Available online at www.sciencedirect.

com

Applied Mathematical Modelling 33 (2009) 648662 www.elsevier.com/locate/apm

A dynamic model for milk fouling in a plate heat exchanger

Youcef Mahdi a,*, Abdelkader Mouheb b,1, Lounes Oufer b,1
b

Centre Universitaire de Medea, Institut des Sciences et de la Technologie, Quartier Ain Dheb Medea 26001, Algeria Universite des Sciences et de la Technologie Houari Boumediene, Faculte de Genie Mecanique et de Genie des Procedes, ` Laboratoire des Phenomenes de Transfert, B.P. 32, El-Alia, Bab-Ezzouar, Algiers, Algeria Received 9 July 2007; received in revised form 27 November 2007; accepted 27 November 2007 Available online 22 January 2008

Abstract A two-dimensional dynamic fouling model for milk fouling in a plate heat exchanger (PHE) is proposed. Emphasis is placed on fouling prediction based on the hydrodynamic and thermodynamic performances of the PHE. A 12-channel PHE with counter-current ows is used in quantication of the milk deposition developed inside the channels. The aggregation rate of unfolded protein is found to increase exponentially with increasing wall temperature and is accompanied by a substantial reduction in the heat-transfer coecient. 2007 Elsevier Inc. All rights reserved.
Keywords: Milk; Plate heat exchanger; Modelling; Fouling; Pasteurization

1. Introduction Food products are subject to quite elaborate transformations in order to present qualities that are in conformity with existing food safety regulations. Therefore, food producers and manufacturers always try to dene the best lines of processing in order to obtain optimal sanitary and economic conditions. This is done by taking into account the physical and biochemical modications of the raw materials along process production. For instance, the phenomenon of fouling constitutes one of the major encountered problems in the concerned processes [1]. This is because it generates limitations in the thermal performance of equipment which increases the cost of production but it can also have a considerable eect on the quality of the product as in the case of the dairy industry [2]. Fouling of plate heat exchangers (PHE) during milk processing is a major problem with a negative impact on operating costs and product quality. It also leads to a rise in pressure drop across the exchanger and to possible deterioration in product quality due to failure of the process uid to reach the required temperature. Another serious problem associated with fouling is the cleaning of fouled surfaces by means of costly and

Corresponding author. Tel./fax: +213 25 58 56 45. E-mail addresses: mahdiyoucef@yahoo.fr (Y. Mahdi), mouhebk@yahoo.fr (A. Mouheb), lounesoufer@yahoo.com (L. Oufer). Tel./fax: +213 21 24 71 69.

0307-904X/$ - see front matter 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.apm.2007.11.030

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

649

Nomenclature b Bi C C* Cp D d E e H h I km Ks Kw L LL m Massp Nav Nu P Pr R Re Sc Sh T t U u V w proportionality constant Biot number bulk protein concentration (kg/m3) protein concentration in thermal boundary layer (kg/m3) specic heat (kJ/kg C) diusion coecient (m2/s) particle diameter (m) activation energy (kJ/mol) gap between the plates (m) height of plate (m) convective heat-transfer coecient (W/m2 C) activity product mass transfer coecient (m/s) deposit constant of calcium phosphate (kg/m2 s) wall reaction rate constant (m/s) plate length (m) solubility product channel number mass of deposit (g/m2) Avogadro constant Nusselt number pressure (Pa) Prandtl number gas constant (kJ/mol C) Reynolds number Schmidt number Sherwood number temperature (C) time (s) overall heat-transfer coecient (W/m2 C) velocity component in x-direction (m/s) molecular volume (m3) width (m)

Greek symbols m kinematic viscosity (m2/s) q density (kg/m3) d dynamic boundary layer thickness (m) thermal boundary layer thickness (m) dT l dynamic viscosity of uid (Pa s) k thermal conductivity (W/m C) Subscripts A aggregate protein amb ambient B terminal block c hot d deposit

