Sie sind auf Seite 1von 13

Food Hydrocolloids 27 (2012) 1e13

Contents lists available at ScienceDirect

Food Hydrocolloids
journal homepage: www.elsevier.com/locate/foodhyd

Review

Toward an integrated modeling of the dairy product transformations,


a review of the existing mathematical models
J. Foucquier a, b, *, E. Chantoiseau c, d, e, S. Le Feunteun a, b, D. Flick c, d, e, S. Gaucel a, b, N. Perrot a, b
a

INRA, UMR782 Gnie et Microbiologie des Procds Alimentaires, F-78850 Thiverval-Grignon, France
AgroParisTech, UMR782 Gnie et Microbiologie des Procds Alimentaires, F-78850 Thiverval-Grignon, France
c
AgroParisTech, UMR1145 Ingnierie Procds Aliments, F-91300 Massy, France
d
INRA, UMR1145 Ingnierie Procds Aliments, F-91300 Massy, France
e
Le Cnam, UMR1145 Ingnierie Procds Aliments, F-91300 Massy, France
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 28 March 2011
Accepted 3 August 2011

The large diversity of dairy products is the consequence of complex processes involving series of unit
operations with a wide range of controls that makes the complete modeling awkward. Due to the variety of
milk components and process conditions, a generic model describing the milk processing can only be
achieved by the integration of various models into a generic modeling framework. Moreover, the building
of such an approach involves the coupling of transformation models describing each unit operation. In this
scope, the present work aims to review existing mathematical models of the dairy products processing
with a special focus on some main process units: thermal treatment, homogenization and coagulation.
For each unit operation, several transformation models are investigated according to their complexities,
relevance to depict actual phenomena and ability to be integrated with others models to represent
complex transformations. As a rst step for the integration of models, a focus on their input parameters
and predicted variables is achieved.
2011 Elsevier Ltd. All rights reserved.

Keywords:
Review
Modeling
Heat treatment
Homogenization
Acid coagulation
Rennet coagulation

1. Introduction
Processed dairy products have attracted extensive research in
the past, ranging from chemical transformations of the product
components through to the nal product properties. However, due
to the complexity of the mechanisms involved, most studies have
focused on a particular process unit and its effects on milk components, including effects of one or two parameters. Few authors
have attempted to study an entire dairy process, which is the entire
conversion of the milk into the nal product by sequence of process
unit, and integrating all process parameter effects on the nal
product properties. Indeed, the nal mechanical properties of the
end-product, such as rmness or sensory properties, have been
poorly modeled to date.
Milk is a lipid in water emulsion, whose aqueous phase, named
whey, is a mixture of water, mineral salts, lactose and proteins. The
proteins constitutes around 3% of the milk mass and allow making
a gel. They are subdivided into two groups: the whey proteins and
caseins. The former are soluble proteins in the aqueous phase of the
* Corresponding author. AgroParisTech, UMR782 Gnie et Microbiologie des
Procds Alimentaires, F-78850 Thiverval-Grignon, France.
E-mail address: jfoucquier@grignon.inra.fr (J. Foucquier).
0268-005X/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodhyd.2011.08.002

milk whereas the latter are mostly found into large structure, the
casein micelles.
The few available modeling-focused reviews are dedicated to
specic process steps, without considering the process as a whole.
For example, Considine, Patel, Anema, Singh, and Creamer (2007)
studied the thermal and high-pressure treatments in order to highlight how the two process units have similar effects on whey proteins.
Donato and Guyomarch (2009) studied the effect of heating on
milk proteins, especially whey proteins and casein interactions.
Raikos (2010) reviewed the effect of heat treatment on milk protein at
the fat droplets interface. Lucey (2002) reviewed the gelation of milk
after a study focusing on acid milk gel (Lucey & Singh, 1997). Milk
homogenization has been reviewed by Hayes and Kelly (2003).
Pugnaloni, Dickinson, Ettelaie, Mackie, and Wilde (2004) studied
different types of simulations found in the literature on competitive
adsorption at the fat droplets interface.
Furthermore, process units are investigated from standpoint of the
dimension scale of the studied phenomena, while other scales are
either ignored or more generally represented by empirical relations.
Very few studies are properly multiscale, i.e. investigating a phenomenon at molecular-scale (microscale) in order to generate values for
a separate model at a higher scale to describe the product structure
(mesoscale) or its macroscopic properties such as viscosity or gelation

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

pH. Moreover, multiscale modeling studies are available only for model
products. Although some authors have developed models to illustrate
microscale phenomena, such as the unfolding of whey proteins at the
fat droplet interface (Dickinson & Euston, 1990), few have built models
considering different scales, ranging from molecular adsorption
to product viscosity (Hkansson, Trgrdh, & Bergensthl, 2009).
The aim of this work is to provide a synthesis of current
modeling knowledge on full end-to-end processes typically
occurring in the dairy industry. Instead of a precise description of
the chemical phenomena involved in individual process units, we
focus on descriptions of the available modeled representations of
the process units involved in milk transformation. We analyze the
models available in terms of size scale, inputs and outputs. Added
focus is given to models coupling to other process units or other
size scales in order to prepare the ground for future couplings.
This literature review has been build in order to highlight the
transformations occurring during each treatment that are relevant
to the descriptions of the nal product properties, such as texture,
structure or mechanical properties, that have to be integrated
into a model that could account for a multistep process. We have
selected relevant transformations and dependencies from over 250
publications covering most of the aspects of these models.
After an introduction setting out the typical dairy process and
associated process units investigated, we will focus on the process
units of interest, starting with thermal treatment of the milk, then
moving on to product homogenization, and ending with the acid
and rennet-induced dairy coagulation.
1.1. Typical dairy processes
The main step in making dairy gel is the coagulation of the protein
network that gives the product its nal structure, and thus its texture.
The coagulation or agglomeration of caseins into a gel can be performed by slow acidication down to around pH 4.6, which destabilizes the structure of the main protein components, or by adding
enzymes that will gel the proteins at neutral pH (Lucey, 2002).
The nal gel structure is highly dependent on the method
used to induce gelation (acid, enzymatic, or methods that couple
approaches). Nevertheless, it also depends on the exact composition of the dairy solution as well as the pre-treatments it has
eventually been subjected to.
For instance, several studies have shown that nal gel structure
depends on the milk heating regime, as the temperature-induced
alterations to the milk proteins change their gelation behavior
(Donato & Guyomarch, 2009). Thermal treatment may involve
agglomeration of proteins, and incorporating such agglomerates in
the nal gel can cause broad modications in product texture. A
proper model of the gelation mechanisms must therefore account
for a possible thermal treatment of the product before the coagulation step.
Moreover, for whole milk, gelation is also preceded by homogenization: the product is pushed through a small gap, where shear
stress breaks the milk fat globules into smaller droplets. This is
a widespread process unit, as the broken milk fat droplets form a more
stable emulsion, improving milk appearance for long conservation
times (Walstra, Wouters, & Geurts, 2006). As the inclusion of smaller
fat droplets in the nal gel changes its mechanical behavior, the effects
of this process unit should be integrated as well to properly describe
the process.
1.2. Investigated process unit
Table 1 summarizes the main process units used for dairy
product transformation and the main effects and phenomena
involved in the component alteration.

Table 1
FunctioneProcess relation.

Denaturation of
whey proteins
Aggregation of
whey proteins
Milk fat fragmentation
Aggregation of caseins

Thermal
treatment

Homogenization

Coagulation

Cooling

: Principal step, : Signicant impact, : Possible effect.

The properties of the nal product and especially the structure


and texture are shaped by the composition, and the operating
conditions of each process unit. Furthermore, there are variable,
indirect and non-linear inuences between these steps.
For each process unit, we will analyze the most widely-accepted
models used to depict the phenomena and the different length
scales concerned. A cognitive map will then be provided to synthesize the inuences of the process units on product components.
The product composition and the interactions between process
units are later treated, mostly by explaining their effect and a way
to account them in the model depicting the main alterations.
2. Thermal treatment
2.1. Aims and mechanisms of the process unit
The rst and main aim of the thermal treatment is to stabilize
the product from a bacterial point of view. Moreover, this allows
controlling the homogeneity of the microbiological population of
the product if it is inoculated further on in the process.
However, many processes resort to longer thermal treatments
than strictly required to control the matrix microbiology. Indeed, as
further described, heating treatments can lead to the denaturation
of the whey protein which results in the formation of whey protein
aggregates that may associate with the milk fat globule membrane
and with caseins. Thus thermal treatment can strongly modify the
physicochemical features of the medium. For this reason, heating
is commonly used to modulate the nal properties of some dairy
products.
2.2. Effects on the milk components
2.2.1. Whey proteins alone
Whey proteins are mainly globular proteins solubilized in
the continuous phase of the milk. Whey contains several types of
protein at variable concentrations, as shown in Table 2, and whose
cumulative mass is about 20% of the total milk protein mass.
Given the compositional distribution of whey proteins, whey
protein behavior is mainly described by b-lactoglobulin, either
alone or with a-lactalbumin. There is extensive literature on the
aggregation of these whey proteins according to thermal process
Table 2
Bovine milk whey protein composition (de Wit, 1990).
Whey proteins

