleaning, namely the swelling, plateau and decay stages, and the overall cleaning time. The length of all three stages and the overall cleaning time decrease signi cantly with increasing temperature up to 35 C and decrease slightly thereafter. It is also evident that the decay phase is the longest stage: the swelling and plateau times are relatively small compared to the decay time. It is also noteworthy that at constant velocity, a temperature of 50 C involves a Reynolds number approximately twice that at 25 C, chie y via the solution viscosity. Cleaning time fraction Although the cleaning time, and hence, individual stage times decrease with temperature (see Figure 8), Figure 9

indicates that swelling takes only 1015% of the total time at all temperatures. The decay stage occupies most of the time involved in cleaning, at 5570%. It appears that increasing the temperature (up to 35 C) leads to a slight reduction in tswell and tplateau and hence longer tdecay. The contribution of temperature is more evident at 0:3 wt% NaOH (see Figure 10(a)) where tswell decreases more noticeably with increase in temperature than at a higher concentration of 0:5 wt% (see Figure 10(b)). This indicates that the temperature of the cleaning solution is more important at lower concentration. Swelling ratio There is also a signi cant decrease in swelling with increasing temperature. It is evident from Figure 7 that during cleaning both Rd and dd increase to a maximum Trans IChemE, Vol 80, Part C, September 2002

CLEANING-IN-PLACE OF WHEY DEPOSITS

207

Figure 8. Effect of temperature on overall cleaning, swelling, plateau and decay time using 0:5 wt% NaOH solution at 0:22 m s1 bulk velocity. Symbols : scaled overall cleaning time, tcs ; D : scaled swelling time, tcs;plateau; F : scaled plateau time, tcs;plateau; &: scaled decay time, tcs;decay ; dashed lines show overall trend.

value before starting to descend. At 22 C, the deposit swells by a factor of ca. 1.8 and both FRd and Fd drop rapidly with increasing temperatures to around 1.2 at 35 C and above (within experimental error). Times to reach maximum swelling, tmax;d and tmax;Rd obtained from the Rd and dd pro les, respectively, decrease rapidly with increasing temperature to around 120 s at 35 C; at 40 C and above, there is hardly any noticeable change. At temperatures above 22 C, the times to reach FRd and Fd agree to within 20 s, suggesting that Rd ;max corresponds to the maximum value of dd;max . Several workers have suggested that the length of the swelling stage is determined by the rate of diffusion of hydroxide to an interface between native and swollen deposit. The thickness measurements allow the calculation of an approximate diffusion time. If an estimate of hydroxyl ions diffusivity of 5 109 m2 s1 (as used by Bird15 and Gillham16) is used, the total time for NaOH solution to diffuse into the deposit structure for these layers of ca. 400 500 mm thickness, would be estimated by Ficks second law as ca. 10 s (Cussler17 ). This is smaller than, but of the same order as, the swelling time above 30 C. However, it would be expected that the diffusivity of NaOH through a protein would be a facilitated diffusion wherein the OH diffusivity is substantially larger than the above value. Hence, it can be concluded that it is likely that the cleaning

solution diffuses into the deposit very quickly and then takes time to react, during which swelling occurs; and/or that there is signi cant interaction between hydroxide and dissolved product affecting diffusion and solubility. Microstructure studies of similar WPC deposits treated with alkali, using a range of surface analysis techniques (SEM, CSEM, X-ray elemental mapping), by Gillham16 also con rmed that the cleaning solution diffused into the deposit very rapidly 3 s. Gillham also reported that the deposit morphology changed from an amorphous structure of protein aggregates of size 110 mm to a fused structure after 3 s and then changed further, over 4 minutes, to a structure resembling a regular network of thin strands 0:20:5 mm, enclosing narrow pores 0:51 mm. Effect of Velocity Cleaning rates are usually found to increase with increased mechanical action, and a frequently used industry guideline is to clean at a uid velocity higher than 1:5 m s1 (Romney18 ). Cleaning studies on a laboratory scale are usually performed at lower ow rates. When studying the removal of whey protein deposits by sodium hydroxide15 , measurable reductions in cleaning times with increase in velocity were found even at low ow velocities 0:170:53 m s1 . This section discusses the effect of

Figure 9. Swelling, plateau and decay times as a fraction of cleaning time using 0:5 wt% NaOH solution at 0:22 m s1 .

