Você está na página 1de 29

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/223591327

Review of thermal conductivity models for


foods

Article in International Journal of Refrigeration · September 2006


DOI: 10.1016/j.ijrefrig.2006.03.016

CITATIONS READS

62 103

1 author:

James Carson
The University of Waikato
64 PUBLICATIONS 1,396 CITATIONS

SEE PROFILE

All content following this page was uploaded by James Carson on 13 November 2017.

The user has requested enhancement of the downloaded file.


Review of effective thermal conductivity models for foods

James K. Carson*

*Department of Materials and Process Engineering, University of Waikato, Private Bag 3105, Hamilton 2020,

New Zealand, j.carson@waikato.ac.nz

Tel: +64 7 8384206, Fax: +64 7 838 4835

Abstract

The literature associated with modelling and predicting the thermal conductivities of food products has

been reviewed. The uncertainty involved in thermal conductivity prediction increases as the

differences between the food components’ thermal conductivities increase, which means that there is

greater uncertainty involved with predicting the thermal conductivity of foods which are porous and/or

frozen, than with unfrozen, non-porous foods. For unfrozen, non-porous foods, a number of simple

effective thermal conductivity models that are functions only of the components’ thermal

conductivities and volume fractions may be used to provide predictions to within ±10%. For frozen

and/or porous foods, the prediction procedure is more complicated, and usually requires the prediction

of porosity and/or ice fraction, which introduces another source of error. The effective thermal

conductivity model for these foods may require an extra parameter (in addition to the components’

thermal conductivities and volume fractions) whose value must often be determined empirically.

Recommendations for selecting models for different classes of foods are provided. There is scope for

more research to be done in this area.

Keywords: effective thermal conductivity models, thermophysical properties

1. Introduction

Historically, the design of thermal processing units for food products (e.g. refrigerators, ovens, dryers,

etc.) has largely been based on experience, but in more recent times analytical approaches have been

employed, in particular the implementation of design methods that have been used by chemical

processing engineers [1]. These methods are largely dependent on mathematical models that are

derived from the physical laws that govern the process (e.g. mass and energy balances, reaction

kinetics, thermodynamics, etc). The accuracy of any model of a thermal process is limited by, amongst

1/28
other factors, the accuracy of data for physical properties [2], which may vary during the thermal

process. One of the most influential physical properties in thermal processing is thermal conductivity.

Thermal conductivity measurement is a relatively complex task, and there are many potential sources

of error [24]. Measurement methods for food products have been reviewed by Rahman [4], Murakami

and Okos [14] and Nesvadba [24]. The line-source method (thermal conductivity probe) appears to be

the most the most widely used method for food products, although the guarded hot-plate, Fitch

apparatus, and several comparative methods have also been employed. A probable explanation for the

popularity of the thermal conductivity probe is its relative simplicity coupled with relatively short

measurement times; however, since the measurements are localised, it will not be suitable for some

foods which have highly non-uniform distribution of phases. The food to be measured needs to be

considered when a measurement device is selected, since not all methods are suitable for all foods.

Databases of thermal conductivities of fresh and minimally processed foods such as fruits, vegetables,

grains, cereals, meat and dairy products may be found in the literature [3–8]. The data are sometimes

compiled as empirical correlations of temperature and/or composition, and a computer program called

COSTHERM has been developed which has condensed much of this data into a predictive thermal

properties database [9]. However, it would be impossible to collate databases with measured properties

of every single food product, especially since new food products are continually being developed. In

the absence of measured physical property data, the best estimate of thermal conductivity may be

obtained from effective thermal conductivity models.

2. Thermal Conductivity Models

The literature contains a large number of models for predicting the thermal conductivities of composite

or heterogeneous materials based on composition. Some reviews have provided lists of such models

[10–18], although none of these is by any means exhaustive. Many of the models that have been

proposed are highly specific to a particular material and contain material-specific parameters. Other

models have more general applicability, but may still contain parameters whose values must be

determined empirically. Several researchers have proposed generic models by deriving a set of

equations, usually based on a conceptual ‘parent’ model that is modified to account for variations in

2/28
composition and structure [19–23], although many of these still include empirical parameters.

However, since new models continue to appear in the literature, it seems that, to-date, no single model

or prediction procedure has been found with universal applicability.

As with most modelling exercises, the prediction of effective thermal conductivity usually involves a

trade-off between simplicity/convenience and accuracy. Due to the inherent biological variation of

food products it is highly unusual to find measured thermal conductivity data having reported

uncertainties of less than ±2%, with ± 3–5% being typical figures. Hence, it is unreasonable to expect

the accuracy of predicted thermal conductivities to be better than ±5%. For design purposes, accuracies

to within ± 10% are usually sufficient for thermal conductivity data [24], which, depending on the food

in question, can often be achieved with relatively simple thermal conductivity models.

Table 1 shows the thermal conductivities of the major food components at 0°C (correlations for the

temperature dependencies of these components may be found in [6] and [71]). Other than ice and air,

the thermal conductivities of food components are of similar magnitude. Several studies have shown

that the uncertainty involved in the prediction of thermal conductivity increases as the difference

between the conductivities of the components increases [25,26]. The difference between components’

thermal conductivities is therefore a useful method of classifying materials from an effective thermal

conductivity perspective. Carson et al. [26] proposed the following classifications for predicting the

effective thermal conductivity of food products:

I. Unfrozen, non-porous foods (kwater/ksolids ≈ 3)

II. Frozen, non-porous foods (kice/ksolids ≈ 12)

III. Unfrozen, porous foods (kwater/kgas ≈ 25)

IV. Frozen, porous foods (kice/kgas ≈ 100)

The selection of thermal conductivity models is discussed separately for each of these classes.