Proteins concentration (g/L)

b-Lactoglobulin
a-Lactalbumin

3.2
1.2
0.4
0.8
0.2
0.03
0
>1

BSA
Immunoglobulin
Lactoferrin
Lactoperoxidase
Enzymes (>50)
Proteoseepeptose

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

conditions or composition, but although general reviews have been


published (Considine et al., 2007; Donato & Guyomarch, 2009;
Raikos, 2010), no general study evoking all the possible operating
conditions is currently available.
The main component involved during heating is b-lactoglobulin.
As temperature increases, small reversible change in the conformation of b-lactoglobulin occurs, and at 65  C at natural milk pH
(pH 6.8), the tertiary structure of the globular protein changes,
leading to the exposure of thiol and hydrophobic groups (Tolkach &
Kulozik, 2007). Denatured b-lactoglobulin then tends to irreversibly aggregate with other b-lactoglobulins, a-lactalbumins, milk fat
globule membranes, or the k-caseins located at the surface of the
micelles or, more rarely, in the whey (Garcia-Risco, Ramos, & LopezFandino, 2002; de Jong & van der Linden, 1998; Oldeld, Singh,
Taylor, & Pearce, 2000).
The denaturation/aggregation kinetics of b-lactoglobulin can be
depicted as a single 1.5-order kinetics reaction representing the
whole process as a single step. However, based on experimental data
and theoretical accounts of the chemical reaction kinetics, most
authors opt to use two successive reactions to depict the denaturationeaggregation pathway (Tolkach & Kulozik, 2007), considering
the aggregation step as a whole. Denoting native b-lactoglobulin
monomers by b and denaturated b-lactoglobulin by b*, they write:

/ b*
)
k1

Denaturation

k2

Aggregation

/b*n

n$b*

k2

This leads to a system of differential equations describing


system component evolutions. Adding reactions with other
components, for example the reaction with free k-casein (de Jong &
van der Linden, 1998) or with a-lactalbumin (Oldeld et al., 2000),
can easily be done simply by adding the corresponding equation.
Due to differences in the activation energies and the temperature evolution of pre-exponential factors, each step in the pathway
limits the overall aggregation rate for its own temperature range.
Up to around 85  C the denaturation step limits the kinetics of the
reaction, whereas above 85  C the aggregation step becomes the
rate-limiting step (de Wit, 1990). Several studies have nevertheless
shows that the exact transition temperature depends on both
the product composition and pH (Oldeld et al., 2000). In order to
depict the aggregates sizes, several authors have given detailed
aggregation pathways, such as Roefs and de Kruif (1994) who
describe the aggregation as a polymerization process:

Denaturation

Propagation

While these chemical models precisely represent the chemical


pathway of the reaction, they fail to depict all the physical
phenomena involved. Indeed, the shear rate effect on the aggregate
highlighted by Walkenstrom, Nielsen, Windhab, and Hermansson
(1999) cannot be accounted for. A solution could be to model
aggregation using the population balance equation proposed by
Smoluchowski (1917). This model determines the variation of the
number of elements of size v at time t, noted n(v,t), by integrating
fragmentation and aggregation by:

dnv;t

dt

mw fv;w gw nw;t dw  gv nv;t


wv

1
2

Zv

bv;w nw;t nvw;t dw


w0

ZN
 nv;t

Where n(w,t) is the number of elements of size w at time t, g(v) and


g(w) are the fragmentation frequency of elements of size v and w
respectively, b(w,v  w) is the coalescence frequency of elements of
size w with elements of size (v  w) (both can be seen as a function
of temperature or shear rate, as their values depend on local ow
and the mean free path of the elements), m(w) is the number of
fragments formed when an element of size w breaks, and among
them, f(v,w) is the fraction of w fragments whose size is in the range
[v,v dv].
The rst term denotes the increase in the number of elements
of size v as larger elements get fragmented. The second term shows
the decrease in the number of elements of size v by breakage.
The third term depicts the creation of elements of size v by union of
smaller elements, and the last term denotes the association of
elements of size v with elements of different sizes.
Both aggregation and breakage of aggregate occurs during
thermal treatment, but for a given set of operation condition, one
phenomenon is usually predominant, and the other can be safely
neglected. In the case of thermal treatment of dairy product, the
aggregation terms largely dominate the breakage ones.
Determining element concentration requires simplications
of the population balance model, as described by Kumar and
Ramkrishna (1996). In the case of aggregation, this involves discretizing the aggregates population into several classes regrouping
particles by number of monomers. The evolution of the number of
particles of size k, noted Nk, is given by:
N
X

i k1

k3

b*k b*nk /bn

Each of the chemical reactions considered is modeled by a rstorder equation. The nal system is easy to solve if initial product
composition and rate constants are known. However, a large set of
parameters depicting the full denaturationeaggregation pathway
can be difcult to determine, and has only been achieved for model
systems (Verheul, Roefs, & de Kruif, 1998).
One advantage of this method is the ease of adding reactions
to the chemical kinetics scheme. Indeed, in order to account for the
b-lactoglobulin reaction with other components, de Jong and van
der Linden (1998) just added the equation depicting the reaction.

i
k
gi Nt
 gk k  1Nt

k1 b
X
k;i
i1

Termination

bv;w nw;t dw
w0

dt

2
b*k b/b*k1

ZN

k
dNt

k1

b / b*

i
ki
k
Nt
 Nt
Nt

N
X
i1

i
bk;i Nt

With g(i) the fragmentation frequency of aggregates of size i,


and b(k,i), the frequency of their aggregation with aggregates of size
k. This kind of model has been used for a wide range of applications,
including the formation of whey protein aggregates by Peron,
Heffernan, Byrne, Rioual, and Fitzpatrick (2007) and Wu, Xie, and
Morbidelli (2005). For pure b-lactoglobulin, Elofsson, Dejmek, and
Paulsson (1996) successfully used a simple Smoluchowski model
for hindered aggregation in order to determine the sizes of the
heat-induced aggregates.
The population balance model is very useful for predicting
the evolution of homogeneous suspensions. However, the reaction
of the studied particles with other species can be difcult to

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

implement in cases involving realistic initial compositions, which


explains why this approach is rarely used for skim milk studies.
a-lactalbumin, the second type of whey proteins by weight,
is more thermosensitive than b-lactoglobulin. At neutral pH and
65  C, it is totally denaturated. However, it does not form pure
aggregates since it can link b-lactoglobulin via disulde bonds. See
Oldeld et al. (2000) and Paulsson and Dejmek (1990) for studies
on a-lactalbumin denaturation kinetics.
The a-lactalbumin effect on b-lactoglobulin can easily be
accounted for by adding its reaction on b-lactoglobulin aggregates
(Chen, Chen, Nguang, & Anema, 1998; Oldeld, Singh, & Taylor, 1998;
Oldeld, Singh, Taylor, & Pearce, 1998; Oldeld et al., 2000) to the
reaction scheme. The chemical reaction formulation is simple, and
makes it possible to easily depict product property inuences on
reaction rates by adding the reaction involving a-lactalbumin,
denoted a, to the reaction involving b-lactoglobulin denoted b.

Denaturation

Fixation

a
a*
a/

kb

b*k a/b*k a

Modeling whey protein behavior during thermal treatment,


apart from determining product composition, is equally useful for
process modeling. Polymerization or population balance models
make it possible to determine the cumulative distribution of the
aggregate in terms of monomer numbers. The link between number
of monomers in an aggregate and its size can be established using
the fractal dimension (D) of the product (Bremer, Bijsterbosch,
Schrijvers, Van vliet, & Walstra, 1990; Byrne, Fitzpatrick, Pampel, &
Titchener-Hooker, 2002; Le Bon, Nicolai, & Durand, 1999; Mahmoudi, Mehalebi, Nicolai, Durand, & Riaublanc, 2007; Vetier,
Desobry-Banon, Eleya, & Hardy, 1997). D lies between 1 and 3 and
denotes that the aggregates become more open as their size
increase. To simplify, it can be said that for D 3, the sphere of radius
R is entirely lled with proteins, whereas for D 2 the proteins are
only located on the sphere surface. This property of the aggregate
depicts its openness. For spherical aggregates containing Np monomers of radius a of a total radius R, we can write:

Np

 D
R
a

(3)

