Thermistor PDF

A Thermistor Based Method for Measurement of Thermal Conductivity and
Thermal Diffusivity of Moist Food Materials at High Temperatures
by
Maarten F. van Gelder
Dissertation submitted to the Faculty of the

Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Biological Systems Engineering
Kenneth C. Diehl, Jr., Chairman

C. Gene Haugh
Joseph E. Marcy
James H. Wilson
Foster A. Agblevor
February, 1998
Blacksburg, Virginia
A THERMISTOR BASED METHOD FOR MEASUREMENT OF THERMAL
CONDUCTIVITY AND THERMAL DIFFUSIVITY OF MOIST FOOD MATERIALS AT
HIGH TEMPERATURES
by
Maarten F. van Gelder
Kenneth C. Diehl, Jr., Chairman
Biological Systems Engineering
(ABSTRACT)
The purpose of this research was to assess the suitability of the thermistor based method for
measuring thermal conductivity and diffusivity of moist food materials at high temperatures.
Research focused on aspects of calibration, thermal contact in solid food materials, natural
convection in liquid media and the performance in moist food materials at high temperatures.
Thermistor probes were constructed in house and calibrated in three materials of known
thermal conductivity and diffusivity, water, glycerol, and a heat transfer fluid, HTF 500.
With few exceptions, the calibrated probe estimated thermal properties with an error of less
than 5%, over the range of thermal properties spanned by the those of the calibration media.
An alternate calibration using two media was also investigated. It was found to give better
accuracy over a more limited range.
Thermal contact in potato and lean beef was investigated through a comparative study that
used a miniature line heat source probe as a reference method. The food materials were
measured at 25, 50 and 100 °C. Good agreement was found between the measurements with
the line heat source probe and the bead thermistor probe, indicating adequate thermal contact
at the thermistor probe.
The effect of fluid viscosity and the magnitude of the temperature step on the occurrence of
natural convection was studied for aqueous solutions of a thickening agent. During a sample
time of 30 seconds, convection was absent in solutions with a viscosity of 25 cp or greater,
when measured with a temperature step of 1.5 and 2.5 °C, and in solutions with a viscosity of
50 cp or greater, when measured with a temperature step of 5.0 °C. A Rayleigh number was
defined to study the notion of a critical Rayleigh number at the onset of convection. This
study found that when the Rayleigh number was below 43, convection could not be
demonstrated. For a Rayleigh number of 84 and higher, convection was observed.
The performance at high temperatures in food materials was studied through tests in tomato
concentrate and in a liquid food supplement. Tomato puree and tomato paste were sampled at
100, 130 and 150 °C. The thermal conductivity of tomato puree at 100, 130 and 150 °C was
measured as 0.638, 0.645 and 0.647 W/m°C respectively. The thermal diffusivity was 1.63,
1.64 and 1.62 10-7m2/s respectively. For tomato paste at 100, 130 and 150 °C, a thermal
conductivity was obtained of 0.590, 0.597 and 0.534 W/m°C respectively. The thermal
diffusivity was 1.63, 1.84 and 2.36 10-7m2/s respectively. With some notable exceptions the
results of this study agreed well with Choi and Okos (1983). A liquid food supplement was
also studied at 95 and 150 °C. The thermal conductivity of the food supplement decreased
with increasing solids content from 0.62 W/m°C at a solids level of 15% to 0.41 W/m°C at a
solids level of 50%.
The results of this study indicate that the thermistor based method was suitable for measuring
thermal conductivity and diffusivity of moist food materials at high temperatures. However,
the type of thermistor used in the research, a glass encapsulated thermistor, was too fragile
for routine work. In particular the high temperature use of the glass thermistor was impacted
by its susceptibility to fracture.
ACKNOWLEDGMENTS
First and foremost I would like to thank the chairman of my committee Dr. Kenneth C.
Diehl, Jr. for his guidance and support. His door was always open. I greatly appreciate his
willingness to support me during the last months of my study after he moved to a new
position well away from Blacksburg.
I thank the staff in the workshop and in particular Mr. Steve Spradlin and Mr. Clyde Adkins
for their assistance. I gave them drawings of sometimes questionable quality and they
returned well made parts for the equipment for my research.
I thank Dr. Steve Thompson, who originally served on my committee but escaped to warmer
climates. Dr. James H. Wilson stepped in to replace him for which I am grateful. I
acknowledge Drs. C. Gene Haugh and Joseph E. Marcy for their assistance. I have kind
words for Dr. William C. Thomas who served on my committee from the start.
Circumstances prevented him from continuing on my committee through the final phases of
my studies. Dr. Foster A. Agblevor kindly stepped in as replacement committee member and
provided thoughtful critique during the completion of my dissertation.
I wish to acknowledge the Center for Aseptic Processing and Packaging Studies, North
Carolina State University (Raleigh, NC), whose financial support made this research
possible. I also wish to thank Van Den Bergh Foods Co. and Abbott Labs/Ross Products
Division for supplying samples.
I wish to thank the interns Erroll Zalmijn (Surinam), Rene Wolters and Rene Sleutel (both
from the Netherlands) for assistance with parts of my research.
Acknowledgments iv
TABLE OF CONTENTS
CHAPTER 1 1
INTRODUCTION 1
CHAPTER 2 8
LITERATURE REVIEW 8
2.1 Steady State Methods for Measurement of Thermal Conductivity 8
2.1.1 Introduction 8
2.1.2 Guarded Hot Plate 9
2.1.3 Radial Heat Flow Method 9
2.2 Transient State Methods 10
2.2.2 Fitch Method 11
2.2.3 Transient Hot Wire 13
2.2.4 Line Heat Source Probe 14
2.2.5 Thermistor Based Method 16
2.3 Natural Convection 24
CHAPTER 3 27
PRINCIPLES OF THE THERMISTOR BASED METHOD 27
3.1 Heat Transfer Model 27
3.2 Solution 30
3.3 Application 30
3.3.1 Measurement of Thermal Conductivity 30
3.3.2 Measurement of Thermal Diffusivity 31
3.4 Calibration: Estimation of Probe Properties 32
3.5 Practical Considerations 33
3.5.1 Sample Dimension 33
3.5.2 Temperature Step 35
3.6 Nomenclature 37
Table of Contents v
CHAPTER 4 38
METHODS AND MATERIALS 38
4.1 Thermistors 38
4.2 Thermistor Assembly 39
4.2.1 Low Temperature Probes 40
4.2.2 High Temperature Probes 40
4.3 Sample Chamber 41
4.4 Controller 44
4.5 Thermal Property Measurement 46
4.6 Effective Probe Properties 46
4.7 Adequacy of Thermal Contact 49
4.7.2 Sample Material 49
4.7.3 Thermistor Based Measurements 50
4.7.4 Measurements with the Line Heat Source Probe 52
4.8.2 Sample Preparation 56
4.8.3 Measurement of Viscosity 57
4.8.4 Measurement of Thermal Properties 57
4.8.5 Rayleigh Number Calculation. 57
4.9 Evaluation at High Temperature 58
4.9.2 Tomato Concentrate 59
4.9.3 Nutritional Supplement 60
4.10 Thermal Standards 61
CHAPTER 5 66
RESULTS AND DISCUSSION 66
5.1 Temperature-Resistance Calibration 66
5.2 Effective Probe Properties 67
5.2.1 Low Temperature Probes 67
5.2.2 High Temperature Probes 69
Table of Contents vi
5.3 Adequacy of Thermal Contact 79
5.3.2 Potato Measurements 79
5.3.3 Beef Measurements 83
5.3.4 Conclusion 85
5.4.1 Viscosity Measurements 86
5.4.2 Effect of the Viscosity 87
5.4.3 Effect of the Temperature Step 93
5.4.4 Rayleigh Number Correlation 98
5.4.5 Validation of Rayleigh Number Correlation 99
5.5 Evaluation at High Temperature 102
5.5.1 Tomato Concentrate 102
5.5.2 Nutritional Supplement 107
5.5.3 Conclusions of High Temperature Measurements 109
CHAPTER 6 110
REMARKS ON CALIBRATION 110
CHAPTER 7 123
SUMMARY AND CONCLUSIONS 123
RECOMMENDATIONS 127
REFERENCES 128
APPENDIX 134
A Density of Glycerol and Dimethyl Phthalate 134
B Measurement Data for the Calibration of the Low Temperature Probe 143
C Specific Heat of Dimethyl Phthalate 148
D Thermal Conductivity of Potato and Beef: Measurement Data 151
VITAE 160
Table of Contents vii

LIST OF FIGURES
Figure 2.1 Glass coated bead thermistor probe. All dimensions in mm. 17
Figure 3.1 Square of voltage drop across thermistor versus the inverse square 32
root of time for a thickened water sample at 100 °C.
Figure 4.1 Probe assembly for low temperature measurements. 40
Figure 4.2 Lid assembly for the high temperature thermal property 42
measurements.
Figure 4.3 Sample chamber for thermal property measurements. 43
Figure 4.4 Control circuit for thermal property measurements. 45
Figure 5.1 Error in estimation of thermal conductivity of the calibration 73
media for Probe #H3 at 100 °C.
Figure 5.2 Error in estimation of thermal diffusivity of the calibration media 73
for Probe #H3 at 100 °C.
media for Probe #H3 at 150 °C.
for Probe #H3 at 150 °C.
media for Probe #H3 at 150 °C, when using reference data for the
thermal conductivity of glycerol from Touloukian (1970).
for Probe #H3 at 150 °C, when using reference data for the
thermal conductivity and specific heat of glycerol from
Touloukian (1970).
Figure 5.7 Thermal Conductivity of Potato at 25, 50 and 100 °C, measured 81
with the Line Heat Source Probe and the Bead Thermistor Probe.
Figure 5.8 Thermal Conductivity of Lean Beef at 25, 50 and 100 °C, 84
measured with the Line Heat Source Probe and the Bead
Thermistor Probe.
Figure 5.9 Rf∆T/Vss2 for the 150 cp solution and temperature steps of 1.5, 2.5 94
and 5 °C. Data originated from linear regression over a 10 second
interval. They are the average of 2 samples.
List of Figures viii

Figure 5.10 S/I for the 150 cp solution and temperature steps of 1.5, 2.5 and 94
5 °C. Data were obtained from linear regression over 10 second
intervals, with an increment in its start time of 1 second. They are
the average of two samples.
Figure 5.11 Simulated Rayleigh numbers for viscosity-∆T combinations that 98
appeared to exhibit signs of convection during measurement of
thickened water of various viscosity.
Figure 5.12 Simulated Rayleigh numbers for measurements in glycerol at 101
temperatures from 100 to 150 °C, with a temperature step of
2.5 °C.
Figure 6.1 Value of R∆T/Vss2 for Probe #H3 at 100 °C for the three 111
calibration materials and a regression interval of 10 seconds. The
average of three samples is shown.
Figure 6.2 Ratio of Slope and Intercept, S/I, for Probe #H3 at 100 °C for the 111
three calibration materials and a regression interval of 10 seconds.
The average of three samples is shown.
Figure 6.3 Difference in the value of R∆T/Vss2 from its value at ST = 10 112
seconds at 100 °C for the three calibration materials. The average
of three samples is shown.
Figure 6.4 Difference in the value of S/I from its value at ST = 10 seconds at 112
100 °C for the three calibration materials. The average of three
samples is shown.
Figure 6.5 Thermal Conductivity of Tomato Puree at 100 °C, measured with 114
a probe (#H3) that was calibrated with water and glycerol.
Figure 6.6 Thermal Diffusivity of Tomato Puree at 100 °C, measured with a 114
probe (#H3) that was calibrated with water and glycerol.
Figure 6.7 Thermal Conductivity of Tomato Puree at 100 °C, measured with 115
a probe (#H3) that was calibrated with water, glycerol and HTF
500.
Figure 6.8 Thermal Diffusivity of Tomato Puree at 100 °C, measured with a 115
probe (#H3) that was calibrated with water, glycerol, and HTF
500.
List of Figures ix
LIST OF TABLES
Table 4.1 Thermophysical Properties of Water 63