650

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

D E f j M N P 0 w

unfolded protein equivalent diameter cold channel number deposited protein native protein plate number initial wall

time-consuming techniques where environmentally oensive chemicals are employed. Given the economic impact of fouling in milk heat exchangers, it is not surprising hat there is a considerable amount of literature available on modelling of the fouling process. There seems to be an agreement that thermal denaturation of whey protein betalactoglobulin plays a major role in the fouling process, certainly when the temperature is below 90 C. A secondary fouling process can also concern salts. A number of authors have modelled fouling occurrence in PHE systems based on simplied 1D representation of the process hydrodynamics [35]. For milk, another paper presented a mathematical model for complex PHE arrangements subjected to fouling by adapting a plug ow model with added axial convection and radial dispersion eects [6]. However, the presented models were based on 1D hydrodynamic performance of the PHE, which is a considerable simplication of the eect of plate geometry used in industry. In previous studies such as those by Georgiadis, and Macchiatto (2000), a 2D model accounting for the hydrodynamics of uid ow was used to predict the temperature distribution of ow with higher accuracy than a one-dimensional model. The latter produced larger prediction errors for PHE with a larger plate aspect ratio. In the authors best knowledge, the study of such problems by a specic 2D modelling including mass and energy balances such as described here in is nearly non-existent. The aim of the present study is to apply the chemical reaction model in a 2D dynamic model to predict the milk deposit patterns on the plate surfaces with more accuracy. This approach is expected to pave the way to organize and optimize the operating conditions for reducing the extra costs involved with fouling.

2. Mechanisms and mathematical model In the present work, a formulation of the fouling problem occurring inside a plate heat exchanger allowing milk pasteurization is proposed. A 2D mathematical model of the process is presented based on a dynamic behaviour of the owing uid (milk). The considered model is based on chemical reaction, mass transfer and various other factors taking place during thermal treatment of milk. The model is coupled with the dynamic models of a plate heat exchanger, hence resulting in a nal model that includes a set of partial differential and algebraic equations. The analysis is carried out to allow parameter evaluation through a dynamic problem optimization. At the end, the model is simulated using a chosen software program (Mathematica 5.0) designed to monitor the dynamic temperature proles of uid at dierent locations.

2.1. System description The considered pasteurization process occurs in a plate heat exchanger recommended for the pasteurization of milk because it oers an important convection coecient and higher turbulence in comparison with other classes of heat exchangers. The process consists of heating milk to a given temperature during a certain time in order to eliminate the pathological action of any possible bacteria. This is a necessary treatment that milk must undergo before any further use. Milk owing from a pre-heating tank enters the heat exchanger where it is heated by hot steam in order to reach the pasteurization temperature.

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

651

2.2. The mechanism of fouling A schematic representation of the proposed fouling model is given in Fig. 1. When the milk is heated to a temperature greater than 65 C, the betalactoglobulin protein is unsteady and becomes the precursor for deposit formation according to two possible mechanisms [7,8]: 1. The betalactoglobulin natural protein (N-protein) experiences a denaturation process (change of structure) and becomes very reactive because of its SH bond (betalactoglobulin denatured form or D-protein). 2. An irreversible polymerisation reaction resulting in insoluble particles as aggregations noted protein A (betalactoglobulin aggregated form). Usually, induction period is extremely short or even instantaneous in plate heat exchangers where intense mixing of uid held due to high turbulence. Fouling decreases with increasing turbulence [1]. The thickness and subsequently the volume of laminar sublayer decreases with increasing velocity and consequently the amount of foulant depositing on the heat-transfer surface decreases. 2.3. Causes of protein aggregation The betalactoglobulin protein has a global structure held together by SS bonds and one non-exposed internal free SH group. When heated, the betalactoglobulin starts to unfold. The free thiol group is therefore exposed and the molecule enters into an activated state making it possible to react with another betalactoglobulin molecule. Therefore, a radical chain may grow to form an aggregate that is able to deposit on the heattransfer surface. It is known that the rate of deposition is related to the concentration of activated molecules in the solution and may be calculated using the model of denaturation and aggregation of the betalactoglobulin [9,10]. The kinetics of the dierent reactions are well known [11]. A mass transfer process of the three forms of the betalactoglobulin protein takes place between the uid and the layer limit. The aggregated proteins form the deposit but it is still essential to know the kinetics and the dierent physical and chemical parameters of the phenomena taking place in order to be able to quantify the extent of this deposit represented by the fouling resistance [911]. 2.4. Mineral deposition In addition to the above mentioned type of fouling takes place the calcium phosphate deposition which presents an inverse solubility relation which temperature. During the pre-heating process, the ionic product becomes high with relation to the solubility concentration limits. Salts deposit in the form of crystals on the surface of the heat exchanger [12]. Calcium phosphate may also precipitate in the core ow. It may then deposit onto the casein micelle surface and/or onto the betalactoglobulin molecules contained in the milk serum phase. In all cases, it will in time ultimately form a deposit onto the stainless steel wall of the heat exchanger plates as shown on Fig. 1.