Trans IChemE, Vol 80, Part C, September 2002

208

TULADHAR et al.

Figure 10. (a) Swelling and decay times as a fraction of cleaning time using 0:3 wt% NaOH solution. Case: um 0:030:30 m s1 ; Re 50010,000. Symbols: swelling stage; D: decay stage. (b) Swelling and decay times as a fraction of cleaning time using 0:5 wt% NaOH solution. Case: um 0:03 0:30 m s1 ; Re 50010,000. Symbols: swelling stage; D: decay stage.

ow velocity, Reynolds number and shear stress on cleaning of WPC deposits using NaOH solution, at constant temperature and concentration. In general, the cleaning pro les observed were similar to those reported above. An increase in bulk ow velocity showed enhancement in cleaning, leading to shorter cleaning time as registered from the Rd and protein pro les. However, the dd cleaning pro le shows only a slight reduction in the cleaning time with increase in bulk ow velocity. This may be due to the fact that the shear stress imposed on the surface by the gauging sensor dominates that imposed by the bulk ow. This was particularly apparent at very slow ow velocities. Cleaning time Figure 11(a) shows the variation in overall scaled cleaning time with temperature for 0:5 wt% NaOH solution at different ow velocities, um , 0:030:30 m s1 . This plot shows that: (i) not only is there a strong effect of temperature on tcs ; but (ii) for a given temperature, the tcs values also depend on um , decreasing as um increased. This behaviour is also evident in all three stages of cleaning as shown in Figures 11(b) and (c), most noticeably between 0.03 and 0:15 m s1 . The degree of in uence of um increased with NaOH bulk temperature. A 10-fold increase in um (0.03 to 0:30 m s1 ) at 22 C and 50 C decreased tcs by a factor of ca. 1.2 and 2.0 respectively (see Figure 11(a)). The sensitivity towards um in the swelling stage appears to decline with increasing temperature.

Figure 11(c) shows a different sensitivity to velocity in the plateau stage, in that the times observed at the lowest velocity, of 0:03 m s1 , are noticeably longer, suggesting that a threshold velocity for the cleaning mechanism may exist, related to surface shear stress (or Re). The shear stresses at 0.03 and 0:15 m s1 were ca. 0.01 and 0:10 Pa. Several workers have reported the existence of a threshold Re value below which mechanical effects in CIP were negligible. Jennings et al.19 reported a threshold Re value of 25,000 in the cleaning of milk deposits. Gillham et al.5 studied the removal of WPC deposit by 0:5 wt% NaOH at similar temperatures to those employed here and found that the cleaning time varied signi cantly even at low Re 500 Re 2000 and was much less sensitive to Re at higher values, up to 5000. These two results suggest the existence of upper and lower threshold values, between which the effect is minimal. Cleaning is often reported as a function of Reynolds number, but Timperley20 concluded that the value of mean velocity (and hence surface shear stress) is more signi cant in cleaning rather than the Reynolds number. Figure 11(d) present the cleaning time replotted as a function of surface shear stress, tw , under various temperature conditions. The shear stress at the wall, tw , was calculated using laminar and the turbulent equations21 assuming a smooth proteinaceous deposit surface and neglecting duct geometry:
8ru2 m Laminar flow Re 0:25 tw 0:0396ru2 Turbulent flow m Re

tw

7 8

Trans IChemE, Vol 80, Part C, September 2002

CLEANING-IN-PLACE OF WHEY DEPOSITS

209

Figure 11. Summary of cleaning time parameters using 0:5 wt% NaOH solution. Effect of (a) temperature and ow velocity on overall cleaning time; (b) temperature and ow velocity on scaled swelling and decay time: Symbolstcs;swell : black; tcs;decay : grey; (c) temperature and ow velocity on scaled plateau time; (d) tw on overall cleaning time.

where um is the mean liquid velocity and r the density of the liquid. The corresponding plot with Re on the x-axis is not reproduced here. The results show a reduction in tcs with increasing tw (and Re). The effect is more noticeable at lower temperatures. These results indicate that when reporting cleaning of such deposits, the temperature at which cleaning takes place is more signi cant than the values of Re or tw. In the range used, there appears to be no incentive to report tw rather than Re, as suggested by others. There was no evidence of the existence of an upper or lower threshold Re value, as cleaning time decreased with increasing Re 400 Re 10;000 for each temperature. Trans IChemE, Vol 80, Part C, September 2002

