Академический Документы
Профессиональный Документы
Культура Документы
Corresponding author. Tel.: +64 9 923 5606; fax: +64 9 373 7463.
E-mail addresses: tajammal.munir@auckland.ac.nz (M.T. Munir), b.young@
auckland.ac.nz (B.R. Young).
Journal of Food Engineering 121 (2014) 8793
Contents lists available at ScienceDirect
Journal of Food Engineering
j our nal homepage: www. el sevi er. com/ l ocat e/ j f oodeng
component library due to fact that food components are mandatory
for food process modeling and simulation. In this work, important
food components for modeling milk were built as hypothetical
components for milk process studies.
One important industrial food process is the milk processing
process, which involves the handling of uids, i.e. milk. For exam-
ple, milk (whole or concentrated) is one of the main raw materials
in milk pasteurization processes, milk powder, and cheese plants.
To develop a milk process simulation, based on a commercial pro-
cess simulator, the milk as a collection of hypothetical components
needs to be developed. For that purpose, data on the properties
(mostly physical properties) of milk (Ruiz et al., 2010; Bisig,
2011; Sindhu and Arora, 2011) is used to predict actual milk
behaviour and properties.
In this work VMGSim was selected as the commercial process
simulator for its simulation capabilities, its ability to incorporate
customized calculations using the spread sheet tool, user friendly
interface and its most recent and updated thermodynamics from
TRC/NIST for the prediction of thermodynamic data (i.e. heat
capacity and thermal conductivity) required for the hypothetical
components simulation. It is one of the latest commercial process
simulators mainly conceived for the chemical and petrochemical
industries (Daz et al., 2011; Satyro et al., 2011; Munir et al.,
2012a,b) and developed by Virtual Materials Group Inc. (VMG)
(Virtual Materials Group Inc., 2012). It is extensively used to design
a new process, troubleshoot an existing process unit or optimize
operations in a process (Saber and Shaw, 2008; Jiang et al., 2011;
Satyro et al., 2011; Motahhari et al., 2012). In addition to oil and
gas, and chemical purposes, it has also been used for biofuel pro-
cess applications (Lee et al., 2011).
The aim of this work was to develop hypothetical components
based on the process simulator to simulate actual milk. For this
purpose the data on the properties of milk needed to develop
hypothetical components were obtained from literature (Bylund,
1995; Bon et al., 2010), and the VMGThermo thermodynamic data-
base in the simulator. This would allow the simulation of milk as a
collection of new components (hypothetical components) in the
simulator, needed to simulate milk processing, and to predict milk
process behaviour closely enough to its real-life operation.
This manuscript is organized as follows. After this general intro-
duction, the materials and methods used in this work are explained
and discussed in Section 2. In Section 3 results are discussed. Final-
ly in Section 4 results are summarized, limitations are discussed,
and conclusions are made.
2. Materials and methods
2.1. Raw material (milk): composition and properties
The typical whole (13 wt.% total solids) and concentrated
(50 wt.% total solids) milk compositions considered in this work
are shown in Table 1 (Bylund, 1995; Bon et al., 2010). A material
stream was built in the simulator in order to develop a pseudo
milk mixture having hypothetical components (represented by a
superscripted asterisk,
). Depending on the component informa-
tion contained in the simulator (e.g. palmitic acid, n hexadeca-
noic acid, oleic acid, sodium chloride (NaCl) and potassium
chloride (KCl) are already present in the simulator component li-
brary) component library, each component of the milk composition
(fat, proteins, lactose and minerals) was further classied into sim-
pler components as shown in Table 2 and Fig. 1. The total solid
components include fat, proteins, lactose and minerals.
The assumptions considered in this work and the further classi-
cation of each component of the milk composition (fat, proteins,
lactose and minerals) into simplercomponents are given as
follows:
2.1.1. Assumptions
The following assumptions were adopted after the following
considerations:
(a) Fat: Milk fat is usually considered as a mix of triglyceride
esters, which are composed of various fatty-acids and glyc-
erol. As each glycerol can bind three fatty-acids and they
are not necessarily the same kind, the number of different
glycerides is extremely large. As a result, fats are usually
characterized by fatty-acids. (Bylund, 1995) In this research
we used fatty-acids instead of fatty-esters in simulating milk
due to the following considerations.
The composition of fatty esters is extremely complex,
none of which can be found in the literature or the VMG-
Sim or other commercial process simulator thermody-
namic databases.