Calcium phosphate

N*

D*

A*

Fig. 1. Model of proteins and salts deposition on the heat exchanger surface.

652

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

Fouling caused by milk constituents is a complex process in which both whey protein aggregation and calcium phosphate formation in the bulk uid are to be accounted for. The whole fouling mechanism can be described in the following steps: 1. Straight adsorption of a protein mono-layer even at room temperature. 2. Formation of activated betalactoglobulin molecules in the bulk solution at temperatures higher than 65 C. The betalactoglobulin aggregates (tenths of nanometers) and calcium phosphate particles are formed. 3. Continuous transport of these foulant particles formed in the bulk to the heated surface. However, some activated molecules can be inactivated during this phase because some given reactions with other molecules in the bulk can make the particles not enough active to create fouling. 4. Deposition of activated molecules by adsorption on the heat exchanger surface. Calcium ions entrapped in the protein deposit may help to stabilize these structures. 5. At relatively high temperatures (above 85 C), the main deposit component is calcium phosphate which oers an open network structure where small protein aggregates can be entrapped. 3. Mathematical model of fouling 3.1. Flow and energy equations The heat exchanger is supplied with hot water with one channel per pass for a total of ve passes. Milk supply is to ve channels per pass for a total of ve passes. The heat exchanger consists of 12 plates. Assuming that the plate surface is at and smooth (Fig. 2), the governing 2D ow equations include the continuity and the momentum equations in the Cartesian coordinates, as given by the following equations: ou ov 0; ox oy  2  ou ou ou 1 oP o u ou2 ; u v m ot ox oy q ox ox2 oy 2  2  ov ov ov 1 oP ov ov2 2 ; u v m 2 ox oy ot ox oy q oy 1 2 3

Fig. 2. Control volume of uid inside the channel in 2D.

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

653

where m is the kinematic viscosity, q the density, P the pressure, t the time, and u and v are the velocity components along the x and y directions, respectively. The transient energy equation for a 2D constant property, incompressible ow can be given by:  ej qj Cpj Cpp qp dp oT j ot     ! oT j oT j uj vj U j T pj1 T pj 2T j ; ox oy 4 5

oT pj U j T j T j1 2T pj ; oT

where Tj is the temperature of the uid in channel j, Tpj is the temperature of plate j, Cpj and Cpp are respectively, the specic heats of uid in channel j and plate p, qj and qp are the respective densities of uid in channel j and plate p, ej is the distance between the plates, dp is the plate thickness, and Uj is the overall heat-transfer coecient in channel j. 3.2. Mathematical model The rate constant expression for the dierent reactions involved in the protein transformations is of the Arrhenius-type form given by: K K 0 exp ! E ; RT 6

where, values of K0 and E are given in Table 1 for dierent temperatures. Proteins react in both the uid bulk and the thermal boundary layer in uid milk. Native protein N is transformed to denatured protein D, in a rst order reaction. The denatured protein then reacts to give aggregated protein A in a second order reaction. Mass transfer between the uid bulk and the thermal boundary layer takes place for each protein. Only the aggregated protein is deposited on the wall (M). The rate of deposition is proportional to the concentration of aggregated protein in the thermal boundary layer. The fouling resistance to heat transfer is proportional to the thickness of the deposit. The mass balances in bulk uid are as follows: !  !  ! oC Nj oC Nj oC Nj EN o oC Nj o oC Nj u v k NO exp C Nj DN DN ot ox oy RT j ox oy ox oy k mN C Nj C : Np dT For the denatured protein: ! !  ! oC Dj oC Dj oC Dj EN ED o oC Dj 2 u v k NO exp C Nj k DO exp C DD ot ox oy RT J RT j Dj ox ox  ! o oC Dj k mD C Dj C : DD Dp oy dT oy

Table 1 Kinetic parameters for the fouling reaction scheme [3] T (C) Native Denatured Aggregation 7090 7090 7090 E (J/mol) 2.614 105 3.37 1037 2.885 105 ln (K0) 86.41 89.40 91.32