3. Class I: Unfrozen, non-porous food products

Pham and Willix [27] performed effective thermal conductivity measurements between -40°C and +30

°C for various meat products and compared the results to the predictions of six simple models (Eqs. (1)

– (5), (11)).

3/28
Series model:

1
ke  (1)
 vi / ki
i

Parallel model:

k e   ki vi (2)
i

Kopelman isotropic model [6,28]:

 1  v22 / 3 (1  k 2 / k1 ) 
k e  k1  1/ 3 
(3)
1  v2 (1  k 2 / k1 )(1  v2 ) 
2/3

Maxwell-Eucken model [29,30]:

2k1  k 2  2(k1  k 2 )v2


k e  k1 (4)
2k1  k 2  (k1  k 2 )v2

Levy’s model [31]

2k1  k 2  2(k1  k 2 ) F
k e  k1 (5)
2k1  k 2  (k1  k 2 ) F

2 / G  1  2v2  (2 / G  1  2v2 ) 2  8v2 / G


where: F
2

(k1  k 2 ) 2
and G
(k1  k 2 ) 2  k1k 2 / 2

They found that for the meat products in their unfrozen state, all the models other than the Series

model provided sufficiently accurate predictions.

Mattea et al. [32] found that the effective medium theory (EMT) equation provided predictions of

sufficient accuracy for thermal conductivity data of fruits and vegetables at 20°C:

EMT model: [32–34]

ki  ke
v
i
i
k i  2k e
0 (6)

which may be rewritten to be explicit in terms of ke for two components:


k e  ¼ (3v2  1)k 2  [3(1  v2 )  1]k1  [(3v2  1)k 2  (3{1  v2 }  1)k1 ]2  8k1k 2 

4/28
Murukami and Okos [14] recommended the use of the Parallel model for un-frozen, non-porous foods

in general (other than meat products), which had been the recommendation of the COST 90 group

[35,36].

Hsu and Heldman [37] compared the thermal conductivities of aqueous starch solutions at

temperatures between 5°C and 45°C to the predictions of six effective thermal conductivity models

including the Kopelman (Eq. (3)), Maxwell (Eq. (4)), and EMT (Eq. (6)) models, and found that while

the Maxwell model provided the best predictions overall none of the prediction errors from any of the

models were greater than 10%.

Figure 1 shows a plot of Eqs. (1) to (6) where component 1 is water and component 2 is a food solids

phase having a thermal conductivity of 0.2 W m-1 K-1. Although the physical structures assumed in the

derivations of each of Eqs. (1) to (6) are very different, apart from the Series model, the model

predictions are all very similar. The inference to be drawn from this is observation is that because the

thermal conductivities of these food components are similar, the influence of material structure on

effective thermal conductivity is minimal. Evidence supporting this conclusion is easily seen in the

studies of Pham and Willix [27] and Hsu and Heldman [37].

The choice of effective thermal conductivity model for non-frozen, non-porous foods is therefore

relatively straightforward since any of the models listed above, other than the Series model, may be

used with sufficient accuracy for most purposes (± 10%). It should be noted: firstly that while the

Parallel model is the simplest computationally, of all these models it is most likely to over-predict (Fig.

1); secondly, the Maxwell-Eucken model requires identification of a continuous phase (component 1 in

Eq. (4)) and a dispersed phase (component 2 in Eq. (4)). Unless there is clear indication to the contrary,

it should be assumed that water, rather than the solids phase, forms the continuous phase of the food

product.

4. Class II: Frozen, non-porous food products

Thermal conductivity predictions involving frozen food require knowledge of the food’s ice content

(xice) which in turn requires knowledge of the initial freezing temperature (Tf). These data may be

5/28
determined experimentally from measured enthalpy-temperature data, or alternatively may be

predicted. The initial freezing temperature for a number of foods may be found in the literature [6,7].

For high-moisture content foods such as fresh fruits and vegetables, and meat and seafood products,

the initial freezing temperature is typically between -0.5°C and -3°C. For lower moisture content foods

such as cheese and egg yolks, the initial freezing temperature can be much lower. Models for

predicting the initial freezing temperature based on the food’s composition may be found in the

literature [17,38–40].

Models for predicting ice fractions have been reviewed by Rahman [17]. One of the most widely used

models (including [27,41,42]) is based on Raoult’s law and the Clausius-Clapeyron Equation:

 Tf 
xice  ( xw  xb )1   (7)
 T 

For their study on the thermal conductivity of meat products, Pham and Willix [27] related the bound

(i.e. unfreezable) water to the protein mass fraction [43]:

xb  0.4 x prot (8)

Tchigeov (cited in [17]), recommended the following empirical ice fraction models for use with meat,

fish, milk, eggs, fruits and vegetables:

xice 1.105
 (9)
xw 0.70138
1
ln(T f  T  1)

xice exp( 9.703 10 3 T f  4.794 10 6 T f2 )  1


 1 (10)
xw exp( 9.703 10 3 T  4.794 10 6 T 2 )  1

Figure 2 shows Eqs. (7), (9) and (10) plotted assuming an initial freezing temperature of -1°C, a

protein mass fraction of 0.2 and a total water mass fraction of 0.65. It is clear from the discrepancies

between the predictions of the different models shown in Fig. (2) that the selection of an ice fraction

model introduces an extra source of uncertainty. For example, the discrepancy between the predictions

of Levy’s model based on ice fractions calculated firstly from Eq. (7) and secondly from Eq. (10) may

by as high 15%, depending on the ice fraction. This error occurs independently of any further error that

may result from the selection of an unsuitable thermal conductivity model. Ideally, a thermal

conductivity model should therefore be tested independently of an ice fraction model, since if the ice

6/28
fraction model over-predicts and the conductivity model under-predicts (or vice versa), thermal

conductivity predictions may appear to be accurate for a given set of data, but may only be so by

coincidence. However, the most common approach when thermal conductivity models are being

compared appears simply to have been to base all thermal conductivity models on a single ice fraction

model.

