Journal of Food Engineering 58 (2003) 267275

www.elsevier.com/locate/jfoodeng

The original Plank equation and its use in the development

of food freezing rate predictions

pez-Leiva *, Bengt Hallstro
m
Miguel Lo
Department of Food Engineering, Lund University, P.O. Box 124, 22100 Lund, Sweden
Received 22 March 2002; accepted 21 September 2002

Abstract
A large number of methods to predict freezing and thawing times for foodstus have been proposed. Normally the original Plank
equation is used as the starting point, but since this equation does not include the times below and above the freezing itself, several
attempts have been made to improve it by adding new terms and parameters, to make it suitable for the entire freezing process.
In this paper, Planks equation and how it has been interpreted and modied by dierent scientists during the years is reviewed
and discussed. We assess as well several of these models, by comparing the values they predict with experimental freezing times
available for the same experimental conditions. Two software programs are also included in the analyses: a commercial package and
an own developed software which follows the entire freezing process by a simple geometrical iterative approach.

2003 Elsevier Science Ltd. All rights reserved.
Keywords: Plank equation; Rjutov equation; Food freezing; Freezing rates; Computer model

1. Introduction
The length of time that a freezing process undergoes
can be divided into:
Pre-cooling time: the time it takes for the food to be
cooled down from its initial temperature to its freezing point.
Freezing time: the time it takes for the freezable water
present in the food to freeze.
Sub-cooling time: the time it takes for the food, after
freezing, to reach the nal temperature (normally a nal temperature in the centre equal to )10 C or )18
C).
Not all the boundaries of these three periods are well
dened.
The initial temperature of the food is of course well
known. It becomes more dicult to decide about the
foods nal temperature, since here we have two alternatives: to dene this temperature as the average temperature of the body or as the temperature of the

Corresponding author. Tel.: +46-46-222-9814; fax: +46-46-2224622.

E-mail address: [email protected] (M. L

opez-Leiva).

thermal centre of the food. This latter being recommended by International Institute of Refrigeration (IIR)
and consequently should be preferred. In any case both
temperatures are related to each other.
The temperature at which a food begins to freeze is a
known parameter, but as the freezing proceeds, the
freezing temperature is decreasing because the free water
left in the food is becoming more and more concentrated
in salts with a consequent decrease in its freezing temperature. The rst portion of water freezes at the foods
initial freezing temperature, but the last portions will do
it at a much lower temperature. Due to this it becomes
dicult to dene the end of the freezing step. To cope
with this problem, Plank considered that all the freezing
period happened at constant temperature (initial
freezing temperature).
It is considered that the freezing of the food is ended
when its centre temperature reaches )10 C.
2. The Plank equation
In the rst paper Plank (1913) presents a formula to
calculate the freezing time for a block of ice. Dierent
geometric shapes are considered: cylinder, quadratic and
rectangular rods as well as slabs. In a second paper
(Plank, 1941) a similar calculation method is used for

0260-8774/03/$ - see front matter
2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0260-8774(02)00385-0

268

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

Nomenclature
k=CpS q foods thermal diusivity (frozen
state) (m2 /s)
a; b
widest and shortest dimensions in a nonregular cylindrical shape (m)
b0
width of a brick-shaped object (m)
CpL
heat capacity (unfrozen state) (J/kg C)
CpS
heat capacity (frozen state) (J/kg C)
d
sample thickness (h0 ) (m)
D
parameter used in Cowells equation ()
De
equivalent diameter for sh shape (m)
F
freezing time according to Plank (s)
Feff
total freezing time (eective freezing time)
(s)
Fsubcooling subcooling time (s)
Fprecooling precooling time (s)
f0
heat transfer area of the body (m2 )
G
parameter used in Cowells equation
h
heat transfer coecient (W/m2 C)
h0
shortest dimension of a body (m)
l0
length of a brick-shaped object (m)
n
correction factor, depends on the Biot number, varies between 1 and 1.21
P; R
parameters dening the geometry of the food
()
Pm; Rm modied Planks parameters ()
Q
total heat to be removed during the freezing
step (J)
a

