Kylee R. Goode, Konstantia Asteriadou, Phillip T. Robbins, and Peter J. Fryer Abstract: Fouling of food process plant surfaces and the subsequent cleaning needed is a signicant industrial problem, and as the cost of water and chemical disposal increases, the problem is becoming more signicant. Current literature on water-based cleaning is reviewed here according to the classication of 3 types of cleaning problems. By doing this, it is hoped that new knowledge can be highlighted applicable to improving industrial cleaning. (i) For type 1 deposits (that can be cleaned with water alone)Cleaning time appears related to Reynolds number and surface shear stress. An increase in Reynolds number seems to decrease cleaning time. Cleaning temperatures greater than 50 C do not appear benecial. (ii) For type 2 deposits (biolms)Removal behavior of biolms seems to be dependent on the microbial aging time on the surface. Keeping a material hydrated on a surface enables easier removal of it with water. a. Water rinsing: Temperature and wall shear stress have varied effects on removal. b. Chemical rinsing: Flow and temperature were seen to have the biggest effect at the start of cleaning, but contact time was more important as cleaning progressed at a given sodium hydroxide solution ow and temperature. (iii) For type 3 deposits (that require a cleaning chemical)For specically, protein-based systems excessive chemical forms a deposit difcult to remove. Increasing wall shear stress and temperature was most benecial to cleaning rather than concentration. The action of temperature can reduce the use of a chemical for type 2 and type 3 soils. The ndings suggest that the right combination of ow characteristics at a given temperature and concentration is crucial to achieving fast cleaning in all cases. There are a number of cleaning monitoring methods at various stages of commercialization that may be capable of monitoring bulk cleaning and cleaning at the surface. To optimize cleaning will require integration of measurement methods into the cleaning process.
Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. Fouling studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.1 Adhesion of microorganisms to surfaces . . . . . . . . . . . . . . . . 4 2.3 Preventing fouling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.3.1 Process surface modication . . . . . . . . . . . . . . . . . . . . . . 5 2.3.2 Process alterations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3. Cleaning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 3.1 Product recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.2 The effect of CIP parameters on type 1 removal . . . . . . . 10 3.2.1 Flow and wall shear stress . . . . . . . . . . . . . . . . . . . . . . . . 11 3.2.2 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.2.3 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.3 The effect of CIP parameters on type 2 and type deposit removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.3.1 Membrane cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.3.2 Water-rinsing of hard surfaces . . . . . . . . . . . . . . . . . . . . 14
3.3.3 Chemical effects on the cleaning of type 2 deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.3.4 Chemical effects on the cleaning of type 3 deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 4. Novel cleaning approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4.1 Increasing boundary layer disruption. . . . . . . . . . . . . . . . . .16 4.2 Alternative cleaners to reduce environmental impact . . . 16 4.3. Other studies related to cleaning behavior . . . . . . . . . . . . 17 4.3.1 Deposit shear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.3.2 Deposit deformation and strength . . . . . . . . . . . . . . . . 17 5. Measuring cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 5.1 Online bulk measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 18 5.2 Online surface measurements . . . . . . . . . . . . . . . . . . . . . . . . 19 5.3 Measuring microbial cleanliness . . . . . . . . . . . . . . . . . . . . . . 19 6. Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Introduction
The success of a branded fast moving consumer goods (FMCGs) business depends fundamentally on product quality and safety conMS 20120483 Submitted 3/29/2012, Accepted 10/14/2012. Authors are with formance at a required level. Poor cleaning or hygiene conforSchool of Chemical Engineering, Univ. of Birmingham, Edgbaston, Birmingham, B15 mance can be a result of fouling layers building up in a plant 2TT, U.K. Direct inquiries to author Fryer (E-mail: p.j.fryer@bham.ac.uk). or other problems. Figure 1 illustrates a typical route by which a large FMCG manufacturer will ensure a hygienic plant. There
C 2013 Institute of Food Technologists doi: 10.1111/1541-4337.12000
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 121
Scienc invesgaon
By site educaon, training and empowerment and recording site CIP condions
NO
YES
Figure 1Flow diagram illustrating the route taken by industry to ensure plant hygiene (Heineken personal communication 2012).
should be fundamental research and development obtained from both practice and science that are integrated and applied in-plant to provide the optimum cleaning protocol. In food and beverage manufacturing operations, cleaning-in-place (CIP) is used to remove residual product, fouling, and microbes that remain in the process line from production. The act of cleaning therefore maintains product quality, safety, and production efciency. During CIP, water and/or chemical solution is circulated around plant process equipment. With large-scale manufacturers, the process is generally fully automated. A typical CIP philosophy in industry is that of Scottish & Newcastle Breweries (2008):
ensure all production, processing, and packaging plant is cleaned by a standard regime and to a schedule which ensures cleanliness and microbiological integrity at all times; with minimum cost, energy, and delay to production in a manner which ensures human, plant, product, and environmental safety.
identify the best way to clean a processing plant from experiments of different plants. The direct selection of cleaning protocols is not always possible. In practice, cleaning protocols can only be developed semiempirically in industry. In most cases, CIP cannot be optimized in situ because of the risk posed of compromising existing cleanliness. Fryer and Asteriadou (2009) suggest a classication of cleaning problems in terms of cleaning cost and soil complexity. A diagrammatic representation of this relationship is presented in Figure 2. This classication enables the nature of a foulant to be related to the type of cleaning employed, and therefore, the cost. This classication also indicates the environmental impact of the type of cleaning employed; complex soils require chemical and thermal cleaning that lead to a high cleaning cost and high environmental impact. Three deposit types were chosen to represent a broad range of cleaning problems seen in food, beverage, and personal care products manufacturing:
(i) Type 1: Viscoelastic or viscoplastic uids such as yogurt and A signicant body of cleaning knowledge exists within individtoothpaste that can be rinsed from a process surface with ual manufacturers, equipment suppliers, and chemical companies; water. however, the determined cleaning regimens have often been kept (ii) Type 2: Microbial and gel-like lms such as biolms and condential and plant-specic. This has resulted in independent polymers removed in part by water and in part by chemical. development of cleaning operations. Organizations such as the European Hygienic Engineering and Design Group (EHEDG) (iii) Type 3: Solid-like cohesive foulants formed during thermal processing such as milk pasteurization and brewery have produced extensive guidelines on the types of surface and wort evaporation. These operations mostly require chemical equipment that are easy to clean, such as detailed in the EHEDG removal. Yearbook (2007). CIP tends to follow a similar series of steps for a prescribed Yang and others (2008) also classied cleaning optimization time and at a prescribed ow rate, temperature, and chemical methods, here into 2 types of investigation: concentration known to give a repeatable level of cleanliness. It (i) Engineering investigations: reducing energy, time, and cost in is not yet possible to predict before an operation how a given established cleaning operations. piece of equipment could foul and be cleaned. It is difcult to
122 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
2013 Institute of Food Technologists
LACK OF FUNDAMENTALS e.g. The eect of CIP parameters on dierent fouling condions
Figure 2Cleaning map; a classication of cleaning problems based on soil type and cleaning chemical use (from Fryer and Asteriadou 2009).
(ii) Scientic investigations: achieving cleanliness or a cleaning time as a function of inuencing factors; for example, wall shear stress, temperature, surface type, and nish. The aim of this review was to provide an overview of current knowledge on cleaning solutions classied by soil cleaning type. This novel classication is hoped to highlight new CIP optimization opportunities for industry and any future research in the eld. Current knowledge of fouling prevention and novel cleaning methods are also discussed here.
r Growth of biolmoften after the formation of a condir Accumulation of material in stagnant or low-ow areas of r Loss of membrane activity.
Fouling is a costly problem in the food, beverage, and other industries, which is often unavoidable due to the heat treatment that often has to be given to products to develop certain colors and avors and ensure safety. By denition, foods are sources of nutrients favorable not only to people but also to microbes that stick to process surfacesso microbial adhesion to surfaces and subsequent growth are important pheFouling Studies Fouling is dened as the unwanted buildup of material on a nomena. The economic penalties of fouling in heat exchanguller-Steinhagen (2000) and can be surface. The fouling process generally involves a number of steps ers were discussed by M summarized: (Epstein 1983): (i) (ii) (iii) (iv) (v) surface conditioning, mass transfer of species to the surface, surface deposition, deposit aging, and possible removal. (i) Capital expenditure, due to: (a) Excess heat transfer surface area compensating for the occurrence of fouling. This has been estimated as an average of 30% additional capital cost. (b) Higher transport and installation costs for bigger and heavier equipment. (c) Cleaning systems, including their installation and maintenance costs. (i) Fuel costIf extra energy (such as steam) is required to keep the fouled heat exchanger operating for the required performance. (ii) Maintenance costOf the heat exchanger, cleaning system, and any ancillary equipment in the process (and cleaning) loop, for example, chemical tank level probes, ow meters, interface probes, and boilers.