Cleaning time fraction An increase in ow rate of the cleaning solution increases mass transfer rate both to and from the surface, and the shear stress at the surface, which may lead to a shortening of the decay stage. The proportion of time occupied by the decay stage was found to be relatively insensitive to ow rate. The swelling time fraction, however, decreased slightly with increasing ow rate (especially at temperature 30 C) suggesting an effect of mass transfer aiding in faster diffusion. It should be noted that the ow velocity used here 0:030:30 m s1 is signi cantly less than that practised in industry18, viz around 1:5 m s1 .

210

TULADHAR et al. The deposit thickness pro le at 2:0 wt% is noticeably different to the thermal resistance pro le. The shorter measurement of decay stage, due to siphon suction, gives a tc;d value of 770 s which is consistent with those observed at lower concentrations. The decay times, as indicated by the heat recovery and mass removal values, however, are markedly longer than those at lower concentrations. This effect was observed at all temperature and ow velocities for 2:0 wt% NaOH and is discussed further below. Cleaning time Figure 13 shows a scaled cleaning time, tcs , as a function of concentration at a mean solution velocity of 0:30 m s1 and different operating temperatures. For 30 and 40 C, as the concentration is increased from 0.3 to 1:0 wt%, cleaning performance is enhanced producing a shorter cleaning time. However, cleaning experiments at the higher concentration of 2:0 wt% involved signi cantly longer cleaning times, with a ca. 3-fold increase compared to 1:0 wt% NaOH. An optimum concentration, Copt, clearly exists at around 0:5 1:0 wt% NaOH. The concentration optimum was also evident in each cleaning stage over this concentration range. The existence of Copt is also evident at 50 C (see Figure 13), but tcs increased only slightly at 2:0 wt%, mainly due to the increase in tcs;decay . There was no noticeable change in tcs;swell and tcs;plateau as compared to other concentrations. For 2:0 wt% NaOH solution, there is a signi cant reduction in tcs when the temperature is raised from 40 C to 50 C, by a factor of 6 as compared to 1:5

Swelling ratio There is no clear evidence of change in FRd and Fd with velocity. A ten-fold increase in ow velocity from 0.03 to 0:30 m s1 reduced both FRd and Fd slightly < 10% using 0.5 wt% NaOH solution at 40 C. The time to reach maximum swelling also dropped somewhat from ca. 150 s to 90 s. It appears that an increase in ow rate may aid in the removal of the upper swollen deposit before all the deposit has been swollen, leading to smaller FRd and Fd . Effect of Concentration The effect of cleaning solution concentration has received considerable attention. A series of cleaning experiments were performed under similar conditions (TNaOH 30 C and um 0:15 m s1 ) and various concentrations. The cleaning curves featured the general shape discussed previously. Figure 12(a) shows that at 0:3 wt%, both Rd and protein pro les indicate that the deposit was removed completely at ca. 1060 s. On increasing the concentration to 0:5 wt%, the cleaning time decreased to 900 s. A further concentration increase to 1:0 wt% produced a slight reduction in the cleaning time to 840 s. However, when the concentration was increased to 2:0 wt% (see Figure 12(b)), tc;Rd increased to 2700 s, with a noticeably longer decay phase. This suggests that with increasing concentration up to 1:0 wt%, the deposit strength is reduced to such an extent that slow-moving uid is suf cient to detach the deposit in the decay phase. At 2:0 wt%, however, the opposite applies.

Figure 12. Effect of concentration on cleaning pro les at 30 C at 0:30 m s1 : (a) 0:3 wt% ; (b) 2:0 wt%. Symbols : normalized Rd ; : normalized dd ; : protein fraction.