The majority of the effective thermal conductivity studies for this class of food have been concerned

with meat products. Hill et al. [44] performed thermal conductivity measurements on fresh and frozen

meat products between 0 and 150°F (-18°C – +66°C), and derived an equation to model the effective

thermal conductivity as a function of temperature, fat and total moisture content, but not explicitly as a

function of ice-fraction. The model assumed that heat was conducted through the meat by three parallel

pathways: through the meat fibre, through the aqueous phase, and through the meat fibre and aqueous

phase in series (see Figs. 2 and 3 of [44]):

Hill’s model [44]:

8k1k 2 (C  C 2 )
k e  k 2 (2C  C 2 )  k1 (1  4C  3C 2 )  (11)
k1C  k 2 (4  C )

where: C  2  4  2v2

and: component 1 is the aqueous phase, component 2 is the meat fibre phase

Pham and Willix [27] found that Hill’s model was not as accurate as others such as Levy’s, and it does

not appear to have received widespread use.

Mascheroni [45] developed a model for frozen and unfrozen meats in which the meat was modelled as

a bundle of partially dehydrated meat fibres surrounded by ice. The meat fibres in turn were assumed

to be a meat solids matrix containing unfrozen water as a dispersed phase. Above the initial freezing

temperature, all the water was assumed to be contained within the fibre; below the initial freezing

temperature it was assumed that ice progressively formed in the extra-fibril region at the expense of

water within the fibres. Heat flows both parallel to, and perpendicular with, the fibre were considered.

The model was relatively complex, and subsequent workers [27,42] have found that a simpler model

such as Levy’s provides predictions that have comparable accuracy.

7/28
Pham and Willix [27] found that Levy’s model Eq. (5) based on ice fractions calculated from Eqs. (7)

and (8) provided good predictions for both the frozen and unfrozen meat products, and was

substantially better for frozen meats than the other models they considered in the study. They were,

however, reluctant to place too much confidence in Levy’s model due to its “… lack of physical

justification, since it was based on mathematical rather than physical arguments.” Wang et al. [23]

have subsequently shown that Levy’s model can have a physical interpretation, which should allay any

further reservations about its use on that basis.

Renaud et al. [46] studied the thermal conductivity of frozen model foods between -40°C and +20°C

and compared the predictions of the Series, Parallel, Maxwell-Eucken models and their own model

(Eq. (12)):

Renaud model:

 1  v2 v2 
k e  f [(1  v2 )k1  v2 k 2 ]  (1  f )   (12)
 k1 k2 

They found that when Eq. (12) was fitted to their experimental data, the empirically determined

weighting factor, f, was strongly dependent on the specific food in question, and hence they concluded

that as a predictive tool the Maxwell-Eucken model was better, even though it produced prediction

errors of up to 28%.

Tarnawski et al. [42] used the data of Pham and Willix to compare the predictions of several models

including those that had been used previously for meats (such as the Mascheroni and Levy Models)

and some models that had been used for soils, such as the models of De Vries [47] and Gori [48].

Some of the more complex models provided slightly more accurate predictions than Levy’s (based on

the root mean square error); however, the differences were marginal, and some, such as De Vries’

model, contained empirically determined parameters, which is undesirable.

Recently, Wang et al. [49] compared the predictions of 31 models (including the Series, Parallel,

Maxwell-Eucken and Levy models) to the thermal conductivity data for 22 meat and seafood products

between -40°C and +40°C [7,27], based on ice fractions calculated from Eqs. (7) and (8). Levy’s

model was again found to be only marginally less accurate than some more complex models.

8/28
It is difficult to give comprehensive, generic guidelines for predicting the thermal conductivity of

frozen, non-porous foods. For meat products, Levy’s model based on ice fractions predicted by Eqs.

(9) and (10) appears to be tried and trusted; however, for other Class II products fewer studies appear

to have been performed and there remains a significant level of uncertainty, both in the selection of

thermal conductivity models and in the selection of ice fraction models. There is scope for more work

to be done in this area.

5. Class III: Unfrozen, porous food products

Many, if not most of the effective thermal conductivity models that may be found in the literature are

concerned with porous materials. The term ‘porous’ may refer to granular or particulate materials in

which the void volume may be occupied by either liquid or gaseous components, or alternatively, it

may refer to a material having a continuous solid matrix that contains pores/bubbles which may be

isolated (sometimes referred to as ‘closed pores’) or interconnected (sometimes referred to as ‘open

pores’) [50]. The measurement and prediction of porosity and pore formation is a field of study in

itself, particularly for drying foods [51].