Table 4.3 Thermophysical Properties of HTF 500 65
Table 5.1 Calibration coefficients and value of the χ2 statistic for the 67
temperature-resistance function of the probes
Table 5.2 Properties for low temperature probe #L4, corresponding to the 68
regression interval 3 to 9 seconds.
Table 5.3 Radius and thermal conductivity of the #H3-probe, calibrated for 71
the measurement of thermal conductivity at 100 °C. The error in
estimation of medium thermal conductivity for the calibration
media is given. Regression interval was 10 seconds.
Table 5.4 Radius and thermal conductivity of the #H3-probe, calibrated for 72
the measurement of thermal diffusivity at 100 °C. The error in
estimation of medium thermal conductivity for the calibration
media is given. Regression interval was 10 seconds.
Table 5.5 Viscosity of sodium alginate solutions 86
Table 5.6 Tukey's W grouping for viscosity for each of the three temperature 88
steps and different values for the start time (ST) of the regression
interval; SAS analysis was performed on the means of R∆T/Vss2
(Table 5.8) corresponding to viscosity (and ∆T); the means of
R∆T/Vss2 originate from linear regression over a 2 second wide
interval.
Table 5.7 Tukey's W grouping for viscosity for each of the three temperature 89
steps and different values for the start time (ST) of the regression
interval; SAS analysis was performed on the means of S/I (Table
5.9) corresponding to viscosity (and ∆T); the means of S/I
originate from linear regression over a 2 second wide interval.
Table 5.8 Values of Rf∆T/Vss2 as related to temperature step and start time of 90
the regression interval. The regression interval was 2 seconds
wide. Values are ordered, per temperature step, in decreasing
value from top down. The corresponding values for viscosity can
be found in Table 5.6.
List of Tables x
Table 5.9 Values of S/I as related to temperature step and start time of the 91
regression interval. Regression interval was 2 seconds wide.
Values are ordered, per temperature step, in decreasing value from
top down. The corresponding values for viscosity can be found in
Table 5.7.
Table 5.10 Earliest start time of regression interval (of 2 seconds wide) at 92
which a viscosity would test different (at the 5% level) of the
more viscous solutions. Times are given for both the analysis of
Rf∆T/Vss2 and S/I
Table 5.11 Estimated onset time of convection in thickened water at 50 °C, 92
[s].
Table 5.12 Tukey's W grouping for 150 cp solution for the effect of the 97
magnitude of the temperature step on the value of R∆T/Vss2 and
S/I
Table 5.13 Estimated onset time of convection in Glycerol sample #1, 100
measured with probe #H1.
Table 5.14 Parameter values for simulation of temperature profiles in 100
Glycerol with program MEDTEMP.PAS
Table 5.15 Average thermal conductivity and coefficient of variation (COV) 103
of tomato puree (solids content 14% w/w), at high temperature
Table 5.16 Average thermal diffusivity and coefficient of variation (COV) of 104
tomato puree (solids content 14% w/w), at high temperature
Table 5.17 Average thermal conductivity and coefficient of variation (COV) 106
of tomato paste (solids content 26% w/w), at high temperature
Table 5.18 Average thermal diffusivity and coefficient of variation (COV) of 106
tomato paste (solids content 26% w/w), at high temperature
Table 5.19 Thermal Conductivity of "TC1", "TC2", "TC3" and "TC4" at 109
High Temperatures, [W/m°C]
Table 6.1 Thermal Properties of Ethylene Glycol at 100, 130 and 150 °C, 118
Table 6.2 Thermal Properties of Dimethyl Phthalate at 100, 130 and 150 °C, 120
List of Tables xi
CHAPTER 1
INTRODUCTION
The primary goal of the food processing industry is to supply the market with safe products.
One common means of attaining this is food preservation by means of thermal processing.
The application of heat can induce favorably product changes, deactivate enzymes, and kill
micro-organisms. The result is a product, hopefully of appetizing quality, with an extended
shelf-life and safe to consume. Unfortunate side effects of product heating involve nutritional
and textural losses. Underprocessing of food may lead to health hazards whereas
overprocessing may decrease a product’s quality as well as constitute a wasteful use of energy
and other resources. Heat transfer studies seek to find processing regimes that satisfy the
markets’ requirement for a high quality product but with an optimal utilization of production
resources. These studies require the knowledge of thermal properties of food materials.
Various methods exist to thermally process food materials. Which method is used for a
particular food material depends on several factors (product characteristics, available
technology, costs, logistics, market factors) most of which are outside the scope of this work.
Two commonly used thermal processes are in-can processing and aseptic processing.
With in-can processing, the conventional method of canning, a non-sterile food product is
filled into a non-sterile container and hermetically sealed. Subsequently, the container and
product are subjected to a heat treatment to obtain a commercially sterile product. Product
heating is usually non-uniform and a temperature gradient will exist between the product near
the can wall and at its cold point. As a result, part of the can contents will be overprocessed
before adequate sterility is obtained at the cold point.
During aseptic processing, product and container are treated separately from each other. The
product is sterilized or pasteurized in a continuous heat-hold-cool system. Rapid heating
takes place in a heating heat-exchanger. The product is then held in the holding tube long
Introduction 1
enough to ensure adequate lethality. A cooling heat-exchanger rapidly cools the product after
which it is transferred to an aseptic filler, where under a sterile atmosphere the product is
packaged in separately sterilized containers. Where this can be applied, aseptic processing
can significantly shorten the process time and yield a higher quality product then can be
obtained with the conventional canning method. The product is also more uniformly
processed than in a conventional system. Aseptic processing of homogeneous liquid products
has become more common in the food industry. Liquid foods containing particles posed
problems in the design of aseptic processing system. As of 1997, only one process has been
filed with the FDA (Palaniappan, 1997).
One of the requirements to file an aseptic process of a low-acid liquid food containing
particulates with the FDA, is to have the process established by a processing authority. This
involves, among other things, that the filing party can demonstrate that the process delivers
the selected lethality to the slowest heating zone (cold spot) of the particle by the time it
reaches the end of the holding tube. Heat penetration measurements and biological methods,
as used for process establishment of conventionally canned foods, are impractical for an
aseptic system. Temperature measurement of the cold spot of a particle while it travels
through a continuous aseptic system encounters tremendous experimental difficulties.
Biological methods to validate the process are likewise complex to implement for an aseptic
process of a liquid food containing particulates (de Ruyter, 1973). A possible approach to
process establishment is through modeling (Chandarana, 1989)
Models seek to predict the lethality delivered to the product, and in particular that to the cold
point of the particle, as it travels through the continuous heat-hold-cool aseptic system. Early
models were developed by De Ruyter and Brunet (1973) and Manson and Cullen (1974).
These models were based on some simplifying assumptions and did not study all the
parameters involved in an aseptic process. Both models assumed temperature independent
thermophysical properties for the particle and an infinite heat transfer coefficient at the
particle boundary. Later models were both mathematically more advanced and more
inclusive in testing for the effect of variation in parameter values. Lee (1988) and Chandarana
Introduction 2
(1989) used their models to study the effect of variation in the thermophysical properties of
the particle on the time to receive adequate lethality at the particle center. Their models
demonstrated an effect of particle diffusivity on the process time. Larkin (1990) performed a
sensitivity analysis of a one-dimensional heat transfer model for a heat-hold-cool aseptic
system of food products containing particulates. The model system consisted of a fixed
processing system with constant heater length and diameter and constant hold tube length.
Particulates were modeled as spheres. The exit temperature of the fluid from the heater was
kept constant by varying the steam temperature of the heater. Eighteen different design,
product and process control variables were considered, and apart from the direct effect of
fluid specific heat, all were found to significantly affect the amount of lethality delivered to
the product. With respect to particle properties he wrote: "Very little work has been done on
measuring the thermal properties of particulates and on any changes in the thermal properties
that may result during the processing of a food product. [...] Considering the importance of
the thermal properties of fluid and particulates and the lack of any research in this area
[measurement of properties at aseptic processing temperatures], work in this area should be
initiated."
The thermal properties that are generally considered in the context of heat transfer are:
specific heat, thermal conductivity, thermal diffusivity, enthalpy, and surface heat transfer
coefficient. The latter is not a thermal property of a material per se but it is influenced by
thermal properties. Thermal conductivity is a measure of the ability of a material to conduct
heat. In foods, the thermal conductivity is primarily dependent on composition, but also on
any factor that affects the heat flow paths such as porosity and shape, size and arrangement of
void spaces, homogeneity, fibers and their orientation (Sweat 1986). Thermal diffusivity
relates the ability of a material to conduct heat to its ability to store heat. It is used in the
analysis of transient heat transfer in solids. For lack of accurate methods to determine it
directly, Sweat (1986) recommended calculation of thermal diffusivity from thermal
conductivity, density and specific heat.
Introduction 3
The majority of the research on thermal conductivity and diffusivity of food materials has
been conducted in the low temperature range. Aseptic processing generally takes place at
high temperatures, above 100 °C but rarely exceeding 150 °C. Extrapolation of data obtained
at lower temperatures may fail to take into account the effect of temperature on thermal
properties, be it related to temperature alone or to temperature induced changes in the product
(moisture loss, protein denaturation, starch gelatinization). Few researchers have measured
thermal properties at temperatures above 100 °C. Baghe-Khandan and Okos (1981) measured
the thermal conductivity of beef through the temperature range 30 - 120 °C. Choi and Okos
(1983) measured the thermal conductivity and thermal diffusivity of tomato concentrate. The
measurements were performed over the range from 20 to 150 °C. Kravets (1988) measured
the thermal conductivity of milk and cream from 25 to 125 °C. The thermal conductivity of
starch gels was measured over the range 80 - 120 °C (Wang, 1992). Gratzek and Toledo
(1993) measured the thermal conductivity of potato and carrot from 25 to 130 °C. As this
brief summary shows, there is a very limited amount of thermophysical property data
available in temperature ranges applicable to aseptic processing.
High temperature thermal property measurement of moist food materials demands a method
that is both applicable to moist foods and adaptable to use at high temperatures. Steady state
methods for thermal conductivity measurement require long equilibration times and are
therefore not suitable for food materials in which moisture migration may take place during
the measurement. Aseptically processed foods fall in that category. Also, these methods only
yield a value for thermal conductivity. Independent measurements of density and specific
heat at aseptic temperatures are required in order to calculate thermal diffusivity. These
measurements are no small task. Transient state methods have the advantage of being rapid.
They are therefore better suited for measuring moist food materials. Measurement above the
atmospheric boiling point of water poses design problems for some of the transient methods.
The direct measurement of thermal diffusivity of food materials at high temperatures has
been done with methods that were based on: heat penetration, the line heat source probe, the
Introduction 4
bead thermistor. These methods will be evaluated with one further constraint in mind. The
method should be capable of measuring small samples.
The thermal diffusivity of carrot cubes was determined at UHT temperatures (130 - 150 °C)
from heat penetration data (Chang and Toledo, 1990). Cubes of 2 cm were subjected to
flowing water at 140 °C. The center temperature was monitored. Analysis of
time-temperature data was done with a finite difference scheme. A Biot number was
assumed. Thermal diffusivity was found by trial and error. The same researchers determined
thermal diffusivity and heat transfer coefficient simultaneously for 1 and 2 cm carrot cubes in
a packed bed, placed in flowing water of 140 °C. Sample temperatures were measured at the
center and at a point midway between the center and the surface. Trial and error solutions
with the finite difference method yielded both heat transfer properties. Thermal diffusivity
found through both measurements were in agreement, 1.88 10-7 and 1.94 10-7 m2/s
respectively. The corresponding thermal conductivities, were 0.74 and 0.77 W/m°C. These
values are somewhat higher than the value of water at that temperature, 0.69 W/m°C. This
method is computation intensive and demands great accuracy in the placement of the
temperature sensor. The latter becomes more critical as the sample size decreases.
Among the transient state methods, the line heat source probe is most commonly used for
measurement of thermal conductivity, both at low and at high temperatures. Sweat and
Haugh (1974) developed a miniature line heat source probe for the measurement of thermal
conductivity of small cylindrical samples. Choi and Okos (1983) and Casada and Walton
(1989), used an additional temperature sensor with known distance to the probe for direct
determination of thermal diffusivity. The drawback of this procedure is that the distance
between the sensor and the probe has to be known accurately. For a solid food this may give
experimental difficulties as insertion into the sample can alter its location relative to the
probe. The additional temperature sensor for thermal diffusivity measurement will increase
the sample diameter, making it less suited for measurement of small particles.
Introduction 5
A method that can simultaneously measure thermal conductivity and diffusivity is the
thermistor probe method. The method was developed by researchers in the biomedical field
for measuring thermal conductivity and diffusivity of human tissue (Chato, 1968;
Balasubramaniam, 1974; Valvano, 1981). Its use for measurement of food thermal properties
is limited to one study; thermal conductivity of liquid milk products was measured over the
temperature range 25 to 125 °C (Kravets, 1988). The thermistor based method utilizes a
small thermistor probe whose active region approximates a sphere. Because of the small
diameter of this region, beads with diameters of 2.54 mm and smaller are commonly used, it
is applicable to small samples. Kravets (1988) concluded that samples could be as small as
5 mm in diameter. The thermistor is most suited for measurement of materials that can close
tightly around or wet the probe, ensuring a low thermal contact resistance (Balasubramaniam,
1975). Fluids and human or animal tissue have been the most frequent test materials.
Woodbury (1984) and Dougherty (1987) also measured the thermal conductivity of
insulation materials. The composition and rheological properties vary between solid foods
and can be temperature dependent for a given material. These food characteristics may affect
the adequacy of thermal contact between probe and food and can ultimately determine the
suitability of the thermistor based method of thermal property measurement of those food
materials.
The thermistor based method is not an absolute method. It requires calibration with materials
whose thermal conductivity and diffusivity are well documented. Available reference
materials whose thermal conductivity is comparable to that of moist food materials are
liquids. A thermal property measurement with the thermistor involves establishing a thermal
gradient in the test medium. For a valid measurement, heat penetration has to occur solely by
conduction. It has been observed that the temperature gradient will induce convection in
certain liquid media. Balasubramiam (1975) noticed the departure of measurement data for
water from those for agar-gelled water and ascribed that to convection. Dougherty (1987)
noted that the thermistor based method may not be suitable for low viscosity fluids because
of convection. Approaches traditionally taken to prevent or minimize the occurrence of
Introduction 6
natural convection use thickeners or flow inhibitors (angel hair). In both cases, an additional
material is introduced with the potential of altering the thermal properties of the material
under investigation. In general, the concentrations of these convection inhibitors are low and
are assumed to have near negligible effect on the measured properties. No studies were found
that attempt to relate fluid properties and heating power to the occurrence of convection and
its onset time.
Thermal property data of food materials are relevant for the establishment of aseptic
processes. However, they are scarce. The thermistor based method holds promise as a method
for simultaneously measuring thermal conductivity and diffusivity at high temperatures. In
particular the latter is an advantage as it avoids the independent determination of density and
specific heat. Whether the method is suitable for use with solid food materials is a question
that merits answer. Adequate thermal contact between the very small thermistor bead and the
food material can be a problem, because of either inherent product characteristics or those
effectuated by heating at high temperatures. Thermal contact is no issue when a fluid is
sampled with a thermistor. Convection can however disrupt a measurement. Its occurence is
related to fluid properties and method variables that have not been the subject of previous
investigations. The following objectives were formulated to address the issues discussed in
the foregoing.
Objectives:
1. Investigate the suitability of the thermistor based method for measurement of thermal
conductivity and diffusivity at temperatures up to 150 °C,
2. Investigate the adequacy of thermal contact during bead thermistor probe thermal
property measurements in solid foods,
3. Assess the effect of medium viscosity and heating power on the quantification of
medium thermal properties,
4. Measure the thermal conductivity and diffusivity of moist food materials at
temperatures above the atmospheric boiling point of water.
Introduction 7
CHAPTER 2
LITERATURE REVIEW
This chapter presents the results of the literature research. Some methods for measuring
thermal conductivity and/or diffusivity of food materials will be discussed. These methods
can roughly be divided in steady state and transient methods. The research reported in this
thesis uses the thermistor probe method. Literature pertaining to this method for
measurement of thermal conductivity and diffusivity will be treated in great detail. However,
the heat transfer model and its application is omitted here and presented in the next chapter.
When the thermistor probe is used with liquid media, natural convection can be an unwanted
phenomenon. The thermistor probe relies on liquid media as calibration materials as they are
the most suited for this purpose, making natural convection of particular interest.
2.1 STEADY STATE METHODS FOR MEASUREMENT OF THERMAL