Another compound in the VMGSim database named TRI-
GLY(C18)3 (C
54
H
105
O
6
) and belonging to the oil family
was attempted to be used to represent the total amount
of fat. However, the approach resulted in signicant dif-
ferences in density (1223 kg/m
3
vs. 1022 kg/m
3
), heat
capacity (1055 kJ/kmol K vs. 79 kJ/kmol K), thermal con-
ductivity (0.23 W/m K vs. 0.55 W/m K) and viscosity
(0.123 Pa s vs. 0.00203 Pa s) for the nal pseudo milk.
Fatty acids are the closest match to milk fat and using
these fatty acids, the pseudo milk showed a close match
of the main physical properties to actual milk.
The main drawback is that the chemical properties of
esters and acids are very different. However since this
research only focuses on the physical properties, the dif-
ferences in chemical properties can be ignored, which
need to be considered in fouling reactions and cheese
making.
The main component of milk fatty acid is palmitic acid (25
29 wt.%) and oleic acid (3040 wt.%), which can both be found in
the simulator component list. Other minor fatty acids include bu-
tyric acid, myristic acid, stearic acid, and so on. For the simplica-
tion purpose and their similar weight percentage in total fatty acid
contents, the fraction of each of these two fats was set to half
(50 wt.% palmitic acid and 50 wt.% oleic acid) and all other types
of fat components were ignored (Assumption 1).
(b) Protein: Caseins are the proteins commonly found in
mammalian milk, making up to 80% of the total proteins
in milk while whey proteins make up the rest 20%. Due
to the unavailability of public domain literature on the
physical properties of whey proteins, and considering
their relatively minor weight percentage in the total pro-
teins, the protein of milk is simplied to be casein
(Assumption 2). The average molecular weight of casein
was set to 23,000 kg/kg mole, and the density of casein
was set to be 1250 kg/m
3
(Karlsson et al., 2005; Choi
et al., 2011). A hypothetical component was created in
the simulator based on these attributes to simulate
proteins.
Table 1
Quantitative milk composition (% in mass) (Bylund, 1995; Bon et al., 2010).
Product Water Fat Proteins Lactose Minerals Total
solids
Whole milk 87.0 4.0 3.4 4.8 0.8 13.0
Concentrated
milk
50.0 16.0 13.0 18.0 3.0 50.0
88 Y. Zhang et al. / Journal of Food Engineering 121 (2014) 8793
(c) Minerals: The mineral fraction in milk, which is a small
fraction of milk (0.8 wt.%), contains cations (calcium,
magnesium, sodium and potassium) and anions (inor-
ganic phosphate, citrate and chloride) (Bylund, 1995;
Gaucheron, 2005). Due to the unavailability of calcium
salts in the VMGSim thermodynamic database, the min-
erals in raw milk were specied as 50% NaCl and 50% KCl
for simplicitys sake (Assumption 3). They can both be
found in the simulator component list.
(d) Viscosity: Since oleic acid, palmitic acid, NaCl and KCl
can be found in the component list in VMGSim, their
viscosity value cannot be manually changed. Also, the
lactose as a hypothetical compound in VMGSim is con-
sidered to have a high solubility in water, so it should
hardly affect the total liquid viscosity. Due to the insolu-
bility of proteins and the availability of changing their
viscosity value in VMGSim, they were selected as the
main contributor of pseudo milk viscosity (Assump-
tion 4).
(e) Lactose: The component list of the simulator does not
include lactose, thus another hypothetical component
was created with these attributes: molecular weight
(342.3), normal boiling point (668.9 C), density
(1525 kg/m
3
) (Herrington, 1934; Zadow, 1984), and
other properties estimated from these parameters and
the selected thermodynamic package.
2.2. Process simulation: Components and thermodynamic model
selections
Before setting up and solving the simulation cases, the pseudo
milk mixture was simulated using the assumptions, composition
and databases (databases from literature and the simulator library)
explained in Section 2.1.
The simulator library contained information for the following
components required for milk simulation: water, fat (palmitic acid
and oleic acid), and minerals (NaCl and KCl). Components not
available in the library were specied using the Hypothetical
compound manager tool. Proteins (Proteins
) and Lactose
(Lactose
and Lactose
Lactose
= hypothetical components).
Y. Zhang et al. / Journal of Food Engineering 121 (2014) 8793 89
From Table 3 it can be observed that the raw milk liquid den-
sity, heat capacity and thermal conductivity showed very little dif-
ference between the literature and simulated values. Considering
the small differences in the raw milk density, heat capacity and
thermal conductivity between literature and simulated data, these
properties were not further considered and compared for concen-
trated milk. However, density, heat capacity and thermal conduc-
tivity can vary to some extent with varying total solids in milk.