Trans IChemE, Vol 80, Part C, September 2002

CLEANING-IN-PLACE OF WHEY DEPOSITS

211

Figure 13. Effect of NaOH concentration on scaled cleaning time at different NaOH bulk temperatures: um 0:30 m s1 .

swelling (or reaction) of the deposit, representing some type of saturation effect. However, the time to reach this maximum swelling ratio drops with increasing ow rate up to 1:0 wt% but increases signi cantly at 2:0 wt% by a factor of around 2. Hence, at 2:0 wt%, although the deposit swells to an extent similar to that at the optimum concentration 0:51:0 wt%, the time to reach maximum swelling is much longer. The values of Fd , tmax;d , FRd and tmax;Rd obtained using NaOH solutions at 0.15 and 0:30 m s1 ow velocity at other temperatures indicate that although the concentration effect on swelling ratio trend is not clear at higher temperatures (40 C and 50 C), the time to reach maximum swelling generally shows similar trends to that obtained for 30 C. ANALYSIS OF CLEANING RATES The insensitivity of cleaning behaviour to gauge suction in the swelling and plateau stages observed at concentrations < 2:0 wt% indicates that the cleaning mechanism in these stages is not surface erosion: the generation of a fragile deposit layer and its initial removal appears to be controlled by mass transfer and=or reaction, which can be studied via rate analysis. As little protein is removed until the onset of the decay stage, mean cleaning, swelling and decay rates were based on total protein coverage as: rc
Pcov ; tc;Rd rs
Pcov ; tswell;Rd rd
Pcov tdecay;Rd 10

for a similar 10 C increase from 30 to 40 C. This notable reduction in cleaning time was also obtained at other ow velocities. These results indicate that improvement in WPC deposit cleaning due to NaOH concentration and temperature are cumulative and inter-related. The existence of an optimal cleaning solution concentration has been also discussed by other workers. Romney18 attributed the optimum to an increase in solution viscosity at higher concentrations of alkali making penetration of the deposit more dif cult, but this is unlikely as the viscosity of a 2:0 wt% caustic solution is only 10% greater than that of water15. It is more likely that increases in caustic concentration induce changes in the deposit structure by chemical means. Protein deposits immersed in 2:0 wt% NaOH were noticeably more sticky to the touch than those immersed at lower concentrations. The optimum has been linked to morphological changes at different alkali concentrations15 . Bird15 recorded morphological changes in SEM studies of protein deposits swollen by hydroxide solutions. Micrographs of structures swollen with 0:5 wt% alkali at 50 C showed more open pore structures than those obtained at higher or lower concentrations. The methodology used in these SEM studies suggests that some of the observed features were affected by sample preparation, particularly rupture by ice crystals, but the variation indicates that signi cant differences in micro-structure, which will affect both mechanical strength and diffusion, are involved. Gillham et al.5 con rmed this artefact during sample preparation of the swollen deposit using a chemical xation technique. Cleaning time fraction Plots of cleaning time fractions, similar to Figure 9, for the different concentrations studied (neglecting the ow rate effect) suggested that the proportion of time occupied by the swelling and decay stage were relatively insensitive to concentration at all NaOH bulk temperatures. At all concentrations, the decay stage was the longest, taking 6070% of the total cleaning time, with swelling taking 1015% of the total cleaning. Swelling ratio Both FRd and Fd decreased with increasing concentration up to 1:0 wt% while a further increase to 2:0 wt% resulted in little change in FRd and Fd . This suggests that above the optimum, NaOH concentration has no in uence on the
Trans IChemE, Vol 80, Part C, September 2002

where rc , rs and rd are mean cleaning, swelling and decay rate in g m2 s1 , respectively. Concentration Effects on Cleaning Rates The relation between the deposit removal rate and NaOH concentration is determined for concentration 1:0 wt% only, since the cleaning behaviour at 2:0 wt% was clearly related to surface shear stress. The reaction orders for overall cleaning, rc , swelling, rs , and decay rates, rd , neglecting the effect of ow velocity, are listed in Table 1. At 22 C, the swelling rate was found to be approximately rst order in NaOH concentration. With measurements made at only three concentrations (0.3, 0.5 and 1:0 wt%), great weight cannot be put on this result, but it is notable that it agrees with that reported by Jennings22 , who treated the initial phase of milk deposit removal as rst order for concentrations lower than 0:6 wt%. The effect of concentration on swelling, however, appeared to be sensitive to the temperature and its reaction order decreased with increasing temperature. At 30 C and 40 C, the swelling reaction order was around 0.60.5.
Table 1. Reaction order at different temperatures for mean overall cleaning, swelling and decay rates up to 1.0 wt% NaOH solution. NaOH bulk temperature C 22 30 40 Reaction order for Overall cleaning rate 0.4 0.4 0.3 Mean swelling rate 0.9 0.6 0.5 Mean decay rate 0.2 0.3 0.3

212

TULADHAR et al.