CONDUCTIVITY
2.1.1 Introduction
Methods that employ steady state measurement of thermal conductivity apply Fourier’s law
of heat conduction:
Q = kA ∆T (2.1)
∆x
where: Q = heat flow [W]

k = thermal conductivity [W/m°C]
A = area perpendicular to heat flow [m2]
DT = temperature difference [°C]
Dx = thickness [m]
Literature Review 8
When rewritten, the thermal conductivity of a material can be expressed as a simple function
of surface temperatures or heat fluxes and sample dimensions. Different methods are based
on steady state heat flow. Underlying all is the attempt to simplify the mathematics by
reducing the heat transfer problem to a one-dimensional problem. The results
are methods that use the infinite slab, the infinite cylinder or sphere as model.
2.1.2 Guarded Hot Plate
The guarded hot plate uses the infinite slab as heat transfer model. Since sample dimensions
are finite, guard heaters are used to facilitate uni-directional heat flow. It is considered the
most accurate and most widely used method for the measurement of thermal conductivity of
poor conductors of heat (Mohsenin, 1980). It is most suitable for dry homogeneous samples
that can be formed into a slab. The sample is sandwiched between a heat source and a heat
sink. Thermal guards are kept at the same temperatures as their adjacent surfaces (heat
source, heat sink) to prevent heat leakage from heat source, sample and heat sink, thus
ensuring uni-directional heat flow. The heat input is monitored. After steady state has
developed, as shown by stable temperatures of the heating and cooling plate, the thermal
conductivity can be calculated from the heat input, the temperature differential across the
sample, and the sample dimensions thickness and heat transfer area. Since steady state
conditions may take several hours to develop, this method is unsuitable for use with material
in which moisture migration may take place. The method has been used for measuring the
thermal conductivity of dried or frozen foods.
2.1.3 Radial Heat Flow Method
Whereas the guarded hot plate is generally used for measuring the thermal conductivity of
samples that can be formed into a slab, radial heat flow steady state methods are more
commonly used with powdered or granular material.
Literature Review 9
A cylindrical test device employs a central line (or cylindrical) heat source. End effects are
assumed negligible due to either the large length to diameter ratio of the test apparatus or the
use of end guard heaters. After steady state has been established, the thermal conductivity can
be calculated from the heating power, the length of the cylinder, the temperature differential
between two internally (to the medium) located sensors and their radial position.
Alternatively, the cylindrical sample can be bordered by the heater on one side and a
reference material on the other side. Temperatures on all surfaces are monitored. The thermal
conductivity of the test material is calculated from the temperatures, the radial position of the
sensors and the thermal conductivity of the reference material.
2.2 TRANSIENT STATE METHODS
2.2.1 Introduction
A disadvantage of steady state methods are the long measurement times involved. Materials
containing moisture may experience moisture migration before steady state conditions
develop. Consequently, a steady state method will fail to measure representative properties.
Therefore, transient methods are preferred for measuring the thermal conductivity of moist
food materials. They are rapid as data are obtained in minutes or even less compared to hours
for a steady state measurement.
Steady state methods are only capable of measuring thermal conductivity. When the
measured properties are to be used in the study of transient heat transfer, density and specific
heat have to be found independently. They are combined with thermal conductivity to find
the thermal diffusivity, the material property of interest. For lack of accurate methods of
direct measurement of thermal diffusivity, this approach is recommended (Sweat in Rao,
1986). Methods based on transient heat transfer have the potential of directly determining
thermal diffusivity, but they are not as accurate as steady state methods with dry materials
(Mohsenin, 1980).
Literature Review 10
Three transient methods will be treated. The Fitch method has applications for small samples
but is not practical for use at high temperatures. It is mainly included here for completeness.
The line heat source probe is widely used for measurement of thermal conductivity of food
materials. This report contains a comparative study between the line heat source probe and
the focus of this research, the thermistor probe.
2.2.2 Fitch Method
The Fitch method was developed by Fitch in 1935 (Mohsenin, 1980) and uses a plane source
of heat. It is applicable to poor conductors of heat that can be formed into a slab. The Fitch
apparatus consists of a heat source and a heat sink. The heat source is a constant temperature
vessel, insulated on the sides and with a highly conductive bottom. The heat sink is a copper
block, insulated on all sides but the one facing the vessel. The heat transfer areas are smooth
to minimize contact resistance. The roles of heat sink and heat source are reversed when the
vessel is maintained at a temperature lower than that of the copper block. Initially the sample
is in thermal equilibrium with the copper block. The vessel, maintained at a temperature
different - typically by about 20 °C - from this temperature, is brought into contact with the
sample. Thermocouples record the temperature history of the copper block, which is assumed
to have a uniform temperature distribution, and the temperature of the bottom of the vessel.
The model assumes a linear temperature profile, negligible heat storage in the sample, and
negligible surface contact resistance. Heat transfer between the copper block and its
insulation is assumed negligible. Using these assumptions, an expression was derived for the
temperature of the copper block:
T0 − T∞
ln( ) = kA t , (2.2)
T − T∞ Lm c c pc
where: A = heat transfer area, [m]

k = thermal conductivity of sample [W/m°C]
L = thickness of sample [m2]
mc = mass of copper block [kg]
cpc = specific heat of copper [J/kg°C]
t = time [s]
T = temperature of copper block [°C]
T0 = initial temperature [°C]
T∞ = constant temperature of vessel bottom [°C]
The plot of the dimensionless temperature against time on semilog paper is linear with an
initial curvilinear part. The thermal conductivity of the sample is calculated from the slope of
linear part of this plot, the heat capacity of the copper block, and the thickness and heat
transfer area of the sample.
Murakami et al. (1984), described a modified Fitch apparatus. The apparatus had a largely
reduced heat transfer area compared to the regular Fitch apparatus. The heat sink and heat
source both were 6.35 mm in diameter. This setup was suitable for measuring small disc
shaped food particles. Measurements of thin, 3.25 mm, samples of cork showed a sensitivity
to variations in sample homogeneity.
Zuritz et al. (1989) used a modified Fitch apparatus very similar to that of Murakami et al.
(1984), to measure the thermal conductivity of kidney beans. Model error analysis showed
that the errors associated with sample thickness and heat transfer area had the strongest
influence on the error in sample conductivity. Ensuring good contact between sample and
heat source and sink was mentioned as being critical.
Rahman (1991), used a modified Fitch apparatus with a 15 mm-diameter heat transfer area
for measurements of moist food materials above and below freezing and reported an accuracy
of 5 and 9% respectively (as compared to measurements with a line heat source probe). The
effect of sample thickness was investigated, and was shown to have a significant effect on the
quantification of thermal conductivity.
The assumptions of linear temperature profile and negligible heat storage favor a small
sample thickness. On the other hand, a larger thickness will have lower measurement error.
Also, Zuritz et al. (1989) mention that a fast response decreased the linear portion of the
graph (for a modified Fitch apparatus). Optimum sample thickness for the Fitch apparatus is
given as 6.4 to 13 mm (Mohsenin, 1980). The modified Fitch apparatus were used with much
thinner samples. Thickness varied from 1.1 mm slabs of kidney bean (Zuritz, 1989) to
4-7 mm for moist food samples (Rahman, 1991). The latter tested samples of 15 mm in
diameter.
The sample times for the regular Fitch apparatus are in the range of 10-20 minutes
(Mohsenin, 1980). Modified Fitch apparati lose linearity for longer sample times probably
due to heat transfer between the copper block and the insulation (Murakami, 1984). These
apparati were generally used with sample times less that 8 minutes.
2.2.3 Transient Hot Wire
This is one of the most commonly used transient methods in particular with granular
materials (Mohsenin, 1980). It is also known as the line heat source method. The theory is
based on an linear heat source of infinite length and infinitesimal diameter. The line heat
source is embedded in the material whose thermal conductivity is to be measured. From a
condition of thermal equilibrium, the heat source is energized and heats the medium with
constant power. The temperature response of the medium is a function of its thermal
properties. The thermal conductivity is found from the temperature rise measured at a known
distance from the heat source.
An often used implementation of the line heat source method for the measurement of thermal
conductivity of food materials is the line heat source probe also known as the thermal
conductivity probe.
2.2.4 Line Heat Source Probe
The line heat source probe is based on the theory of the line heat source method. Probes are
constructed around a rigid straight rod like, e.g. a metal tube. A heating element and a
thermocouple are placed inside, or on the outside of, the tube or rod. Internal placement
makes the probe more robust. External placement of the heater and thermocouple has been
reported to yield data of higher accuracy and with lower standard deviation (Baghe-Khandan,
1981b) than internal placement for a comparable sized probe. Large probes of about 47 cm
length and 0.47 cm in diameter were used for measurements in soil (Hooper, 1950). For
measurement of thermal conductivity of small food samples Sweat and Haugh (1974) used a
miniature probe. Their probe was made from a narrow tube like a hypodermic needle. The
probe was 38 mm long and had an outside diameter of 0.81 mm. An insulated heater wire
(0.0762 mm diameter with 0.0762 mm teflon coating) was inserted along the full length of
the tube. Constantan wire was used because of its small temperature dependence of
resistance. An insulated thermocouple (Type E, chromel-constantan, 0.0762 mm diameter)
was located inside the tube at the halfway point. The tube was bonded to an Omega miniature
thermocouple connector.
To make a measurement, the probe was embedded in the material of interest. From thermal
equilibrium, the probe heater was energized and heated the medium with constant power. The
temperature rise at the heat source was monitored. Following a brief transient period, the plot
of the temperature versus the natural logarithm of time became linear. The slope was equal to
Q/(4πk). The thermal conductivity could be calculated from the relation:
ln [(t 2 − t 0 )/(t 1 − t 0 )]
k=Q , (2.3)
4π(T 2 − T 1 )
where: k = thermal conductivity of sample [W/m°C]

Q = power dissipation by heater wire [W/m]
t1 = time since probe heater was energized [s]
t2 = time since probe heater was energized [s]
t0 = time correction factor [s]
T1 = temperature of probe thermocouple at time t1 [°C]
T2 = temperature of probe thermocouple at time t2 [°C]
The time correction factor t0 compensates for the finite size of the probe and for differences
in properties between sample and probe material. Its value is calculated from the data. Sweat
and Haugh (1974) did not require time correction factors when measuring the thermal
conductivity of meat at temperatures above freezing. At those temperatures, the heat capacity
of the probe was reasoned to be similar to that of food. Also the small diameter probe they
used (0.81 mm) closely approximated a line heat source.
Another source of error with the line heat source probe is axial flow error due to its finite
length. Hooper (1950) recommend a length to diameter ratio of no less than 100. Sweat and
Haugh (1974) found no statistical difference, at the 10% level, between probes with ratios
varying from 93 down to 31 when testing glycerin. In line with these findings were those
obtained by them when measuring agar-water samples that were shorter than the probe
length. Samples of different length were measured. They were centered on the location of the
thermocouple junction. The samples were bounded on both ends by teflon which has
approximately half the thermal conductivity of water. No statistical effect was noticed for
samples as small as half the probe's length.
The line heat source probe has the advantage of being a rapid method. Sweat (1986) reports
measurements times from 3 seconds for liquid to 10 to 12 seconds for most solid foods.
Besides the advantage of being rapid, the probe requires only a small sample size. Although
the theory requires the medium to be of infinite size, in practice the temperature field is
limited to a narrow range around the probe. Vos (1955) showed that when the expression
4αt/d2 is greater than 0.6, an error will be noticeable due to the small sample diameter. In the
above expression, α is the thermal diffusivity of the sample, t is the time since the heater was
energized and d is the shortest distance between the probe and the sample boundary. For the
probe of 38 mm length and 0.81 mm in diameter a sample diameter of 19 mm will suffice
when measuring a thermal conductivity close to that of water.
Although often reported as an absolute method, it is recommended to calibrate the probe.