An unfrozen porous food product may be thought of as a binary mixture of a ‘condensed phase’

containing immobile water and food solids (which, for the purposes of effective thermal conductivity

prediction, may be treated as a Class I food), and a gaseous phase such as air or carbon dioxide. The

simplest method for determining a food’s porosity is from its apparent density and the density of the

solid and gaseous components:

 c   ap  ap
  1 (13)
 c   gas c

Figure 3 shows a plot of Eqs. (1) to (6) for a theoretical porous, high-water content food in which the

thermal conductivity of the condensed phase is 0.48 W m-1 K-1. By contrast with Fig. 1, the predictions

of the different models vary significantly, and the effect of the physical structure that each model is

based on becomes significant. Since thermal conductivity data for porous foods may lie anywhere

between the predictions of the Series and Parallel models (the so-called “Wiener Bounds” [52]), no

9/28
single model which is a function solely of the components’ thermal conductivities and volume

fractions (including Eqs. 1 to 6) will be suitable for all types of porous foods [26,53].

Due to differences in their structures, a foam and a particulate material may have different effective

thermal conductivities, even if they have identical void fractions and component thermal conductivities

[53]. Hence problems may arise when a model that has been shown to work well for one type of

porous material is assumed to be applicable to another type, simply because both materials have been

described as ‘porous’. Carson et al. [53] proposed that porous materials (including foods) should be

divided into “external porosity” materials (i.e. grains and particulates) and “internal porosity” materials

(i.e. foams and sponges), because the mechanism for heat conduction in a granular material is different

from that in a foam. Upper and lower bounds were proposed for the effective thermal conductivity of

two types of isotropic porous materials: for isotropic external porosity materials it was proposed that

the effective thermal conductivity is bounded above by the EMT model (Eq. 6), and below by the

Maxwell-Eucken model (Eq. 4) with air as the continuous phase; for isotropic internal porosity

materials, the upper bound is provided by the Maxwell-Eucken model with air as the dispersed phase

and the lower bound is provided by the EMT model (refer to Fig. 6 of [53]). The merit of these bounds

is the significant reduction in the range of possible thermal conductivity values for the two types of

porous foods, by comparison with the range constrained by the Wiener bounds. However, since data

will most likely lie between the bounds, they are not in themselves a complete solution to the problem.

The literature contains some correlations of thermal conductivity as functions of porosity and moisture

content for selected fruits and vegetables [54–56]; however, the applicability of these models is clearly

limited to the dataset from which the correlation was calculated. Neural network techniques have also

been employed [57,58], but are considerably more complex to implement than the simple algebraic

models that are available.

A semi-empirical approach to the prediction of the thermal conductivity of porous materials was

introduced by Krischer [59]. He reasoned that since the thermal conductivity of any two-component

material must lie between the Wiener bounds, its structure could be modelled as a mixture of the Series

10/28
and Parallel structures. He proposed that the effective thermal conductivity of the combined structure

should be the weighted harmonic mean of the Series and Parallel conductivities (Eq. (14)):

Krischer model [59]:

1
ke  (14)
 1 f  1  v2 v2 
  f   
 (1  v 2 ) k1  v 2 k 2  k1 k 2 

The value of the parameter f was determined empirically. By suitable adjustment of the f parameter

between 0 and 1, Krischer’s model may predict a thermal conductivity anywhere within the Wiener

Bounds. A number of studies of thermal properties of foods have used Krischer’s model, and hence

values for f (often referred to as the “distribution factor”) may be found in the literature for some

foods, including granular foods between 20°C and 70°C [14,60], dried fruits and vegetables between

5°C and 100°C [61,62], and partially baked French bread between -35°C and +25°C [63,64].

Several other models (including Eq. 12) are similar to Krischer’s in that they may predict thermal

conductivities anywhere between the Series and Parallel models by suitable adjustment of the f, Z, n

and j parameters:

Chaudhary-Bhandari model [65]:

(1 f )
 1  v2 v2 
k e  [(1  v2 )k1  v2 k 2 ] 
f
  (15)
 1 k k2 

Kirkpatrick model (modification of Eq. 6) [66]:

ki  ke
v
i
i
ki  ( Z / 2  1)k e
0 (16)

Hamilton model (modification of Eq. 4) [67]:

(n  1)k1  k 2  (n  1)(k1  k 2 )v2


k e  k1 (17)
(n  1)k1  k 2  (k1  k 2 )v2

Carson’s modified Maxwell model [18,26]:

 j2   j2 
  k  k   (k s  k a )
1  j 2  1  j 2 
s a

ke  k s   (18)
 j2 
 k  k a  (k s  k a )
2  s
1 j 

11/28
Carson’s modified EMT model [18,26]:

ki  ke
v i
 j 
0 (19)
k i   k e
i

1 j 
Of these models (Eqs. (12), (14) – (19)), Krischer’s model appears to have been employed most

frequently.

However, the fact that the f-value of Krischer’s model (and equivalent parameters of other models)

cannot be determined mechanistically is a significant shortcoming, since its determination may require

a thermal conductivity measurement – which might defeat the purpose of the prediction (although the j

parameters of Carson’s models may be related to the thermal conductivity bounds for internal and

external porosity materials [26]). Ideally all parameters in the model should be related to some physical

property. Carson et al. [68] concluded that after the food components’ thermal conductivities and

volume fractions the next most important variable was the extent (or quality) of thermal contact

between particles (in the case of particulate materials) or pores (in the case of sponge or foam-like

materials). However, as yet there does not appear to be any suitable method for predicting or

measuring this property other than for regular arrangements of regularly shaped objects. Unfortunately,

there remains an inevitable degree of empiricism involved in the thermal conductivity prediction of

porous foods, particularly for those with external porosity [26], and there is scope for more work in

this area.

A separate issue to the influence of structure on the thermal conductivity of a porous food is the

potential increase in the apparent thermal conductivity of the gaseous phase, due to the evaporation and

condensation of moisture across the pores. While this phenomenon is more likely to be an issue above

temperatures likely to be encountered in refrigeration applications, Hamdami et al. [63] observed this

behaviour even at sub freezing temperatures. They employed a model introduced by Sakiyama et al.