At all temperatures, the reaction order for the decay rate is low (0.20.3), suggesting a less sensitive effect of concentration on the nal stage, as expected. Inevitably, the reaction order for the overall cleaning rate falls between that of swelling and decay rates, lying beneath half order (0.30.4). Corrieu23 reported the cleaning of milk protein deposit from steel and glass surfaces to be zeroth order with respect to hydroxyl ion concentration for NaOH concentration above 1:0 wt% based on data reported by Schlussler24. Temperature Effects on Cleaning Rates Figure 14(a) shows the results of a basic lumped-parameter approach, where the mean cleaning rate, rc , is presented in pseudo-Arrhenius form for 0:5 wt% NaOH solutions. Although there is noticeable scatter in the data due to the variation in ow rate (um 0:03 0:30 m s1 ; Re 400 10,000, the combined data sets yield an activation energy parameter, Ec , of 43 kJ mol1 . This value is greater than that reported by Gillham16; 23 34 kJ mol1 , and less than that reported by Bird and Fryer4 ; 80 kJ mol1 . The absolute magnitude suggests reaction-control, or mass transfer reaction control. Gillham16 hypothesized that the (low) magnitudes of the activation energies observed in such studies are linked to the partial denaturation which occurs during fouling: these activation energies observed in cleaning are related to the energy required to break the remainder of the denatured structure. The importance of hydroxide diffusion can be assessed by inspecting the effect of temperature on swelling behaviour. The mean swelling rates, rs , yielded a slightly higher activation energy, Es , of 46 kJ mol1 (see Figure 14(b)). This activation energy is higher than that which would be obtained for simple diffusion control, but broadly consistent

with activation energies for protein re-arrangement reactions25 . This value indicates that swelling is a reaction-controlled process, rather than a purely diffusional process. Figure 14(b) also shows that the mean decay rate data exhibit pseudo-Arrhenius behaviour, yielding an activation energy parameter, Ed , of 41 kJ mol1 . These values indicate that the decay stage is also strongly affected by temperature. As the temperature of the deposit lies between that of NaOH solution and the wall temperature, the activation energy of these processes was also calculated using the wall temperature. The activation energies obtained using both bulk and wall temperatures for other cleaning solution concentrations studied, neglecting any ow rate effect, are presented in Table 2. Each energy parameter exhibits a minimum within the same concentration interval, namely between 0:51:0 wt%. The optimum concentration, therefore, represents a condition which is least sensitive to temperature effects. That the swelling is a reactiondominated process was con rmed by studies at different temperatures, which yield apparent activation energies of 4676 kJ mol1 , as given in Table 2. Low activation energy, E , values 20 80 kJ mol1 have also been reported for typical enzyme, cellular and life-related reactions, occurring at around room temperature26 . The activation energy calculated for each bulk ow velocity used did not differ much from that tabulated in Table 2. This may be due to the fact that the velocity range of 0:030:3 m s1 is small. However, the activation energy calculated using cleaning, swelling and decay times obtained from the dd pro les for rate determination in equation (10) clearly demonstrate the mechanical effect in the decay stage. Inspection indicates that the extra surface shear stress created by the gauging sensor had negligible in uence on the E values for concentrations between 0:31:0 wt%, but had a signi cant effect for 2:0 wt% where the E values for all stages were reduced substantially. This raises questions about the applicability of laboratory studies, often performed at low velocities (small shear stresses) to high velocity (and shear stress) applications, as found in industrial CIP applications.

Table 2. Activation energies calculated for WPC cleaning studies at different concentrations using: (i) Rd pro le data (normal font); and (ii) dd pro le data (italics). First and second values are calculated from bulk and wall temperature respectively. Note that the dd pro le data also include the effects of gauge siphon force. Activation energy for NaOH concentration wt% 0.3 0.5 Overall cleaning rate, Ec kJ mol 1 4954 5962 4345 4345 4357 4661 7590 4857 Mean swelling rate, Es kJ mol 1 7076 7580 4650 5460 4761 5573 6274 3136 Mean decay rate, Ed kJ mol 1 4043 4953 4142 3234 3951 3749 7894 4654

Figure 14. (a) Pseudo-Arrhenius plot of mean overall cleaning rates for 0:5 wt% NaOH solutions. Symbols show data obtained at different ow velocities. (b) Pseudo-Arrhenius plot of mean swelling and decay rates for 0:5 wt% NaOH solution. Symbolssolid: mean swelling rates; open: mean decay rates.