Calibration also takes care of inaccuracies in measured resistance of heater wire and of
temperature effects on resistance of wire. Agar-gelled water and glycerin are commonly used
as thermal references. The line heat source probe is probably no more accurate than 2%
(Sweat, 1986). Sweat and Haugh (1974) estimated the accuracy for measurements above 0 °C
to be about 5%.
The line heat source probe is mainly used for measurement of thermal conductivity. Choi and
Okos (1983) used an additional temperature sensor with known distance from the probe to
measure thermal diffusivity of tomato juices. Using only a single probe, thermal diffusivity
was estimated by numerical methods with heating and/or cooling data by Lan and Sweat
(1992).
2.2.5 Thermistor Based Method
A thermistor is a temperature dependent resistor. Those used for thermal property

measurement have a negative temperature coefficient of resistance, indicating that their
resistance will decrease when their temperature is increased. Being resistive devices, a
thermistor will heat when a current is passed through it. Both phenomena are exploited in the
thermistor based method for thermal property measurement. The thermistor is used both as a
temperature sensor and as a point heating source. The relationship between temperature and
resistance is non-linear and is usually described by the empirical Steinhart-Hart equation
(Thermometrics, 1987):
1 = a + a ln (R) + a [ln (R)] 3 (2.4)

0 1 2
T
where: T = temperature, [K]
R = resistance, [Ω]
a0, a1 and a2= derived constants.
The constants a0, a1 and a2 are obtained in a calibration procedure.
In most studies concerning the measurement of thermal conductivity and diffusivity, a glass
coated probe was used. Figure 2.1 shows a typical probe as was used in this research.
Figure 2.1 Glass coated bead thermistor probe. All dimensions in mm.
Early work on the use of a thermistor for measuring thermal conductivity took place in the
biomedical field. Chato (1968) developed a transient method for measuring thermal
conductivity and thermal inertia, k ρ cp, of biological materials. The thermistor was modeled
as a spherical heat source, with negligible thermal capacitance and negligible surface contact
resistance, embedded in an infinite homogeneous and isotropic medium. The medium was
treated as a semi-infinite solid. He solved the heat conduction equation in the medium for the
case in which, starting from an equilibrium situation, the surface temperature of the
thermistor is suddenly elevated with a predetermined increment and held constant at the
elevated temperature. The temperature solution for the medium is given by:
T − T0 a  
= r erfc  r − a  (2.5)
Ta − T0  2 αt 
and the heat dissipation by the medium is given by:
q = − k 4πa 2  δT  = 4πak(T a − T 0 ) + 4a 2 (πkρc p ) (T a − T 0 ) t −1/2 (2.6)
δr r=a
where: a = radius of thermistor bead [m]

cp = specific heat of medium [J/kg°C]
k = thermal conductivity of medium [W/m°C]
q = heat dissipation function [W]
r = radial coordinate [m]
t = time [sec]
T = temperature in medium [°C]
Ta = surface temperature of bead for t > 0 [°C]
T0 = initial temperature [°C]
a = thermal diffusivity of medium [m2/s]
r = density of medium [kg/m3]
The heat dissipation by the thermistor is thus seen to be linear in the inverse square root of
time. Linear regression of measurement data can be used to find the thermal conductivity
(from the intercept of regression), and the thermal inertia (from the slope of regression).
This method required calibration with a medium of known thermal conductivity to find the
effective bead radius. Beads were calibrated with agar-gelled water (1.5 and 1.75%) and
paraffin. Chato (1968) was successful only in measuring thermal conductivity for which he
reports an accuracy of 20%. Thermal inertia could not be determined due to inconsistent
calibration results. Furthermore, the slopes of regression were much more sensitive to
experimental error than the intercept. The method was rapid. Thermal properties could be
measured within two minutes.
Balasubramaniam (1974, 1975, 1977) and Bowman (1976) improved on Chato's heat transfer
model by acknowledging thermal gradients in the thermistor bead: both thermistor and
medium were treated as distributed thermal masses. The thermistor was modeled as a sphere
embedded in an infinite homogeneous and isotropic medium with no surface contact thermal
resistance. This coupled heat transfer problem was solved for the case where starting from a
thermal equilibrium, the temperature of the thermistor was raised instantaneously to, and
maintained at, an elevated temperature. The required heat generation was assumed to take
place uniformly throughout the bead. This heat generation was experimentally determined to
be of the form Γ + β t-1/2. The temperature distributions of thermistor bead and surrounding
medium were presented as indefinite integrals which had to be solved numerically. At steady
state a simple equation was obtained for measuring thermal conductivity, k m:
1 = 4πa ∆T − 0.2 (2.7)

km P ss k b
In the above equation, DT/Pss is the ratio of temperature step and the steady state power
dissipation in the thermistor required for maintaining DT. Calibration with two materials of
known thermal conductivity was necessary to find values for the effective probe radius, a,
and effective probe thermal conductivity, kb. Medium thermal diffusivity was determined
simultaneously through a somewhat involved procedure which required knowledge of the
bead thermal diffusivity (Balasubramaniam, 1974, 1975, 1977), Bowman (1976). Sample
times were short, 20 to 60 seconds. The method was reported to be capable of estimating
thermal conductivity and diffusivity with an accuracy of 1 - 2% and 10% respectively.
Valvano (1981) mathematically showed the heat generation function to be of the form of
Γ + β t-1/2. Closed form approximations for the temperature distributions were presented
where Balasubramaniam had given them as indefinite integrals (see the next chapter for a
more detailed treatment of the heat transfer model). Valvano also showed that while the
thermal conductivity of the bead plays a role in determining the medium thermal
conductivity, the thermal diffusivity of the medium could be calculated without knowledge of
the bead thermal diffusivity. However, an additional calibration with two media of known
thermal diffusivity was required. Agar-gelled (1.5%) water and glycerol were used as
calibration media to find the probe properties. Valvano (1984) tested a thermistor probe as an
interpolating instrument. After calibration with agar gelled water and glycerol, the thermistor
was used to estimate the thermal conductivity and diffusivity of three mixtures of water and
glycerol. The thermal property measurement was found to be accurate within 2% and
reproducible to 1%. Valvano (1984, 1985) used a sample period of 20 seconds with data
analysis by linear regression over 4 - 18 seconds. Tests with longer periods (30, 60 and 120
seconds) yielded no different thermal properties (Valvano, 1985). Tests with different levels
of agar (0.5 - 5%) showed no discernible effect of concentration on thermal properties
(Valvano, 1985). Tests over the temperature range 3 - 45 °C (every 7 °C) on biological
tissues showed that temperature dependence of the tissues followed that of water (except for
fat, lung and cancer tissues) (Valvano, 1985).
Woodbury (1984) used self-heated thermistors to measure the thermal conductivity of

building insulation with varying degree of wetness. He employed teflon coated probes with a
diameter of 2.79 mm. Balasubramaniam and Valvano had used smaller glass encapsulated
(and some bare) probes. Small probes heat up faster, more closely mimicking a step change
in temperature. Large probes increase the sample volume, reduce the effect of imperfect
contact at the thermistor medium interface and are more durable. Woodbury noticed two
linear parts in the plots of power versus the inverse square root of time, an observation not
reported by previous researchers. The first linear part was initially used as calibration interval
(10 to 20 seconds). It was followed by a concave part. For long times, larger than 60 sec. for
glycol and larger than 200 sec. for the other two materials (both insulation materials), the plot
became linear again. Regression over the first linear part gave smaller values for the
intercept, and thus for the 'steady state' power consumption, than regression over the second
linear part. Woodbury considered the intercept as found from this latter part of the data the
true estimator of steady state power consumption. The difference between the intercepts
found with the two parts of the data was related to the thermal conductivity of the medium.
Larger differences were found for media with smaller thermal conductivities. Long term tests
were done with all probes in all calibration media and correction ratios (CR's) were found.
The CR is the ratio between 'true' Vss2 and Vss2 found from regression over 10 - 20 seconds. A
linear relationship was assumed between CR and km. Linear regression on the CR-km data for
several probes showed that for a medium thermal conductivity of 0.40 - 0.50 [W/m°C] the
value of CR would be unity. For larger medium thermal conductivities, the graph should
display a convex part instead of a concave part. Tests with agar-gelled water at 25 °C (km is
0.61 W/m°C) confirmed this. Probe parameters were recalculated with the 'true' values of
Vss2. Calibration error (error in estimating the thermal conductivity of the calibration media in
a 3-media calibration) was given as 4.6%.
Larger probes and stem losses were put forward to explain the observed two-part graph.
Balasubramaniam and Valvano had used small glass encapsulated probes; Woodbury used
large teflon encapsulated probes. Heat loss through the probe stem was probably larger in his
study because the thermal conductivities of the test media (0.03 - 0.25 W/m°C) were far
lower then encountered by the previous researchers, who dealt with animal and human tissue
(0.3 - 0.6 W/m°C), and the probes had larger diameter lead wires and a metal support rod in
the stem.
Dougherty (1987) tested glass and teflon encapsulated probes with a variety of materials,
ranging from low viscosity fluids like toluene to insulation materials. Glass encapsulated
thermistors appeared to perform better in fluids whereas the teflon encapsulated thermistor
performed better with insulation materials. Based on visual observations of test data, the
researcher noted that the method may have a limited applicability due to either convection
losses in low viscosity fluids or stem losses in low thermal conductivity materials. He
concluded that the thermistor method is most suited for viscous liquids of higher thermal
conductivity like castor oil and glycerol. An alternate approach to the probe operation was
investigated. Previous researchers had used analog circuitry, developed by Balasubramaniam,
to control the thermistor's temperature at the preset level. Voltage was the controlled variable.
Dougherty's instrumentation used digital control to regulate the current to the probe.
Kravets (1988) was the first to use self-heated thermistors at elevated temperatures. Glass
encapsulated probes were used to measure thermal conductivity of milk and cream over the
range of 25 to 125 °C. This temperature range was divided in four sub-ranges because the
thermistors could be used only over certain resistance - and thus temperature - range. The
control circuitry limited the allowable thermistor resistance to the range 500 and 3500 Ω. The
assumption of an infinite boundary, which was made in the development of the heat transfer
model for the thermistor based method, was experimentally tested. A sample diameter of
5 mm was found to satisfy this assumption for a probe of 1.5 mm in diameter.
From error analysis of the measurement of thermal conductivity, Kravets concluded that: a)
error in estimation of medium thermal conductivity decreases with a higher thermal
conductivity of the probe; b) a temperature step of minimally 2.5 °C has to be used to
minimize errors in medium thermal conductivity; c) more accurate thermal conductivity
reference materials are needed to minimize the errors in estimated bead parameters.
Of the methods discussed, a self-heated thermistor promises to be the best for measurement
of thermal conductivity and thermal diffusivity. The line source probe in combination with an
additional temperature sensor is capable of measuring thermal diffusivity but requires a much
larger sample volume than a thermistor. A disadvantage of the thermistor based method is
that it is not absolute. Calibration with at least two reference materials is necessary. The
calibration is required at every measurement temperature and demands thermal standards for
both thermal conductivity and thermal diffusivity.
Alternate Modes of Operation of Thermistor Probes
The mode of operation of the thermistor in the foregoing was one of fixed temperature step:
the thermistor is maintained at a fixed elevated temperature and the required power is
monitored. Alternate modes of operation have been investigated and are briefly discussed in
the following. They were developed to address the issue of perfusion in human tissue.
In the biomedical field, research into the use of thermistors for measurement of thermal
conductivity focused not only on the tissue thermal conductivity itself, but also on thermal
conductivity as an estimator of tissue perfusion. The thermistor based method that maintains
a fixed elevated bead temperature requires a no-flow calibration in order to be able to extract
intrinsic tissue thermal conductivity from the effective thermal conductivity that is measured
in the presence of perfusion. Alternate modes of probe operation were investigated that
would forego the no-flow calibration. Using a sinusoidal varying voltage supply, Valvano et
al. (1987) , Anderson et al. (1988), and Anderson and Valvano (1989) simultaneously
measured intrinsic and effective tissue thermal conductivity. Accuracy in measurement of
stationary media was no better than when the probe was operated with a fixed temperature
increment. Determination of thermal diffusivity was not performed.
Chen and Rupinskas (1977) measured thermal conductivity invasively with a thermistor
probe by supplying the thermistor with a heating pulse and monitoring the subsequent
temperature decay. The pulse-decay method is considered to be absolute, calibration should
not be required. Limited results are given in the 1977 paper for three fluids. These indicate
the necessity for further research before this method can be used to determine thermal
conductivity with a large degree of accuracy. The pulse-decay method was later extended to
predict perfusion in biological tissues as well (Chen and Holmes, 1980, Arkin et al., 1986).
The invasive probe technique causes some tissue trauma due to insertion of the probe. To
measure tissue properties and perfusion non-invasively, Patel et al. (1987) developed a
surface probe. A glass encapsulated thermistor was partly embedded in an insulation material
to direct the heat flow downward. The bottom half was brought into contact with the medium
of interest. The probe was not operated with a fixed temperature increment. Instead, a
precision voltage powered the thermistor in series with a precision resistance. Power
dissipation by and temperature response of the thermistor were monitored. Thermal
conductivity and diffusivity could be determined simultaneously. The accuracy of the
thermal properties of water-glycerin mixtures, measured with a probe that was calibrated in
agar gelled water and glycerin, was comparable to that obtained with an invasive probe.
Experiments with rat liver tissue, revealed a wide scatter in the data. This was attributed to
poor contact between probe and surface. Valvano (1990) commenting on the surface probe
method stated that it was "unreliable due to poor contact with tissue and uncertain boundary
conditions at the contact surface".
2.3 NATURAL CONVECTION
One of the assumptions when using the thermistor based method is that heat transfer from the
thermistor bead occurs solely by means of conduction. In a liquid medium, this assumption
can be violated through natural convection. Localized heating by the thermistor bead induces
density gradients in the initially isothermal and quiescent medium. The resulting buoyancy
force will be opposed by a combination of inertia and viscous forces. Convection then may
occur depending on the interplay of these forces. Their magnitude depends on the properties
of the fluid, the size of the probe, and on the imposed temperature step. It is of interest to
know if there is a means of assessing the potential for convection during a thermal property
measurement with a thermistor. A literature study was undertaken to develop insight into the
phenomenon of transient natural convection.
The onset of natural convection was first experimentally studied by Bénard for thin
horizontal fluid layers heated from below (Chandrasekhar, 1961). Bénard’s experimental
work motivated Rayleigh to undertake an analytical study for the case of an infinite wide
fluid layer heated from below. For a fluid layer between two horizontal boundaries and
heated from below, cellular convection will be initiated once the Rayleigh number exceeds a
critical value of approximately 1108. The Rayleigh number was defined as:
gβ∆Td 3
Ra = αν (2.8)
where: g = gravitational acceleration [m/s2]