There is also a classication of fouling mechanisms demonstrated by Bott (1990) detailed in Table 1. Fouling problems that have been reported in the food and beverage industries include (but this is by no means a complete list):
r r r r r
Protein and mineral deposition in heat exchangers. Ice buildup in freezers. Scale buildup in cooling water systems. Fat burn-on in ovens. Product solidication.
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 123
Formation of crystals on the surface formed from solutions of dissolved substances when the solubility limit is changed. Cooled surfaces are subject to fouling from normally soluble salts, fats, and waxes. Inversely soluble salts, such as calcium phosphate deposits on heated surfaces. Where the uid or components of the uid solidify onto the surface, this is called solidication fouling (Sharma and others 1982). Particulate deposition Small suspended particles such as clay, silt, or iron oxide deposit onto heat transfer surfaces. Where settling by gravity is the determining factor, this is then called sedimentation fouling. Biological growth (biofouling) The deposition and growth of organic lms consisting of microorganisms and their products, called biolm. Chemical reaction at Reaction of some part of the ow to generate insoluble material. The deposit formed on the surface (particularly heat uid/surface interface transfer surfaces) has a different composition to the process uid (for example, in petroleum rening, polymer production, and dairy plants). Corrosion The material of the heat transfer surface is involved in reactions with components of the uid to form corrosion products on the surface, a specic type of chemical reaction fouling. Freezing Deposit formed from a frozen layer of the process uid, for example, ice from water or solid fats from a food uid.
(iii) Cost due to production lossCost of continuous pro- organisms usually follows the formation of a conditioning layer duction (without shut-down for cleaning or maintenance) of protein (Lorite and others 2011) that makes subsequent adhesion and biolm formation easier. The sequence of events that as compared to the actual production cost. occur during lm formation is discussed by Busscher and others Accurate measurement of the effects of fouling and the efciency (2010) and Chen and others (2010) who show the kinetics of lm is critical. Changes in heat transfer efciency have widely been formation. recorded. Most common is following the change in heat transfer Other researchers have studied yeast adhesion and proliferation during fouling by including a fouling resistance, Rf , in the equation on processing surfaces, critical in brewing operations. Reynolds relating the initial clean heat transfer coefcient, (U0 ), to that at and Fink (2001) proved that Bakers yeast can initiate biolm fortime t, (U): mation on plastic when in a low-glucose environment. Mozes and 1 1 others (1987) found that yeast could attach and form a dense layer + Rf (1) = of cells on stainless steel and aluminum at pH 3 and pH 5 and 6. U U0 The authors also determined that a dense layer of yeast cells would And the extent of fouling may be expressed by a Biot number attach to glass and plastics if the negative charge was reduced by (Bi), which accounts for deposit thickness (x) and thermal con- treatment with ferric ions. The system pH will determine the ductivity (): Bi = Rf .U0 , where Rf = x/ for the deposit. Deposit surface charge of both the substrate and the adhering species. The resistance during cleaning can be described as the reverse process isoelectric point, the pH where the material carries no charge, will to (1) (Tuladhar 2001) as also vary with surface and organism. Yeast has also been found by other authors to readily attach to stainless steel, plastics, elastomers 1 1 Rd = (2) (Guillemot and others 2006), and glass (Mercier-Bonin and othUc Ut ers 2004), all of which are used extensively in FMCG industries. where Ut is the heat transfer coefcient at time t and Uc the heat The effect of cleaning parameters on yeast removal from process transfer coefcient of the nal clean system, so the rate of change surfaces is discussed in later sections. of this is a measure of cleaning. Product contact surface nishes with a roughness (Ra ) value The rate and extent of fouling and cleaning is often classied of up to 0.8 m are recommended (Lelieveld and others 2005), in terms of uid ow, either in terms of the Reynolds number which is often called 2B nish of stainless steel. Surface rough(Re = vd/, where and are the density of viscosity of a ness exists in 2 principal planes, one perpendicular to the surface uid owing at mean velocity v through a system of characteristic described as height deviation and one in the plane of the surlength d, such as pipe diameter) or the surface shear stress. In face described by spatial parameters. The effect of average surface this paper, many correlations in terms of Reynolds number are roughness height, Ra , and surface topography on microbial rediscussedto convert to velocity requires knowledge of density tention has been investigated most thoroughly. Hilbert and others and viscosity of the uid, which is simple for water but may be (2003) investigated the effect of stainless steel roughness (Ra 0.9 to more complex for cleaning solutions. 0.01 m) on retaining various microbes. The surfaces also had a conditioning layer. The retention of microbes (measured by indiAdhesion of microorganisms to surfaces rect conductometry) on the conditioned surfaces was similar over The principal factors responsible for adhesion between surface the range of Ra tested. and foulant include: (i) van der Waals forces, (ii) electrostatic forces, Cluett (2001) investigated the effect of stainless steel surface and (iii) contact area effects; the larger the area, the greater the nish on the fouling and cleaning of a beer fermenter. Surface total attractive force (Bott 1995). Microbes have a natural afnity nishes investigated included 2B milled stainless steel and meto surfaces. Numerous authors have reported the adhesion of bac- chanically polished 120 grit, 240 grit, and electropolished (EP) teria to processing surfaces (for example, Geesey and others 1996; stainless steel. The top surface of the fermenter was half EP, half B en ezech 2001; Zhao and others 2007). If left to proliferate, in- 240 grit, and the cone was EP. The cylinder of the vessel had dividual microbes can grow into biolms (adhesive and cohesive all nishes, one quarter of the vessel from top to bottom reprecommunities of microbes) that become difcult to remove from sented by each surface nish. After lager beer fermentation lasting a surface (Jefferson 2004). Garrett and others (2008) summarize 12 d, Cluett (2001) found that all surfaces fouled similarly and the occurrence of biolms in industry, fouling mechanisms and the level of deposition was heavy. He also found that all the surmethods of observing and probing structures. The adhesion of faces cleaned similarly using a similar CIP regime with a spray ball
124 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
Preventing fouling If fouling were not to occur, there would be little need for cleaning. Broadly speaking, 2 methods for preventing fouling have been approached in the literature:
(i) Functional surfacesFor example, smooth surfaces with specic nish, topography, hydrophobicity, or surface charge. Nonstick surfaces are designed to have a specic surface energy to minimize fouling. (ii) Processing alterationsFor example, changing product ow characteristics, holding times, transient times, and other process parameters designed to minimize fouling.
Process surface modication. A hygienic surface needs to be smooth, easy to clean, able to resist wear, and retain its hygienic qualities. Stainless steel is the most common food contact material used in the industry, being stable at a variety of temperatures, inert, relatively resistant to corrosion, and it may be treated mechanically or electrolytically to obtain a range of nishes (Akhtar and others 2010). The wettability of a surface is dependent on its surface energy. A surface with a high surface energy is hydrophilic and a drop of cleaning uid will spread over the surface. A low-energy surface is hydrophobic and a drop of water will not spread. Water partially wets glass and acrylic and does not wet Teon (PTFE) surfacesbut surfactants are often added to commercial cleaning agents to improve wetting. Wetting is determined by the nature of both the liquid and the solid substrate. The cleanability (and disinfectability) of stainless steel has been compared with those of other materials, and is comparable to glass when cleaning microbes, and signicantly better than polymers, aluminum, or copper (Akhtar and others 2010). Microbes are known to readily attach to SR. Everaert and others (1998) absorbed long uorocarbon chains (Ar-SR-C8 F17 ) to SR used in prosthetics in an attempt to reduce the number of adhering microbes. They found that the initial adhesion rate of Streptococcus bacteria to the treated rubber was signicantly reduced, from around 2500 to 900 cm2 s1 , without a conditioning lm of saliva and 400 cm2 s1 with a conditioning lm of saliva. The adhesion rate of Candida species to treated rubber was also reduced compared to untreated rubber. Dhadwar and others (2003) investigated the effect of oligopeptide treatment of glass (hydrophilic) and plastic (hydrophobic) on
2013 Institute of Food Technologists
1 2
LW D +
LW F
(3)
LW LW LW , D , and F are the Lifshitzvan der Waals (LW) where S surface free energy of the surface, deposit, and uid, respectively, and which can be quantied from contact angle measurements (Zhao and others 2004). Liu and others (2006) studied the interactions of 316 L stainless steel with baked and unbaked tomato deposit: a minimum removal energy range of 20 to 25 mN/m was found in both cases. Either side of this surface energy range, the adhesive strength of the deposit on the surface increased. Zhao and others (2005a) found that stainless steel surfaces coated with Ag-PTFE reduced Escherichia coli attachment by 94% to 98%, compared with silver coating, stainless steel, or titanium surfaces. A surface with energy of 24.5 mN/m roughly matching the theoretical minimum adhesion energy of E. coli, 28.3 mN/m, was achieved. Composite coatings using nickel, phosphorus, copper, and PTFE were also used by Zhao and others (2005b) and Zhao and Liu (2006) to create surfaces with specic energies shown to reduce biofouling. A major EU project (MODSTEEL) developed and studied a wide range of surfaces and how they might reduce fouling from milk (see Santos and others 2004 and Rosmaninho and others 2007). Work by Pereni and others (2006) conrmed the effect of surface free energy in minimizing P. aeruginosa adhesion to a range of coatings including silicone, polished and nonpolished stainless steel, PFA (peruoroalkoxy polymer) and PTFE nickel, phosphorus, and aluminum composite coatings. The total surface free energy was in the range 17.2 to 48.3 mN/m, as shown in Figure 3(A). Minimum retention of bacteria was found at 20 to 27 mN/m. Silicone had a surface free energy of around 20 mN/m and the lowest colony forming units (CFUs) count. Surface free energy has been shown to be the parameter dominating E. coli adhesion over a range of metalpolymer coatings, and a minimum adhesion energy of 25 mN/m was found (Zhao and others 2007), as shown in Figure 3(B). Parbhu and others (2006) used a transient treatment to modify a stainless steel surface. The treatment was present during the processing cycle and removed at high pH during alkaline cleaning.