[70] to account for the increase in thermal conductivity:

k ap  k gas  kevap (20)

D P dP
k evap  Law s (21)
RT P  aw Ps dT

12/28
Similarly, radiation in the gaseous phase may also alter the apparent thermal conductivity of the

gaseous phase [69], although it is unlikely to be significant in the range of temperatures typically

encountered in refrigeration.

As for Class II foods, it is difficult to give comprehensive, generic guidelines for model selection. If

the f-value for the food in question is known, then Krischer’s model should be used, otherwise the

algorithms of Maroulis et al. [62] or Carson et al. [26] should be consulted.

6. Class IV: Frozen, porous food products

Class IV foods combine the difficulties associated with Class II and Class III foods, but, conveniently,

the effects of ice formation and porosity may be dealt with sequentially rather than simultaneously, as

demonstrated in the studies performed by Cogné et al. [41] with ice cream and Hamdami et al. [63]

with partially baked French bread.

The study by Cogné et al. was very thorough, complete with a micrograph of ice cream which allowed

the selection of thermal conductivity models to be based on the food’s microstructure. The micrograph

(Fig. 7 of [41]) showed that ice cream is composed of a continuous aqueous phase in which are

suspended discrete air bubbles, ice crystals and fat globules, and since the physical basis of the

Maxwell-type models is a structure comprised of distinct continuous and dispersed phases, this family

of models was an obvious choice for this application. Cogné et al. calculated the ice cream’s thermal

conductivity in three steps: firstly the Parallel model was used to combine the thermal conductivities of

the liquid water and food solids components (consistent with the discussion on Class I foods);

secondly, the resulting thermal conductivity from the first step was combined with the thermal

conductivity of ice using a modification of Maxwell’s model (De Vries model) to produce the thermal

conductivity of the condensed phase; thirdly, the thermal conductivity of the condensed phase was

combined with the thermal conductivity of air using the Maxwell-Eucken model (Eq. 4) to give the

overall thermal conductivity. Hence, in the terms used in this review, the Class IV food (ice cream)

was treated successively as a Class I food, a Class II food and a Class III food, as outlined

diagrammatically in Fig. 9 of [41]. Using this technique they were able to predict the thermal

conductivity of ice cream to within 8% mean relative uncertainty.

13/28
Hamdami et al. [63] used successive application of the Maxwell-Eucken model, as outlined

schematically in their Fig. 2b. They also considered another approach in which they combined the

thermal conductivities of all the components in one step using Krischer’s model, which, unlike the

Maxwell-Eucken model, is capable of handling multi-component materials. The predictions of both

approaches were compared to experimental data, and it was found that Krischer’s model could provide

more accurate predictions; however, it was acknowledged that this was because the f-value for

Krischer’s model was determined from their own experimental data, and hence the model had

essentially been fitted to the data, whereas the Maxwell-Eucken model had provided genuine

predictions.

Models such as the Series, Parallel and EMT models and their derivatives (e.g. Eqs. (8), (14) – (16)

and (19)) are capable of handling multi-component materials and hence the thermal conductivity

prediction may be achieved in one step. However, in these models all components are assumed to have

identical structures and spatial distributions, and so it is questionable whether this approach is as sound

as the sequential approach which allows for different structural models to be applied to the different

phases of the mixture.

It should be noted that when thermal conductivity is predicted by sequential application of one or more

models, the order in which the food’s components are included is significant; i.e. for a food comprised

of components A, B and C, the result of combining A and B first to give AB and then combining AB

with C, will not necessarily be the same as the result of combining A and C first etc. For some foods,

such as ice cream, it may be apparent from a visual inspection of the food’s structure not only which

effective thermal conductivity models are appropriate, but also the order in which the components

should be combined. However, there may be many situations where very little may be inferred from

examining the food’s structure. The literature does not appear to contain any definitive guidelines on

this issue, but, consistent with previous discussion, unless there is clear indication from the material’s

structure that a different order should be used, as a general rule components with the most similar

thermal conductivities should be combined first, followed by the component with the next most similar

thermal conductivities and so on.

14/28
7. Conclusion

The uncertainty involved in thermal conductivity prediction increases as the differences between the

food components’ thermal conductivities increase, which means that foods which are frozen and/or

porous require greater consideration than unfrozen, non-porous foods for which thermal conductivity

prediction is a relatively straightforward exercise. Recommendations for selecting models for different

classes of foods are summarised in Fig. 4, although it should be noted that they are only basic

guidelines. Greater understanding and characterisation of the effects of ice and/or porosity on food

structures would help improve the prediction accuracy for frozen and/or porous foods.

Acknowledgments

The author would like to acknowledge the assistance of Dr. Milan Houška of the Food Research

Institute Prague and Dr. Paul Nesvadba of Rubislaw Consulting Ltd in compiling this review. The

majority of the work was performed while the author was at AgResearch Ltd, New Zealand.

References:

[1] S. Thorne. Mathematical Modelling of Food Processing Operations, Elsevier Applied Science,

London, 1992

[2] Proceedings of a Conference - Workshop - Practical Instruction Course on Modelling of Thermal

Properties and Behaviour of Foods during Production, Storage and Distribution, Food Research

Institute Prague, Prague (Czech Republic), 1997

[3] M. Houška et al., Thermophysical and rheological properties of foods: Milk, milk products and

semiproducts, Food Research Institute Prague, Prague, 1994

[4] M.S. Rahman. Food Properties Handbook, CRC Press, Boca Raton, 1995, Chapter 5.