β = coefficient of thermal expansion [1/K]
∆T = temperature difference between surface and fluid [°C]
d = characteristic length [m]
α = thermal diffusivity [m2/s]
ν = kinematic viscosity = µ/ρ [m2/s]
µ = dynamic viscosity [Pa·s]
ρ = fluid density [kg/m3]
The onset of natural convection around a suddenly heated horizontal wire was investigated
by Vest and Lawson (1972). Experiments were performed in air and water. An interfero
meter registered the shape of the isotherms after application of the current. The earliest
distortion of the isotherms - an indication of convection effects - were observed above the
wire. Based on this observation the heating situation was taken to resemble that of the Benard
problem of a fluid layer heated from below. The Rayleigh number at the onset of convection
was assumed to be 1100. Vest and Lawson defined the Rayleigh number in terms of the
strength of the heat source and an instantaneous heat penetration depth. The latter was found
from an approximate solution to the heat transfer problem. A relationship was developed
between the onset time (called the delay time) and the energy dissipation in the wire, with the
fluid properties as parameters. The delay times for water agreed fairly well with the
prediction. For air, though following the predicted trend, the relationship consistently
underpredicted the delay time.
Parsons and Mulligan (1980) investigated the onset of natural convection in air surrounding a
cylinder. A steel rod (6.3 mm diameter) was suddenly heated with an electrical current.
Experimentally observed onset times of convection compared favorably against those that
were predicted. The latter were calculated from the solution to the heat transfer model, and an
assumed value of 1100 for the Rayleigh number at onset of convection. The Rayleigh number
was numerically evaluated from the exact solution to the heat transfer problem using an
effective temperature gradient, (Tw - Tw)e. The effective temperature gradient was taken as the
linearized temperature function that integrated to the same value - over the heat penetration
depth - as the integrated exact solution. Although the problem was approached as that of a
fluid layer heated from below, the researchers observed that localized convection currents
near the vertical surfaces of the rod occurred prior to the onset of global convection.
Genceli (1980) investigated the onset of natural convection in air surrounding suddenly
heated nickel plated Plexiglas cylinders with diameters of 15, 25 and 40 mm and found
Rayleigh numbers at the onset of convection that ranged from about 20 for the 15 mm
diameter cylinder to about 30 for the 40 mm diameter cylinder.
Wang et. al (1991) numerically studied transient natural convection from horizontal cylinders
under diverse surface boundary conditions. They observed that "the onset of motion results
from the natural convection along the approximately vertical portions of the cylinder surface
[...], and not from a Benard type convective instability in the statically unstable conduction
temperature profile near the top of the cylinder".
Jia and Gogos (1996) studied transient laminar natural convection heat transfer from
isothermal spheres with a numerical method. They covered a Grashof number range of 10 to
107 for a fluid with Prandtl number of 0.72. The focus was mainly the developing plume. An
observation pertaining to the onset of convection was that "the temperature gradient in the
radial direction at q = 90° [the equator of the sphere] seems to lead the onset of convection".
Critical Raleigh numbers were calculated at the onset time of convection. The penetration
depth was defined as the depth at which the medium temperature was reduced to 1% of the
surface temperature. The temperature differential was taken as Ts - Tw. The critical Rayleigh
numbers increased with the Grashof number from about 139 for a Grashof number of 10 to
3950 for a Grashof number of 108. The authors concluded that the use of the critical Rayleigh
Number from the Benard problem (1100) was completely inappropriate [for the geometry of
their study].
In conclusion, the literature on transient natural convection is rather limited. The above cited
studies comprise about all that seemed relevant to the present study. The closest resemblance
to the specific geometry of this study, that of a more or less spherical object at the tip of a
downward pointing vertical rod, was the study of Jia and Gogos (1996) for isothermal
spheres. As far as the expected onset times of convection are concerned, the literature was of
little help. Nor could it suggest a value for the Rayleigh number that coincides with the onset
of convection. That of the Benard problem is clearly not applicable because thermistor based
measurements don't resemble that of a fluid layer heated from below. The idea of a critical
Rayleigh number, however, was intriguing.
CHAPTER 3
PRINCIPLES OF THE THERMISTOR BASED METHOD
This chapter will discuss the principles of the thermistor based method, and the assumptions
made in its development. The reality as manifested in the measurement system deviates to
some extent from the model as will be pointed out. The equations used for estimation of
thermal conductivity and diffusivity will be presented. These equations contain parameters
which are found by calibration. The approach taken for the calibration will be outlined. Some
considerations when using the thermistor based method will be discussed as they relate to the
choice of samples and measurement procedure.
3.1 HEAT TRANSFER MODEL
The thermistor based method utilizes a thermistor as a point heating source. Heat emanates
from the bead into the surrounding medium. The thermistor is operated with a constant
temperature step: heat is generated inside the bead to maintain its temperature at a constant
value above the initial equilibrium temperature. Provided some conditions are met, the power
dissipated inside the thermistor bead is related to the thermal properties of the medium.
The heat transfer model acknowledges space and time dependency of temperature in both the
thermistor bead and the surrounding medium. This coupled heat transfer model is based on
the assumptions that:
1. the thermistor is of spherical shape,
2. the thermistor is made of isotropic and homogeneous material,
3. heat is being generated uniformly in the thermistor,
4. the surrounding medium is isotropic, homogeneous and infinite in extent,
5. no contact resistance exists between thermistor and medium,
6. the mode of heat transfer is pure conduction.
Principles of Thermistor Based Method 27

The assumptions simplified the development of the model. The reality in this research
deviated from the model in several ways. Rather than being spherical, the thermistors
resembled a prolate spheroid, and had a stem attached to it. Unlike the assumption, they were
not isotropic and homogeneous, nor had heat generation uniformly throughout their volume.
Instead, a glass shell insulated the semi-conductor core from the environment. Lead wires ran
down the glass stem and attached to one side of the semi-conductor core. The electrical
pathways inside the core were many with heat generation probably not completely uniform
throughout. The probes used in this research were hand picked under a microscope for
minimal deviations from the model. Hayes and Valvano (1985) assessed the validity of the
first three assumptions with a steady state finite element model. They found "that the
empirical linear relationship (between the power consumption of the probe and the inverse
square root of time) remains valid when all three assumptions are significantly relaxed".
As mentioned in Chapter two, the thermally affected region is small. Kravets (1988)
experimentally determined this region to be no more than 5 mm in diameter for a thermistor
with a bead diameter of 1.5 mm. A sample larger than this region can be considered infinite.
Whether a sample material can be seen as homogeneous and isotropic depends in general on
the scale of its internal variations compared to the scale of the measurement volume. With the
thermistor this volume is very small as mentioned above. The degree of homogeneity has to
be evaluated with this small volume in mind.
Media that wet the surface of the thermistor minimize its contact resistance. Patel et al.
(1987) developed a two dimensional transient finite element model to investigate the effect of
imperfect contact between the probe and medium. A decoupler medium was introduced to
model contact resistance. The error in estimated medium thermal conductivity was
determined for different probe sizes and different values for decoupler thickness and
decoupler thermal conductivity. They concluded that the error decreases with: a larger probe,
longer heating time, thinner decoupler layer, smaller difference between medium and
decoupler thermal conductivity.

The assumption of pure conduction heat transfer can be challenged when measuring the
thermal properties of fluids. Localized heating by the thermistor induces temperature and
density gradients in the fluid. Convection will take place if the buoyancy force resulting from
the density gradient exceeds the viscous drag from the fluid. Low viscosity fluids are thus
more prone to natural convection than high viscosity fluids. The low viscosity fluids are
treated with either a thickener - like agar-agar in water - or a flow inhibitor, like glass fiber.
The treatment level is kept to a minimum as not to change the thermal properties of the
medium. In addition to that, small beads and low heating power are used in an attempt to
delay the onset and minimize the effect of natural convection.
Using the assumptions the heat transfer problem can be described by the following
one-dimensional heat diffusion equations:
1 δ  r 2 δT b  + q(t) δT b
= α1 0≤r≤a (3.1)
r 2 δr  δr  k b b δt
1 δ  r 2 δT m  = α1
δT m
r≥a (3.2)
r 2 δr  δr  m δt
Symmetry dictates the boundary condition at the center of the thermistor to be:
δT b
=0 r = 0, t≥0 (3.3)
δr
The assumptions of pure conduction, absence of surface contact resistance and infinity of the
medium gives as boundary conditions:
Tb = Tm r = a, t>0 (3.4)
δT b δT
kb = km m r = a, t>0 (3.5)
δr δr
T(r, t) = T i r → ∞, t>0 (3.6)
The initial condition is given by the requirement that the thermistor and the surrounding
medium should be at a thermal equilibrium at the start of a test.
T(r, t) = T i 0 ≤ r ≤ a, t=0 (3.7)

T(r, t) = T i r > a, t=0 (3.8)
3.2 SOLUTION
The solution to this coupled heat transfer problem is taken from Valvano (1981). In order to
obtain the solution, the form of the heat generation function, q(t), has to be known. A heat
generation function of the form q(t) = G + b f(t) is assumed. Γ is the steady state term, f(t) is
the transient term which will approach zero for long times. The form of the heat generation
function was experimentally verified (Balasubramaniam, 1975). The mathematics involved in
solving this heat transfer problem can be found in Valvano (1981). The resulting two
equations for the temperature distribution in the thermistor bead and the medium are:
2 k  2
 Γa 3 f(t)
T b (r, t) − T i = a  b + 1 1 −  ar   (Γ + βf(t)) − (3.9)
k b  3k m 6   3k m α m π
 (a − r)f(t)  βa 3 f(t) Γa 4 f(t)

T m (r, t) − T i = a Γ 
3
+ − (3.10)
3k m r  α m π  3k m r 3k m r α m π
Equation (3.9) describes the temperature distribution inside the bead and is used to derive
equations that form the basis for the thermistor based measurement technique.
3.3 APPLICATION
3.3.1 Measurement of Thermal Conductivity
The test procedure consists of controlling the temperature of the thermistor bead at a
specified increment above the initial temperature. The experimentally measured temperature
of the bead, as observed from the resistance of the thermistor bead, is a spatial average over
its volume. The integral of equation (3.9) over the volume of the bead will give an equation
for the average bead temperature as function of time:

 a2  k b  2
 Γa 3 f(t)  2
+ 1 1 −  ar   (Γ + βf(t)) −
a
∫0
T b (t) − T i = 1
   4πr dr
(4/3)πa 3  k b  3k m 6   3k m α m π 
Γa 3 f(t)
= a  b + 1  (Γ + βf(t)) −
2 k
(3.11)
k b  3k m 15  3k m α m π
The transient heating function f(t) will approach zero for long times. When the limit is taken
of equation (3.11) with the time variable going to infinity an expression for the average
steady state temperature of the thermistor bead is obtained:
T b ss − T i = Γa  b + 0.2 
2 k
(3.12)
3k b  k m 
The constant internal heat generation in the bead is:

2
steady state power V ss /R f
Γ= = (3.13)
volume of bead (4/3)πa 3
Substituting this expression into equation (3.12) gives, after rearranging:
1 = 4πaR f ∆T b − 1 (3.14)
km V 2ss 5k b
This expression shows that when the bead parameters, a and kb, are known, the medium
thermal conductivity can be calculated from the experimentally obtained values for Rf , DT,
and Vss. The thermistor probe requires calibration with media of known thermal conductivity
to estimate the bead parameters, a and kb.
3.3.2 Measurement of Thermal Diffusivity
The temperature of the thermistor bead, as observed from its resistance, reaches a constant
value shortly after energizing. As shown above, this temperature is related to the steady state
heat generation term only. The terms involving the transient heat generation function f(t) in
equation (3.11) must therefore sum to zero:
Γa 3 f(t)
0 = a  b + 1  βf(t) −
2 k
(3.15)
k b  3k m 15  3k m α m π

This expression can be rewritten to single out the medium thermal diffusivity, am, on the left
side of the equal sign:
2
 
 a 
αm =   (3.16)
 π (β/Γ)  1 + k m  
  5k b  
The values for β and Γ of the heat generation function are found from the experimentally
obtained power response of the thermistor. In the process of solving the heat transfer
problem, Valvano showed that the time-function had to be of the form t-1/2. The heat
generation function can thus be described as: q(t) = Γ + β t-1/2. Values for Γ and β are found
through linear regression. The resistance of the thermistor reaches a constant value within a
few tenths of a second after startup. The voltage drop across the thermistor can thus be
described by V(t)2 = Vss2 + s t-1/2. Figure 3.1 shows the results of an experiment in thickened
water at 100 °C. The graph depicts the square of the voltage drop across the thermistor
against the inverse square root of time. The values for the probe radius and thermal
conductivity in equation (3.16) are not identical as those used for estimation of thermal
conductivity and have to be determined through calibration with media of known thermal
diffusivity.
3.4 CALIBRATION: ESTIMATION OF PROBE PROPERTIES
Equations (3.14) and (3.16) are the working equations for the thermistor based method. The
probe parameters, effective radius and effective thermal conductivity, are found through
calibration with reference media. These media have well documented thermophysical
properties over the temperature range of interest . Calibration is required at each temperature
of future use. The results of measurements in two media can be used to solve for the probe's
effective properties. Ideally, any two suitable media would yield identical probe properties.
However, as observed by earlier researchers (Balasubramaniam, 1975; Dougherty, 1987;
Kravets, 1988), this rarely is the case. Some of the explanations that are brought forward