1.0 2.0

Trans IChemE, Vol 80, Part C, September 2002

CLEANING-IN-PLACE OF WHEY DEPOSITS Q10 Values The Q10 value of a reaction is often used for reporting the temperature dependence of biological reactions. It is de ned as the number of times a reaction rate changes with a 10 C change in temperature27. The Q10 value is temperaturedependent and should not be used over a wide range of temperatures27. Q10 values of around 1:71:9 were obtained for the overall cleaning rate for concentration below 1:0 wt%, which agrees reasonably well with those cited in the literature for similar studies: 1.6 (Jennings28 ) and 1.8 (Gallot-Lavalle e et al.29 ). At all concentrations, the swelling rates appeared to double with a 10 C increase in temperature. A Q10 value of around 2 has also been reported for reactions such as enzymatically induced colour or avour change in foods, degradation of natural pigments and microbial growth27 . The decay rates appeared to increase by a factor of ca. 1.7 for every 10 C rise.
d dt E Ec Ed Es h H m Pcov q Q10 r rc rd rs R2 Rd Rd ;0 Rd ;max Re t tc;d tc;P tc;Rd tcs tcs;decay tcs;plateau tcs;swell tdecay;Rd tmax;Rd tplateau;Rd tswell;Rd tdecay;d tmax;d tplateau;d tswell;d TNaOH TH2 O um Uc Ut

213

NOMENCLATURE
tube inside diameter, m nozzle throat inside diameter, m activation energy, kJ mol1 activation energy for overall cleaning rate, kJ mol1 activation energy for decay rate, kJ mol1 activation energy for swelling rate, kJ mol1 clearance between the nozzle tip and the surface (deposit), m siphon head, m siphon discharge mass ow rate, g s1 total protein coverage, g m2 heat ux, W m2 factor by which reaction rates change with 10 C change in temperature reaction rate, g m2 s1 mean cleaning rate, g m2 s1 mean decay rate, g m2 s1 mean swelling rate, g m2 s1 regression coef cient deposit thermal resistance, m2 K W1 deposit thermal resistance at time t 0, m2 K W1 maximum deposit thermal resistance, m2 K W1 Reynolds number time, s total cleaning time from dd pro le, s total cleaning time from protein pro le, s total cleaning time from Rd pro le, s scaled cleaning time tc;Rd =Pcov , m2 s g1 scaled decay time tdecay ;Rd =Pcov , m2 s g1 scaled plateau time tRd ;plateau =Pcov , m2 s g1 scaled swelling time tRd ;swell =Pcov , m2 s g1 decay time from Rd pro le, s time to reach Rd ;max , s plateau time from Rd pro le, s swelling time from Rd pro le, s decay time from dd pro le, s time to reach dmax , s plateau time from dd pro le, s swelling time from dd pro le, s NaOH bulk temperature, C hot water bulk temperature, C NaOH bulk ow velocity in the duct, m s1 overall heat transfer coef cient of the clean surface, W m2 K1 overall heat transfer coef cient of the fouled surface at time t, W m2 K 1

CONCLUSIONS The technique of dynamic gauging has been implemented successfully in duct ow con gurations. A cleaning cell, incorporating this dynamic thickness gauging sensor and a commercial heat ux sensor, was used to study the cleaningin-place of WPC fouling deposits by 0:3 2:0 wt% NaOH solutions under well-de ned ow conditions over a range of temperatures 15 50 C and ow velocities 0:03 0:30 m s1 : Re 50010;000. A chemical protein assay was also performed to measure protein removal rate. Several features previously reported were observed, while the data generated allow the extractton of more, and new, information such as swelling rates from a single experiment. Cleaning of the protein deposits occurred in 3 stages, namely swelling, plateau and decay. The time taken to clean the deposit was strongly determined by the length of the decay stage, taking around 5070% of the total cleaning time, and swelling involved 1020% of the total cleaning time. Temperature and ow velocity had evident effects on the cleaning process and an optimum cleaning solution concentration was observed. Temperature was found to be the most important parameter, and a transition in swollen protein properties was observed between 1.0 and 2:0 wt% NaOH. Concentration dependency was less marked at higher temperature > 40 C. The results also suggest that the optimal concentration of the cleaning solution depend on the surface shear stress as the behaviour of swollen deposit obtained at 2:0 wt% was more sensitive to the shear regime underneath the gauge compared to the conditions imposed by bulk ow. The activation energy of swelling 4676 kJ mol1 indicates that swelling involves reaction (or simultaneous reaction and diffusion) control as opposed to being a purely diffusion-controlled process. The results reported here cannot yet be related directly to industrial practice, which normally feature temperatures higher than 60 C and ows in the turbulent regime. This work does, however, relate more closely to fouling deposits in dead zones or areas of low owwhich hygienic design seeks to eliminate. Work is currently under way30 to identify the shear stresses associated with the gauging ows, which will allow the mechanical effects noted here to be related to conditions arising in faster ows. Trans IChemE, Vol 80, Part C, September 2002