[5] M. Houška et al., Thermophysical and rheological properties of foods: Meat, meat products and

semiproducts, Food Research Institute Prague, Prague, 1997

[6] ASHRAE Handbook of Refrigeration, American Society of Heating, Refrigeration and Air-

conditioning Engineers, Atlanta GA, 2002

[7] J. Willix, S.J., Lovatt, N.D. Amos. Additional thermal conductivity values of foods measured by a

guarded hot plate, J. Food Eng., 1998;37(2): 159-174

15/28
[8] Z. Mayer, M. Houška, M. Cigáník, Luong Thi Cam Tu, J. Šesták, P. Nesvadba, Thermophysical

properties of food: Selected fruits and vegetables, Food Research Institute Prague, Prague, 2004

[9] C.A. Miles, M.J. Morley, Estimation of the thermal properties of foods : a revision of some of the

equations used in COSTHERM, in : Proceedings of a Conference - Workshop - Practical

Instruction Course on Modelling of Thermal Properties and Behaviour of Foods during

Production, Storage and Distribution, Prague (Czech Republic), 1997, pp. 135-146

[10] P. Harriott, Thermal Conductivity Of Catalyst Pellets And Other Porous Particles - 1. Review Of

Models And Published Results, Chem. Eng. J., 1975;10(1): 65

[11] R.C. Progelhof, J.L. Throne, R.R. Reutsch. Methods for Predicting the Thermal Conductivity of

Composite Systems: A Review, Polym. Eng. Sci., 1976, 16; 615-625

[12] O.T. Farouki. Thermal Properties of Soils, Series on Rock and Soil Mechanics Vol II, Trans Tech

Publications, 1986

[13] E. Tsotsas, H. Martin. Thermal conductivity of packed beds: A review, Chem. Eng. Process.,

1987; 22: 19-37.

[14] E.G. Murakami, M.R. Okos. Measurement and prediction of thermal properties of foods in: R.P.

Singh, A.G. Medina (Eds.), Food Properties and Computer Aided Engineering of food Processing

Systems (pp. 3-48). Kluwer Academic Publishers, New York, 1989

[15] A.G. Leach, Thermal conductivity of foams. I. Models for heat conduction, Journal of Physics D:

Applied Physics, 1993;26(5): 733-739

[16] P. Cheng, C-T. Hsu., Heat Conduction, in: D.B. Ingham, I. Pop (Eds.), Transport Phenomena in

Porous Media, Pergamon Press, 1998; 57-76

[17] M.S. Rahman. Thermophysical Properties of Foods, in: Sun, D.-W. (Ed.) Advances in Food

Refrigeration, Leatherhead Publishing, Surrey, England, 2001

[18] J.K. Carson, Prediction of the thermal conductivity of porous foods, PhD thesis, Massey

University, New Zealand, 2002

[19] D.A.G. Bruggeman. Berechnung verschiedener physikalischer Konstanten von heterogenen

Substanzen: I. Dielektrizitätskonstanten und Leitfähigkeiten der Mischkörper aus isotropen

Substanzen, Ann. Phys., 1935;24: 636-664

[20] E. Behrens. Thermal conductivities of composite materials, J. Composite Materials, 1968;2: 2-17

16/28
[21] T.H. Bauer. A general analytical approach toward the thermal conductivity of porous media, Int.

J. Heat Mass Transfer, 1993;36: 4181-4191

[22] E.E. Gonzo, Estimating correlations for the effective thermal conductivity of granular materials,

Chem. Eng. J., 2002;90: 299-302

[23] J.F. Wang, J.K., Carson, M.F., North, D.J. Cleland, A new approach to the modelling of the

effective thermal conductivity of heterogeneous materials, Int. J. Heat Mass Transfer, (in press)

[24] P. Nesvadba. Methods for the measurement of thermal conductivity and diffusivity of foodstuffs,

J. Food Eng., 1982;1(2): 93-113

[25] S.C. Cheng, R.I. Vachon. Prediction of thermal Conductivity of two and three phase solid

heterogeneous mixtures, Int. J. Heat Mass Transfer, 1969;12:249-264

[26] J.K. Carson, S.J. Lovatt, D.J. Tanner, A.C. Cleland. Predicting the effective thermal conductivity

of unfrozen, porous foods, J. Food Eng. 2006;75(3):297-307

[27] Q.T. Pham, J. Willix. Thermal Conductivity of Fresh Lamb Meat, Offals and Fat in the Range –40

to +30°C: Measurements and Correlations, J. Food Sci., 1989;54(3): 508-515

[28] I.J. Kopelman, Transient heat transfer and thermal properties in food systems, PhD Thesis,

Michigan State University, USA

[29] J.C. Maxwell, A Treatise on Electricity and Magnetism 3rd Ed., Dover Publications Inc., New

York, reprinted 1954.

[30] A. Eucken. Allgemeine Gesetzmäßigkeiten für das Wärmeleitvermögen verschiedener Stoffarten

und Aggregatzustände, Forschung Gabiete Ingenieur, 1940;11(1): 6-20

[31] F.L. Levy. A modified Maxwell-Eucken equation for calculating the thermal conductivity of two-

component solutions or mixtures, Int. J. Refrig., 1981; 4(4): 223-225

[32] M. Mattea, M.J. Urbicain, E. Rotstein. Prediction of thermal conductivity of vegetable foods by

the effective medium theory, J. Food Sci., 1986;51(1): 113-115, 134

[33] R. Landauer, The electrical resistance of binary metallic mixtures, J. Appl. Phys. 1952;23: 779-

784

[34] H.T. Davis, L.R. Valencourt, C.E. Johnson. Transport processes in composite media, J. Am.