654

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

For the aggregated protein:

!  !  ! oC Aj oC Aj oC Aj ED o oC Nj o oC Nj 2 u v k DO exp C DA DA ot ox oy RT j Dj ox ox oy oy k mA C Aj C : Ap dT

The mass balances for the various proteins in the boundary layer are as follows: !  ! oC oC oC oC Np EN o Np Np Np u v k NO exp C DN ot ox oy RT p Np ox ox  ! oC Np o k mN C Np C Nj ; DN oy dT oy ! ! oC oC oC EN ED Dp Dp Dp u v k NO exp C k DO exp C 2 ot ot oy RT p Np RT J Dp  !  ! oC Dp oC Dp o o k mD DD DD C Dp C Dj ; ox oy dT ox oy !  ! oC oC oC oC Ap ED o Ap Ap Ap u v k DO exp C 2 DA Dp ot ox oy RT p ox ox  ! oC Ap o k mA C Ap C Aj k w C ; DA Ap oy dT oy oC kw Mp C : ot dT Ap

10

11

12 13

Modelling the fouling phenomenon in the plate heat exchanger requires knowledge of several properties both at bulk and at wall conditions. Moreover, the thickness of the hydrodynamic boundary layer, d, is related to that of the thermal boundary layer, dT, by the well known solution obtained by Eckert [13]: dT Pr1=3 : d 14

The use of the Von Karman integral theory can drive to this type of relation between dT and d knowing the boundary conditions along with the velocity and temperature proles. Furthermore, the protein mass transfer coecient, kmi, is related to the diusion coecient, Di, by: Di 15 k mi ; d where Di can be calculated as follows [14]: Tj Di 1:31017 lV 0:6 i

16

The mass of protein and salt material deposited on a plate heat exchanger, expressed in g/m2, can then be calculated as follows: ! kd Bip x; yqd I : 17 tk s log Massp x; y U0 LL The overall heat-transfer coecient for clean conditions, U0, is calculated according to the following equation [10]: Nu where De 2ej 19 hDe 0:214Re0:662 3:2Pr0:4 ; k 18

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

655

and 1 1 1 Pj : U 0 hc hf kp The overall heat-transfer coecient for the fouled conditions, U, is given by [1]: U U0 : 1 Bi 21 20

The rate of deposition is related to the rate of change of heat transfer, Bip, as follows [15]: oC oBip Mp bdT ; oT oT 22

where b is a constant. The dynamic boundary layer thickness, d, can be calculated from the Sherwood number, Sh, by the relation [16]: d De ; Sh 23

where the Sherwood number is determined by: Sh 0:214Re0:662 3:2Sc0:4 : 24

Table 2 summarizes the methods by which physical and thermo-physical properties used in the calculations are determined. Moreover, Table 3 gives the heat exchanger data and operating conditions used in the study.

Table 2 Evaluation of the thermo-physical properties used in the simulation [16] Thermal conductivity of milk (W/m C) Density of milk (kg/m3) Dynamic viscosity of milk (Pa S) Specic heat of milk (J/kg C) Thermal conductivity of plate (W/m C) Specic heat of plate (J/kg C) Density of plate (kg/m3) Thermal conductivity of deposit (W/m C) Density of deposit (kg/m3) 0.00133T + 0.539911 1033.70.2308T 0.00246T2 (0.60445T + 0.947)103 1.68T + 3864.2 16.3 450 7200 0.5 1030

Table 3 Operating conditions and technical details of the PHE Parameter Flow rate (kg/s) Diameter (m) Width of plates (m) Height of plates (m) Thickness of plates (m) Gap between plates (m) Constant b Native protein diameter (m) Wall reaction rate (m/s) Symbol De l l Pj E DN kw Value 0.074 0.0022 0.1 0.1 8 104 4 103 129 9.92 1011 107