80
70
Voltage Drop Squared (V )
2
60
50
40
30
20
Vss2
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4
-1/2
Inverse Square Root of Time (s )
Figure 3.1 Square of voltage drop across thermistor versus the inverse square root of time
for a thickened water sample at 100 °C.
mention inaccurate property data of the reference media, experimental error and stem losses
(heat transfer through the stem of the probe). A three material calibration seeks to minimize
the effect of the various errors. The probe's parameters are estimated by taking the average
over the solutions for each two material subset.
3.5 PRACTICAL CONSIDERATIONS
When developing the procedures for the experiment, some decisions have to be made
regarding the samples used and the heating power. Two of them, the required sample
dimensions and the size of the temperature step used, will be discussed.
3.5.1 Sample Dimension
One of the assumptions made in the development of the heat transfer model is that the
medium is infinite in extend. In reality, the temperature field generated by the thermistor

probe reaches only a short distance into the medium. Probe size, temperature step, medium
thermal properties and sample time will influence the size of the thermally affected region.
Balasubramaniam (1975) determined the size of the temperature field theoretically. The
transient solution for the temperature distribution of the medium was used to model the
temperature field in agar-gelled water, for a heating power function that caused a 3 °C
temperature rise in a thermistor bead of 0.8 mm radius. A temperature rise of less than 1% of
the temperature step after 60 seconds was taken as defining the infinite boundary. Under
these conditions, the infinite boundary assumption was satisfied by a sample with radius of
0.25" (6.35 mm) or about eight times the bead radius. For lower medium thermal
conductivity, the bead surface temperature increased (for a given value of ∆T), leading to a
temperature field that extended farther into the medium.
Valvano (1985) reported an effective sample volume of 5 to 10 bead radii. This was
evaluated with a steady state finite element model. The effective sample volume was taken to
be bordered by a temperature contour of 10% of the average bead temperature. The effective
measurement depth of a surface probe was assessed in thin layers of glycerol on agar-gelled
water (Valvano, 1987). The surface probe was made from a 1.6 mm diameter glass
encapsulated probe with insulating material covering the top hemisphere. Effective
measurement depth was evaluated as having a measured thermal conductivity corresponding
to a mixture of 95% glycerol and 5% water. Effective depth was found to increase from 0.9
radii at 20 sec., 1.3 radii at 40 sec., to 4.3 radii after 80 seconds. For a probe of 1.5 mm
diameter, the effective sample volume is given as about 0.5 cc (corresponds to a sphere with
radius of 5 mm). The effective sample volumes are quite a bit smaller than the 8 to 10 bead
radii given by Balasubramaniam and Bowman.
Kravets (1988) determined the minimal sample size that satisfied the infinite boundary
condition through experimentation. Holes of increasing diameter were drilled in an aluminum
block and filled with glycerol. Sampling time and temperature step are not specified but
probably similar to those used in other parts of his research: 20 seconds and 2.5 - 4 °C

respectively. The infinite boundary condition was evaluated from measured conductivities in
the different size holes. Tests with two different diameter probes yielded a minimal
acceptable sample diameter (infinite boundary) of 5.0 mm for the thermistor with a nominal
diameter of 1.52 mm and 7.16 mm for the thermistor with a nominal diameter of 2.54 mm.
Expressed in bead radii, the infinite boundary assumption is satisfied by sample volumes
having a radius of 3.3 and 2.9 bead radii respectively. These findings are consistent with the
observations made with the surface probe by Valvano (1987).
3.5.2 Temperature Step
The size of the effective sample volume is related to the magnitude of the temperature step,
∆T. For very localized measurements, a small temperature step and short measurement times
can be used. A small step will be attained fast and more closely mimicking a step change in
the (average) bead temperature. Because of the possibility of natural convection, low
viscosity fluids favor a small value. A larger value for ∆T will improve the accuracy of the
temperature measurement and increase the accuracy of estimated thermal conductivity. A 1%
error in ∆T will give a 1+0.2 km/kb % error in km (Balasubramaniam, 1975), were km/kb is the
ratio of medium and bead thermal conductivity. For a probe with a thermal conductivity of
0.1 W/m°C, a 1% error in ∆T will give an error in the thermal conductivity of water, glycerol
and castor oil of 2.2, 1.6 an 1.4% respectively.
Chato (1968) tested different values for the temperature step. A range of 3 - 7 °C was given
as preferred with 5 °C being optimal. From tests in water and agar-gelled water at about
30 °C, Balasubramaniam (1975) concluded that a temperature step in the range 1.4 - 2.1 °C is
best suited for water-like media. For solid biomaterials and viscous liquids, the temperature
step of 5 to 10 F (2.8 to 5.6 °C) should generally work well.
Valvano (1981) investigated the effect of the temperature step on the magnitude and accuracy
of thermal conductivity and diffusivity. Measurements were done in 1.5% agar-gelled water
at 37 °C, with nine values for the temperature step ranging from 1 to 9 °C. It was concluded

that the magnitude of the temperature step had little effect on the quantification of the
thermal properties. The reproducibility was significantly better at higher values for ∆T.
Valvano (1985) reports on using a value of 4 °C. Dougherty (1987) used 2 or 3 °C, without
finding an effect on the measured value of thermal conductivity.

3.6 NOMENCLATURE
a = radius of thermistor bead [m]

k = thermal conductivity [W/m°C]
q = heat generation [W/m3]
r = radial coordinate [m]
R = resistance [W]
s = slope of voltage squared versus
inverse square root of time [V2s1/2]
t = time [s]
T = temperature [°C]
V = voltage drop across thermistor [V]
α = thermal diffusivity [m2/s]
β = slope of heat generation function [Ws1/2/m3]
Γ = steady state power dissipation of bead
per unit volume [W/m3]
Subscripts
1 = thermal conductivity calibration (e.g. a1)
2 = thermal diffusivity calibration (e.g. a2)
b = bead
f = final
i = initial
m = medium
ss = steady state

CHAPTER 4
METHODS AND MATERIALS
The first part of this chapter will present the thermistors that were chosen for this research
and the instrumentation that was built for making the thermal property measurements.
Calibration is a requirement with this method and the way that this was accomplished will be
outlined. The selection of the calibration materials is discussed.
The evaluative part of this research is presented as several studies: a) a comparative study
that uses a miniature line heat source probe as reference for the thermistor, b) an investigative
study into the effect of viscosity and temperature step on the quantification of thermal
conductivity, c) two evaluative studies at high temperatures, one with tomato concentrate and
the other with a nutritional supplement.
4.1 THERMISTORS
The thermistors in this research were glass coated probes from the P60-series of
Thermometrics (Edison, NJ). They had a nominal bead diameter of 1.52 mm and an overall
length of 12.7 mm. The leads were 0.203 mm in diameter. A representation of a typical
thermistor used in this research was given earlier in Chapter 2 (Figure 2.1). For optimal
performance of the instrumentation, to be discussed later, the resistance of the thermistor
should fall in the range 500 to 3500 Ω. The temperatures encountered in this study varied
from 25 to 150 °C. Thermistors exhibit a strong temperature dependence of resistance, and
no single thermistor could be found for use over the entire temperature range. Experiments at
low temperature, 25 and 50 °C, were performed with P60DA202M thermistors, with a
nominal resistance of 2 kΩ at 25 °C. For experiments in the range 100 to 150 °C, thermistors
with a nominal resistance of 50 kΩ at 25 °C, model P60DB503M, were chosen.
Methods and Materials 38

The thermistors were calibrated for resistance versus temperature against a thermal reference
with 0.001 °C resolution (Guildline 9540 Digital Platinum Resistance Thermometer, Smith
Falls, Ontario) traceable to National Institute of Standards and Technology (NIST). The
probes for low temperature measurements were calibrated from 20 to 62 °C at 3 °C
increments, in a circulating bath (Haake A82, Haake Buchler Instruments Inc., Saddle Brook,
NJ) filled with a water-ethylene glycol mixture. The thermistors for high temperature
measurements were calibrated from 95 to 155 °C at 5 °C increments, in a circulating bath
(Exacal EX-251 HT, NESLAB Instruments, Inc., Newington, NH) that contained the heat
transfer fluid HTF 500 (Union Carbide, Danbury, CT). For calibration, the thermistors and
thermometer were placed inside a metal box 3.5 by 3.5 by 9 cm, that restricted the flow, and
greatly improved the stability of the thermometer and thermistor readings. The resistance of
the thermistors was determined indirectly by measuring the voltage drop resulting from a
precision current of 0.1076 mA. The self heating by this current was negligible. Voltage was
measured with a multimeter (Keithley 197 Autoranging Microvolt DMM, Keithley
Instruments Inc., Cleveland, OH) with 0.01 mVolt resolution. Temperature was related to
resistance using the following equation:
1 = a + a ln (R) + a [ln (R)] 3 (2.4)

0 1 2
T
The coefficients, a0, a1, and a2, were determined by non-linear regression analysis in SAS
(Statistical Analysis Software, Cary, NC).
4.2 THERMISTOR ASSEMBLY
The thermistors had a stem, approximately 12 mm long, from which two bare wires
extended. The bare wires were covered with insulation that was removed from wire wrap
wires. Teflon coated wires were soldered to the thermistor's lead wires and the junctions were
insulated with heat shrink tubing. Completion of assembly was determined by the intended
use, low or high temperature measurements.

4.2.1 Low Temperature Probes
Stainless steel tubing with an outside diameter of 6.35 mm and suited for use in a pipe fitting
(Swagelock, Solon, OH) was used as a holder for the low temperature probes. They were
bonded inside this tubing with epoxy (J-B Weld, J-B Weld Company, Sulphur Springs, TX),
with the thermistor's bead extending about 6.5 mm from the holder. Figure 4.1 shows the
probe assembly for the low temperature measurements.
Figure 4.1 Probe assembly for low temperature measurements.
4.2.2 High Temperature Probes
A significant amount of time and effort was spent on the design of a moisture and pressure
resistant thermistor assembly. The high temperature probes were to be used at temperatures
up to 150 °C and pressures of up to about 500 kPa. The seal around the thermistor's stem had
to withstand these conditions.
Initially, a similar approach was taken as for the low temperature probes, only instead of J-B
Weld, a special high temperature epoxy (Omegabond 200, Omega Engineering Inc.,
Stamford, CT) was used. The epoxy was cured at 150 °C for 6 hours. In use, the repeated
heating and cooling, in combination with the harsh conditions to which the probes were
subjected took their toll on the epoxy. It deteriorated to the point that moisture could reach
the wires coming from the stem, ending the useful life of the probe. The holder was
redesigned to one where the exposed surface of the epoxy was greatly reduced. However this
design too failed after repeated use. Again the holder was redesigned. A thermistor-guide was

constructed from a solid stainless steel rod. A hole was drilled along the axis with a 3/16 inch
diameter drill bit. The hole was drilled almost to the other end. A 1.5 mm drill bit was then
used to make the channel go all the way through. It allowed the thermistor to pass through
the thermistor-guide as it was inserted from the side of the guide with the 3/16 inch diameter
opening. The outside of the guide was partially fitted with threading. The lid to the sample
chamber was designed to accept the guide. The hole in the lid was threaded at the upper end
and machined to a level bottom. A narrow opening in the bottom provided access for the
thermistor into the sample chamber. The bottom was only 1.5 mm thick. A rubber gasket was
used to provide the seal between the thermistor, the guide and the lid. A stainless steel
washer was placed between the rubber gasket and the bottom of the hole in the lid to
facilitate disassembly after a measurement. Figure 4.2 shows the complete lid assembly for
the high temperature measurements. This design gave an adequate seal around the thermistor.
However, disassembly after a measurement, and in particular removing the rubber gasket,
carried the risk of damaging the thermistor. The rubber gasket had to be replaced after almost
every test. Otherwise, the gasket bonded too strongly and disassembly would damage the
thermistor.
4.3 SAMPLE CHAMBER
The sample chamber consisted of three parts: a cylindrical body with top and bottom flanges,
a bottom plate, and the lid. The internal diameter was 21.5 mm and the depth 44.5 mm. The
lid for the low temperature measurements had a pipe fitting (Swagelock, Solon, OH) for
fastening the probe. The lid for the high temperature measurements was described above and
is shown in Figure 4.2. It contained a cavity to allow expansion of the test fluid. The lid and
the bottom plate were bolted to the body of the sample chamber with three bolts each.
O-rings provided the seal on the contact surfaces. A representation of the sample chamber is
given in Figure 4.3

Figure 4.2 Lid assembly for the high temperature thermal property measurements.

Figure 4.3 Sample chamber for thermal property measurements.