Ceram. Soc., 1975;58: 446-452

17/28
[35] C.A. Miles, G. van Beek, C.H. Veerkamp, .Calculation of thermophysical properties of foods, in:

R .Jowitt, F. Escher, B. Hallström., H.F.T. Meffert, W.E.L. Spiess, G. Vos (Eds). Physical

Properties of Foods, Applied Science Publishers, London, 1983

[36] H.F.T. Meffert, History, aims, results and future of thermophysical properties within COST 90, in:

R. Jowitt, F. Escher, B. Hallström., H.F.T. Meffert, W.E.L. Spiess, G. Vos (Eds). Physical

Properties of Foods, Applied Science Publishers, London, 1983

[37] C.-L. Hsu, D. R. Heldman. Prediction of the thermal conductivity of aqueous starch, Int. J. Food

Sci. Tech., 2004;39(7): 737-743

[38] E.G. Murakami, M.R. Okos. Calculation of initial freezing point, effective molecular weight and

unfreezable water of food materials from composition and thermal conductivity data, J. Food Proc.

Eng., 1996;19(3), 301-320

[39] C.A. Miles, Z. Mayer, M.J. Morley, M. Houška, Estimating the initial freezing point of foods

from composition data, Int. J. Food Sci. Tech., 1997;32(5): 389-400

[40] R.G.M. van der Sman, E. Boer, Predicting the initial freezing point and water activity of meat

products from composition data, J. Food Eng. 2005;66(4): 469-475

[41] C. Cogné, J. Andrieu, P. Laurent, A. Besson, J. Nocquet. Experimental data and modelling of

thermal properties of ice creams, J. Food Eng., 2003;58(4): 331-341

[42] V.R. Tarnawski, D.J. Cleland, S. Corasaniti, F. Gori, R.H. Mascheroni. Extension of soil thermal

conductivity models to frozen meats with low and high fat content, Int. J. Refrig., 2005;28(6):

840-850

[43] Q.T. Pham. Calculation of bound water in food, J. Food Sci., 1989;54(3): 508-515

[44] J.E. Hill, J.D. Leitman, J.E. Sunderland. Thermal conductivity of various meats, Food Tech.,

1967;21: 1143-1148

[45] R.H. Mascheroni, J. Ottino and A. Calvelo. A model for the thermal conductivity of frozen meat,

Meat Sci., 1977;1: 235–243.

[46] T. Renaud, P. Briery, J. Andrieu, M. Laurent. Thermal properties of model foods in the frozen

state, J. Food Eng., 1992;15(2): 83-97.

[47] D.A. DeVries, Thermal properties of soils. In: W.R. van Wijk, (Ed.) Physics of plant environment,

Wiley, New York ,1963, 210–235.

18/28
[48] F. Gori, A theoretical model for predicting the effective thermal conductivity of unsaturated

frozen soils, Proceedings of the fourth international conference on Permafrost, Fairbanks (Alaska),

National Academy Press, Washington, DC, 1983, 363–368

[49] J.F. Wang, J.K. Carson, D.J. Cleland, M.F. North. A comparison of effective thermal conductivity

models for frozen foods, Proceedings: Innovative Equipment and Systems for Comfort and Food

Preservation, Auckland, 2006, 336-343

[50] M.S. Rahman. Mechanism of pore formation in foods during drying: Present status, in: J. Welti-

Chanes, G.V. Barbosa-Canovas, J.M. Aguilera (Eds) Proceedings of the Eighth International

Congress on Engineering and Food, Technomic Publishing Co., Lancaster, PA, 2001: 1111-1116

[51] M.A. Hussain, M.S. Rahman, C.W. Ng. Prediction of pores formation (porosity) in foods during

drying: generic models by the use of hybrid neural network, J. Food Eng., 2002;51(3): 239-248

[52] O. Wiener, Die theorie des Mischkorpers für das Feld der Statonären Stromüng I. Die

Mittelwertsatze für Kraft, Polarisation und Energie, Der Abhandlungen der Mathematisch-

Physischen Klasse der Königl. Sachsischen Gesellschaft der Wissenschaften, 1912;32: 509-604

[53] J.K. Carson, S.J. Lovatt, D.J. Tanner, A.C. Cleland. Thermal conductivity bounds for isotropic,

porous materials. Int. J. Heat Mass Transfer, 2005;48(11): 2150-2158

[54] M.S. Rahman, Thermal conductivity of four food materials as a single function of porosity and

water content. J. Food Eng. 1992;15: 261–268.

[55] M.S. Rahman, X.D. Chen, A general form of thermal conductivity equation as applied to an apple:

effects of moisture, temperature and porosity. Drying Technol., 1995;13: 1–18

[56] M.S. Rahman, X. D. Chen, C.O. Perera. An improved thermal conductivity prediction model for

fruits and vegetables as a function of temperature, water content and porosity, J. Food Eng.,

1997;31(2): 163-170

[57] S.S. Sablani, O.-D. Baik, M. Marcotte, Neural networks for predicting thermal conductivity of

bakery products, J. Food Eng., 2002;52(3): 299-304

[58] S.S. Sablani, M.S. Rahman. Using neural networks to predict thermal conductivity of food as a

function of moisture content, temperature and apparent porosity, Food Res. Int., 2003;36(6): 617-

623

[59] O. Krischer, Die wissenschaftlichen Grundlagen der Trocknungstechnik (The Scientific

Fundamentals of Drying Technology), Springer-Verlag, Berlin, 1963, cited in English in: Keey, R.