The fractal dimension of the aggregate is useful for structural


characterization of the product, as it gives a parameter describing
the whey protein aggregates, such as the total volume fraction of
the aggregates. The latter value allows the determination of product
viscosity using the appropriate relation from among the broad
empirical selection developed and commented in the literature to
depict the evolution of the suspension viscosity with the solid
volume fraction (Derkach, 2009; de Jong & van der Linden, 1998).
As whey protein aggregates are involved in the network of dairy
gels, they play a role in the nal product properties. Determining
whey protein aggregate structure is thus important in order to
understand the nal product properties.
2.2.2. Caseins and whey proteins
Caseins form the main group of milk proteins. Natively associated
on a supra-molecular structure called casein micelles, they
represent 80% of milk protein content (Qi, 2007). The micelles
are large structures (between 50 and 200 nm in diameter). Once
destabilized, either by acidication or enzyme action, they form
the protein network of the dairy gels. Despite intensive study, the
structure of the micelle is not yet entirely depicted. Even though
several models have been proposed for their internal structure, all of

them agree on two facts: i) the micelles are held together by calcium
phosphate nanocluster and hydrophobic interactions, and ii) the
micelle stability is provided by a surface layer of k-caseins that
prevents their aggregation by steric and electrostatic repulsions.
While heat treatment does not alter the casein micelles themselves, k-caseins at the surface of the micelles react with denaturated
b-lactoglobulin and whey protein aggregates to form disulde bonds
(Donato & Guyomarch, 2009). This affects b-lactoglobulin aggregation, since the k-caseins x the denatured b-lactoglobulin and limit
the extent of aggregates, as depicted by Corredig and Dalgleish (1996)
and de Jong and van der Linden (1998). This reaction is dependent on
temperature, pH, and ionic strength. The mechanism specic to the
formation of k-caseind;b-lactoglobulin complexes is not clearly
established, as conicting data exist on whether association occurs in
whey or at the micelle surfaces (Donato & Guyomarch, 2009).
This association also reduced the repulsion between micelles,
which tend to link with each other or with other molecules to
form aggregates, precipitates, or gels. Lee and Sherbon (2002) also
underlined that for higher temperatures, caseins (mostly k-caseins)
can also associate with proteins included in the milk fat globule
membrane via k-caseindb-lactoglobulin complexes.
Moreover, micelle behavior during gelation, either by renneting
or by acidication, is strongly dependent on the degree of destabilization by b-lactoglobulin, so that the nal gel properties are
partly governed by pre-heating (del Angel & Dalgleish, 2006).
2.2.3. Milk fat globules
Heated milk fat globule membranes can undergo several
chemical transformations, including denaturation of the composite
proteins, but no size change is reported for the thermal treatment
alone, suggesting an absence of globule breakage or coalescence.
As most available studies on thermal treatment are based on whey
protein concentrate or skim milk rather than whole milk, there is little
data on milk fat globule membrane modications during heating.
Nevertheless, as for casein micelles, it is know that the main milk fat
globule evolution during heating is due to the aggregation of denaturated whey proteins. Ye, Singh, Oldeld, and Anema (2004) published
a study on associations between whey proteins and milk fat globule membranes via quantitative polyacrylamide gel electrophoresis
to measure the compositional evolution of the milk fat globule
membrane during thermal treatment of whole milk. They showed that
whey proteins associate with milk fat globule membranes following
a site lling model (Sharma & Dalgleish, 1994) where parameters
can be determined according to temperature. The concentration C(t)
(mg proteins per g of membrane) of either b-lactoglobulin or
a-lactalbumin at time t, and at 80  C, can be described by:



cN  ct
kt
ln
cN

(4)

where k is the rate constant that follows an Arrhenius law, and CN is


the nal concentration. Above 85  C, the concentration of whey
proteins can be depicted using:



cN  ct 1
1 k0 t
cN

(5)

Splitting the kinetics at 85  C is coherent with observations on


the b-lactoglobulin denaturation pathway and its rate-limiting step
reported earlier.
Apart from the adsorption of the whey proteins on the milk fat
globule membrane, Lee and Sherbon (2002) showed changes in
interfacial membrane composition occurring above 70  C as glycoproteins in the milk fat globule membranes are also denatured,
xing the denatured whey proteins.

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

2.3. Inuencing factors


Once denaturated, b-lactoglobulins form aggregates of sizes that
depend on pH and temperature, as well as on lactose, salt, and calcium
concentrations, as described by Donato and Guyomarch (2009).
2.3.1. Component interactions
The interactions between the milk components during thermal
treatment have been heavily studied but, as pointed earlier, most
published works mainly focus on a given component for a narrow
operating conditions range.
For example, Spiegel (1999) showed that lactose has a stabilizing
effect on b-lactoglobulin during heating. By protecting the ternary
structure of the globular proteins, lactose limits on b-lactoglobulin
denaturation.
Studies on denaturation kinetics according to the calcium
concentration have also been conducted by Parris, Hollar, Hsieh, and
Cockley (1997) and Spiegel and Huss (2002). Both show a decrease
of the aggregates size when the Ca2 concentration is limited.
Indeed, large protein aggregates of b-lactoglobulin are linked by
calcium bridges between small aggregates. The less calcium is
present the less large aggregates can be formed.
If denaturation is modeled using chemical kinetics, the inuence
of component interactions on the denaturation and aggregation
processes can easily be taken into account by modifying the rate
constant depending on the interaction, as performed by Spiegel (1999)
2.3.2. pH
Numerous studies have addressed the inuence of pH on whey
protein behavior (Elofsson et al., 1996; Spiegel & Huss, 2002;
Verheul et al., 1998). These studies showed that for a pH around
the pHi of b-lactoglobulin, i.e. a pH between 4 and 4.5, globular
proteins are more stable, thus limiting denaturation by heat. A
similar pattern is observed when the whey has a high ionic
strength, as in the case of high salt concentrations.
Corredig and Dalgleish (1996) and Guyomarch (2006) reported
a decrease in the b-lactoglobulin reaction on casein micelles from
pH 5.2 to pH 10.5, as the k-caseins are expelled into the whey.
In the reaction kinetics framework, the effects of pH can be
accounted for by using pH-dependent parameters. Thus, Spiegel
and Huss (2002) showed that pH effects can be taken into
account by a variation in the pre-exponential factor.
Table 3 lists a summary of the relevant publications for descriptions of thermal treatments for process modeling, giving the
variables investigated, model types, and length scale involved. Most
of these publications investigate the product at the microscopic level
to depict whey protein aggregation in order to predict the concentrations of native and aggregated proteins at macroscopic level. Note
the extensive use of global chemical kinetics to describe the complex
protein aggregation process that is poorly modeled at a microscale.
As stated previously, the selected models cover the main factors
involved in the thermal treatment of dairy product. Thus the effects
of other components are seen as negligible compared to the ones
described previously. The application of such models to realistic
products could be achieved by setting the initial composition of the
actual product for the pertinent components.
3. Homogenization
3.1. Aim and mechanisms of the process unit
Homogenizing consists of pouring a pressurized (at severaldozen MPa) milk through a small hole. The uid velocity and the
Reynolds number locally increase, and together with turbulences
create large shear stresses that break the milk fat globules and

destroy the large pre-existent protein aggregates. As the resulting


product contains smaller fat droplets, it is less subjected to coalescence, thus improving product stability.
There is ongoing debate over the physical mechanisms involved
in droplet breakage. The main mechanism seems to be the effects
of turbulence forces (Floury, Desrumaux, & Lardires, 2000)
that deform the droplets until they break. Nevertheless, effects of
laminar shear have been pointed by Walstra (1993), and effects
of cavitations in the homogenizer have been cited as well. While
several relations have been proposed for predicting the nal
suspension properties, the most commonly accepted is:

d32 zDP q

(6)

That links the Sauter mean diameter of the droplets (d32) to the
drop in operating pressure drop (DP) by a negative constant q. More
precise predictions of the process unit, such as evolutions in droplet
size distribution, require more complex models.

3.2. Effects on milk fat globules


Homogenization leads to total or at least partial disruption of
the fat globules and to the adsorption of milk proteins, and thus
increases the interfacial surface area of fat in solution by increasing
the number of fat droplets. Researchers have worked on mathematically modeling these phenomena. Some articles deal with drop
size distribution, for either fat drops or other types of drops.
Chen, Pruss, and Warnecke (1998) tackled drop size distribution
in emulsion via a modeling approach based on the assumption that
coalescence can be ignored if the volume fraction of the dispersed
phase is small enough and emulsier concentration is large
enough. They dened f(x,t)dx as the number of droplets per unit
volume of the dispersion at time t at a size between volume x and
x dx. The resulting population balance equation is expressed as:

vf x; t v
1
Gf x; t f x; t  fF x; t Ex; t  Dx; t
s
vt
vx

(7)

where x is particle volume, G is the particle growth rate,


v=vxGf x; t is convective ux along the size axis, s is residence
time, f(x,t) is drop size distribution, fF(x,t) is feed size distribution,
and E(x,t) and D(x,t) are the birth and death rate functions,
respectively. The difference between the birth and death rate
functions denes the accumulation rate. This expression is only
available for a mixed suspension.
Breakage rate s(x) depends on the volume of the globules x, local
energy dissipation 3 , surface tension s, viscosity of the oil hd, and oil
density rd, and is expressed as:

sx k1 exp

!
k2 s1 f2 k3 hd 1 f

rd x5=9 3 2=3
rd x4=9 3 1=3

(8)

where k1, k2 and k3 are parameters to be determined from experimental data.


Rabe, Verdes, and Seeger (2011) published a review focusing on
the modeling of protein adsorption. It seems that the Langmuir
adsorption model could be considered as a reference. In this model,
the adsorption occurs to distinct available surface sites.



q
dq
 koff q
kon cs 1 
qmax
dt

(9)

where q refers to the protein coverage, qmax is the maximum


coverage level at which no more binding site is available, kon and
koff are, respectively, the on-rate and off-rate constants and cs is the
protein concentration directly above the surface.