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 125
Figure 3(A) P. aeruginosa AK1 retention on investigated surfaces compared with the total surface free energy (adapted from Pereni and others 2006). (B) Effect of surface free energy on E. coli adhesion (adapted from Zhao and others 2007).
The treatment was shown to reduce the interaction potential between stainless steel and phosphate anions resulting in signicant reductions in fouling rates. Akhtar and others (2010) compared adhesion of a range of food and personal care foulants to different surfaces. Particle tips of different materials were attached to an atomic force microscope (AFM) cantilever to study the detachment from toothpaste and some confectionery components: Turkish delight, caramel, and sweetened condensed milk (SCM). The study did reveal signicantly different detachment forces for the same deposit from different surface types (see Figure 4). Caramel and SCM seemed to be more difcult to detach from glass than stainless steel. It was possible to relate data from the AFM to measurements taken on a millimeter scale using micromanipulation probes (Liu and others 2002, 2006, 2007). Akhtar and others (2012) describe further research using AFM to study food adhesion to process surfaces. Process alterations. Dror-Ehre and others (2010) tested the effect of biolm development of P. aeruginosa when pretreated in an aqueous solution of molecularly capped silver nanoparticles (MCNPs). Under specic conditions, cells and surfaces incubated for 39 h at 37 C, Ag-MCNPs retarded biolm formation even when a high percentage of planktonic P. aeruginosa cells survived pretreatment with Ag-MCNPs. At the various incubation times, a stable, low value of biomass was formed that could be easily removed. The authors found, from micrographs of pretreated cells, that the intracellular material was pushed toward the peripheral parts of the
Cleaning
No economically viable fouling prevention method is yet to be demonstrated in industry. Should one of the modied surface methods prove economic, then the problem will be greatly reduced. Understanding the cleanability of surfaces requires combining understanding of surface chemistry and engineering, the deposit and the cleaning uid (for a recent review of cleanability, see Detry and others 2010). Further research discussed here considers the ndings of studies relevant to optimizing cleaning. One signicant issue is determination of the correct cleaning time. A deposit that has aged on a surface is more difcult to remove than fresh material on a surface, so cleaning is encouraged after production. Aging of a particular soil type could make a deposit harder to remove from a process surface. For example, a type 1 soil could become a type 2 soil over time and heating may result in a type 3 soil. Goode and others (2010) found that in
2013 Institute of Food Technologists
126 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
0.01
0.001
0.0001
Figure 4Force of attraction between stainless steel, PTFE (uorinated low energy surface), and glass particles and different food materials obtained using AFM (from Akhtar and others 2010). F/R is the force/probe radius with units of N/m.
beer fermentation vessels, there were 2 distinct deposit types to be and Asteriadou 2009), is a useful cleaning problem classication tool and forms the basis for the structure of this review. Examples cleaned, classied as type A and type B foulants: (i) Type AFormed during fermentation above the beer level of each deposit type include: at the top of the vessel, (ii) Type BResidual yeast attached to the vessel wall and cone below the beer level during emptying. Type B fouling is shown by Salo and others (2008), as seen in Figure 5(A), while an example of type A foulant viewed from a fermenter man way door at the top of the vessel is given in Figure 5(B). Type B fouling has a shorter aging time than type A fouling. As such, type B foulant can be removed by the falling lm in a tank, whereas type A foulant may require a larger impact force for removal or a combination of water and chemical rinses for complete removal (Goode and others 2010). Similarly, Liu and others (2002) found that the force required to remove a tomato deposit from a surface increased with time until after about 200 min of heating it remained constant. Automated CIP has been widely applied in dairies, food processing, brewing, and wine processing for the last 50 y to return the plant to a clean state (Stewart and Seiberling 1996). During CIP, water and chemicals are circulated around the plant for a prescribed duration (Tamine 2008). The CIP factors found to determine cleaning can be described by Sinners circle, a circle of the cleaning parameters: mechanical action, chemical action, time, and temperature (Lelieveld and others 2005). Cleaning can also be dependent on geometry. In a pipe, the contribution of the cleaning factors is equal. In a pipe dead leg, time determines cleaning (Lelieveld and others 2005). A number of attempts have been made to try to incorporate computational models into the design process, as shown by Asteriadou and others (2006) and Jensen and Friis (2005). This approach will become more important as understanding of the processes in cleaning increases. Rheological characterization of materials enables their classication. Materials within a similar class may have similar cleaning behavior, according to the classication by Fryer and Asteriadou (2009). Vinogradov and others (2004) characterized the rheology of a dental plaque biolm. Biolm rheology has been viscoelastic, temperature-dependent and/or time-dependent (Rao 1999). Characklis (1980) compared the elastic and viscous modules obtained for a biolm and a cross-linked protein gel, brinogen. The elastic modulus was the same order of magnitude for the protein gel and the biolm. The cleaning map, presented in Figure 2 (Fryer
2013 Institute of Food Technologists
(i) Type 1: toothpaste, tomato paste, yogurt, shampoo, beer, wine, milk, and yeast. (ii) Type 2: microbes and microbial lms of bacteria, spores, and yeast species. (iii) Type 3: milk, whey protein concentrate (WPC), cooked SCM, starch, boiled wort, and egg albumin. Some of the research that has considered the inuence of cleaning parameters in owing systems on the removal behavior of deposits is listed in Table 2 to 4. Table 2 details type 1 deposit removal studies, Table 3 details type 2 deposit removal studies, and Table 4 details type 3 deposit removal studies. The cleaned geometry, effect of CIP parameters, and the method of determining cleaning effectiveness are listed in each Table. The effect of ow has been studied both in terms of the Reynolds number (Re) and the surface shear stress. Both may provide further insight into the effect of removal behavior on ow velocity. Milk processing is a large industry and fouling is a signicant problem, as both protein aggregates and minerals are deposited; Burton (1967) classied the proteinaceous deposit seen in pasteurizers as type A and the mineral deposit seen at UHT temperatures as type B. Reviews of dairy fouling research are presented by Changani and others (1997) and Bansal and Chen (2006). Proteins have been identied as a major source of fouling deposits. Fickak and others (2011) found that increasing the protein concentration of whey protein increased the amount of fouling on a pilot-scale heat exchanger. Holding of milk before heating sections has been shown to aggregate -lactoglobulin in the holding sections rather than the heating sections (de Jong and van der Linden 1992). Christian and others (2002) found that increasing the mineral content of whey protein decreased the extent of fouling on a PHE. WPC is often used in research studies to represent a milk fouling deposit, because it is easier to handle and store than milk, and the fouling composition is thus easier to control and replicate. Robbins and others (1999) compared the cleaning of milk and WPC from a PHE. They found that in the pasteurization and UHT sections of the PHE, both materials fouled heavily. However, in the intermediate section, WPC also fouled excessively, whereas milk did
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 127
Gasket
T Type A fouling on the he gasket v vessel and th Wall fouling Contact t agar d been (s samples had s scraped from the vessel) v
Fermente er cone
Figure 5(A) 80 L stainless steel tank (0.8 m 0.4 mm) with residual yeast fouling attached to the wall and the cone. The wall was also sampled by contact agar (adapted from Salo and others 2008). (B) Type A deposit seen at the top of a fermenter around the man way door and the gasket (Goode 2012). Table 2Some CIP studies of type 1 deposit. Deposit Toothpaste Geometry 1 m long, 2 OD, 316 L ss pipe (horizontal) Effect of ow or w Increase ow velocity (1 to 3 m/s, decrease cleaning time. Effect of temperature Increase temperature (from 20 C), decrease cleaning time (to a point approximately 40 C). (31 to 51 C), removal of shampoo layers faster at higher temperatures as cleaning proceeds. Effect Re Increase Re (4000 to 250000) decreases cleaning time. Cleaning determinant Turbidity reaches 4 ppm. Reference Cole and others (2010)
Shampoo
Mustard
(0.14 to 0.47 m/s) higher ow velocity, more efcient removal at the start of cleaning. Increase ow velocity (1 to 1.88 m/s) increase removal rate. Increase w , decrease number of cells. (i) linearly for plastics (ii) as a curve for glass (0.7, 1.5, 2.3 L/min) increase ow rate, the effect of temperature on cleaning time decreases
Above a certain Reynolds number, the recirculation zone length becomes constant.