656

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

3.3. Initial and boundary conditions The plate heat exchanger was preceded at steady state with a non-fouling uid. This is acceptable since pasteurizers are pre-heated with hot water before the milk uid is processed. The following initial conditions are then imposed and given by: T j x; y; t T 0 ; 8j; 8x; 8y; t 0 T p x; y; t T amb ; 8p; 8x; 8y; t 0 C Nj x; y; t C Dj x; y; t C Aj x; y; t 0; 8j; 8x; 8y; t 0 C x; y; t C x; y; t C x; y; t 0; Np Dp Ap C Nj x; o; t C x; o; t 5 kg=m3 Np 8j; 8p; 8t; x 2 0; 0:1 C Dj x; o; t C x; o; t 0kg=m3 Dp 8j; 8p; 8t; x 2 0; 0:1 C Aj x; o; t C x; o; t 0 kg=m Ap 8j; 8p; 8t; x 2 0; 0:1 Top and bottom wall boundaries:
oC Nj oC Dj oC Aj oC Np oC Dp oC Ap 0; oy oy oy oy oy oy 3

8p; 8x; 8y; t 0

8j; 8p:

Left and right wall boundaries:

oC Nj oC Dj oC Aj oC Np oC Dp oC Ap 0 8j; 8p: ox ox ox ox ox ox

4. Results and discussions 4.1. Numerical simulation The model described in this paper includes a set of integral partial dierential and algebraic equations. The solution method is based on a two phase method of lines approach. In the rst phase, the spatial dimensions are discretized in terms of nite dimensional representations, leading to a reduction of the integral, partial dierential and algebraic equations into sets of dierential algebraic equations with respect to time. The second phase, the dierential algebraic equations are integrated with respect to time by a modied Eulers technique [4]. Hierarchical model construction is employed. Each channel is modelled as a sub-model. All the sub-models are eventually connected to a general model through appropriate boundary conditions. The plate heat exchanger thermal model is also dened in parallel and suitably connected with all the other sub-models to dene the temperatures required by the protein reaction scheme. Dierent numerical methods are used for discretization, according to the ow direction. A nite dierence method with 40 elements is used if the ow is in the two directions from zero to the exit of the channel. The plate heat exchanger thermal model is approximated by a second order centered nite dierence method with 40 elements. This discretization schemes and orders where chosen to obtain results. Parameter estimation analysis are performed and based on the solution of a dynamic optimization problem. A regular, structured grid of plate domains in the 2D of hexahedral mesh elements was created to divide the domain into 66,000 discrete elements. The domain was meshed using tetrahedral scheme. The validation data were collected in an industrial plant with a capacity of approximately 2000 kg/h. The plant consisted of a pre-heating section, a heating, and cooling sections. In the following, models and the related variables are discussed. At rst, the validation of the calculated values was done by comparing the

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

657

deposition rates with data from the literature, then, in turn, indirectly by comparison of the simulated results with temperatures measured in an industrial heating plant. The knowledge obtained from the experiments was integrated manually into the fuzzy systems. The operating time interval considered for both simulation and validation data was set to 8 h, since this was the approximative average operation time between two cleaning cycles. 4.2. Evolution of milk temperature The calculate vector velocity pattern for milk in PHE is realised by simulations were performed with Dx = 10 mm, Dy = 25 mm, and discretized by considering Dt = 0.5 s (<0.1 s), which satised the stability criterion for the explicit solution of the model. Since uid milk ows upward through the plate surface from bottom left to top left, the model predicts a lesser amount of uid ow on the right of plate surface, leading to an excessive temperature rise. Fig. 3 shows the milk temperature in the channel 4 of the PHE predicted after a eight (08) hours operation. The uid enters the canal at 70 C which then increases progressively the x and y two directions of the channel in an exponential fashion as the milk gains heat from the hot steam. The peaks in the temperature, recorded along the height direction are the result of thermal exchange on both sides of the considered channel. The temperature is calculated while xing the width. At the entry of the channel adjoining plate 4, milk enters by the orice and creates a turbulence that generates a better thermal exchange on this side in relation to the other side of the plate. This results in a more rapid progression of the temperature as opposed to the other side of the channel. Once the pasteurization temperature is reached, the thermal exchange enters the two uids is stopped. 4.3. Evolution of mass deposition Fig. 4 shows the distribution of the foulant deposition from milk in the plates of the 4th channel. The potential foulant mass per unit area is calculated to be 16 g/m2 at the most. The peaks of fouling quantities,

Fig. 3. Simulated temperature of milk in side the 4th channel (C).