t
Te
Tf
Ti
Ta
U

V0
DH

DHPR
DHf
b1 ; b2
k
q
Pk
Fo
Bi
Ste
Ko

time (s)
nal temperature (C)
initial freezing point (C)
initial temperature of the food (C)
temperature of the freezing medium (C)
parameter in Hung and Thompsons approach dening a new temperature gradient
()
volume of the food (m3 )
enthalpy change between initial freezing temperature (Tf ) and nal centre temperature
()10 C)
enthalpy change during the whole process
(J/m3 )
enthalpy change during freezing step (J/kg)
geometrical parameters dening a brickshaped object ()
thermal conductivity of the frozen food
(W/m C)
food density (kg/m3 )
CpL Ti  Tf =DH Plank number
Fourier number at=d 2
Biot number hd=k
Stefan number CpS Tf  Ta =DH
Kossovitch number 1=Ste

food products. In this latter paper the following approximations are made:

Planks theoretical analysis concludes that P and R

are functions of the geometry of the body. P becomes

the entire foodstu has freezing temperature right

from start
heat transfer through the foodstu is by thermal conduction
physical properties are independent of temperature
volume change is neglected

V0
f0 h0

where V0 is the total volume of the body, f0 is the area of

the body through which heat is being removed, h0 is the
thickness of the body.
The relationship for R becomes more complicated,
and will not be given here.
Eq. (2) applies for any type of regular geometry
(sphere, cylinder, cube, slab, brick,. . .).
For the specic case of a brick-shaped geometry, P
can also be expressed as function of the two geometric
parameters b1 and b2

For the general case, and starting from basic heat

transfer equations (Fouriers equation for heat transfer
by conduction and the Prandtl concept of boundary
layer for heat transfer by convection), Plank derived the
following equation:


q DHf
d
d2
F
P R
1
h
Tf  Ta
k

It is worthwhile to note that this is a theoretically derived equation where both parameters P and R appear as
result of the mathematical derivation; they are not empirical entities added with the purpose of tting experimental results.

where l0 is the length of the brick, b0 is its width and h0

its thickness.
From the original equations given by Plank, the following values for P and R can be calculated (Table 1).

b1 b2
2b1 b2 b1 b2

b1

l0
b0
b
h0 2 h0

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

Table 1
Values of Planks parameters P and R for some regular geometries
Geometry

Heat transfer

Sphere
Cube
Cylinder
(length radius)
Cylinder
(length radius)
Slab

1=6
1=6
1=6

1=24
1=24
1=24

Radial
From all sides
From all sides

1=4

1=16

Radial

1=2

1=8

Two sides

As the calculations were quite lengthy, especially with

regard to the calculation tools that were available in
those days, some authors tried to simplify the procedure
developed by Plank, in most cases proposing graphical
solutions for the calculation of P and R.
3. D.G. Rjutov
Already in 1936, Rjutov presented an equation,
which can be considered to be the predecessor of Planks
equation:


DH
V0 1 d
F

4
Tf  Ta f0 h 4k
We see that this expression is equal to Planks equation
for the special case P =R 4 (regular geometries, e.g.,
sphere, cube, cylinder, innite slab).
Using the denition for Biot number, this equation
becomes


Q
Bi
1
F
5
hf0 Tf  Ta
4
For the prediction of the pre-cooling time, running experiments with slabs of meat, Rjutov found the following correlation
tTi t1 C0:0053Ti

In this equation: Ti is the initial food temperature (C),

tTi is freezing time when initial temperature of
meat Ti (C), t1 C freezing time when initial
temperature 1 C (can be equalled to freezing time
according to Plank).
Plank (1941), rearranged this equation to the following expression:
Fpre-cooling F  0:0053Ti  Tf

The constant 0.0053 was experimentally obtained by

Rjutov for a meat block, 68 mm in thickness.
For the subcooling time, Rjutov makes a theoretical
analysis, based on the development of the temperature
prole by Fourier series, of a slab being frozen from two
opposite sides, arriving to the following expression:



0:2333n
4k
Tf  Ta
Fsubcooling
d d
ln
 0:0913
a
h
Te  Ta
8

269

n is a correcting factor, function of Biot number, and

which is normally very near to 1 (Plank, 1941). A simple
curve tting gives n 1:14Bi0:05
From Eqs. (7) and (8), the total freezing time becomes