Visual
Visual
Tomato paste
(30, 50, and 70 C) increase temperature, decrease the time to remove deposit
Christian (2004)
OD = outer diameter, ID = inner diameter, ED = equivalent diameter, ss = stainless steel, MSS = mechatronic surface sensor, MHFS = microfoil heat ux sensor.
not. Compositional analysis revealed protein fouling from both materials in the pasteurizer section. Increasing to UHT temperatures revealed milk fouling to become more mineral-based, whereas the WPC fouling remained predominantly protein-based, suggesting comparison of milk fouling and WPC fouling is not wise at UHT temperatures. Yeast can exhibit type 1 (if in contact with glass) and type 2 (if in contact with stainless steel) cleaning behavior. Guillemot and
others (2006) found that yeast cells could be wholly removed from glass using water but that yeast cells had strong adhesion to stainless steel. The wall shear stress required to remove 50% of the attached cells from stainless steel, denoted as w 50% , was 30 Pa, while for plastics w 50% ranged from 1 to 2 Pa. The effect of CIP parameters on the removal of different deposit types is discussed in the following sections. Even though there is clear evidence that different deposit types are removed
C
128 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
B. cereus spores
316 L ss pipe (20 cm long, 2.37 cm ID) and 2 way valve (entry to exist 18 cm long, 3.5 cm ID) 316 L ss, (210 90 mm long) in horizontal ow cell 304 L ss pipes (15 102 m long, 2.3 102 m ID) (horizontal)
Progressive-cavity pump (with axial or tangential exit pipe). Tangential was best. In the axial setup, the number of CFU was greater than 10 CFU/cm in the pump body and gaskets.
Increase w , decrease number of cells barely for stainless steel (10%). Increase w (17.45 to 68.95 Pa, that is, 1.61 to 3.29 m/s) decrease number of spores (after 5 min). Contact time was more important in reducing spores.
NaOH 0.5% (w/w) at 60 C, 2200 L/h, up to 30 min. The % residual spores decreased as cleaning time increased.
Visual
Prerinse 0.5 m/s (6 min); 0.2% NaOH at 1.5 m/s, 60 C (10 min); intermediate rinse 0.5 m/s (6 min); 0.2% HNO3 at 1.5 m/s, 60 C (10 min); nal rinse 0.5 m/s (6 min).
OD = outer diameter, ID = inner diameter, ED = equivalent diameter, ss = stainless steel, MSS = mechatronic surface sensor, MHFS = microfoil heat ux sensor.
from surfaces differently, the approach to cleaning is typically the Product recovery same: At the end of a process, there can be a signicant amount of material left in pipes and tanks. This product may be saleable, in (i) Prerinse (or product recovery stage); to remove loosely which case it should be recovered, or it may be considered waste. In bound soil and product. both cases, the bulk of this material should be removed (generally (ii) Detergent phase (alkali or acid); to remove the fouling layers. in the rst rinse phase) prior to the cleaning phase. Type 1 (iii) Intermediate rinse; to remove chemical. deposits will generally be saleable for products like toothpaste or (iv) Sanitization/disinfection step (chemical and/or thermal); to shampoo, and recovery should be maximized. Type 2 and type 3 kill viable microbes and restore the hygienic condition of deposits will generally be formed as thin layers at the wall of a the system. different composition to product. The layers need to be removed, (v) Final water rinse; so product can be reintroduced to the to return the plant to a clean state, and tend not be recovered at system. the end of a process. Palabiyik and others (2012) investigated the effect of the prodAlthough disinfection is done after the deposit has been removed uct recovery (using water) on overall cleaning time of toothpaste from a process surface, this stage is often included as part of the from a 1 m pipe. They determined that the amount of toothCIP operation in industry. The purpose of this stage is to make the paste removed during product recovery was not a function of pipe surface free of product spoilage microbes rather than to remove the Reynolds number. A similar mass fraction was removed over the foulant. Product may be recovered before cleaning, depending on Re range 5000 to 25000. They did, however, nd that prodthe type of product, its value, and the geometry used in processing. uct recovery conditions had a profound effect on overall cleaning This is discussed in the following section. time. The cleaning phase was conducted at the same condition:
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 129
Continuous and Local cleaning N/A, constant Re > 25000 N/A. Constant abrupt time has a investigated. 0.5%. expansions minimum (ID 26 mm, where the wss expanding shows a from 26 to 38 maximum and mm) vice versa An increase in An increase from Cooked SCM 316 L ss coupons Increase ow An increase in Re (6500 to 0.5 to 1.5% (sweet (square: 30 velocity from temperature 27500) NaOH did not condensed milk) 30 mm long) 0.25 0.5 (40, 60, and revealed a signicantly m/s, decrease 80 C) revealed a decrease in affect cleaning time linear cleaning time cleaning time at all decrease in according to at higher ow temperatures cleaning time Power law. velocities. Increase Re No cleaning at Egg albumin 316 L ss coupons Increase ow 30 C did not (1090 to clean. 0.1 wt% (circular: 26 less 4840) Increase NaOH. mm D) signicant at decreases temperature, Concentration higher cleaning time decrease in 0.25% to 3% chemical con (at 50 C, 0.1 cleaning time. (at 50 C, 2.3 centrations. to 1% NaOH). L/min) However, 50 C removed At 70 C decreases more deposit 0.1%, cleaning time. at 1% NaOH increase Re, Most than 70 C. increase signicant at cleaning time. low ow (0.7 L/min) Increase WPC 316 L ss coupons Limited benet Increasing Re Limited benet temperature (circular: 26 to increase (1090 to to increase (30 to 70 C) mm D) ow velocity 4840) only concentration decrease horizontal at 70 C and benecial at above 0.5%. 1% NaOH. cleaning time ow cell 0.1% NaOH. Benet if at all ow increase ow rates and at low concenchemical contration. centrations (0.7, 1.5, 2.3 L/min, 0.1%, 0.5%, 1% NaOH) WPC 10 cm sections Increasing ow Wall Re 500 to 6500 N/A. Constant of sstubes (6 rate does not temperature investigated. 0.5%. mm ID 0.15 necessarily did not affect As Re mm thickness) decrease the plateau. increases fouled in cleaning time. Increasing the cleaning time counter It is important bulk decreases current heat to decay temperature generally. exchanger phase time. decreases cleaning time.
Aziz (2008)
Christian (2004)
Thermal Gillham and resistance using others (1999) MHFS and mass
OD = outer diameter, ID = inner diameter, ED = equivalent diameter, ss = stainless steel, MSS = mechatronic surface sensor, MHFS = Microfoil heat ux sensor, wss = wall shear stress.
0.55 m/s at 50 C. High ow velocity and low temperature in the product recovery stage revealed the fastest cleaning times. The results suggested that the structure of the toothpaste lm after the product recovery stage is important in determining the overall cleaning time. Product recovery can be done by pigging, in which uid is expelled from a system by the pig which could be solid, liquid, or gas. Solid pigs tend to be used in long sections of straight pipe work where complex geometries do not need to be navigated; for example, in crude oil pipelines to remove parafn wax (Guo and others 2005). The use of crushed ice (with a freezing point depressant) in pigging systems has been developed and researched at the Univ. of Bristol to remove starchwater mixes (Quarini 2002). The void fraction of the ice is controlled so that the pig can navigate bends and T-pieces as well as straight pipe work. Application of this technology in the food and beverage industries is currently limited. The ice is expensive to make and store. A company called Aeolus promotes a Whirlwind technology that uses compressed air to remove soft deposits like fruit juice from pipe work with
bends (see www.aeolustech.co.uk). Application of this technology in the food and beverage industry is also limited as the cost of compressed air is considerable. However, use of air in cleaning is likely to increase in the future as water becomes more precious.