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

Table 3
Selected references on thermal treatment modeling.
Reference

Step

Input variable

Output variable

Model type

Model scale

Oldeld, Singh, Taylor,


& Pearce, 1998

Whey protein denaturation


and aggregation

Temperature, time

Chemical kinetics

Macroscale

Spiegel, 1999

Whey protein denaturation


and aggregation

Temperature, lactose
concentration, pH

Chemical kinetics

Macroscale

Roefs & de Kruif, 1994

Whey protein denaturation


and aggregation

Initial concentration

Chemical kinetics

Macroscale

de Jong & van der


Linden, 1998

Whey protein aggregation


and micelle destabilization

Temperature,
concentration, time

Chemical kinetics,
polymerization

Micro and
macroscale

Anema, Abby, Mike, &


Harjinder, 2008

Micelle destabilization

Time, pH

Empirical model

Micro and
macroscale

Oldeld, Singh, &


Taylor, 1998

Micelle destabilization

Time, temperature

Chemical kinetics

Macroscale

Elofsson et al., 1996

Whey protein aggregation

Population balance

Macroscale

Le Bon et al., 1999

Whey protein aggregation

pH, concentration,
temperature
Temperature, time

Chemical kinetics

Micro and
macroscale

Oldeld et al., 2000

Whey protein aggregation

Temperature, b-lg and


a-la concentration

Chemical kinetics

Macroscale

Ye et al., 2004

Whey proteins associating


with MFGM

Temperature, product
composition

Macroscale:
Residual native
whey proteins
Mesoscale:
Aggregate size
Microscale
Aggregate structure
Macroscale
Residual native protein
concentration
Mesoscale
Size of the aggregates
Macroscale
Aggregate size
distribution, viscosity
Microscale
Micelle surface state
Macroscale
Viscosity
Microscale
Micelle surface state
Microscale
b-lg, a-la association
with micelles
Macroscale
Aggregate size distribution
Macroscale
Gel time, gelation
temperature, aggregate
concentration
Macroscale
Residual native protein
concentration
Microscale
MFGM composition

Chemical kinetics

Macroscale

Other models have been developed to describe the protein


adsorption (Rabe et al., 2011) as the RSA (Random sequential
adsorption) model which seems to be more realistic. In this model,
the adsorption occurs to random available surface sites.
In the specic case of fat droplet dispersion, such as in whole milk,
Hkansson et al. (2009) developed a model to describe the fragmentation, adsorption, coalescence and turbulence in a high-pressure
homogenizer based on the population balance equation (see Eq. (1)).
To express droplet fragmentation, turbulent fragmentation is
split into two classes depending on the size of active eddies,
i.e. regime TI (turbulent inertia) and regime TV (turbulent viscous).
Thus, total fragmentation rate is expressed as the sum of fragmentation in TI, gTI(3 ,d), and fragmentation in TV, gTV(3 ,d):

g3 ; d gTI 3 ; d gTV 3 ; d

(10)

where d is the drop diameter and 3 is the rate of turbulent energy


dissipation.
This same team went on to develop a model describing the
adsorption of macromolecular emulsier to the drop interface.
In this modeling approach, adsorption in a ow eld is considered
as a collision-driven phenomenon. It can be described, in terms of
the change in surface load G(r,t) (in kg/m2), as the sum of adsorption due to Brownian and turbulent forces:

vG
vt


 
vG
vG

vt turbulent
vt brownian

8
r
3
C t
>
>
<
0:163
d d3 E 2 aads if d < 2l
vG
v E
p$d

vt turbulent >
C t
>
: 0:272p3 1=3 dE d3 E aads if d  2l
p$d2

(11)

(12)

 


vG
2kB v
1 1 CE t
a

dE d

vt brownian
3mC
dE d p$d2 ads

(13)

where n is the kinematic viscosity (m2/s), aads is the number of


adsorptions per collision, kB is the Boltzmanns constant (J/K), mC is the
dynamic viscosity of the continuous phase, dE is the hydrodynamic
diameter of macromolecular emulsiers in meters, CE(t) is the continuous phase concentration of macromolecular emulsiers in kg/m3.
Thus, Hkansson et al. (2009) developed a complete modeling
approach extending from the adsorption of emulsiers to the
size distribution of fat droplets, and accounting for Brownian and
turbulent forces. However, this model requires ne parameters,
which are tricky to acquire on real emulsions.
Other mathematical models have dealt with the motion of particles under forces at the interface (attractive energy, repulsive energy,
external energy) for theoretical solutions of homogeneous and equalsize particles at an interface. Based on Langevin equations, two
complementary approaches are proposed (Pugnaloni et al., 2004):
- The Molecular Dynamics (MD) simulation makes it possible to
obtain the position of a particle depending on the forces applied to
it. This consists in solving the motion equations of the entities
involved in forming the system. The equation can be expressed as:

mi

N
X


d2 ri t
Fiext ri t
Fi;j ri t; rj t
2
dt
jsi

(14)

where mi is the mass of particle i, ri(t) is its position at time t, and


Fiext and Fi,j are the forces applied on i by any external eld and by
particle j, respectively.

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

Dickinson (1992) used this method to simulate the ratio of a real


fraction of interface components q1 =q2 . For competitive adsorption
between adsorbates of equal size, this ratio is expressed as:

q1
S h
1 1
q2
S2 h 2

(15)

where Si is a measure of the degree of stickiness for the adsorbent


surface of species type i, and hi is the bulk volume fraction of
component i.
Within the limits of strong adsorption, the ratio will be Boltzmanns ratio:

3 
1

exp
q1

kT

3
q2
exp 2

(16)

kT

where 3 i is the maximum well depth for component i at a distance


z 0, k is the Boltzmanns constant and T is the temperature.
- The Brownian Dynamics (BD) simulation is almost the same
as the MD simulation, but here the viscous force of the
solvent and a random force are added. Thus, the equations are
expressed as:

mi

N
X


d2 ri t
dr t
ext

F
r
t

Fi;j ri t; rj t  3phs i FiR t


i
i
dt
dt 2
jsi

(17)
where FiR t represents the random force, h is solvent viscosity, and
s is particle diameter.
Several studies have focused on the unfolding of particles on
an interface, and more specically on simulating the unfolding of
a protein. These studies are based on Monte Carlo simulations,
which make it possible to propose a ne-grained theoretical
representation of the protein at the interface.
Dickinson and Euston (1992a) used the Monte Carlo method to
simulate the number of molecules adsorbed at the interface when
dynamic equilibrium is reached. In this approach, each deformable
particle consists of a linked group of m identical occupying sites
on a lattice. In each iteration of a Monte Carlo algorithm, surfaceesubunit interactions, U(y), and attraction interactions, U(r), are
calculated, respectively, by:

8
< N y  0 or y  Ly
1
y 1 or y Ly
Uy
: EA

0 1 < y < Ly

(18)

8
< Nr 0
Ur
E r 1
: I
0 r>1

(19)

where EI is the interparticle subunit interaction energy, EA is an


adsorption energy, and r is the subunit pair separation in units of
lattice spacing.
Euston and Abu Naser (2005) then adapted the Dickinson
and Euston algorithm (Dickinson & Euston, 1992b) to simulate the
number of adsorbed molecules at the interface when dynamic
equilibrium is reached, taking into account surface pressure
and then particle conformation. The surface pressure Pr is then
written as:

Pr

Dp
2pRg x; yL

(20)

Where Dp is a measure of the force applied to the globule,


2pRg(x,y) is a cross-sectional circumference, and L is a weighted
average thickness.
The self-consistent eld-based calculations (Ettelaie, 2003)
are an example of model used as a tool to study the equilibrium
properties of polymers at interface. The usefulness of this method is
the application to adsorbed layers of as1-casein and b-casein at
solid interfaces and the calculations of steric interactions mediated
by such layers. This model makes it possible to represent the
protein molecules in which a distinction was made between polar,
apolar and the charged ionic amino acid groups comprising the
protein sequences.
Leermakers, Atkinson, Dickinson, and Horne (1996) used selfconsistent eld theory to compare the effects of pH and ionic
strength on the predicted equilibrium interaction potential between
identical surfaces with the two different caseins.
All these models are listed in Table 4 giving predicted variables,
parameters, model type, and scales. To conclude, approaches
have been conducted at various scales, but except for the work of
Hkansson et al. (2009), multiscale approaches are rare. Moreover,
at micro/nanoscales, these models require a lot of input variables
which need to be measured. Hence, these micro/nanoscale
approaches remain mainly theoretical, and need to be validated
with real solutions. At macroscopic scale, the variables are less
numerous (Chen, Chen, et al., 1998; Chen, Pruss, et al., 1998) but also
less informative.
The presented models cover the main phenomena occurring
during an homogenization but there is currently no article that
account for the fact that dairy products contain two types of
proteins (casein and whey proteins). Indeed, the structure, and
hence properties, of dairy products partly depend on the composition of the adsorbed proteins at the fat droplet interface. A future
and interesting continuation for such approaches would therefore
consist in modeling the competition between the adsorption of
whey proteins and caseins at the fat droplet interface.
3.3. Inuencing factors
As stated earlier, after homogenization, the newly-created
surface of the milk fat globules is populated with either whey
proteins or caseins. As both proteins are greatly altered by heating,
the milk fat globule membranes after repopulation vary strongly
depending on whether there was preliminary thermal treatment.
As highlighted in several studies (Cano-Ruiz & Richter, 1997;
Garcia-Risco et al., 2002; Lee & Sherbon, 2002), for non-heated
milk the surface is mainly populated by caseins, whereas for
heated milk the denatured whey proteins are the main component
associated with the newly-created interface. As the modied
behavior of whey proteins is due to their denaturation, proper
modeling of the milk fat globule membrane population after
homogenization can integrate the pre-heating factor by considering denaturated whey protein apart from the native population,
and thus include the effects of thermal treatment in the population
competition modeling.
As the milk fat globules are broken up during homogenization,
viscosity progressively increases. There are few studies on midprocess dairy product viscosity, but Lee and Sherbon (2002) show
that homogenization entails only a slight increase in the viscosity of
unheated milk. Processes coupling thermal treatment and homogenization (whatever the order) led to a greater increase in viscosity,
while the viscosity increase following heating was reported as
negligible under their operating conditions. Moreover, the previously heated product contains protein aggregates, which are largely
altered by homogenization, leading to breakage due to high shear
rates (Iordache & Jelen, 2003).