 2
d
1
4
Feff F 1 RjTi  Tf  0:266
1
Bi
a Bi0:05


Tf  Ta
 ln
 0:0913
9
Te  Ta
Rj is Rjutovs coecient which he found that for meat
was equal to 0.0053.
By tting this equation to experimental values for
other products we can widen the scope of this equation.
For instance, for Tylose (a 23% methylcellulose gel, also
known as Karlsruhe test substance), Rj is equal to
0.026 (data from Hung & Thompson (1983) and Pham
& Willix (1990)). The same value is obtained with four
foods (lean beef, mashed potatoes, carp and ground
beef, data from Hung & Thompson (1983)).
For these ve products the dierence between predicted and experimental time, is, as average, 0.5% with a
standard deviation of 6.9%.
There is a large dierence between this value of Rj
(0.026) and the one found by Rjutov (0.0053 for meat),
but we have tted experimental values considering the
whole freezing process, while Rjutovs constant was
obtained from pre-cooling experiments.
4. J. Nagaoka
Working with sh, Nagaoka, Takaji, and Hohani
(1955), use Rjutovs equation (Eq. (6)) together with
Planks equation (Eq. (1)) for the calculation of the
pre-cooling freezing times. They use the original
Plank equation with the values of P and R for cylinders,
plus the denition of an equivalent diameter De, to account for the non-cylindrical transversal shape of the
sh. This expression is then multiplied by a factor similar to the one found by Rjutov for meat, with a factor
0.008 instead of 0.0053.
 2

q DHf
De
De
Fsub-cooling freezing 1 0:008Ti

Tf  Ta 16k 4a
2ab
10
De
ab
a and b are the widest and shortest dimensions of the
sh transversal section.
5. F.L. Levy
Levy (1958), based in the work of Nagaoka, developed the following equation for the calculation of
the total freezing time (Feff ). Note that a factor 1
0:008Ti  Tf is used instead of 1 0:008Ti , which is in
accordance with what Plank proposed (Eq. (7)).

270

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

In this equation, the enthalpy change during the

freezing step of Planks original equation (DHf in
Eq. (1)) is replaced by the total enthalpy change (precooling freezing sub-cooling). A textbook (Andersen & Risum, 1989), proposes the use of this equation
for any kind of food.


DHPR
Feff F
11
1 0:008Ti  Tf 
DHf
6. A.J. Ede
Ede (1949) developed a graph which is probably the
rst attempt to simplify the use of Planks equation. He
plotted the parameters R and P of Eq. (1) as function of
b1 and b2 , (Fig. 1). In spite of its limitations, this graph
has been widely used in the technical literature during
the last 50 years (Andersen & Risum, 1989; Brennan,
Butters, Cowell, & Lilly, 1976; Heldman & Lund, 1992;
zilgen, 1998). The graph is limited to values of b and
O
1
b2 equal or less than 10 and is not very easy to use.
One interesting point concerning Edes article (Ede,
1949) is that the mention of this graph takes only one
line of his article, showing that he himself did not consider this graph as being the most important part of the
paper. In eect the article is mainly devoted to the development of a geometrical, iterative method to calculate freezing times for unidirectional heat ow. The
method is based on Schmidts work (1953) about geometrical evolution of temperature proles.
Ede (1949), uses this tool to calculate some examples
and compare the results with those obtained using
Planks formula. He concluded that the dierences were
so small that it was of little value to use a complicated
geometrical analysis instead of the compact and easy-to-

use formula developed by Plank. We are later using

Edes method when comparing dierent calculation approaches. Evidently Ede should be better recognised for
this contribution to the calculation of freezing times
than for the graphical determination of P and R.

7. M. Backstrom
Backstr
om (1970) presented a diagram intended to
simplify the calculations of freezing times. The diagram
is essentially based on Planks equation but is restricted
to 25 C temperature dierence between freezing point
and freezant temperature and to values of the physical
properties normal for foodstus. The author proposed
the following equation for the calculation of the total
freezing time:


Ti  Ta
Feff F 1 0:0017Tf  Ta ln
12
Te  Ta
where F is the time according to Plank as obtained
from the diagram (Backstr
om, 1970).
The original diagram is in a way very complete, it
considers both freezing from all 6 sides of the body as
well as from two sides. The author also discusses appropriate values for freezing velocities for dierent types
of food products, a discussion which was of great interest at that time. The Backstr
om diagram, in spite of
its completeness was never widely used.