The effect of CIP parameters on type 1 removal Schl usser (1976) compared cleaning behavior of 3 type 1 soils; beer, wine, and milk, illustrated in Figure 6. The products themselves were not heated. The cleaning proles of each product were different. Type 1 products can have a complex rheology, but are often shear thinning, that is, they have an effective viscosity that is a function of shear rate. The shear-thinning rheology of yogurt was determined by Henningsson and others (2007) who also studied the use of water to displace the yoghurt. For ow velocities of 0.05 to 0.25 m/s, yogurt was observed and predicted to ow as a plug. If the process was set up so that yogurt ows as a plug, at changeover, the mixing zone between the 2 yogurts would be smaller and yield reduced losses. Prediction of the mixing zone of a HershelBulkley material with and without wall slip at 0.19 m/s
130 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
Figure 6Cleaning characteristics of 3 type 1 products, beer, red wine, and milk with water (Schl usser 1976).
was also done by Henningsson and others (2007). With wall slip, it was predicted that the material would have a larger plug ow region. However, predicting the ow of a high-viscosity plug or wall layer is very difcult in practice. Flow and wall shear stress. Flow rate has an effect on the removal rate of type 1 materials. The rheology of tomato paste has been represented by the Carreau model (Bayod and others 2008), and the cleaning behavior of tomato paste in a ow cell has been investigated by Christian (2004). At 30 C, it was found that by increasing the cleaning water ow rate from 0.7 to 1.5 to 2.3 L/min (Re 750 and 4840), the cleaning time decreased. The relationship appears linear. This was also true at 50 and 70 C. Shampoo (SUNSILK R color radiant, viscosity quoted as 7000 cP at 24 C) was rinsed by water from a stainless steel plate in a vertical ow cell by Pereira and others (2009), and they found that the faster the initial ow rate (in the range of 0.14 to 0.47 m/s), the more shampoo was removed from a duct. The same effect was found for removing toothpaste (a HershelBulkley uid with a yield stress) from a pipe (Cole and others 2010). The effect of wall shear stress ( w ) in the range of 0.5 to 10 Pa on toothpaste removal was studied. Shear stress is affected by both uid density and Re that are both affected by temperature. Toothpaste cleaning time is governed by 2 removal phases by Cole and others (2010): (i) Core removalwhere most of the product is removed as a slug of product that can be recovered. (ii) Thin-lm removalwhere the remaining annular wall lm of toothpaste is removed. More recently, 3 phases were dened by a further investigation of the effect of product recovery on cleaning of toothpaste using water (Palabiyik and others 2012): (i) Core removalthe rst few seconds (a time comparable to the residence time of the uid in the system) where approximately half the product mass is removed and the remaining toothpaste coats the pipe wall. Also called the product recovery stage. (ii) Film removalfurther product is removed up to about 1000 s according to a process that is 1st order in deposit weight/thickness, leaving a thinner but continuous lm of toothpaste remaining on the pipe wall. (iii) Patch removalgreater than 1000 s, the continuous lm is broken up and only patches of toothpaste are left on the
2013 Institute of Food Technologists
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 131
Figure 7Downstand geometry used for investigating the inuence of different ow rates during CIP (ow was from left to right) (from Jensen and others 2007).
that: (i) Increasing the ow velocity increased removal rate. However, the authors suggested that this was more likely due to greater acceleration of the water at 1.88 m/s into the T-piece. At the lower ow velocities, ow had not fully developed before entering the T-piece. (ii) Some areas of the T-piece were harder to clean than others. The position in the downstand most difcult to clean was always located in the same position (see Figure 8, shown as a downstand). As expected, the top of the downstand was difcult to clean. However, an additional area located on the downstand pipe was always the last part to be cleaned in all the experiments, regardless of velocity. Jensen and others (2007) used computational uid dynamics (CFD) simulations to predict the wall shear stress in the 4-cm downstand. Their CFD ndings are illustrated in Figure 8(A) to 8(C) where blue is low wall shear stress (0 Pa) and red is high wall shear stress (5 Pa). As the ow velocity was increased, the blue area decreased in size. Within these simulations, the area most difcult to clean, the center of the downstand, is identied. Increasing the ow rate does not improve cleaning of this area. The wall shear stress achieved at this position is low at all 3 ow velocities. The other areas hardest to clean are circled. Jensen and others (2007) examined the effect of pulsed ow in the downstand. They found that pulsing ow only affected the cleaning time of the 4-cm-depth T piece, not the 6-cm-depth T piece. They compared cleaning at 1 m/s (v1 ) and 2 m/s (v2 ) and pulsing at 15 s (p1 ) and 30 s (p2 ). The cleaning time of the 4-cm downstand was longer when rinsed at 1 m/s than when the ow was pulsed. However, rinsing the downstand at 2 m/s gave the quickest cleaning time. The authors concluded that at turbulent Re, the area of the recirculation zone in the T-piece did not change. A recirculation zone is typically located after a pipe expansion and depends on Reynolds number and the expansion ratio. At lower Re (less than 10000 in this case), the length of the recirculation zone may change; hence, cleaning times are shorter for pulsed ow at 1 m/s than using constant ow at 1 m/s. Jensen and Friis (2005) used CFD simulations to predict the cleanability of a mix proof valve fouled with B. stearothermophilus spores in accordance with the EHEDG standard cleanability test (EHEDG 1992; Timperley and others 2000). In the EHEDG
132 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
Figure 8CFD simulations of the ow eld in 4 cm downstand T piece at (A) 0.5 , (B) 1, and (C) 2 m/s. Blue is low wall shear stress (0 Pa) and red is high wall shear stress (5 Pa). White represents wall shear stress in excess of 5 Pa. Water enters the section from the right and exits the T section on the left represented by the arrow in (a) (adapted from Jensen and others 2007).
Figure 9Commercially available (A) spray ball static device (GEA, Warrington, U.K.), (B) rotary spray head dynamic device (Alfa Laval, Minworth, U.K.), and (C) rotary jet head dynamic device (Alfa Laval, U.K.).
Morrison and Thorpe (2002) dened the wetting rate at the mass ow rate (kg/s) required to completely wet a surface of width W (in meter). Wetting rates achieved by single jets from a spray ball were 0.1 to 0.3 kg/ms. The act of removing deposit from a vessel involves initial wetting and subsequent softening (or dissolution) of the deposit, followed by complete removal by further impingement. Morrison and Thorpe (2002) measured the dimensions of the wet area by the impaction of single water jets onto a sheet of painted acrylic for a range of pressures and distances from spray balls of different sized holes. They found that if the jet directly impacted the area to be cleaned, then this area was cleaned within 60 s. The point of impact was smaller than the total area being wetted; however, certain areas were not cleaned by the spray ball. The width of the falling lm from the point of impact remained the same size throughout rinsing. Jet breakup was observed at 45 C, which increased the distribution of the jet and cleaned a larger area. An interesting model for the ow behavior of jets is given by Wilson and others (2012).