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

Table 4
Summary table listing relevant articles dealing with phenomena occurring during and/or after homogenization.
Reference

Step

Competitive adsorption of
proteins and lowmolecular-weight
surfactants: computer
simulation and
microscopic imaging
(Pugnaloni et al., 2004)

Adsorption of
surfactants at the
interface, at
microscopic and
mesoscopic scales

Input variables

Nano/microscopic scale:
- Perimeter length-dependent
- Position of particles
interaction energy ED
- m: number of subunits
at a given law
in a deformable particle
- Adsorption energy EA
- Interparticle subunit energy EI
- r: center-to-center separation
- di,N 1: cross-term diameter

Output variables

Adsorption of sticky hard


spheres: relevance to
competitive protein
adsorption
(Dickinson, 1992)

Adsorption of
surfactants at the
interface at
nanoscopic scale

- r: center-to-center separation
- di,N 1: cross-term diameter

A statistical model for


simulating adsorption
from a mixture of
deformable particles
(Dickinson & Euston,
1992b) (M1) simulating
the equation of state of
model globular protein
adsorbed at a surface
(Euston & Abu Naser, 2005)
(M2)
Dynamic simulation of
emulsion formation in
a high-pressure
homogenizer
(Hkansson et al., 2009)

Adsorption of
surfactants at the
interface, at
nanoscopic scale

Nano/microscopic scale:
- Perimeter length-dependent
Stochastic model
- Number of adsorbed
interaction energy ED
(MC simulation)
- m: number of subunits
molecules at the interin a deformable particle
face when dynamic
- Adsorption energy EA
equilibrium is reached
- Interparticle subunit energy E
- For model M2, they
- Force applied to the globule D3 P
added surface pressure
(M2)
and then the impact of
protein conformation

Nanoscopic and
microscopic scales

Fragmentation,
adsorption,
coalescence and
turbulence in
a high-pressure
homogenizer at
mesoscopic and
macroscopic scales

- Rates of reactions of coales- Macroscopic scale:


cence and fragmentation
- Number of drops of
- Fragmentation in a turbulent
volume v at time t
inertia regime
- Total fragmentation
- Fragmentation in a turbulent
rate
viscous regime surface load
Mesoscopic scale:
- Continuous phase
concentration of
macromolecular
emulsiers

- Population equation
balance
- Empiric modeling
- Stochastic modeling

Mesoscopic and
macroscopic scales

- Population equation
balance
- Empiric model
(breakage rate)

Macroscopic scale

Nano/microscopic scale:
- Volume fraction ratio
of components at the
interface

Macroscopic scale:
- Accumulation rate
(evolution in drop size
distribution, convective
ux along the size axis,
residence time, feed
size distribution)
- Breakage rate

A population balance model Drop disruption


for disperse systems:
during emulsion at
drop size distribution
mesoscopic scale
in emulsion
(Chen, Pruss et al., 1998)

Understanding protein
adsorption phenomena
at solid surfaces
(Rabe et al., 2011)

- Protein coverage
Mesoscopic scale:
- Maximum coverage level at
- Evolution of protein
which no more binding site is
coverage during time
available
- Description of the
- On-rate and off-rate constants
surface covered by
- Protein concentration
proteins
- Protein coverage
- Saturation coverage
Microscopic scale:
- Equilibrium properties
of polymers at interface

Adsorption of protein
at solid surfaces

Computer simulation and


Equilibrium
modeling of food colloids properties of
(Ettelaie, 2003)
the emulsion

Birth rate function


Death rate function
Critical energy Ec
Turbulent kinetic energy Ek

4. Coagulations
4.1. Aims and mechanisms of the process unit
4.1.1. Acid coagulation
Acidication is classically performed by two different ways
(Lucey & Singh, 1997). One is to use bacterial cultures which
ferment lactose to lactic acid. The second one, which is more often
used for research, is to add Glucono-d-Lactone (GDL) which is
progressively hydrolyzed in gluconic acid.
From pH 6.7 to 6.0 (the initial pH of milk being 6.7), there is
a decrease in the net negative charge on the micelles, as the isoelectric

Model type
- Stochastic model
(Monte Carlo (MC)
simulation)

Model scales
Nanoscopic and
microscopic scales

Molecular modeling
(Molecular dynamics (MD)
simulation and Brownian
Dynamics (BD) simulation)
Molecular modeling (MD
Nanoscopic and
simulation)
microscopic scales

- Empiric modeling
Mesoscopic scale
(Langmuir adsorption
model)
- Stochastic modeling
(RSA (Random sequential adsorption) model)

- Self-consistent eld
modeling

Microscopic scale

pH of caseins is of about 4.6 (Horne, 1999). This initiate a reduction of


the electrostatic repulsions between nearby particles but the structural features of the casein micelles remain relatively unchanged.
From pH 6.0 to 5.0, there is a decrease in both electrostatic and steric
stabilization, due to the collapse of the k-caseins surface layer.
Moreover, the dissolution of the Colloidal Calcium Phosphate (CCP),
which starts as soon as acidication is in progress, becomes faster
beyond pH w 6.0 and is fully completed at pH w 5.0. This necessarily
affects the internal structure of the casein particles, and, for pH  5.0,
the attractive forces (hydrophobic interactions and van der Waals
forces) overcome the repulsive forces and casein particles aggregate.
In unheated milk gels, gelation occurs around pH 4.

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

Many articles studying acidication have focused on characterizing the gel after acidication. Under these approaches, few methods
make it possible to obtain macroscopic features from microscopic/
nanoscopic data, or broadly speaking, to recover textural informations from the structure. Bremer, van Vliet, and Walstra (1989) used
the fractal method to model the overall volume fraction of particles in
an acid casein gel. They considered that the particles form a gel as
soon as the network homogeneously occupies the total volume.
The overall volume fraction ft is expressed as:

ft

aD3

RD
i

(21)

R3i

where R is the radius of the aggregates, a is the radius of a given


lattice site, and D is the fractal dimension.
Moreover, they expressed permeability B as:

 2
2
a
fD3
K

(22)

where K is a constant of tortuosity. The storage modulus of the gel


was expressed as:
2

G K 0 fD3

(23)

K0

where
is a constant. The fractal dimension is accessible via
turbidity measurements, but these measurements are only
accessible for a simple product such as a pure casein gel.
Another method is the percolation theory, which models how
a structure develops in a gel through aggregation (Horne, 1999). This
method depends on the fraction of reacted bonds p. Indeed, above
a certain threshold fraction, pc, the clusters grow in size to form
a large cluster extending through the whole sample. In rheology,
this threshold corresponds to the gel point, and the storage modulus
can be estimated by:

G p  pc

p>pc

(24)

where g 1.9.
The percolation model demonstrates how gel strength grows as
more material is incorporated into the network since more bonds
are formed. However, the model does not provide access to the
pre-factors in the proportionality relationships.
If the two latter modeling approaches remain only theoretical,
other types of model have been proposed.
Kristo, Biliaderis, and Tzanetakis (2003) developed an empirical
model to predict the simultaneous effect of fermentation temperature, milk solids level, and total inoculum concentration on the
acidication process. This was achieved by second-order polynomial-response surface models, where the independent variables
fermentation temperature (FT), total solids level (TS) and total
inoculum level (TI) were implemented for each response variable,
i.e. G0max , tandmin, onset of gelation, IE (rate of gelation), Vm
(maximum acidication rate), Tm (time at which Vm is reached), Te
(time to reach the end of fermentation), as follows:

Y b0 b1 X1 b2 X2 b3 X3 b11 X12 b22 X22 b33 X32


b12 X1 X2 b13 X1 X3 b23 X2 X3 3

(25)

Where Y is one of the given response variable, b0, b1, ., b23


represent the estimated regression coefcients, with b0 as the
constant term, b1, b2, b3 as linear effects, b11, b22, b33 as quadratic
effects, and b12, b1, b23 as interaction effects, 3 is random error and
X1, X2, X3 are the independent coded variables (FT, TS, TI, respectively). All of the models that include Eq. (25) appear to robustly
explain the data. Nevertheless, this approach is tted in a given
conguration, and cannot be extrapolated to others.