Greek symbols dd deposit thickness, mm dd ;0 initial deposit thickness, mm dd ;max maximum deposit thickness, mm DT thermal driving force TH2 O TNaOH , C or K Fd maximum swelling ratio dd ;max /dd ;0 FRd maximum swelling ratio Rd ;max /Rd ;0 ld thermal conductivity of the deposit, W m K 1 r uid density, kg m3 tdecay decay time fraction tdecay;Rd =tc;Rd tplateau plateau time fraction tplateau;Rd =tc;Rd tswell swelling time fraction tswell;Rd =tc;Rd tw surface shear stress, Pa Acronyms CIP CSEM d.b. HTC HTST i.d. MHFS RO SEM UHT w.b. and abbreviations cleaning-in-place cryogenic scanning electron microscopy dry basis overall heat transfer coef cient heat transfer surface temperature inside diameter micro-foil heat ux sensor reverse osmosis scanning electron microscopy ultra high treatment wet basis

214 REFERENCES

TULADHAR et al.
18. Romney, A. J. D., 1990, CIP: Cleaning in Place (The Society of Dairy Technology, Cambridge, UK). 19. Jennings, W. G., Mckillop, A. A. and Luick, J. R., 1957, Circulation cleaning, J Dairy Sci, 40: 14711479. 20. Timperley, D., 1981, The effect of Reynolds number and mean velocity of ow on the cleaning-in-place of pipelines, In: Fundamentals and Applications of Surface Phenomena Associated with Fouling and Cleaning in Food Processing Proceedings, Hallstro m, B., Lund, D. B. rdh, A. C. (eds.) (Reprocentralen, Lund, Sweden), pp. 402 and Tra ga 412. 21. Coulson, J. M. and Richardson, J. F., 1977, Chemical Engineering, 3rd edition (Pergamon Press, Oxford, UK), Vol. 1, pp. 318. 22. Jennings, W. G., 1965, Theory and practice of hard-surface cleaning, Advances in Food Research, 14: 388419. 23. Corrieu, G., 1981, State-of-the-art of cleaning surfaces, In: Fundamentals and Applications of Surface Phenomena Associated with Fouling and Cleaning in Food Processing Proceedings, Hallstro m, B., Lund, D. B. and Tra ga rdh, A. C. (eds.) (Reprocentralen, Lund, Sweden), pp. 90114. 24. Schlussler, H. J., 1970, Zur reinigung fester ober a chen in der lebensmittelindustrie, Milchwissenschaft, 25: 133145. 25. Middelberg, A. P. J., Radke, C. J. and Blanch, H. W., 2000, Peptide interfacial adsorption is kinetically limited by the thermodynamic stability of self association, PNAS, 97(10): 50545059. 26. Levenspiel, O., 1979, The Chemical Reactor Omnibook (OSU Book Stores, Oregon, USA), p. 1.4. 27. Toledo, R. T., 1999, Fundamentals of Food Process Engineering, 2nd edition (Aspen Publishers, Inc. USA), pp. 311314. 28. Jennings, W. G., 1959, Circulation cleaning III. The kinetics of a simple detergent system, J Dairy Sci, 42: 17631771. 29. Gallot-Lavalle e, T., Lalande, M. and Corrieu, C., 1984, Cleaning kinetics modelling of holding tubes fouled during milk pasteurization, J Food Proc Eng, 7: 123142. 30. Chew, J. Y. M., 2002, CFD of dynamic gauging ows, CPGS Report (Department of Chemical Engineering, University of Cambridge, UK).