The effect of CIP parameters on type 2 and type 3 deposit removal Type 2 deposits can be viscoelastic, temperature-dependent, and/or time-dependent (Rao 1999). Type 3 deposits tend to be thermally induced and precipitate from the process stream onto the heat exchanger surface over time. For example, wort is a complex uid with several components that change structure and solubility upon heating, including carbohydrates, proteins, vitamins, minerC
als, and lipids. The deposits formed during wort boiling are solid and dissimilar to the process stream (Tse and others 2003). Membrane cleaning. There are many types of ltration processes in food and beverage manufacturing operations. The fouling of membranes alters permeability and selectivity and can be characterized by increased pressure differential and decreased membrane ux. Membranes used in the food and bioprocess industries include reverse osmosis (RO), nanoltration (NF), ultraltration (UF), and microltration (MF) (Cui and Muralidhara 2010). In the brewing industry, beer is claried using MF in which yeast readily fouls the membranes. Membrane cleaning is complex as it is necessary to both remove the surface layers and open the pores in the structurethis must be done without the cleaning agent damaging the material. G uell and others (1999) found that when yeast cells were present on cellulose acetate membrane (CAM) as a layer (yeast cake), the yeast was believed to have formed a secondary membrane. Increasing the thickness of the yeast cake reduced permeate ux and protein transmission through the membrane. Increasing the yeast concentration in the feed solution resulted in lower uxes and protein transmission through the CAM. Hughes and Field (2006) discussed the fouling of MF and UF membranes with yeast at subcritical uxes where fouling is negligible. For the MF membrane, the rate of fouling increased with increasing feed concentration, increasing membrane pore size, and decreasing shear stress. The UF membrane could not be cleaned effectively. Mores and Davis (2002) examined the effect of pulsing ow through a CAM to clean it. They found that the ux increased
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 133
Figure 10The fractional importance of different factors: time, coverage, physical action (impact), temperature, and chemical action (chemistry) required for effective tank cleaning by (A) spray ball and (B) rotary jet head (adapted from Tamine 2008).
with increasing shear rate, back pulse pressure, and back pulse duration. At higher shear rate and back pulse pressure, multiple short back pulses were more effective in cleaning the membrane. At low shear rate and back pulse pressure, fewer longer back pulses were more effective. Longer, weaker back pulses led to the highest recovered uxes. Shorrock and Bird (1998) fouled a MF membrane (hydrophilic polyethersulfone, 0.1 m pore diameter), with yeast cake. Water rinsing was found to remove most of the deposit and an increase in temperature from 30 to 60 C was found to decrease fouling resistance (at 0.74 m/s cross-ow velocity (CFV)). At 40 C, using NaOH as optimum concentration, there was optimum ux through the membrane, 0.01% to 0.025%. Formulated sodium hydroxide solution was found to restore membrane ux completely. Cleaning of MF membranes with WPC was considered by Bird and Bartlett (2002) using a at plate stainless steel membrane and by Blanpain-Avet and others (2009) using a tubular ceramic membrane. An optimum alkaline detergent concentration of 0.02% NaOH was found to give maximum ux after cleaning of the stain-
134 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
when moving from laminar to turbulent ow. Disruption of the boundary layer is further discussed in Section 4.1. Generally increasing the temperature decreases the cleaning time. Gillham and others (1999) found that removal of whey protein deposits from stainless steel pipes was strongly dependent on temperature (less so the swelling phase). SCM is an intermediate in the manufacture of some confectionary products, made by evaporating water from milk and adding sugar to lower the water activity of the product. SCM has 70% to 74% total solids of which 40% to 45% is sucrose (Fisher and Rice 1924) leaving 29% to 30% milk solids. In the study of Othman and others (2010), SCM was cooked for 4 h at 85 to 90 C on stainless steel coupons and washed by chemical cleaning in a ow cell. It was found that increasing the ow velocity from 0.25 to 0.5 m/s decreased the cleaning time at all temperatures. An increase in temperature from 40 to 80 C decreased the cleaning time linearly. Interestingly, the authors found that increasing the NaOH concentration from 0.5% to 1.5% did not signicantly affect the cleaning time at each temperature. This agrees with ndings for WPC cleaning that quote 0.5% NaOH as the optimum concentration. It was the increase in temperature rather than the increase in chemical concentration that decreased cleaning time. Cleaning time was plotted compared with Re for SCM at 1% NaOH (40, 60, and 80 C) by Othman and others (2010) and at 0.1%, 0.5%, and 1% NaOH (at 30, 50, and 70 C) for WPC by Christian (2004) as shown in Figure 11(A) and 11(B). For WPC, the range of investigated Re was around 800 to 4840. There were separate groups of data at each temperature that could not be plotted on one master curve. This suggests that temperature was the dominant parameter in controlling cleaning time. Christian (2004) concluded that an increase in Re was only benecial to cleaning time at low concentration. Jennings and others (1957) suggested the existence of a threshold Re of 25000 for cleaning a pipe surface of dry milk deposit before an increase in Re resulted in increased cleaning rate. For SCM, the Re range investigated was much higher, from 6500 to 27000. All the data collapsed onto one curve. As the Re increased, the cleaning time decreased, suggesting that Re was the dominant parameter controlling cleaning time. Othman and others (2010) did nd, however, that the effect of Re on cleaning time became less signicant as the temperature was increased. Gillham and others (1999) found that tc was proportional to Re n , where n was in the range of 0.2 to 0.35 for 0.5% NaOH. For SCM at 1% NaOH, tc was again proportional to Re n where n = 1.28 (R2 = 0.92). The cleaning of egg albumin was characterized by Aziz (2008). Generally, an increase in temperature decreased the cleaning time. However, at 1% NaOH, cleaning time was faster at 50 C than at 70 C. Deposit was not removed at 30 C at any ow velocity or NaOH concentration investigated. An increase in NaOH concentration from 0.25% to 3% NaOH decreased the cleaning time (at 50 C, 2.3 L/min); however, increasing the ow rate had a less signicant impact on cleaning time at higher chemical concentration. The author concluded a high temperature, mid to high ow velocity, and mid-range chemical concentration appeared to be the optimum, similarly to WPC cleaning optima. For egg albumin, the range of Re investigated was 1090 to 4840. There were separate groups of data at 50 and 70 C that could not be plotted on one master curve. This suggests that temperature was the dominant parameter in controlling cleaning time similarly to WPC. In Christian (2004), WPC cleaning experiments were conducted using 0.5% NaOH at 30, 50, and 70 C and 0.7, 1.5, and
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 135
Figure 12Stationary ow and oscillating components of ow, wos symbolizes an oscillating uid movement, wos,max is maximum oscillating uid velocity, wstat is mean uid velocity, and t is frequency of oscillation (Augustin and others 2010).
an oscillating uid movement, wos , is superimposed, as illustrated in Figure 12. The intensity of a superimposed pulsation (wos,max + wstat ) can be quantied using waviness, W, the ratio of the maximum oscillating (wos,max ) and the stationary or mean ow velocities, w: W= wos,max w (4)
Figure 11Re compared with visual cleaning time of (A) SCM at 40, 60, and 80 C using 1% NaOH (from Othman and others 2010) and (B) WPC at 30, 50, and 70 C, using 0.1%, 0.5%, and 1% NaOH (from Christian 2004). Shading denotes the ow temperature: Black: 30 C, Gray: 50 C, and Open: 70 C.
w for an oscillation interval is dened by Augustin and others (2010) as: w= 1 tos w (t ) d t with w (t ) = wstat + wos tos 0 = wstat + wos,max . sin(w t )
(5)
2.3 L/min. Rd (the fouling resistance) is a measure of resistance to the ow of heat to the sensor. Rd measured at the same ow rate reduced Rd more rapidly as the temperature was increased from 30 to 70 C. An increase in ow rate from 1.5 to 2.3 L/min revealed similar Rd proles, suggesting that temperature dictated the cleaning time in this case. For all type 3 deposits detailed here, temperature seems to be the dominant contributor to cleaning time at both low and high ow velocity and low and high concentration. Reaction rate, solubility, and possible phase transitions (such as in fats) will all be affected by temperature.