The liquid milk viscosity showed around a 1719% difference
between the simulation results and literature data as shown in
Table 3. As a result the liquid milk viscosity was further considered
to compare simulation and literature viscosity data for raw and
concentrated milks. The milk viscosity was therefore further
optimized by regressing literature viscosity data using the Model
Regression option in the simulator.
Model regression was used to estimate the parameters required
to calculate the viscosity of the hypothetical components. The sim-
ulators model regression tool uses an optimizer to manipulate the
regressed variables and minimizes the total error of a model based
on the input data.
In this work, model regression was used to modify the viscosity
of the milk mixture stream. The viscosity of the raw milk mixture
stream was modied by manipulating the adjustable viscosity
parameters (A, B, C, and D in Eq. (1)) of the hypothetical compo-
nents. Eq. (1) can be used to regress the viscosity of any hypothet-
ical component and represents the inuence of temperature on the
viscosity of milk (viscositytemperature correlation). This is al-
ready in place in the thermo physical property database. However
it does not consider the effect of total solids on the viscosity of milk
(limitation of Eq. (1)).
lnl A B=T CT DT
2
1
where l is the viscosity (Pa s), T the Temperature (K), A, B, C, and D
is the adjustable viscosity parameters.However the viscosity of milk
is inuenced by the total solid content along with temperature
(Bakshi and Smith, 1984; McCarthy, 2002). Eq. (1) represents only
the inuence of temperature on the viscosity of milk. To re-calibrate
and to consider the combined effect of temperature and total solids
on milk viscosity, a Fernndez-Martn (1972) and Minim et al.
(2002) type viscosity model was employed as shown in Eq. (2). It in-
ter-relates viscosity, temperature and concentration (total solids). It
provides a theoretical viscosity value to compare with the simulator
viscosity results based on Eq. (1).
logg A
0
A
1
t A
2
t
2
B
0
B
1
t B
2
t
2
s
C
0
C
1
t C
2
t
2
s
2
2
where s is the total solids content (% in mass), A
i
, B
i
and C
i
are
dimensionless coefcients calculated by the least squares method,
and are given in (Fernndez-Martn, 1972).
The Fernndez-Martn (1972) and Minim et al. (2002) viscosity
model (Eq. (2)) was used to t the viscosity results calculated in the
simulator using Eq. (1) as explained in Section 3.2.
Total solids content and temperature inuence the viscosity of
milk (Bakshi and Smith, 1984; McCarthy, 2002). It is well known
that of the solids components (fat, proteins, lactose and minerals),
proteins are the main contributor to milk viscosity (Reddy and
Datta, 1994; Bienvenue et al., 2003; Herceg and Lelas, 2005; Karls-
son et al., 2005). Hence, in this work, protein was selected
(assumption 4) as the main contributor (hypothetical compound)
to milk viscosity and was the compound that the viscosity of which
was actually re-calibrated. The re-calibration steps were as
follows:
(a) Obtain whole milk viscosity data from VMGSim, from 1
to 70 C.
(b) Obtain whole milk without protein data from VMGSim,
from 1 to 70 C.
(c) Calculate the weighted natural logarithm difference of
the two sets of data.
(d) Regress the data set obtained from step c with Eq. (1).
(e) Fine-tune the adjustable viscosity parameters A, B, C, and
D in Eq. (1).
3.2. Simulation results validation
Simulation results validation was conducted to determine the
impact of different independent variables on a particular depen-
dent variable and the differences were identied between the liter-
ature and the simulated data under a given set of assumptions.
Considering the small differences in the raw milk density, heat
capacity and thermal conductivity shown between simulation
and literature, these properties were not further compared with
actual milk data. Only the milk viscosity (which showed around
a 1719% difference between simulation and literature) was
further compared with measured milk viscosity data.
In this work, the simulation of the physical properties of
density, heat capacity, and thermal conductivity (q, Cp, and k,
respectively) were validated using the thermo-physical properties
models available in Minim et al. (2002), shown as following
equations.
Cp 3744:48 1:15T 3:93E
3
T
2
; R
2
0:982 3
q 1042:01 0:37T 0:36E
3
T
2
; R
2
0:993 4
k 0:49 2:23E
3
T 1:08E
3
T
2
; R
2
0:991 5
where T is the Temperature (C), and R
2
is the squared residuals.
The simulation of the milk viscosity was validated using viscos-
ity models available in Fernndez-Martn (1972) and Minim et al.
(2002), shown as Eq. (2). Measured milk viscosity data was also
used to see actual differences in milk viscosity data between the
literature, simulated and actual data.
Table 3
Comparison of the physical properties between simulated raw milk and literature data.
Physical properties RML
RMV
Difference (%) WL
WV
RMVR