8. N.D. Cowell
Cowell (1969) uses Planks equation in the following
dimensionless form:


Fo
1
D
G
13
Ko
Bi
Here Ko is the Kossovitch number ( 1=Ste) and the
parameters D and G relate to Planks original R and P
parameters by:
G R=P
DP

Fig. 1. Chart giving P and R in Planks equation, by Ede.

Cowell (1969) plots G as a function of b1 and b2 . This

plot adds nothing to the one that Ede had already presented; even more, from Cowells graph it is possible to
obtain only one parameter (G R=P ), while Edes plot
gives both P and R separately. But to Cowells credit we
must say that he realised that in reality one does not
need to obtain from a plot the value of P (or D in his
case) since this value can much more easily and accurately be obtained from its denition (Eq. (2)).
Both Edes and Cowells plots have the disadvantage
that they are limited for values of b1 and b2 smaller than
10.

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

271

Table 2
Modied P and R values for innite slabs, obtained by Cleland and Earle and by Hung and Thompson

Pm
Rm

Cleland and Earle

Hung and Thompson

0:513 0:290Pk Ste 0:115Pk 0:0091=Bi 0:0525

0:15 0:426Pk 0:0917Ste

0:7306  1:083Pk Ste 15:4U  15:43 0:01329 Ste=Bi

0:2079  0:2656U Ste

9. A.C. Cleland, R.L. Earle, Y.C. Hung and D.R.

Thompson
Cleland and Earle (1976, 1977, 1979a,b) and Hung
and Thompson (1983) used a very similar approach to
the problem. Using Planks equation in dimensionless
form as a basis (Eq. (14)), they make some changes to it
to account for the precooling and subcooling periods.
The enthalpy, for instance, is now the whole enthalpy
change between the initial and nal temperatures.
Both research groups developed relations to dene
new values of P and R (Pm, Rm) as functions of Bi, Ste
and Pk numbers. In addition, Hung and Thompson
(1983), dened a new temperature gradient (DT , expressed in the equations via the parameter U ) valid for
the complete process from initial to nal temperature,
while Cleland and Earle (1976) used a coecient
(EHTD number of equivalent heat transfer dimensions) to account for the shape of the food. 1 With all
these changes made to the shape of the original Plank
equation, empirical equations were then obtained by
multiple linear regression of numerous experimental
results with a model food (Tylose). The nal expressions
(for slabs) are presented in Table 2.
Fo Pm

1
1
Rm
Bi Ste
Ste

14

Here Fo is the Fourier number, Bi is Biot number, Ste is

Stefan number, Pm and Rm are modied shape parameters and to account for the pre-cooling period, the
Plank number is dened.
Pk

CpL Ti  Tf
DH

15

Due to the many steps necessary to perform the calculations, Cleland and Earle presented three graphs from
where the values of Pm, Rm and EHTD can be obtained.
Hung and Thompson applied their own equation to
the calculation of freezing times for some real foods and
report that the dierence between the experimental and
predicted values varied between )4.5 and 6.14%.
The method of Cleland and Earle predicts their own
experimental values within )8.7 and 8.4%.

10. Q.T. Pham

Pham (1984) starts with what is considered to be a
modication of the Plank equation, but which in reality
was already derived by Rjutov (1936) (Eq. (5)).


Q
BiS
1
F
16
hf0 Tf  Ta
4
Here Q is the total heat to be removed during the
freezing step V0 q DHf , Bis hd=k.
Using the F -expression given above, Pham introduces
similar expressions for the precooling and subcooling
periods, giving the following set of equations:


Qpre
Bi1
Fprecooling
1
hf0 DTm1
6


Q
BiS
1
F
17
hf0 Tf  Ta
4


Qsub
Bi3
Fsubcooling
1
hf0 DTm3
6
DTm1 and DTm3 are the logarithmic mean temperature
dierences during the precooling and subcooling periods. The Biot values are dened as follows
Bi1 BiS BiL =2 BiL referring to unfrozen state
Bi3 BiS
The number 4 dividing Biot number in the freezing step
is in accordance with the Plank equation (Eq. 1) for the
special case R 1=4 (or Rjutovs equation (5)). The
number 6 in the two other equations is found by comparison with a theoretical cooling curve.
In a later paper Pham (1986) introduces some approximations resulting in a simplied version of the
original equation