19/28
B. Drying of Loose and Particulate Materials, Hemisphere Publishing Corporation, New York,

1992, Chapter 7.

[60] M.K. Krokida, Z.B. Maroulis, M.S. Rahman. A structural generic model to predict the effective

thermal conductivity of granular materials, Drying Tech., 2001;19(9): 2277-2290

[61] X.D. Chen, G.Z. Xie M.S. Rahman. Application of the distribution factor concept in correlating

thermal conductivity data for fruits and vegetables. Int. J. Food Properties, 1998;1(1): 35–44.

[62] Z.B. Maroulis, M.K. Krokida, M.S. Rahman. A structural generic model to predict the effective

thermal conductivity of fruits and vegetables during drying, J. Food Eng., 2002;51(1): 47-52

[63] N. Hamdami, J-Y. Monteau, A. Le Bail. Effective thermal conductivity of a high porosity model

food at above and sub-freezing, Int. J. Refrig., 2003;26(7): 809-816

[64] N. Hamdami, J-Y. Monteau, A. Le Bail. Thermophysical properties evolution of French partly

baked bread during freezing, Food Res. Int., 2004;37(7): 703-713

[65] D.R. Chaudhary, R.C. Bhandari. Heat transfer through a three-phase porous medium, J. Phys. D

(Brit. J. Appl. Phys.), 1968;1: 815-817

[66] S. Kirkpatrick, Percolation and conduction, Rev. Mod. Phys. 1973;45: 574-588.

[67] R.L. Hamilton, O.K. Crosser. Thermal conductivity of heterogeneous two-component systems.

Ind. Eng. Chem. Fundamentals, 1962;1(3): 187-191

[68] J.K. Carson, S.J. Lovatt, D.J. Tanner, A.C. Cleland, An analysis of the influence of material

structure on the effective thermal conductivity of porous materials using finite element

simulations, Int. J. Refrig., 2003;26(8): 873-880.

[69] A.J. Loeb. Thermal conductivity: VIII, A theory of thermal conductivity of porous materials, J.

Am. Ceramic Soc., 1954;37(2): 96-99

[70] T. Sakiyama, M. Akutsu, O. Miyawaki, T. Yano, Effective thermal diffusivity of food gels

impregnated with air bubbles, J. Food Eng. 1999;39(3): 323-328

[71] J.P. Holman, Heat Transfer, McGraw-Hill Book Company, Singapore, 1992

20/28
Nomenclature

a activity

C, F, G intermediate variables defined within the text

D diffusivity of water vapour in air

f empirical weighting factor (sometimes referred to as the “distribution factor”)

j relative weighting factor

k thermal conductivity of a component (W m-1 K-1)

L enthalpy of vaporisation

n weighting parameter

P pressure

T temperature (°C)

v volume fraction

x mass fraction

Z weighting parameter

 porosity

 density (kg m-3)

Subscripts

1 component 1

2 component 2

ap apparent density

b bound water

c condensed phase

e effective thermal conductivity

evap evaporation-condensation effect

f initial freezing temperature

gas gaseous phase

i ith component

ice ice

21/28
prot protein

s saturation

w total water (ice + unfrozen free water + bound water)

22/28
Table 1: Thermal Conductivities of Food Components at 0 °C [6,71]

Figure 1: Plots of Eqs. (1) to (6) for k1/k2 = 3 (equivalent to an unfrozen, non-porous food)

Figure 2: Plots of three different ice fraction prediction models (Eqs. (7), (9) and (10)), with Tf = -1,

xprot = 0.2, xb = 0.08, xw = 0.65

Figure 3: Plots of Eqs. (1) to (6) for k1/k2 = 20 (equivalent to an unfrozen, porous, high-water-content

food)

Figure 4: Summary of recommendations for selecting models for different classes of foods.

23/28
Food Component k/W m-1 K-1
protein 0.18
fat 0.18
soluble carbohydrate 0.20
fibre 0.18
ash 0.33
liquid water 0.57
ice 2.2
air 0.024

Table 1:

24/28
1

0.8

0.6
k e /k 1

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1

v2
Series Parallel
Maxwell-Eucken Levy
Effective Medium Theory Kopelman Isotropic

Figure 1:

25/28
0.7
Eq. (7)
0.6 Eq. (9)
Eq. (10)
0.5

0.4
x ice

0.3

0.2

0.1

0
-15 -12 -9 -6 -3 0

T (°C)

Figure 2:

26/28
1

0.8

0.6
k e /k 1

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1

v2
Series Parallel
Maxwell-Eucken Levy
Effective Medium Theory Kopelman Isotropic

Figure 3:

27/28
Start

Sufficient to treat food as


Class I foods binary mixture of water
unfrozen, non-porous and solids. Use any of Eq.
(2) to (6) to determine k e .

Select ice fraction model


from Eqs. (7) to (10), and
Class II foods
ascertain T f from
frozen, non-porous
predictive model (e.g.
[39]).

Use Levy's model (Eq. 5)


to predict k e .

Treat food as binary


mixture of condensed
Class III foods
phase and air. Determine
unfrozen, porous
k e of condensed phase as
for Class I foods.

Estimate porosity. Use Eq.


(13); alternatively refer to
[62]

If f -value available in
literature, use Krischer's
model (Eq. 14) to
determine k e ; alternatively
refer to [26] for model
selection guidelines

Deal with ice content (as


for Class II food) and
Class IV foods
porosity (as for Class III
frozen, porous
food) sequentially. Follow
example of [41].

Finish

Figure 4

28/28

View publication stats

Você também pode gostar