658

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

Fig. 4. Fouling distribution of uid milk in the plate adjacent to the 4th channel (g/m2).

in the height direction, are due to the thermal exchanges on both sides of the considered channel (channel 4) knowing that the activation of the fouling process depends mainly on temperature. Such heat exchange becomes more important as the uid advances in the channel. The results remain very similar to those obtained by other authors with dierent operating conditions but with similar type of geometry [7,17,18]. The deposit is primarily controlled by the protein aggregation reaction. Once heated, the proteins aggregate and stick on the wall. It is believed that the deposit of salts remains a minor phenomenon at the pasteurization temperature of 90 C. The faster progression of milk temperature in the entry side activates the process of fouling more quickly through the deterioration of the proteins and other saturation of salts on this side in comparison with the side opposite of the plate. This result is more important deposit of the entry of the channel of the side of the entry in relation to the contrary. 4.4. Evolution of the overall heat-transfer coecient The variation of the overall heat-transfer coecient between the two uids in the 4th channel of the heat exchanger is represented in Fig. 5. It is clearly established that the overall heat-transfer coecient decreases gradually along the length of the channel following the deposit formation on the plate walls which constitutes an additional layer resistant to heat exchange between the two uids [13,19]. Fouling growth tends to be dominant over the entire plate surface leading to a decrease of the ow rate caused by a reduction of the section crossing of ow. After a given time, the mass deposited does not aect any more the thermal exchange between the two uids a constant evolution of the exchange coecient is observed. This is explained by the fact that, the value of the thickness of the deposit passes through a certain critical value after which the resistance of the fouling deposit remains appreciably constant.

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

659

Fig. 5. Evolution of the overall heat-transfer coecient ratio inside the 4th channel.

4.5. Evolution of the betalactoglobulin protein concentration The form of the betalactoglobulin protein concentration along the exchanger is shown in Fig. 6. The protein concentration decreases following the denaturation process induced by the increase in the milk temperature during the heating operation. A reduction of 30% with respect to the initial concentration of betalactoglobulin is observed. The altered (active form) protein gives birth to aggregations that deposit on the heat exchanger surface. From the 5th channel, the pasteurization temperature pasteurization (90 C) is

Fig. 6. Evolution of the betalactoglobulin protein concentration in the heat exchanger.

660

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

Fig. 7. Evolution of the betalactoglobulin protein concentration inside the 11th channel.

reached, and the heating is stopped. Thus, it appears that the fouling is governed by the deposit of salts and not by the aggregation of the proteins; the measure of the protein concentration is realized by Kjeldahl method for example. The protein concentration prole along the 11th channel of the heat exchanger is shown in Fig. 7. The concentration remains constant in the uid along the channel. The reaction of denaturation is stopped and there is no more deposit of protein. This result conrms the one already observed (Fig. 6). 4.6. Eect of Reynolds number The mass of deposit accumulated on the plate adjacent to the 4th channel of the heat exchanger surface as a function of the Reynolds number is displayed in Fig. 8a,b and c. Observation of this gure shows that the mass of foulant increases along the channel as mentioned. Furthermore, it also appears that the mass deposition decreases with an increase in the Reynolds number. This result is in agreement with various previous ndings [1821] which is explained by the higher removal rate of foulant as shear forces increase. However, it should be reminded that higher Reynolds numbers values also mean higher pressure drops, which may be of considerable impaction pumping costs. 5. Conclusions A 2D fouling model, coupled with a 2D dynamic model using material balance equations for the various interactions and chemical reaction taking place during pasteurization of milk are applied to predict the performance of a plate heat exchanger subject to fouling. The results showed fouling is highly dependent on the various process operating conditions. The dierent parameters seem to aect the phenomenon more specically in the ow rather than in the with direction. The mass of deposit depends mainly on milk temperature and time of processing. Moreover, it is observed that the fouling extent is strongly related to the Reynolds number value and is inversely proportional to the latter. For a better control of the problem, which passes by a minimization of production expenses and an optimal quality of the product, a number of solutions are recommended. For instance, a good quality of the water used

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

661

Fig. 8. Foulant mass deposition (in g/m2) in the plate adjacent to the 4th channel for: (a) Re = 6000, (b) Re = 8000 and (c) Re = 10000.