1. Visser, H. and Jeurnink, T. J. M., 1997, General aspects of fouling and cleaning, In: Fouling and Cleaning of Heat Treatment Equipment (Bulletin of the International Dairy Federation), Visser, H. (ed.) (IDF, Brussels, Belgium), Vol. 328: 5. 2. Wilson, D. I., Fryer, P. J. and Hasting, A. P. M. (eds.), 1999, Fouling and Cleaning in Food Processing 98 (EU, Brussels, Belgium). 3. Burton, H., 1968, Deposit from whole milk in heat treatment plantA review and discussion, J Dairy Res, 35: 317330. 4. Bird, M. R. and Fryer, P. J., 1991, An experimental study of the cleaning of surfaces fouled by whey proteins, Trans IChemE Part C Food Bioprod Proc, 69(1): 1321. 5. Gillham, C. R., Fryer, P. J., Hasting, A. P. M. and Wilson, D. I., 1999, Cleaning-in-place of whey protein fouling deposits: Mechanism controlling cleaning, Trans IChemE Part C Food Bioprod Proc, 77(2): 127136. 6. Gillham, C. R., Fryer, P. J., Hasting, A. P. M. and Wilson, D. I., 2000, Enhanced cleaning of whey protein soils using pulsed ows, J Food Eng, 46: 199209. 7. Tuladhar, T. R., Paterson, W. R., Macleod, N. and Wilson, D. I., 2000, Development of a novel non-contact proximity gauge for thickness measurement of soft deposits and its application in fouling studies, Can J Chem Eng, 78: 935947. 8. Tuladhar, T. R., 2001, Development of a Novel Sensor for Cleaning Studies, PhD Thesis (University of Cambridge, UK). 9. Gotham, S. M., Fryer, P. J. and Paterson, W. R., 1988, The measurement of insoluble proteins using a modi ed Bradford assay, Analytical Biochem, 173: 353358. 10. Delplace, F. and Leuliet, J. C., 1995, Modelling fouling of a plate heat exchanger with different ow arrangements by whey protein solutions, Trans IChemE Part C Food Bioprod Proc, 73(C2): 112120. 11. Davies, T. J., Henstridge, S. C., Gillham, C. R. and Wilson, D. I., 1997, Investigation of whey protein deposit properties using heat ux sensors, Trans IChemE Part C Food and Biomod Proc, 75(C2): 106 110. 12. Rose, I. C., Watkinson, A. P. and Epstein, N., 2000, Testing a mathematical model for initial chemical reaction fouling using a dilute protein solution, Can J Chem Eng, 87(1): 511. 13. Bridge, S. P., Robbins, P. T., Paterson, W. R. and Wilson, D. I., 2001, A pneumatic gauging sensor for measuring the thickness of soft lms, Trans IMechE, Part E, 215: 1927. 14. Vaishnav, R. N., Patel, D. J., Atabek, H. B., Deshpande, M. D., Plowman, F. and Vossoughi, J., 1983, Determination of the local erosion stress of the canine endothelium using a jet impingement method, J Biomech Eng, 105: 7783. 15. Bird, M. R., 1992, Cleaning of Food Process Plant, PhD Thesis (University of Cambridge, UK). 16. Gillham, C. R., 1997, Enhanced Cleaning of Surfaces Fouled by Whey Proteins, PhD Thesis (University of Cambridge, UK). 17. Cussler, E. L., 1984, Diffusion: Mass Transfer in Fluid Systems (Cambridge University Press, Cambridge, UK), pp. 391426.

ACKNOWLEDGEMENTS
TRT wishes to acknowledge Cambridge Overseas Trust, Schlumberger Cambridge Research Limited and the Committee of Vice Chancellors and Principals, UK, for funding towards this research.

ADDRESS
Correspondence concerning this paper should be addressed to Dr Ian Wilson, Department of Chemical Engineering, New Museums Site, Pembroke Street, Cambridge CB2 3RA, UK. Email: ian_wilson@ cheng.cam.ac.uk The manuscript was received 16 January 2002 and accepted for publication after revision 26 April 2002.

Trans IChemE, Vol 80, Part C, September 2002

Anda mungkin juga menyukai