4.4 CONTROLLER
Thermal property measurement was performed with a fixed temperature step of the
thermistor. The temperature of the thermistor bead had to be raised almost instantaneously to
a temperature above the initial equilibrium temperature and kept at this elevated temperature
for the duration of the test. An electronic circuit performed this task. It was taken from
Valvano (1981) who had modified the original design of Balasubramaniam (1975). The
circuit (Figure 4.4) is best described as a resistance equilibration circuit. Once activated, it
will match the resistance of the thermistor to that of a resistor bank, provided that entails a
resistance reduction (the circuit is incapable of increasing the resistance of the thermistor).
The resistor bank consisted of nine resistors (metal film 1% precision resistors, 1/4-W
rating). Eight resistors could be individually short circuited by computer controlled relays
(EAC D1A05A). The value of these resistors were chosen to be powers of two, with the
lowest value being 1 W. The resistors were hand picked as close approximations of their
intended value. All eight resistors in series would give a resistance of approximately 255 W.
The ninth resistor was permanently part of the circuit. This resistor had to be changed
manually to accommodate a thermistor resistance from 500 to 3500 W. The resolution of the
resistor bank allowed the actual temperature step to be within 0.03 °C of its intended value.
Zener diodes were added to the circuit to attenuate the output of the circuit and prevent
overpowering the thermistors. They reduced the maximum output of the operational
amplifiers (OpAmps) from 12 to 9 Volt. The circuitry was interfaced to a Personal Computer
(PC) with an input-output board (PIO-12, MetraByte Corporation, Taunton, MA), and a data
acquisition board (DAS-20, MetraByte Corporation. Taunton, MA). A passive low-pass filter
consisting of a capacitor (0.047 µF) and a resistor (1000 Ω) was used at the signal
connections to the data acquisition board. The resistor was placed between one of the signal
leads and the binding post. The capacitor was placed across signal leads. Measurement
control and data acquisition took place through software written in Turbo Pascal 4.0 (Borland
International, Scotts Valley, CA) running on the PC.

Figure 4.4 Control circuit for thermal property measurements.

4.5 THERMAL PROPERTY MEASUREMENT
A thermal property measurement started from thermal equilibrium. The temperature was
measured by placing the circuit in resistance measurement mode. A precision current of
0.1076 mA was fed through the thermistor. The resulting voltage drop was measured with the
ADC. The resistance was calculated and converted to a temperature with equation (2.4). The
chosen temperature step, 2.5 °C in this research, was added to the equilibrium temperature
and this final temperature was converted to the corresponding resistance of the thermistor.
The resistor bank was set to the closest match of this resistance by switching the appropriate
relays. The resistance was measured so the realized temperature step could be calculated. The
circuit was then placed into thermal property measurement mode: the master switch was
thrown, which connected the output of the OpAmps to the thermistor and the resistor bank;
voltages of equal size but different sign were applied across the thermistor and the resistor
bank as shown in Figure 4.4 with -V(t) and +V(t). The initial unbalance in resistance caused a
non-zero voltage at the junction of thermistor and resistor bank. This voltage formed the
(error signal) input for the integrator. The circuit reduced the error signal to zero which was
accomplished by matching the thermistor's resistance to the value of the resistor bank. The
circuit response was very fast, with equilibration typically taking place within 0.1 second,
closely approximating a step change in temperature.
The voltage across the thermistor was sampled at a rate of 100 Hz for a period of 30 seconds
and stored to disk. The realized temperature step and the final resistance of the thermistor
were stored for each measurement.
4.6 EFFECTIVE PROBE PROPERTIES
The bead properties effective radius and effective thermal conductivity of equations (3.14)
and (3.16) were estimated from measurements in reference media. They have to be
determined at each temperature of intended use. Water, glycerol (Fisher Scientific,
Pittsburgh, PA) and a heat transfer fluid, HTF 500 (Union Carbide, Danbury, CT), were used

as reference media. The choice of these media will be discussed later in this chapter ("4.10
Thermal Standards"). Convection in water was controlled with a gelling agent, Kelset (Merck
Inc., Kelco division, Rahway, NJ) in 0.9 % concentration (w/w). For the glycerol and
HTF 500 samples fiberglass pipe wrap insulation (Frost King, Thermwell Products Co., Inc.,
Paterson, NJ) in a concentration of 1.5 % (w/v) served to control convection. This material
worked better than glass wool (Fisher Scientific, Cat. No. 11-390, Pittsburgh, PA), which
could not give a uniform fill of the sample chamber. The use of glass fiber in a low
concentration to control convection is common practice. At such a low level it is assumed
that the thermal properties of the test medium are unaffected.
The water samples required a more involved sample preparation than samples of the two
other calibration materials. A 0.9 % (w/w) Kelset solution was made in de-ionized water as
described in the following. Kelset was added to 80 °C water in a blender. After blending, the
solution was heated to 85 °C, and placed under vacuum (600 mg Hg) to expel entrapped air.
The solution was allowed to cool. The sample chamber was filled with the lukewarm
solution. With glycerol and HTF 500, the glass fiber insulating material was placed in the
sample chamber. The fluid was heated to 80 °C to reduce its viscosity, and introduced in the
sample chamber. The sample was placed under vacuum (600 mm Hg) to remove any
entrapped air. The prepared samples were hermetically sealed into the sample chamber with
the lid which held the probe.
The low temperature probes were calibrated at 25 and 50 °C, with three samples of two
calibration fluids. Since the research for which probe #L4 was calibrated involved materials
with thermal properties that were known to fall between those of water and glycerol, a
deviation from the three-fluid calibration used in the rest of this research was made in that a
two-fluid calibration was used instead. This expedited the pace of the work and was deemed
appropriate and acceptable in light of the objective of this part of the research. Also, a
calibration with water and glycerol only, was not unprecedented, as Valvano (1981) had done
exactly this in his research. The calibration took place with the sample chamber submerged in

a circulating bath (Haake A82, Haake Buchler Instruments Inc., Saddle Brook, NJ), filled
with a water-ethylene glycol mixture, and maintained at either 25 or 50 °C.
Calibration of the high temperature probes was performed over the range 100 to 150 °C at
10 °C increments with samples of water, glycerol and HTF 500. Each sample was used over
the entire temperature range. The sample chamber was submerged in the circulating bath
(Exacal EX-251 HT, NESLAB Instruments, Inc., Newington, NH). The bath was started and
heated the sample to the starting temperature. With the gelled-water samples, the first test
started at 100 °C after which the temperature was incremented 10 °C until the final
temperature was attained, 150 °C. The glycerol and HTF 500 samples were started at 150 °C
after which the temperature was decreased. This approach was taken because of experience
gained in earlier tests. Prior experiments had revealed a risk of probe failure. The epoxy,
though rated for high temperature, deteriorated as a result of the harsh conditions to which it
was subjected. Extra protection was supplied by a coat of polyurethane varnish that was
applied at the base of the glass stem. The coating would not hold up long and was reapplied
frequently. By starting at 150 °C, the worst conditions were met by a coating that was still
fresh, thus stronger. The water samples however could not be started at 150 °C. Kelset
browns at about 130 °C. It does however form a sort of rigid matrix and regains it convection
controlling characteristics albeit not as adequate as at lower temperatures (under 130 °C).
With the final design of the thermistor assembly, that used the replaceable rubber gaskets as
seal, all calibration sessions were started at 100 °C.
The standard protocol at each temperature was to take three measurements (repeats) with
each sample. The repeats were separated by adequate time to allow the system to return to
equilibrium, typically 8 minutes. Sampling was performed at a rate of 100 Hz for 30 seconds.
As mentioned before, a temperature increment of 2.5 °C was used in all the tests. About 70
min. were allowed between successive temperatures for the bath and sample to equilibrate.

4.7 ADEQUACY OF THERMAL CONTACT
4.7.1 Introduction
The small size of the thermistor allows measurement of small particles, or localized
properties (see the discussion of the infinite boundary condition in chapter 3 ("3.5.1 Sample
Dimension")). It also indicates a very small heat transfer area. Good thermal contact on this
area is critical. Homogeneous liquid media can be measured with good repetition, provided
convection is effectively controlled. With solid media, imperfect contact between thermistor
and medium may give inaccurate findings.
The suitability of the thermistor based method for measuring the thermal conductivity of food
materials of different texture was investigated. The thermistor based method was compared
to a line heat source probe, a widely used method for measuring the thermal conductivity of
food materials. The line heat source probe has a sampling volume many times that of the
thermistor and can deal more adequately with slightly inhomogeneous media or media that
do not provide a perfect contact. Two materials were chosen for this study, potato and lean
beef. Potato is a very homogeneous material. Lean beef is less homogeneous and due to its
fibrous nature probably not isotropic, and may not close well around the thermistor bead.
Temperature will affect the properties of the test materials. The methods were compared at
25, 50 and 100 °C.
4.7.2 Sample Material
Two brands of potatoes were used, Idaho and California White. The potatoes were obtained
from a local grocery store (Kroger). Measurements at 25 and 50 °C used both brands. Only
Idaho potatoes were tested at 100 °C. Six potatoes were sampled at each temperature. The
thermal conductivity of each potato was measured with both methods, using separate
samples.

Two slabs of lean beef (top round steak), 18 and 50 mm in thickness, were procured from a
local grocery store (Kroger). They were sequential cuts to minimize variability in
composition of the slabs. Six samples were used for each method at each temperature.
4.7.3 Thermistor Based Measurements
Two thermistors were used in the study. Low temperature measurements were done with a
2-kΩ probe, designated #L4. For the measurements at 100 °C, a 50-kΩ probe, designated
#H2, was used.
Sample Preparation, Potato
All samples were cut with a cylindrical open bore, that gave samples with a diameter of
20.5 mm. The samples were taken, parallel to the longitudinal axis of the potato and at least
10 mm removed from the skin. Separate samples were taken from each potato for
measurements with the thermistor and the line heat source probe. Samples were measured at
only one temperature.
Samples for measurement at 50 °C were cut after the whole potato had undergone a hot water
treatment at 50 °C for 60 minutes. The samples for use at 25 and 50 °C were trimmed to a
length of about 15 mm. They were lightly dried with a paper towel. The thermistor probe was
carefully inserted along the longitudinal axis of the sample. The sample was immobilized on
the probe with dental floss and beeswax - through a dip in molten wax. The beeswax also
sealed in the moisture, and prevented the heat transfer fluid from reaching the thermistor
bead.
Samples for measurement at 100 °C were taken from fresh potato. They were trimmed to a
length equal to the depth of the sample container, 45 mm. Prior to measurement they were
sealed in the sample container and subjected to a heat treatment of 60 minutes at 100 °C.
After the heat treatment, the container was removed from the bath and its lid was replaced by
the lid with the thermistor. Guideposts assured thermistor insertion along the sample axis.

The guideposts minimized lateral movement of the thermistor, thereby reducing the effect of
the insertion procedure on thermal contact. The sample chamber was returned to the bath for
the measurement of thermal conductivity.
Sample Preparation, Beef
The meat was frozen to ease the process of cutting samples. Samples were taken parallel to
the muscle fibers, with the same cylindrical bore as was used for the potato samples. Samples
for measurement at 25 and 50 °C were taken from the 18 mm thick slab, those for
measurement at 100 °C from the 50 mm thick slab. Care was taken to avoid the inclusion of
fat in the samples. The samples were wrapped in Saran wrap and stored overnight in a sealed
container in a refrigerator.
Samples for measurement at 50 and 100 °C were heat treated immediately prior to
measurement. They were sealed in the sample chamber and submerged in a circulating bath
for 60 minutes at 50 or 100 °C. The fresh and 50 °C treated beef samples were attached to the
thermistor probe with the same procedure as described for the low temperature potato
samples. After the heat treatment at 100 °C, the sample chamber was removed from the bath
and its lid was replaced by the lid with the thermistor. The heat treatment at 100 °C made the
samples shrink. Some pennies were used to prop them up. The sample chamber was returned
to the bath for the measurement of thermal conductivity.
Measurement of Thermal Conductivity
Thermal conductivity of the potato and beef samples was measured in a circulating bath. The
bath was maintained at the temperature at which the thermal conductivity was to be evaluated
(25, 50 or 100 °C). The sample was submerged in the circulating bath. Adequate time was
allowed for thermal equilibration, as could be observed from reading the thermistor, about 20
minutes. Thermal conductivity was measured according to a standardized protocol. A single
measurement lasted 10 seconds. Data were sampled at a rate of 100 Hz. A temperature step

of 2.5 °C was used. Three repeats were done with the described protocol. Adequate time was
allowed between the repeats for the sample to return to thermal equilibrium, typically 8
minutes. After the measurement of thermal conductivity, the samples were dissected. The
contact area of the thermistor and potato was inspected for quality of contact.
4.7.4 Measurements with the Line Heat Source Probe
Equipment
A miniature line heat source probe was used for measuring thermal conductivity. The probe
is described in Chapter 2 ("2.2.4 Line Heat Source Probe"). The sensor part was 38 mm long
and had an outside diameter of 0.81 mm. The probe was connected to a cold junction
compensator (OMEGA CJ, Omega Engineering, Stamford, CT). The output of the CJC was
connected a circuit that was designed to amplify and filter the CJC signal. This circuit was
interfaced to a PC through an I/O board (PIO-12, MetraByte Corporation, Taunton, MA), and
a data acquisition board (DAS-20, MetraByte Corporation. Taunton, MA). Measurement
control and data acquisition were performed with software (Turbo Pascal 4.0, Borland
International). Heating power was supplied by a precision power supply (B & K Precision
1630), capable of maintaining constant current levels between 0 and 3 Ampere at a voltage
between 0 and 30 volt. The voltage setting was adjusted for each test material to a level that
gave a 6 °C temperature rise in 30 seconds. For agar-gelled water and glycerol the voltage
setting were 2.9 and 2.3 Volt respectively. The line heat source probe was evaluated in 1.6%
agar-gelled water and glycerol. A correction factor was deemed unnecessary (Wolters, 1992).
Sample Chamber
The sample chamber consisted of a sample container and a lid. The container was machined
from solid brass, to an internal diameter of 21.5 mm and a depth of 44.5 mm. The top had a
flange that contained an o-ring which sealed against the lid. The lid was also machined from
solid brass. The one-piece lid consisted of a cylindrical plate with a thickness of 3.2 mm with