As he proposed for homogenization (protein adsorption),


Dickinson (2000) used Brownian dynamics to simulate the properties of particle gels. Dickinsons model also makes it possible to
obtain the position of a particle, expressed as follows:

X
d2 r
m 2i
Fij Ri  Hi
dt
j

(26)

where ri is the coordinate of particle i, Fij is the force on particle i


due to particle j which depends on the distance between them, Ri is
the stochastic (diffusive) force acting on particle i, and Hi is the force
on particle i arising from viscous dissipation.
Table 5 summarizes all these acid coagulation model. Overall,
the existing models dealing with acid coagulation are either
theoretical or empirical, but cannot be applied to other models.
Moreover, most of the time, excluding the fractal approach, only
one scale is taken into account.
4.1.2. Rennet coagulation
The main active component in rennet, chymosin, is an enzyme
that specically hydrolyzes the k-caseins which largely contribute
to the colloidal stability of the casein micelles. Upon rennet
addition, the k-caseins located at the surface of the particles are
thus progressively split off. This reduces the steric and electrostatic
repulsions between the casein particles and, when the enzyme
action is sufciently advanced, the renneted casein particles
aggregate via hydrophobic interactions. In unheated milk, the
percentage of k-caseins proteolysis is of about 90% when the system
gels (Lucey, 2002).
The next three models have been used to track the gelation
mechanism of chymosin-induced gelsin predicting the storage
modulus G0 :
- The Scott Blair model (Scott Blair & Burnett, 1963)

G0 G0N expf  s=t  ts g

(27)

where G0N is the value of G0 when t / N, s is the time-constant of


the model, and ts is a time-shift coefcient.
- The Douillard model (Douillard, 1973):

G0 G0N 1  expf  k=t  t0 g

(28)

where k is the rate constant and t0 is the time elapsed from rennet
addition to detectable onset of gelation.
- The Carlson model (Carlson, Hill, & Olson, 1987), which takes
into account the two phases of the rennet coagulation
(enzymatic reaction and aggregation) process:

3


 k
k1
2
k
k


6
7
2
2
e tt0
G GN 41 
 1 e tt0 5
k2  k1
k2  k1


(29)

where k1 and k2 are rate constants.


Esteves, Lucey, and Pires (2001) used these models to evaluate
the formation of enzymatic gels prepared from plant and animal
coagulants. They tested 3 coagulant types: Cynara cardunculus L.,
Cynara humilis L., and chymosin. The Scott Blair model was the most
efcient at describing the gelation mechanism for each of the
coagulants. The Carlson model worked well for the chymosin
curves, whereas the Douillard model could not predict the start of
gelation.

10

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

Table 5
Summary table of the articles dealing with the acidication.
Reference

Step

Output

Model type

Model scales

Theoretical and
experimental study
of the fractal nature
of casein gel
structure
(Bremer et al., 1989)

Formation of the
acid casein gel

Input
- Radius of the fractal
- Radius of one lattice site
- Fractal dimension

Mesoscopic scale:
- Overall volume
fraction of the particles
Macroscopic scale:
- Permeability
- Rheology

Empirical model
(Fractal method)

Mesoscopic and
macroscopic scales

Formation and
structure of acidied
milk gels
(Horne, 1999)
Modeling of the acidication
process and rheological
properties of milk
fermented with a
yogurt starter culture
using response surface
methodology
(Kristo et al., 2003)

Formation of an
acidied milk gel

- Threshold fraction of
reacted bonds

Macroscopic scale:
- Rheology

Macroscopic scale

Impact of ferments
on gel formation

- Fermentation temperature
- Total solid level
- Total inoculum level

Macroscopic scale:
- G0max
- tandmin
- Rate of gelation
- Max. acidication rate
- Time at which max
acidication rate
is reached
- Time to reach the
end of fermentation

Empirical model
(Percolation model possible to obtain the
rheology from the structure)
Empiric modeling
(polynomial modeling)

Structure and rheology of


simulated gels formed
from aggregated colloidal
particles
(Dickinson, 2000)

Simulation of the
properties of
particle gels

- Force on particle i due


to particle j
- Stochastic force acting
on particle i
- Force on particle I arising
from viscous dissipation

Nano/microscopic scale:
- Position of a particle

Zhong and Daubert (2004) developed a mathematical model


to analyze the rheological data of rennet casein gelation at different
cooling rates. They expressed the complex shear modulus G*
(G* G0  iG00 ) as

1
G*
1 *
GN

!n



d G* =G*N
dT=T0



Enf T0
exp 
a
RT0 T

k0 T0

(30)

where G*N is the shear modulus as time t heads toward innity, T


is the absolute temperature, T0 is the initial temperature, k0 is
a prefactor, a is the cooling rate, R is the universal gas constant, and
Enf is the energy spent as ocs form a network.
Furthermore, population equation balance models have been
developed to describe enzymatic coagulation (Hyslop, 1993). The
base of the models makes it possible to predict the molar concentration of the product aggregates Ck using the Von Smoluchowski
expression (1917):
N
X
X
dCk
1
Ci Cj  Kkj Ck
Cj
Kij
dt
2
j1

(31)

ij k

where Kij and Kkj are the collision rates of partially-created aggregates that contain i, j and k integrated subunits, and Ci, Cj and Ck
represent molar concentrations of these created aggregates.
Drake (1972) converted the Von Smoluchowski equation to
a moment equation:

PN

k1

dt

kn Ck

N X
N

dmn
1X
i jn in  jn K2 Ci Cj
v
2
dt

Molecular modeling
(Brownian Dynamics
simulation)

Macroscopic scale

Nanoscopic and
microscopic scale

represents the diffusion-controlled process of milk coagulation,


according to which the velocity of milk casein coagulation in the
rst phase depends on the velocity of the enzymatic hydrolysis of
k-casein, whilst the second phase proceeds without activation
energy and is exclusively velocity-dependent.
Hyslop (1993) and Hyslop and Qvist (1996) later stated four
different models for enzymatic coagulation which were applied by
Novakovic, Petrak, Kordic, and Slacanac (2000), with some modications, to acid coagulation. These models are listed in Table 6.
Note that all these models mentioned so far make it possible to
obtain quantitative data at macroscopic and mesoscopic scales,
but none of them make it possible to obtain data at nanoscopic
or microscopic scales. Moreover, it is possible to obtain some
properties of product texture (with rheological data) and some
properties of product structure (with the aggregate concentrations), but these models cannot be used to obtain the texture from
the structure.
The previous models about acid and enzymatic coagulation
cover the main operating factors involved in the coagulation of dairy
product. However, to apply these models to realistic products, the
inuences of other components must be accounted. The majority of
these effects can be depicted by changes of the aforementioned
model parameters according to experimental data. However, some
measurements are difcult to do on complex products, and so it
could be difcult to adapt models to these experimental data.
Thereby, modication of the product behavior according to initial
composition change would be more easily investigated.

4.2. Selected inuencing factors

i1 j1

(32)
where mn is the nth moment for a distribution of partly-aggregated
particles, K2 is the occulation rate constant, and n represents the
velocity of primary phase of k-casein hydrolysis. This equation

4.2.1. Heat treatment effects on acid coagulation


Acid gels made from heated milk have a higher pH at gelation
and produce considerably rmer gels than unheated milk (Lucey &
Singh, 1997). Many of the effects of heat treatment on acid milk gel
microstructure were explained by suggesting that the aggregation

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

11

Table 6
Summary table of the articles dealing with enzymatic coagulation.
Reference

Step

Output

Model type

Model scales

Mathematical modeling of the


formation of rennet-induced
gels by plant coagulants
and chymosin
(Esteves et al., 2001)
Enzyme-induced coagulation
of casein micelles: a number
of different kinetics models
(Hyslop, 1993)
Application of numerical analysis
to a number of models for
chymosin-induced coagulation
of casein micelles
(Hyslop & Qvist, 1996)
Kinetics of rennet gelation
at different cooling rates
(Zhong & Daubert, 2004)

Gelation mechanism
of milk gels made
using chymosin

Input
- G0 when t->N

Macroscopic scale:
- G0

First-order kinetics
(Scott Blair, Douillard
and Carlson models)

Macroscopic
scale

Gel formation

- Rate constants for collisions


of partially-created aggregates
- Molar concentrations of
the created aggregates

Mesoscopic scale:
- Change of molar
aggregate concentration
- Change of the moment
of molar aggregate
concentration

Population equation
balance

Mesoscopic
scale

Rennet casein
gelation at different
cooling rates

Macroscopic scale:
Evolution of the complex
shear modulus as a function
of temperature