is essential. Besides, a possible decrease in the pasteurization temperature is believed to allow an additional energy savings without aecting the product quality. Also, the use of food surfactants in the pasteurizer would probably contribute to create a competition during the adsorption process between the surfactant and the protein molecules which will avoid deposition on the heat exchanger surfaces. Finally, addition of an emulsier would eventually protect the proteins from aggregating. References
[1] S.D. Changani, M.T. Belmar-Beiny, P.Y. Fryer, Engineering and chemical factors associated with fouling and cleaning in milk processing, Exp. Therm. Fluid Sci. 14 (1997) 392406. [2] J.P. Tissier, M. Lalande, G. Corrieu, A study of milk deposit on a heat exchange surface during ultra-high-temperature treatment in engineering and food, in: B.M. McKenna (Ed.), Engineering Sciences in the Food Industry, vol. 1, Applied Science Publishers, 1984. [3] K. Grijspeerdt, L. Mortier, J. De Block, R.V. Renterghem, Applications of modelling to optimize ultra high temperature milk heat exchangers with respect to fouling, Food Contr. 15 (2004) 117130. [4] P.K. Saho, I.A. Ansari, A.K. Datta, Milk fouling simulation in helical triple tube heat exchanger, J. Food Eng. 69 (2005) 235 244.

662

Y. Mahdi et al. / Applied Mathematical Modelling 33 (2009) 648662

[5] I.A. Ansari, M. Sharma, A.K. Datta, Milk fouling simulation in a double tube heat exchanger, Int. Comm. Heat Mass Transfer 30 (2003) 707716. [6] M.C. Georgiadis, S. Macchiatto, Dynamic modeling and simulation of plate heat exchangers under milk fouling, Chem. Eng. Sci. 55 (2000) 16051619. [7] M. Lalande, J.P. Tissier, G. Corrieu, Fouling of a plate exchanger used in ultra-high-temperature sterilisation of milk, J. Dairy Res. 51 (1984) 557568. [8] M. Lalande, J.P. Tissier, G. Corrieu, Fouling of heat transfer surfaces related to b-lactoglobulin denaturation during heat processing of milk, Biotechnol. Prog. 1 (1985) 131139. [9] M. Lalande, F. Reno, Fouling by milk and dairy product and cleaning of heat exchange surfaces, in: L.F. Melo, T.R. Bott, C.A. Bernardo (Eds.), Fouling Science and Technology, NATO ASI Series E, Kluwer, Amsterdam, Netherlands, 1988, pp. 557573. [10] F. Delaplace, J.C. Leuliet, J.P. Tissier, Fouling experiments of a plate heat exchanger by whey protein solutions, Trans. IChemE. 72 (C) (1994) 163169. [11] J. Visser, T.J.M. Jeurnink, Fouling of heat exchangers in the dairy industry, Exp. Therm. Fluid Sci. 14 (1997) 407424. [12] T.J.M. Jeurnink, Fouling of milk with various calcium concentrations, Presented at Fouling and Cleaning in Food Processing, Jesus College, Univ. Cambridge, Cambridge, UK, 1994. [13] E.R.G. Eckert, Introduction to the Transfer of Heat and Mass, Mc Graw-Hill Inc, New York, 1950. [14] R.H. Perry, D. Green, Perrys Chemical Engineering Handbook, McGraw-Hill Book, 1984. [15] I. Toyoda, P.J. Fryer, A computational model for reaction and mass transfer in fouling from whey protein solutions, Fouling Mitigation of Industrial Heat Exchange Equipment, Begell House, New York, 1997. [16] P. De Jong, Modelling and Optimisation of Thermal Treatments in the Dairy Industry. Ponsen and Loojin, NIZO Research Report V341 (p. 165) Ede, The Netherlands, 1996. [17] P.K. Nema, A.K. Datta, A computer based solution to check the drop in milk outlet temperature due to fouling in a tubular heat exchanger, J. Food Eng. 71 (2005) 141156. [18] T.R. Bott, L.F. Melo, Fouling of heat exchangers, Exp. Therm. Fluid Sci. 14 (1997) 315321. [19] C. Rivero, V. Napolitano, Estimation of fouling in a plate heat exchanger through the application of neural networks, J. Chem. Tech. Biotechnol. 80 (2004) 594600. [20] A.S. Jerey, W.W. Nazaro, Predicting particle deposition on HVAC heat exchangers, Atmos. Environ. 37 (2003) 55875596. [21] M.T. Belmar-Beiny, S.M. Gotham, W.R. Paterson, P.J. Fryer, A.M. Pritchard, The eect of Reynolds number and uid temperature in whey protein fouling, J. Food Eng. 119139 (1993).