V0 DH1 DH2
BiS
Feff

1
18
hf0 DT1 DT2
4
where DH1 CpL Ti  Tfm , DH2 DHf CpS Tfm  Te ,
DT1 Ti  Tfm =2  Ta , DT2 Tfm  Ta and Tfm is a
mean freezing temperature which Pham obtains by
curve-tting diverse experimental data (Cleland & Earle,
1977, 1979a,b; Hung & Thompson, 1983), equal to
Tfm 1:8 0:263Tc 0:105Ta

1
To determine the value of EHTD, the additional calculation of
two other parameters, W 1 and W 2 is needed.

temperatures in C

It should be observed, as well, that Pham uses dierent

denitions for the characteristic length d to be used with
the Biot number, depending on the shape of the body;
for instance for brick geometry:

272

d 1:46h0 b0

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

1=2

where h0 is the shortest side and b0 the second shortest

side of the brick.
We see that even this simplied procedure produces a
series of formulas not very friendly to work with.

11. This work

Looking at the values of the parameters P and R given in the technical literature (Plank, 1941), for the well
dened shapes slab, innite cylinder and sphere, one
realises that the quotient P =R is constant and equal to 4
(see Table 1). We see as well, that this value holds even
for other geometric shapes like cube, innite square rod,
and cylinder with length equal to diameter. To see what
happens with P =R for other, non-singular geometries,
we have calculated P =R as a function of P and plotted
the results in Fig. 2. The original equations dening P
and R have been used in the calculations (Plank, 1941).
We see that P =R varies between 4 (for the singularities) and 3.5 (for P 0:25 and b1 2).
As an immediate conclusion we can suggest that for
practical purpose, there is really no need of using a
graph at all (like Fig. 1), since an average P =R value of
let us say 3.7 can be adopted for all the cases outside
singularities (in which cases we know that P =R 4),
with an error of no more than around 3%, well inside the
error we inevitably make when calculating the thermal
parameters used in the equation. In this way the procedure becomes much simpler: P is calculated from the
geometry of the body (Eq. (2)) and R is equal to P =3:7.
Both Ede (1949) and Cowell (1969), saw that values
P =R (or other similar) where not constant in all the
cases, and they explained this saying that this was due to
the simplications made by Plank. In reality these uctuations are inherent to the theoretical treatment and
have nothing to do with the assumptions made in the
mathematical treatment.

Replacing R P =3:7 and using the denition of P

( V =Ad), the Plank equation becomes


q DHf d
Bi
P 1
19
F
hTf  Ta
3:7
which is simply equal to Planks equation (or Rjutovs
equation) where the factor 4 is replaced by 3.7.
A change in the value of this factor from 4 to 3.7
produces a dierence of not more than 0. 6% in the value
of F . Consequently this equation can be used for any
kind of geometry, without loosing accuracy.
The comparison of this simplication with the original Planks equation (Eq. (1)) for 45 cases with real
foods presented by Hung and Thompson (1983), gives a
dierence of between )1.6 and 3.3%, with an average
dierence of 0.1%.
Eq. (19) predicts only the freezing time (freezing
time according to Plank). To predict the complete
process time one needs to use this equation together
with an expression that takes into account the precooling and subcooling periods. Here we have tested two
equations:
(a) The expression derived by Rjutov (Eq. (9)).
 2
d
Feff F 1 RjTi  Tf  0:266
a



1
4
Tf  Ta
 0:05 1
ln
 0:0913
Bi
Bi
Te  Ta
(b) The expression of Levy (1958), which is based on the
work of Rjutov, Plank and Nagaoka (Eq. (11)).