Differential equation

Macroscopic
scale

Complex shear modulus


Cooling rate
Initial temperature
Absolute temperature

of denatured whey proteins occurs during the acidication of


heated milk.
Denatured whey proteins associated with casein micelles
could act as bridging material by interacting with other denatured
whey proteins associated with micelles. This is likely to increase
the number and strength of bonds between protein particles.
Moreover, the concentration of gelling protein would be increased
due to the active participation of denatured whey proteins in the
gel structure (Lakemond & van Vliet, 2008).
The presence of denatured whey proteins on the surface of
casein particles would also hinder the close approach of other
casein particles and lessen the likelihood of dense clusters of casein
particles forming.
4.2.2. Action of transglutaminase (TG)
The literature features numerous articles dealing with the
description of the action of TG on acid and enzymatic coagulation.
TG catalyzes an acyl transfer reaction between the g-carboxyamide
group of a peptide-bound glutaminyl residue and a variety of
primary amines, including the amino group of lysins, resulting in
the formation of an isopeptide bond.
However, there is only one modeling study focused on this action
(Schorsch, Carrie, Clark, & Norton, 2000). The authors used the
cascade theory, as follows, to model gelation, which is divided into
two steps: aggregation prior to the gel point, and network formation:

Nf a1  v2 1  b
aRT
2

(33)

where f is the number of equivalent sites of each biopolymer


available for bond formation, a is the fraction of these cross-linking
sites in the system having reacted at any time, N is the number of
moles of polymer per unit volume initially present, and v and b are
functions of f and a, and are key elements of the cascade approach.
5. Conclusion
This work aimed to describe the state-of-the art in mathematical models dealing with dairy product transformations. Specic
focus was given to thermal treatment, homogenization, and coagulation. Looking at thermal treatment, the key point is the behavior
of b-lactoglobulin: denaturated by heat, its agglomeration or
association with casein micelles determines a large part of subsequent coagulation. Looking at homogenization, studies appear to

have focused essentially on the behavior of adsorbed proteins.


Coagulation studies consist in linking process conditions to the
casein network.
Our analysis illustrates that, to date, studies have focused on
isolated phenomena and are either too generic and/or theoretical to
be applied to real complex solutions, or in contrast are too narrowly
applicable to particular cases and difcult to apply to other solutions. This is explained by the complexity of both the product
components and the chemical reactions employed, which involve
different scales with non-linear evolutions and non-linear relations
between the different length and time-scales involved. However,
the large number of studies performed to make it possible to
highlight the underlying transformations.
Some mathematical studies led on other applications (Baudrit,
Sicard, Wuillemin, & Perrot, 2010; Perrot, Trelea, Baudrit, Trystram,
& Bourgine, 2011) have shown nevertheless that it is possible to
make some coupling model integrating the knowledge extract from
simple product and not totally representative of the reality to obtain
the modeling of a complex product. Indeed, for now, some studies
are ongoing in link with the complex system community to make it
possible to obtain a product only with the use of a mathematical
model.
Acknowledgments
The research leading to these results has received funding from
the European Communitys Seventh Framework Program (FP7/
2007-2013) under the grant agreement no. FP7-222 654-DREAM.
References
Anema, S. G., Abby, T., Mike, B., & Harjinder, S. (2008). The whey proteins in milk:
thermal denaturation, physical interactions and effects on the functional
properties of milk. In Milk proteins (pp. 239e281). San Diego: Academic Press.
del Angel, C. R., & Dalgleish, D. G. (2006). Structures and some properties of soluble
protein complexes formed by the heating of reconstituted skim milk powder.
Food Research International, 39(4), 472e479.
Baudrit, C., Sicard, M., Wuillemin, P. H., & Perrot, N. (2010). Towards a global
modelling of the Camembert-type cheese ripening process by coupling
heterogeneous knowledge with dynamic Bayesian networks. Journal of Food
Engineering, 98(3), 283e293.
Bremer, L. G. B., Bijsterbosch, B. H., Schrijvers, R., Van vliet, T., & Walstra, P. (1990).
On the fractal nature of the structure of acid casein gels. Colloids and Surfaces,
51, 159e170.
Bremer, L. G. B., van Vliet, T., & Walstra, P. (1989). Theoretical and experimentalstudy of the fractal nature of the structure of casein gels. Journal of the Chemical Society-Faraday Transactions I, 85, 3359e3372.

12

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13

Byrne, E. P., Fitzpatrick, J. J., Pampel, L. W., & Titchener-Hooker, N. J. (2002). Inuence of shear on particle size and fractal dimension of whey protein precipitates: implications for scale-up and centrifugal clarication efciency. Chemical
Engineering Science, 57(18), 3767e3779.
Cano-Ruiz, M. E., & Richter, R. L. (1997). Effect of homogenization pressure on
the milk fat globule membrane proteins. Journal of Dairy Science, 80(11),
2732e2739.
Carlson, A., Hill, C. G., & Olson, N. F. (1987). The kinetics of milk coagulation. 4. The
kinetics of the gel-rming process. Biotechnology and Bioengineering, 29(5),
612e624.
Chen, X. D., Chen, Z. D., Nguang, S. K., & Anema, S. (1998). Exploring the reaction
kinetics of whey protein denaturation/aggregation by assuming the denaturation step is reversible. Biochemical Engineering Journal, 2(1), 63e69.
Chen, Z., Pruss, J., & Warnecke, H. J. (1998). A population balance model for disperse
systems: drop size distribution in emulsion. Chemical Engineering Science, 53(5),
1059e1066.
Considine, T., Patel, H. A., Anema, S. G., Singh, H., & Creamer, L. K. (2007). Interactions of milk proteins during heat and high hydrostatic pressure treatments a review. Innovative Food Science & Emerging Technologies, 8(1), 1e23.
Corredig, M., & Dalgleish, D. G. (1996). Effect of temperature and pH on the
interactions of whey proteins with casein micelles in skim milk. Food Research
International, 29(1), 49e55.
Derkach, S. R. (2009). Rheology of emulsions. Advances in Colloid and Interface
Science, 151(1e2), 1e23.
Dickinson, E. (1992). Adsorption of sticky hard-spheres - relevance to protein
competitive adsorption. Journal of the Chemical Society-Faraday Transactions,
88(24), 3561e3565.
Dickinson, E. (2000). Structure and rheology of simulated gels formed from
aggregated colloidal particles. Journal of Colloid and Interface Science, 225(1),
2e15.
Dickinson, E., & Euston, S. R. (1990). Simulation of adsorption of deformable
particles modeled as cyclic lattice chains - a simple statistical-model of protein
adsorption. Journal of the Chemical Society-Faraday Transactions, 86(5),
805e809.
Dickinson, E., & Euston, S. R. (1992a). Computer-simulation model of the adsorption
of protein polysaccharide complexes. Food Hydrocolloids, 6(4), 345e357.
Dickinson, E., & Euston, S. R. (1992b). A statistical-model for the simulation of
adsorption from a mixture of deformable particles. Journal of Colloid and
Interface Science, 152(2), 562e570.
Donato, L., & Guyomarch, F. (2009). Formation and properties of the whey
protein/kappa-casein complexes in heated skim milk - a review. Dairy Science
and Technology, 89(1), 3e29.
Douillard, R. (1973). Rheological analysis of curd formation. Journal of Texture
Studies, 4(1), 158e165.
Drake, R. L. (1972). The scalar transport equation of coalescence theory: moments
and kernels. Journal of the Atmospheric Sciences, 29, 537e547.
Elofsson, U. M., Dejmek, P., & Paulsson, M. A. (1996). Heat-induced aggregation
of beta-lactoglobulin studied by dynamic light scattering. International Dairy
Journal, 6(4), 343e357.
Esteves, C. L. C., Lucey, J. A., & Pires, E. M. V. (2001). Mathematical modelling of the
formation of rennet-induced gels by plant coagulants and chymosin. Journal of
Dairy Research, 68(3), 499e510.
Ettelaie, R. (2003). Computer simulation and modeling of food colloids. Current
Opinion in Colloid & Interface Science, 8(4e5), 415e421.
Euston, S. R., & Abu Naser, M. (2005). Simulating the equation of state of model
globular proteins adsorbed at a surface. Langmuir, 21(9), 4227e4235.
Floury, J., Desrumaux, A., & Lardires, J. (2000). Effect of high-pressure homogenization on droplet size distributions and rheological properties of model oil-inwater emulsions. Innovative Food Science & Emerging Technologies, 1(2),
127e134.
Garcia-Risco, M. R., Ramos, M., & Lopez-Fandino, R. (2002). Modications in
milk proteins induced by heat treatment and homogenization and their
inuence on susceptibility to proteolysis. International Dairy Journal, 12(8),
679e688.
Guyomarch, F. (2006). Formation of heat-induced protein aggregates in milk as
a means to recover the whey protein fraction in cheese manufacture, and
potential of heat-treating milk at alkaline pH values in order to keep its rennet
coagulation properties. A review. Lait, 86(1), 1e20.
Hkansson, A., Trgrdh, C., & Bergensthl, B. (2009). Dynamic simulation of
emulsion formation in a high pressure homogenizer. Chemical Engineering
Science, 64(12), 2915e2925.
Hayes, M. G., & Kelly, A. L. (2003). High pressure homogenisation of raw whole
bovine milk (a) effects on fat globule size and other properties. Journal of Dairy
Research, 70(3), 297e305.
Horne, D. S. (1999). Formation and structure of acidied milk gels. International
Dairy Journal, 9(3e6), 261e268.
Hyslop, D. B. (1993). Enzyme-induced coagulation of casein micelles - a number of
different kinetic-models. Journal of Dairy Research, 60(4), 517e533.
Hyslop, D. B., & Qvist, K. B. (1996). Application of numerical analysis to a number
of models for chymosin-induced coagulation of casein micelles. Journal of Dairy
Research, 63(2), 223e232.
Iordache, M., & Jelen, P. (2003). High pressure microuidization treatment of heat
denatured whey proteins for improved functionality. Innovative Food Science &
Emerging Technologies, 4(4), 367e376.