A higher value of waviness is believed to result in a separation of the viscous sublayer and the formation of eddy currents; ow reversal is critical. This effect can decrease the thickness of the laminar sublayer at the pipe surface when applying a turbulent ow. The temporary maximum velocity, wos,max , can occur near the pipe wall resulting in large shear rates and high wall shear stresses. Bode and others (2007) characterized a linear decrease of cleaning time with increasing waviness from 1 to 5. Below a waviness of 1, there is limited ow reversal in the pipe and the cleaning time increases. Augustin and others (2010) compared the cleaning rate of deposit at a ow velocity of 1 m/s (Re 25000), where waviness was 0, and pulsing the ow, where waviness was 1. They found that the amount of deposit in the pipe became asympNovel Cleaning Approaches There have been various methods reported in the literature to totic at 4 min using the pulsed-ow regime and 6 min using the stationary ow regime. Augustin and others (2010) have validated improve cleaning. CFD models accompanying their cleaning data. Phosphorescent Increasing boundary layer disruption zinc sulde crystals were used as an optical tracer enabling accurate Various authors have considered pulsing ow in pipes to en- characterization of ow patterns. hance wall shear stress at lower average ow velocities to enhance cleaning. Gillham and others (2000) showed that pulsing ow at Alternative cleaners to reduce environmental impact a relatively low frequency (2 Hz) enhanced cleaning of a tubuThe efcacy of enzymes in cleaning has been investigated over lar heat exchanger compared to the same steady-ow velocity. A the last decade. Grasshoff (2002) investigated the efcacy of Savpulsed ow creates high periodic accelerations of the liquid ow. inase, a protease (optimum temperature 50 C, pH 9.5), in cleaning The directional change in the ow can increase mass transfer of a milk pasteurizer following a 15-min acid wash. Increasing the the cleaning uid to the surface, thus decreasing cleaning times. concentration of the enzyme from 0.0025% to 0.05% at 60 C A pulsed ow is characterized by a stationary base ow on which showed that residual soil was removed faster. In the plant, the
136 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
2013 Institute of Food Technologists
Other studies related to cleaning behavior Various authors have quantied parameters related to cleaning behavior by other methods rather than in a ow cell or pilot plant. Some techniques used to infer cleaning data are discussed here. Deposit shear. Sim oes and others (2005) fouled stainless steel cylinders with P. uorescens biolm in a bioreactor. Three cylinders were rotated at 500, 1000, 1500, and 2000 min1 sequentially (ReA 2400 to 16100) for 30 s each in phosphate buffer to assess the effect of rotation speed on biolm removal. There was an optimum ReA of 8100 where 45% of biolm was removed. Either side of this ReA removal was less and similar at around 15%. Cylinders were also submerged in different chemical solutions and rotated at 300 min1 for 30 min and then were rotated at 500, 1000, 1500, and 2000 min1 sequentially in phosphate buffer to test cleaning effectiveness. NaOH and sodium hypochlorite (NaClO) representing a CIP detergent and CIP sterilant were investigated. Irrespective of ReA , NaOH and NaClO removed a similar percentage of biolm at the same concentrations (50, 200, and 300 mg/L); however, at the highest concentration, 500 mg/L, NaOH removed approximately 5% more biolm. Demilly and others (2006) characterized the removal of yeast cells as a function of wall shear stress from different stainless steel surfaces using a radial ow chamber, in which ow trajectory was from the center of the plate outwards toward the edges of the plate. This is similar to the impingement of a water jet on a surface. Radial ow investigations may relate to the action of a cleaning head in a tank, to provide a microscale method of investigating water jets, and deposit removal. The steel surfaces were micropolished and electrochemically etched with different sizes of grain and depths. The number of cells remaining from the original number of cells was dened. Interestingly, the authors found a threshold shear stress above which cells were removed regardless of topography; that detachment of yeast cells was faster from etched steel than mirror-polished steel. Deposit deformation and strength. Two methods have been independently developed at Birmingham and Cambridge to determine the strength and deformation behavior of soft-solid fouling layers on hard surfaces immersed in liquid: the micromanipulation technique and the uid dynamic gauging (FDG) technique, reC
Measuring Cleaning
Ensuring good performance of the cleaning process is vital in maintaining reliable product shelf life and quality. Figure 1 demonstrates implementation of a CIP standard in industry, revealing that CIP measurement and control are fundamental to ensuring hygiene on a daily basis. Processes used to establish and run CIP operations include: (i) Validationdetermining the right method of cleaning and setting it as a standard; it is done before implementation of
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 137
Figure 13Potential locations of online CIP measurement techniques that could be used in a process line.
a new method and after alterations in an existing operation. It should always be up-to-date. (ii) Vericationare the results correct and accepted? Checks that the system behaves in the predetermined and expected way, after validation. (iii) Monitoringcontinuous monitoring of specic points of a process determining that the process is under control, after verication. When defects in the system have been identied, appropriate action should always be taken as detailed in Figure 1. Typical CIP monitoring tools include visual detection methods, microbial enumeration of CIP rinse water, and in-line probes that measure temperature and the proportion of hydrogen ions (pH) or electrolytes (conductivity) in the cleaning uid. The conductivity of acid and alkali is different to that of water so that the chemical phases of cleaning can be clearly identied and the chemical recovered. The rst step in optimizing CIP is ensuring that the used monitoring techniques are giving reliable data. Yang and others (2008) described the importance of forecasting models in the approach to optimize cleaning. Data obtained at the beginning of CIP (where the condence limit of variation is high) are used to predict the end point of cleaning (where the condence limit of variation is low). Online application of such a tool would be valuable in optimizing CIP. A range of measurement devices have been considered in the literature and have been used to assess either the cleaning uid (bulk measurements) or the surface. Figure 13 illustrates where in a process of line bulk and surface measurements, CIP measurements may be taken. Flow, temperature, conductivity, and pH are bulk measurements.
Online bulk measurements Online bulk measures monitor the uid during cleaning. Typically temperature, ow, pressure, conductivity, and pH are monitored during CIP to ensure that the process is under control. Pressure drop across a system ( P) is dened as: P = PI PO , where PI and PO are the inlet and outlet pressures. PO increases with time during fouling and decreases during cleaning generally. The rate of change of pressure can be used as a measure of cleaning. Fryer and others (2006) illustrated that pressure
138 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
A 1.5
Exp. 4.14.2 - 50C (Re 2340) Exp. 4.23.1 - 70C (Re 3160)
R d (m K/kW)
1.0
0.5
0.0 -100
400
900
1400
1900
2400
2900
3400
3900
4400
Time (s)
B
2.0
Exp. 5.9- 30C (Re 1600) Exp. 5.33- 50C (Re 2340) Exp. 5.57- 70C (Re 3160)
Rd (m K/kW)
1.5
Online surface measurements Various authors have reported heat transfer measurements during cleaning to assess the effect of cleaning parameters on the heat transfer coefcient and fouling resistance, Rd , including Gillham and others (1999, 2000), Christian (2004), and Aziz (2008). An example of Rd measured during the cleaning of (a) egg albumin and (b) whey protein using 0.5% NaOH at 1.5 L/min is illustrated in Figure 14. Rinsing egg albumin at 30 C did not clean the surface. Deposit remained on the surface even after 3 h of rinsing. Klahre and others (2000) used differential turbidity to monitor biofouling on the pipe walls of water systems. The technique could be considered to study biofouling removal during cleaning. The use of a Mechatronic Surface Sensor (MSS) was being tested to monitor milk components such as calcium phosphate and whey proteins (Pereira and others 2006) and shampoo (Pereira and others 2009). The sensor measures changes in the vibration properties of surfaces due to the buildup or removal of fouling layers. Measuring microbial cleanliness Microbiological enumeration techniques tend to be ofine retrospective techniques. Rinse water samples and surface swabs are plated on selective agar and viable microbes will present themselves within 3 to 7 d, depending on the organism. Most informative swab and plate methods include contact agar method where the agar plate is pressed directly onto the surface and microbes enumerated directly (Salo and others 2008). The cleanliness of a surface can be veried more quickly using ATP bioluminescence that indicates the presence (or absence) of microbes that are alive. However, the measure of ATP does not indicate microbe specicity. Microbes have been identied on surfaces using infrared spectroscopy (Fornalik 2008) and Raman spectroscopy (R osch and others 2003). Visual assessment, image analysis, and mass can all be used to determine the amount of deposit removed (or remaining) on a surface after cleaning; however, these methods are ofine. Janknecht and Melo (2003) published a review discussing online biolm monitoring measurement techniques, known detection limits, and applicability to practical situations. The techC
1.0
0.5
Time (s)
Figure 14Rd proles for (A) egg albumin gel (from Aziz 2008) and (B) whey protein (from Christian 2004) with different ow temperatures: 30, 50, and 70 C using 0.5 wt% NaOH and a ow rate of 1.5 L/min.
2000),
Goode and others (2010) were able to monitor yeast cleaning from stainless steel coupons using a heat ux sensor and the image analysis technique. Less commercially applied techniques discussed by Janknecht and Melo (2003) include measuring radiation signals (spectroscopy, uorometry, and photoacoustic spectroscopy) and electric and mechanical (vibration) signals. The critical part of optimizing a cleaning process would be the incorporation of cleaning measurements into the process schedule. At present, it is usual that cleaning times are set automatically and are not changed in operation. If measurements made at the start of cleaning could be used to determine the end point of cleaning with high condence, it would be possible to develop some form of a process control strategy that could minimize the cleaning cost (Yang and others 2008). This approach can be successful if measurements are robust, inexpensive, and taken at the dirtiest point within a system.
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 139
Any probe use to monitor cleaning needs to be robust and of low cost because cleaning cost is believed to be a relatively low cost of the total cost of production. The robustness and applicability of such novel measurement systems again need to be determined in an industrial setting.