DHPR
Feff F
1 0:008Ti  Tf 
DHf

12. Comparison among the dierent prediction approaches

In this section, freezing times obtained using some of
the more relevant prediction methods are compare with
each other and with existing experimental values.
First we use the results obtained by three research
groups that have used the same model food (Tylose),
and compare their experimental values with those obtained using ve dierent predicting approaches.
In a similar way we will compare published results for
four real foods (lean beef, mashed potatoes, sh and
ground beef, Hung & Thompson, 1983) with those calculated from nine dierent approaches.

13. Experiments with slabs of Tylose

Fig. 2. Planks equation. Generalised diagram P =R vs P , calculated
from the original work of Plank.

Three research groups have published experimental

results for the freezing of slabs of Tylose (Cleland &

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

Earle (1976), 43 experiments, Hung & Thompson (1983),

23 experiments and Pham & Willix (1990), 32 experiments). We use the models developed by these researchers to predict each others results. We also
compare these experimental results with the values calculated using two other approaches: the equation presented by Levy (1958) and the approach developed by
Rjutov, and which Plank himself presented in his paper
from 1941 (Eq. (9)).
The results are presented in Fig. 3. Two parameters
are used for the comparison: the percentage dierence
between predicted and calculated values, and the standard deviation of these dierences.
We arrive at two conclusions: the obvious one that
each of the three models represents better their own
experimental values (of course, they have built the

273

Fig. 4. Numerical example no. 1 (lean beef) solved using Edes geometrical approach. Temperature in the centre vs. time.

models to better t these experimental values). The

second conclusion is that the other experimental results
are not equally well predicted. This means that even
though the experimental substance is the same, the other
variables are not equally well controlled in the experiments; and here the most probable source of error is the
value of h, the heat transfer coecient used in the calculations. This quantity can not be measured directly
during the experiment, but instead values from tables,
empirical formulas or previous experiments under similar conditions are used, and in this procedure errors may
be introduced (Fig. 4).

14. Numerical examples with four real foods

Fig. 3. Freezing of Tylose slabs. Comparison among ve prediction

methods. 1: Hung and Thompson, 2: Levy, 3: Plank and Rjutov, 4:
Pham 5: Cleland and Earle. () average of the percentual dierence
between predicted and experimental values (%); ( ) standard deviation of the average (%).

In the following we calculate the total freezing time

(down to )18 C), using some of the approaches reviewed here, and compare the results with experimental
values. The examples are experiments with four real
foods: lean beef, mashed potatoes, a sh (carp) and
ground beef (Hung & Thompson, 1983).
We have also calculated these examples using two
software packages: the calculation package Food
Product Modeller (FPM), developed at the Meat Industry Research Institute of New Zealand (MIRINZ),
and an own-developed software that makes use of the
geometrical approach suggested by Ede (1949).
The experimental conditions are summarised in Table
3 and the results are presented in Table 4.The best result
is given by the method by Hung and Thompson, but
since the comparison is made against their own experimental results, this accordance might be biased. There
are four other methods that perform equally well: The
approach by Levy (method 3, Eq. (11)), Plank and
Rjutov (method 4, Eq. (9)), and Phams method
(method 7, Eq. (17)). The methods by Cleland and Earle
(method 5) and by Backstr
om (method 6), give very

274

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

Table 3
Experimental conditions in the comparison of experimental and predicted total freezing times
Foodstu
Heat transfer area (cm2 )
Thickness d (mm)
Initial temperature Ti (C)
Freezing point Tf (C)
Freezants temperature Ta (C)
Heat transfer coecient h (W/m2 C)
Thermal conductivity of liquid kL (W/m C)
Thermal conductivity of frozen kS (W/m C)
Specic heat of liquid CpL (J/m3 C)
Specic heat of frozen CpS (J/m3 C)
Latent heat q DH (J/m3 )
Thermal diusivity of liquid aL (mm2 /s)
Thermal diusivity of frozen aS (mm2 /s)

Lean beef

Mashed potato

Fish (carp)

Ground beef

400
14
29.8
)1
)25.8
68
0.51
1.55
3.65E6
1.90E6
2.09E8
0.140
0.816

400
28
17.5
)0.6
)20.2
67
0.53
1.90
3.66E6
1.95E6
2.35E8
0.145
0.974

400
46
18.2
)0.8
)23.9
105.5
0.70
1.73
3.55E6
2.23E6
2.18E8
0.197
0.776

400
31
17.8
)1.2
)20.3
67
0.40
1.62
3.38E6
2.20E6
1.88E8
0.118
0.736

Table 4
Total freezing time for four real foods: comparison between experimental and predicted values using nine dierent methodsResults
Food
Lean beef
Predicted
time (min)