de Jong, P., & van der Linden, H. (1998). Polymerization model for prediction of
heat-induced protein denaturation and viscosity changes in milk. Journal of
Agricultural and Food Chemistry, 46(6), 2136e2142.
Kristo, E., Biliaderis, C. G., & Tzanetakis, N. (2003). Modelling of the acidication
process and rheological properties of milk fermented with a yogurt starter
culture using response surface methodology. Food Chemistry, 83(3), 437e446.
Kumar, S., & Ramkrishna, D. (1996). On the solution of population balance equations
by discretization. 1. A xed pivot technique. Chemical Engineering Science, 51(8),
1311e1332.
Lakemond, C. M. M., & van Vliet, T. (2008). Acid skim milk gels: the gelation process
as affected by preheating pH. International Dairy Journal, 18(5), 574e584.
Le Bon, C., Nicolai, T., & Durand, D. (1999). Kinetics of aggregation and gelation of
globular proteins after heat-induced denaturation. Macromolecules, 32(19),
6120e6127.
Lee, S. J. E., & Sherbon, J. W. (2002). Chemical changes in bovine milk fat globule
membrane caused by heat treatment and homogenization of whole milk.
Journal of Dairy Research, 69(4), 555e567.
Leermakers, F. A. M., Atkinson, P. J., Dickinson, E., & Horne, D. S. (1996). Selfconsistent-eld modeling of adsorbed [beta]-casein: effects of pH and ionic
strength on surface coverage and density prole. Journal of Colloid and Interface
Science, 178(2), 681e693.
Lucey, J. A. (2002). Formation and physical properties of milk protein gels. Journal of
Dairy Science, 85(2), 281e294.
Lucey, J. A., & Singh, H. (1997). Formation and physical properties of acid milk gels:
a review. Food Research International, 30(7), 529e542.
Mahmoudi, N., Mehalebi, S., Nicolai, T., Durand, D., & Riaublanc, A. (2007). Lightscattering study of the structure of aggregates and gels formed by heatdenatured whey protein isolate and beta-lactoglobulin at neutral pH. Journal
of Agricultural and Food Chemistry, 55(8), 3104e3111.
Novakovic, P., Petrak, T., Kordic, J., & Slacanac, V. (2000). Application of mathematical models in milk coagulation process during lactic acid fermentation - I.
Relation between enzymatic and acidic milk coagulation. Acta Alimentaria,
29(3), 241e254.
Oldeld, D. J., Singh, H., & Taylor, M. W. (1998). Association of beta-lactoglobulin and
alpha-lactalbumin with the casein micelles in skim milk heated in an ultra-high
temperature plant. International Dairy Journal, 8(9), 765e770.
Oldeld, D. J., Singh, H., Taylor, M. W., & Pearce, K. N. (1998). Kinetics of denaturation and aggregation of whey proteins in skim milk heated in an ultra-high
temperature (UHT) pilot plant. International Dairy Journal, 8(4), 311e318.
Oldeld, D. J., Singh, H., Taylor, M. W., & Pearce, K. N. (2000). Heat-induced
interactions of beta-lactoglobulin and alpha-lactalbumin with the casein
micelle in pH-adjusted skim milk. International Dairy Journal, 10(8), 509e518.
Parris, N., Hollar, C. M., Hsieh, A., & Cockley, K. D. (1997). Thermal stability of whey
protein concentrate mixtures: aggregate formation. Journal of Dairy Science,
80(1), 19e28.
Paulsson, M., & Dejmek, P. (1990). Thermal-denaturation of whey proteins in
mixtures with caseins studied by differential scanning calorimetry. Journal of
Dairy Science, 73(3), 590e600.
Peron, N., Heffernan, S. P., Byrne, E. P., Rioual, F., & Fitzpatrick, J. J. (2007). Characterization of the fragmentation of protein aggregates in suspension subjected to
ow. Chemical Engineering Science, 62(22), 6440e6450.
Perrot, N., Trelea, I. C., Baudrit, C., Trystram, G., & Bourgine, P. (2011). Modelling and
analysis of complex food systems: state of the art and new trends. Trends in
Food Science & Technology, 22(6), 304e314.
Pugnaloni, L. A., Dickinson, E., Ettelaie, R., Mackie, A. R., & Wilde, P. J. (2004).
Competitive adsorption of proteins and low-molecular-weight surfactants:
computer simulation and microscopic imaging. Advances in Colloid and Interface
Science, 107(1), 27e49.
Qi, P. X. (2007). Studies of casein micelle structure: the past and the present. Lait,
87(4e5), 363e383.
Rabe, M., Verdes, D., & Seeger, S. (2011). Understanding protein adsorption
phenomena at solid surfaces. Advances in Colloid and Interface Science, 162(1e2),
87e106.
Raikos, V. (2010). Effect of heat treatment on milk protein functionality at emulsion
interfaces. A review. Food Hydrocolloids, 24(4), 259e265.
Roefs, S., & de Kruif, K. G. (1994). A model for the denaturation and aggregation of
beta-lactoglobulin. European Journal of Biochemistry, 226(3), 883e889.
Scott Blair, G. W., & Burnett, J. (1963). An equation to describe the rate of setting of
blood and milk. Biorheology, 1, 183e191.
Schorsch, C., Carrie, H., Clark, A. H., & Norton, I. T. (2000). Cross-linking casein
micelles by a microbial transglutaminase conditions for formation of transglutaminase-induced gels. International Dairy Journal, 10(8), 519e528.
Sharma, S. K., & Dalgleish, D. G. (1994). Effect of heat treatments on the incorporation of milk serum proteins into the fat globule membrane of homogenized
milk. Journal of Dairy Research, 61(3), 375e384.
Smoluchowski, M. (1917). Versuch einer mathematischen Theorie der Koagulationskinetik kolloider Lsungen. Zeitschrift fr physikalische Chemie, 92,
129e168.
Spiegel, T. (1999). Whey protein aggregation under shear conditions - effects of
lactose and heating temperature on aggregate size and structure. International
Journal of Food Science and Technology, 34(5e6), 523e531.
Spiegel, T., & Huss, M. (2002). Whey protein aggregation under shear conditions effects of pH-value and removal of calcium. International Journal of Food Science
and Technology, 37(5), 559e568.

J. Foucquier et al. / Food Hydrocolloids 27 (2012) 1e13


Tolkach, A., & Kulozik, U. (2007). Reaction kinetic pathway of reversible and
irreversible thermal denaturation of beta-lactoglobulin. Lait, 87(4e5), 301e315.
Verheul, M., Roefs, S., & de Kruif, K. G. (1998). Kinetics of heat-induced aggregation
of beta-lactoglobulin. Journal of Agricultural and Food Chemistry, 46(3),
896e903.
Vetier, N., Desobry-Banon, S., Eleya, M. M. O., & Hardy, J. (1997). Effect of temperature and acidication rate on the fractal dimension of acidied casein aggregates. Journal of Dairy Science, 80(12), 3161e3166.
Walkenstrom, P., Nielsen, M., Windhab, E., & Hermansson, A. M. (1999). Effects of
ow behaviour on the aggregation of whey protein suspensions, pure or mixed
with xanthan. Journal of Food Engineering, 42(1), 15e26.
Walstra, P. (1993). Principles of emulsion formation. Chemical Engineering Science,
48(2), 333e349.

13

Walstra, P., Wouters, J. T. M., & Geurts, T. J. (2006). Dairy science and technology.
Dairy Science and Technology, 782.
de Wit, J. N. (1990). Thermal-stability and functionality of whey proteins. Journal of
Dairy Science, 73(12), 3602e3612.
Wu, H., Xie, J. J., & Morbidelli, M. (2005). Kinetics of cold-set diffusion-limited
aggregations of denatured whey protein isolate colloids. Biomacromolecules,
6(6), 3189e3197.
Ye, A. Q., Singh, H., Oldeld, D. J., & Anema, S. (2004). Kinetics of heat-induced
association of beta-lactoglobulin and alpha-lactalbumin with milk fat
globule membrane in whole milk. International Dairy Journal, 14(5),
389e398.
Zhong, Q., & Daubert, C. R. (2004). Kinetics of rennet casein gelation at different
cooling rates. Journal of Colloid and Interface Science, 279(1), 88e94.

Das könnte Ihnen auch gefallen