References Akhtar N, Bowen J, Asteriadou K, Zhang Z, Fryer PJ. 2010. Matching the nano- to the meso-scale: measuring depositsurface interactions with atomic force microscopy and micromanipulation. Food Bioprod Process 88:3418. Akhtar N, Bowen J, Robbins PT, Fryer PJ, Asteriadou K, Goode KR. 2012. The effect of temperature on adhesion forces between surfaces and model foods containing whey protein and sugar. J Food Eng. submitted. Allie Z, Jacobs EP, Maartens A, Swart P. 2003. Enzymatic cleaning of ultraltration membranes fouled by abattoir efuent. J Membr Sci 218:10716. Asteriadou K, Hasting APM, Bird MR, Melrose J. 2006. Computational uid dynamics for the prediction of temperature proles and hygienic design in the food industry. Food Bioprod Process 84:15763. Augustin W, Fuchs T, F oste H, Sch oler M, Majschak J, Scholl S. 2010. Pulsed ow for enhanced cleaning in food processing. Food Bioprod Process 88:38491. Aziz NS. 2008. Factors that affect cleaning process efciency. [PhD thesis]. School of Chemical Engineering, University of Birmingham. Baier RE. 1980. Substrate inuences on adhesion of microorganisms and their resultant new surface properties. In: Bitton G, Marshall KC, editors. Adsorption of microorganisms to surfaces. New York: John Wiley. p. 59104. Bansal B, Chen XD. 2006. A critical review of milk fouling in heat exchangers. Compr Rev Food Sci Food Safety 5:2733. Bayod E, Willers EP, Tornberg E. 2008. Rheological and structural characterization of tomato paste and its inuence on the quality of ketchup. LWT Food Sci Technol 41:1289300. Beech IB, Smith JR, Steele AA, Penegar I, Campbell SA. 2002. The use of atomic force microscopy for studying interactions of bacterial biolms with surfaces. Colloids Surf B Biointerfaces 23:23147. B en ezech T. 2001. A method for assessing the bacterial retention ability of hydrophobic membrane lters. Trends Food Sci Technol 12:368. B en ezech T, Leli` evre C, Membr e JM, Viet AF, Faille C. 2002. A new test method for in-place cleanability of food processing equipment. J Food Eng 54:715. Bird MR. 1992. Cleaning of food process plant. [PhD thesis]. UK: University of Cambridge. Bird MR, Bartlett M. 2002. Measuring and modelling ux recovery during the chemical cleaning of MF membranes for the processing of whey protein concentrate. J Food Eng 53:14352. Bird MR, Fryer PJ. 1991. An experimental study of the cleaning of surfaces fouled by whey proteins. Food Bioprod Process 69:1321. Blanpain-Avet P, Migdal JF, B en ezech T. 2009. Chemical cleaning of a tubular ceramic microltration membrane fouled with a whey protein concentrate suspensioncharacterization of hydraulic and chemical cleanliness. J Membr Sci 337:15374. Bode K, Hooper RJ, Augustin W, Paterson WR, Wilson DI, Scholl S. 2007. Pulsed ow cleaning of whey protein fouling layers. Heat Transf Eng 28:2029. Bott TR. 1990. Fouling notebook. Rugby, UK: Institution of Chemical Engineers. Bott TR. 1995. Fouling of heat exchangers. New York: Elsevier. Bowen WR, Lovitt R W, Wright CJ. 2001. Atomic force microscopy study of the adhesion of Saccharomyces cerevisiae. J Colloid Interface Sci 237: 5461. Bremer PJ, Fillery S, McQuillan AJ. 2006. Laboratory scale clean-in-place (CIP) studies on the effectiveness of different caustic and acid wash steps on the removal of dairy biolms. Int J Food Microbiol 106:25462. Burton H. 1967. Deposits from whole milk in heat treatment plant: a review and discussion. J Dairy Res 34:13743. Busscher HJ, Norde W, Sharma PK, van der Mei HC. 2010. Interfacial re-arrangement in initial microbial adhesion to surfaces. Curr Opin Colloid Interface Sci 15:5107.
140 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
Fisher RC, Rice FE. 1924. Sweetened condensed milk II. A comparative study of methods for determining total solids. J Dairy Sci 7:497502. Fornalik M. 2008. Detecting biofouling in food processing systems. Photonics Food 42:5861. Fryer PJ, Asteriadou K. 2009. A prototype cleaning map: a classication of industrial cleaning processes. Trends Food Sci Technol 20:22562. Fryer PJ, Christian GK, Liu W. 2006. How hygiene happens: physics and chemistry of cleaning. Int J Dairy Technol 59:7684. Gallardo-Moreno AM, Gonz alez-Martna ML, Bruque JM, P erez-Giraldo C. 2004. The adhesion strength of Candida parapsilosis to glass and silicone as a function of hydrophobicity, roughness and cell morphology. Colloids Surf A 249:99103. Garrett TR, Bhakoo M, Zhang Z. 2008. Bacterial adhesion and biolms on surfaces. Prog Nat Sci 18:104956. Geesey GG, Gillis RJ, Avci R, Daly D, Hamilton M, Shope P, Harkin G. 1996. The inuence of surface features on bacterial colonisation and subsequent substratum chemical changes of 316L stainless steel. Corros Sci 38:7395. Gillham CR. 1997. Enhanced cleaning of surfaces fouled by whey protein. [PhD thesis]. U.K.: University of Cambridge. Gillham CR, Fryer PJ, Hasting APM, Wilson DI. 1999. Cleaning-in-place of whey protein fouling deposits: mechanisms controlling cleaning. Food Bioprod Process 77:12738. Gillham CR, Fryer PJ, Hasting APM, Wilson DI. 2000. Enhanced cleaning of whey protein soils using pulsed ows. J Food Eng 46:199209. Goode KR. 2012. Characterising the cleaning behaviour of brewery foulants to minimise the cost of cleaning in place operations. [EngD thesis]. University of Birmingham. Goode KR, Asteriadou K, Fryer PJ, Picksley M, Robbins PT. 2010. Characterising the cleaning mechanisms of yeast and the implications for cleaning in place (CIP). Food Bioprod Process 88:36574. Gordon PW, Brooker ADM, Chew YMJ, Wilson DI, York DW. 2010. Studies into the swelling of gelatine lms using a scanning uid dynamic gauge. Food Bioprod Process 88:35764. Grasshoff A. 1997, Cleaning of heat treatment equipment. IDF Monograph, Fouling and Cleaning in Heat Exchangers, Brussels, Belgium: International Dairy Federation. Grasshoff A. 2002. Enzymatic cleaning of milk pasteurisers. Food Bioprod Process 80:24752. G uell C, Czekaj P, Davis RH. 1999. Microltration of protein mixtures and the effects of yeast on membrane fouling. J Membr Sci 155:11322. Guillemot G, Vaca-Medina G, Martin-Yken H, Vernhet A, Schmitz P, Mercier-Bonin H. 2006. Shear-ow induced detachment of Saccharomyces cerevisiae from stainless steel: inuence of yeast and solid surface properties. Colloids Surf B 49:12635. Guo B, Song S, Chacko J, Ghalambor A. 2005. Offshore pipelines. Chapter 16: pigging operations. Oxford, UK: Elsevier. Henningsson M, Regner M, Ostergren K, Tr ag ardh T, Dejmek P. 2007. CFD simulation and ERT visualization of the displacement of yoghurt by water, on industrial scale. J Food Eng 80:16675. Hilbert LR, Bagge-Ravn D, Kold J, Gram L. 2003. Inuence of surface roughness of stainless steel on microbial adhesion. Int Biodeterior Biodegrad 52:17585. Hooper RJ, Liu W, Fryer PJ, Paterson WR, Wilson DI, Zhang Z. 2006. Comparative studies of uid dynamic gauging and a micromanipulation probe for strength measurements. Food Bioprod Process 84: 3538. Hughes D, Field R W. 2006. Crossow ltration of washed and unwashed yeast suspensions at constant shear under nominally sub-critical conditions. J Membr Sci 280:8998. Jackson AT, Low WM. 1982. Circulation cleaning of a plate heat exchanger fouled by tomato juice: III. The effect of uid ow rate on cleaning efciency. J Food Technol 17:74552. Janknecht P, Melo LF. 2003. Online biolm monitoring. Rev Environ Sci Bio/Technol 2:26983. Jefferson K. 2004. What drives bacteria to form a biolm? FEMS Microbiol Lett 236:16373. Jennings WG, Mckillop AA, Luick JR. 1957. Circulation cleaning. J Dairy Sci 40:14719. Jensen BBB, Friis A. 2005. Predicting the cleanability of mix-proof valves by use of wall shear stress. J Food Process Eng 28:89106. Jensen BBB, Friis A. 2004. Critical wall shear stress for the EHEDG test method. Chem Eng Process 43:83140.
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 141
142 Comprehensive Reviews in Food Science and Food Safety r Vol. 12, 2013
Vol. 12, 2013 r Comprehensive Reviews in Food Science and Food Safety 143