Dierence with
experimental
time (%)

Mashed potatoes

Fish (carp)

Predicted
time (min)

Predicted
time (min)

Dierence with
experimental
time (%)

Experimental valuesa

34.8

88.5

Method
Hung and Thompsona
Planks equationb
Levyc
Plank and Rjutovb
Cleland and Earled
B
ackstr
ome
Phamf
FPM softwareg
Edes geometrical
approachh

35.2
16.9
35.7
34.9
23.5
18.6
28.8
29.5
34.1

1.0
)51.1
2.5
0.14
)32.3
)46.7
)17.3
)15.2
)2.0

88.3
52.6
86.0
95.2
65.1
58.8
88.6
93.5
98.3

)0.2
)40.6
)2.8
7.6
)26.4
)33.5
0.1
5.6
11.1

Ground beef
Dierence with
experimental
time (%)

103
100.7
60.1
102.9
106.1
82.2
66.2
96.8
94
94.6

Predicted
time
(min)

Dierence with
experimental
time (%)

91.3
)2.2
)41.7
)0.1
3.0
)20.2
)35.7
)6.0
)8.7
)8.2

91.2
59.7
95.2
95.8
80.3
80.6
95.4
97.5
102.9

)0.1
34.6
4.3
4.9
)12.0
)11.7
4.5
6.8
12.7

Hung and Thompson (1983).

Plank (1941).
c
Levy (1958).
d
Cleland and Earle (1982).
e
B
ackstr
om (1970).
f
Pham (1984).
g
Food Product Modeller (1993).
h
Ede (1949).
b

unsatisfactory results with deviations as large as 30

40%.
It is remarkable to see that Edes geometrical approach gives quite satisfactory values (Fig. 4), sometimes with results that are even better than those from
the commercial software package FPM.

15. Conclusions
A survey about some of the analytical methods developed to calculate freezing times for foodstus has

been done, starting from the original Plank equation

from 1913 to 1941. A modied version of Planks equation is presented, simplifying the calculation process
enormously. Using this quick method, the freezing time
according to Plank is predicted within
3% compared to
the values obtained from the original Plank equation.
Several approaches to predict total freezing times (that
also take into account both pre-cooling and sub-cooling
periods) have been reviewed. In this case we do not come
to any clear conclusion when comparing the results obtained when using these prediction methods with existing
experimental values. There are at least four analytical

M. L
opez-Leiva, B. Hallstrom / Journal of Food Engineering 58 (2003) 267275

methods that compare well (Pham, Hung and Thompson, Levy, Plank and Rjutov). Levys method is the
easiest to use, while the Pham and Hung and Thompson
methods are the most complicated, requiring elaborated
long calculations, with several intermediate parameters.
The analytical modelling of a food freezing process is
a very complex task, due mainly to the change of phase
that is involved. Besides the large dierence in the values
of the foods density, thermal conductivity and heat
capacity between frozen and non-frozen states, the use
of an analytical solution is always inuenced by the lack
of reliability of the values used for these thermal parameters, and also by the uncertainty in measuring h, the
heat transfer coecient between food and freezant, since
it will very much depend on several geometrical characteristics of the specic equipment: racks form and
position, air prole, position of the food pieces in relation to each other, etc.
Two other factors that may further inuence the prediction of freezing times, and which have not been taken
up in the reviewed works, are the swelling and water
losses that foodstus are subjected to during freezing.
The conclusion is therefore that there is little gain in
accuracy by using complex analytical solutions. A proof
for this is the fact that in the industrial world the prediction of the size of industrial freezers is generally made
based on past experience (and actual experiments) and
not on theoretical methods.
Summarising the work done in this area, we want to
stress the following:
In spite of the large amount of research done in this
area during the last half century, the equations rst
developed by Plank and Rjutov have not lost its original relevance.
The work by Rjutov has not received the credit it deserves. Plank himself made large use of it in his own
work.
Ede should be better recognised for his work on the
iterative geometrical solution to the problem than
on the graphical way of calculating P and R.
In the rst stages of the design of a freezing process,
the use of the simplied equations presented here give
satisfactory results.

References
Andersen, P. E., & Risum, J. (1989). Levnedsmiddelteknologien, Bind 1
bis-Konserveringberegninger. Food Engineering, 1st part: Food
Preservation Methods (4th ed.), Polyteknisk Forlag, Copenhagen.

275

Brennan, J. G., Butters, J. R., Cowell, N. D., & Lilly, A. E. V. (1976).

Food engineering operations (2nd ed.). London: Applied Science
Publishers Ltd.
Backstr
om, M. (1970). Kylteknikern (The refrigeration technician) (3rd
ed.). Stockholm: Almqvist & Wiksell.
Cleland, A. C., & Earle, R. L. (1976). A comparison of freezing
calculations, including modication to take into account initial
superheat. Bulletin de lInstitut internationale du froid, Annexe 1,
369376.
Cleland, A. C., & Earle, R. L. (1977). A comparison of analytical and
numerical methods for predicting the freezing times of foods.
Journal of Food Science, 42, 13901395.
Cleland, A. C., & Earle, R. L. (1979a). A comparison of methods for
predicting the freezing times of cylindrical and spherical foodstus.
Journal of Food Science, 44, 958963.
Cleland, A. C., & Earle, R. L. (1979b). Prediction of freezing times for
foods in rectangular packages. Journal of Food Science, 44, 964
970.
Cleland, A. C., & Earle, R. L. (1982). Freezing time prediction for
foodsa simplied procedure. International Journal of Refrigeration, 5(3), 134140.
Cowell, N. D. (1969). The calculation of food freezing times.
Proceedings of the 12th International Congress on Refrigeration, 2,
667676.
Ede, A. J. (1949). The calculation of the rate of freezing and thawing of
foodstus. Modern Refrigeration, 17, 5255.
Food Product Modeller (1993). Meat Industry Research Institute of
New Zealand, version 1.01.868.
Heldman, D. R, & Lund, D. B. (Eds.). (1992). Handbook of food
engineering. Marcel Dekker.
Hung, Y. C., & Thompson, D. R. (1983). Freezing time prediction for
slab shape foodstus by an improved analytical method. Journal of
Food Science, 48(2), 555560.
Levy, F. L. (1958). Calculating the freezing time of sh in airblast
freezers. Journal of Refrigeration, 1, 5558.
Nagaoka, J., Takaji, S., & Hohani, S. (1955). Experiments on sh
freezing in air blast freezers. Proceedings of the Ninth International
Congress on Refrigeration, Paris, paper 4.321.
zilgen, M. (1998). Food process modelling and control. Chemical
O
engineering applications. Gordon and Breach Science Publishers.
Pham, Q. T. (1984). Extension to Plancks (sic) equation for predicting
freezing times of foodstus of simple geometrical shapes. International Journal of Refrigeration, 7(6), 377383.
Pham, Q. T. (1986). Simplied equation for predicting the freezing
time of foodstus. Journal of Food Technology, 21, 209219.
Pham, Q. T., & Willix, J. (1990). Eect of Biot number and freezing
rate on accuracy of some food freezing time prediction methods.
Journal of Food Science, 55(5), 14291434.
Plank, R. (1913). Die Gefrierdauer von Eisbl
ocken (Freezing times
of ice blocks). Zeitschrift fur die gesamte Kalte-Industrie, XX(6),
109114.
Plank, R. (1941). Beitrage zur Berechnung und Bewertung der
Gefriergeschwindigkeit von Lebesnmitteln (Calculation and validation of freezing velocities in foods). Beihefte zur Zeitschrift fur
die gesamte Kalte-Industrie, 3(10), 22pp.
Rjutov, D. G. (1936). Die Anwendung k
unstlicher Kalte in der
UdSSR. Berichte fur den 7. Intern. Kalte-Kongress. Moskau und
Leningrad: Verlag der Nahrungsmittel-Industrie 1936, S. 69. (cited
by Plank, 1941).
Schmidt, E. (1953). Einfuhrung in die Technische Thermodynamik.
Berlin: Springer Verlag.