a cylinder attached to it. The cylinder was 48 mm long and had internal and external
dimensions of 19.1 and 27 mm respectively. It was centered on the plate. A hole in the center
of the plate allowed the probe to be inserted into the sample container. The purpose of the
cylinder was two-fold. It provided support for the probe which was set tight with two set
screws towards the top of the cylinder and it shielded the probe from the heat transfer
medium in the circulating bath.
Experiments
The potato and beef samples were taken with the same bore as described earlier. The beef
samples were taken from the 50 mm slab only. The potato samples for measurement at
100 °C, and the beef samples at 50 and 100 °C were cooked in the sample chamber for 60
minutes at the evaluation temperature. To make a measurement, the lid used during cooking,
was replaced with the lid for the probe. The lid was tightened on the sample container and the
probe was inserted through the hole and set tight with the screws. The chamber was returned
to the bath. The bath was set to a temperature 3 °C below the evaluation temperature to allow
for a temperature rise of approximately 6 °C during the measurement. After temperature
stabilization, the test would start. The probe was energized for 30 seconds with sampling
taking place over that period at a rate of 100 Hz. The heater voltage was sampled
simultaneously with the temperature data enabling exact calculation of the power dissipation
in the heater wire. Each sample was measured three times. These repeats were done about
eight minutes apart to allow the temperature to stabilize.
4.8.1 Introduction
Whether convection will occur within certain time depends both on the fluid properties and
on the imposed temperature step. The effect of one fluid property, viscosity, on the onset of

convection was studied for different values of the temperature step. It was accomplished by
comparing test data for fluids that differed only in viscosity. Water was a suitable medium
for this purpose. It can be thickened with minimal change in its thermal properties. Generally
accepted is the use of the thermal conductivity and diffusivity of pure water for aqueous
solutions of up to 2% thickener (by mass). When convection effects are absent, and assuming
negligible effect of the thickening agent on the thermal properties of interest, measured
values of thermal conductivity and diffusivity are independent of viscosity. The
test-hypothesis was: no effect of viscosity for a given magnitude of the temperature
increment.
As viscosities were selected, 1, 5, 25, 50, 100 and 150 cp. The 150 cp solution would not
experience convection. Prior tests with this viscosity gave results that were identical to those
obtained with the much more viscous Kelset solution that was used for calibration. The
values for the temperature step were chosen to fall within a practical range (see "3.5.2
Temperature Step"), yet be different enough to expect a demonstrable effect; as values for the
temperature step were selected: 1.5, 2.5 and 5.0 °C.
The study design for the thermal property experiments was a split-plot design with two
independent factors. Factor A, the viscosity was the whole plot factor. It had 6 levels. Factor
B, the temperature step, had 3 levels. The value of each level for both factors was mentioned
above. For each level of factor A (viscosity), two independent samples were made. Hence a
total of 12 samples were tested. With each sample all three temperature steps were used and
each temperature step was done in triplicate. The nine tests on each sample were randomized
with aid of a random table (Ott, 1988, Appendix table 8). Sequencing the 12 samples was
done in like manner.
Under model conditions, when heat transfer is by conduction only, the power dissipation in
the bead was shown to be linear in the inverse square root of time (Chapter 3). Therefore,
linear regression of the square of the voltage drop across the thermistor against the inverse
square root of time will yield identical values of slope, S, and intercept, V ss2, regardless of

what part of the data (e.g. 5 to 10 seconds, 20 to 30 seconds) is used. Convection will distort
this linearity. This is seen as an upward curvature in the power dissipation towards increased
duration. It can be demonstrated by comparing regression results from data late in a
measurement to those from data earlier in the same measurement. Alternatively, two samples
of the same material can be compared, whereby convection in one sample is controlled. The
increased heat transfer due to convection will result in an increase in the intercept of
regression, Vss2, and a decrease in the ratio of slope and intercept, S/I.
The data files were compared on the values for the terms RfDT/Vss2 and S/I. These are the
terms that were used to calculate thermal conductivity and diffusivity (equations 3.14 and
3.16). Vss2 is the intercept, I, of linear regression of V2 against t-1/2 and S is its slope. In the
absence of convection, the values of RfDT/Vss2 and S/I, calculated from an identical
regression interval, should be comparable for all values of viscosity (since their thermal
conductivity and diffusivity are the same). Convection in a less viscous solution would result
in values for RfDT/Vss2 and S/I that are significantly lower (indicative of an apparent higher
thermal conductivity and diffusivity) than for the more viscous solutions that experienced no
convection. Estimation of onset time was done by performing successive regressions over a
short interval and using the results in a multiple-comparison procedure. A regression interval
of 2 second length was used. This interval slid down the time axis with 1 second increments
until a start time of 12 seconds was attained and with increments of 2 seconds from thereon.
A short interval was necessary in order to estimate the onset time of convection. A shorter
interval than 2 seconds would have been too sensitive to the noise in the data; a longer
interval would have reduced the resolution with which the onset time of convection could be
estimated. The results of the regression analysis (RfDT/Vss2 and S/I) for each regression
interval formed the input for a multiple-comparison procedure, Tukey’s W, with SAS
(Statistical Analysis Software, Cary, NC). A type 1 error of 0.05 was used.

4.8.2 Sample Preparation
Solutions were made from de-ionized water with a gelling agent. Preliminary testing was
done to find a gelling agent which would yield solutions with Newtonian fluid behavior. The
test results were negative. All solutions were non-Newtonian (pseudo-plastic). Two sodium
alginates, Keltone LV and Keltone HV (Merck & Co. Inc., Kelco Division, Clark, NJ) gave
the best results, yielding solutions that were close to Newtonian. Solutions were made with
these alginates, with consistency coefficients, K, as close as possible to the chosen
viscosities. For ease of writing, the word viscosity will be used from hereon in place of
consistency coefficient.
The 1 cp fluid was in fact pure de-iononized water. The solutions with viscosities 5, 25 and
50 cp were made with Keltone LV. Keltone HV was used in preparation of the 100 and 150
cp solutions. All solutions were made with an identical procedure. De-ionized water was
heated to 80 °C on a hot-plate and 400 grams were poured into the cup of a blender (Waring
commercial blender). While blending, the appropriate amount of gelling agent, weighed on a
digital analytical balance (Sartorius model 1712 MP8 silver edition, Brinkmann Instruments
Inc., Div. of Sybron, Westbury, NY), was gradually added. After thorough mixing, the
solution was poured into a 1000 ml beaker. The solution was heated on the hot-plate to 80 °C
under continuous stirring. It was then placed under 600 mg Hg of vacuum to deaerate. Three
samples were taken for viscosity tests, and one for measurement of thermal conductivity and
diffusivity. To account for evaporative and transfer losses (blender cup to beaker), weights
were taken throughout the process. These measurements allowed calculation of the
concentration of the solutions, but more importantly assured proper replication. All weighing,
except of the amount of thickener, was done with a Mettler balance (Mettler PM6, Mettler
Instrument Corporation, Hightstown, New Jersey), with a resolution of 1 g.

4.8.3 Measurement of Viscosity
Viscosity was measured with a Brookfield LV viscometer (Brookfield Engineering

Laboratories, Stoughton, MA). The type of cup and spindle was determined by the viscosity
of the sample. For the solutions of 50 cp and below, the UL adapter was used. The 100 and
150 cp solutions required the small sample adapter (SC4-18/13R). All viscosity
measurements were done at 50 °C. To accomplish this, the UL adapter was lowered into a
circulating bath, maintained at 50 °C. The small sample adapter had a flow jacket. It was
connected to the ports of the circulating bath through insulated tubing. The viscometer was
programmed to take data at shear rates over its full operational range. An up and down ramp
was used to check for shear thinning or thickening behavior. The consistency coefficient and
flow behavior index were calculated from the data with a spreadsheet program (SuperCalc
5.0, Computer Associates International, Inc.). Viscosity was measured in triplicate.
4.8.4 Measurement of Thermal Properties
Thermal conductivity and diffusivity were measured with a 2-kW probe, designated #L4. A
sample time of 30 seconds was used. All measurements were performed at 50 °C.
Temperature control was attained by submersion of the sample chamber in a constant
temperature bath (Haake A82, Haake Buchler Instruments Inc., Saddle Brook, NJ), filled
with a water-ethylene glycol mixture, maintained at 50 °C. The measurement procedure was
identical as described earlier in this Chapter ("4.6 effective Probe Properties").
4.8.5 Rayleigh Number Calculation.
Rayleigh numbers were calculated by numerical evaluation of the exact solution for the
temperature distribution in the medium as given by Balasubramaniam (1975). The evaluation
was done with software written in Turbo Pascal 4.0 (Borland International, Scotts Valley,
CA). The infinite integrals of the solution were evaluated with Romberg integration. The
procedures for the integration were taken from "Numerical Recipes in Pascal" (Press, 1989).

The boundaries of integration were set to 10-30 and 1030. Program input consisted of the bead
parameters, the medium properties and the size of the temperature step. For the bead
parameters, the calibrated values were used except for the bead thermal diffusivity, which
could not be found from the calibration. A value of 1.3 10-7 m2/s, given by Balasubramaniam
(1975), was used. Balasubramaniam had used that value for the estimation of thermal
diffusivity (his method required knowledge of the bead's thermal diffusivity, whereas the
present method does not). The bead thermal diffusivity was calculated from the thermal
conductivity of the bead and its density and specific heat. Balasubramiam's value for the
thermal diffusivity falls between those experimentally found by Valvano (1981). Valvano
obtained the bead's thermal diffusivity by monitoring its temperature as the bead was quickly
moved from a well stirred bath at 17.5 °C to one at 38 °C. Four thermistors were tested in
this manner. Bead thermal diffusivity ranged from 0.958 to 1.636 [10-7m2/s].
Temperature profiles in the medium were generated with a resolution in the radial variable of
1/50 of the bead radius. They were generated at 0.5 second intervals and the Rayleigh
number was calculated with the time dependent heat penetration depth and effective
temperature gradient. The criterion for the penetration depth was chosen as the distance at
which the temperature rise in the medium extinguished to below 1% of the surface
temperature of the thermistor bead. The effective temperature gradient over the penetration
depth was calculated as the linearized temperature gradient that integrated to the same value
as the integrated simulated temperature profile. Rayleigh numbers were calculated for those
combinations of temperature step and viscosity that exhibited signs of convection.
4.9 EVALUATION AT HIGH TEMPERATURE
4.9.1 Introduction
Few food materials have documented values of thermal conductivity and diffusivity above
100 °C. Data for tomato concentrate had been collected by Choi and Okos (1983) with the

line heat source probe over the range 25 to 150 °C. Being a paste like material, it was thought
to be a good material for evaluation of the thermistor based method. A paste was expected to
provide good surface contact with the thermistor bead.
Thermal conductivity of liquid food is strongly influenced by its water content. Liquid
nutritional supplements with various solids content were tested. These materials came
without reference data. They were used to test if the probes would measure thermal
conductivity that followed the expected downward trend with solids content.
4.9.2 Tomato Concentrate
Thermal conductivity and thermal diffusivity of the tomato puree and tomato paste were
determined with the thermistor probe. The tomato concentrate was supplied by Van Den
Bergh Foods Co. (Roxbury, CT) in size 10 cans. Three temperatures were chosen for thermal
property evaluation; 100, 130 and 150 °C. Two cans per material (puree, paste) were used
and from each can two samples were taken for measurement of thermal conductivity and
thermal diffusivity. Each sample was used at all temperatures. A sample of the tomato
product was carefully placed in the sample chamber. Care was taken not to create air pockets
which would alter the thermal properties. The sample chamber was hermetically sealed with
the lid with the probe (#H3) mounted to it. The sample chamber was then submerged in a
constant temperature bath (Exacal EX-251 HT, NESLAB Instruments, Inc., Newington, NH)
and bath and sample were heated to 100 °C. Measurements were performed with the protocol
as outlined in "4.6 Effective Probe Properties". Three measurements were taken at 100 °C.
After the third measurement, the sample was heated to 130 °C for three measurements at that
temperature and subsequently to 150 °C for a final set of measurements. After this final set,
the sample was allowed to cool to room temperature and visually inspected for heating
induced changes.
In order to compare the results of the thermal property measurement to the data reported by
Choi and Okos (1983), the solids content of the material needed to be known. Factory

supplied values for solids content were checked. The refractive index was obtained with a
refractometer (ABBE MARK II digital refractometer, Reichert-Jung). Tomato paste was
diluted with water in the ratio 1:2 before measuring the refractive index. The refractive index,
which is a measure of soluble solids, was related to total solids with aid of a table (Pearson,
1973, Table 9.5).
4.9.3 Nutritional Supplement
A fluid nutritional supplement was supplied for testing by Abbott Labs/Ross Products
Division (Columbus, OH). Products of different solids content and different composition
(e.g. fat, protein, carbohydrates) were tested. Some of these products were viscous enough
that natural convection was deemed unlikely. For the others, fiberglass pipe insulation was
used at 1 - 1.5% (w/v) as a convection inhibiter/reducer. These low concentrations were
assumed to have no influence on the thermal properties. Care was taken in the sample
preparation to limit air inclusion in the samples as much as possible. The samples that used
the pipe insulation were heated to reduce their viscosity. Using a pipette, the sample was
introduced at the bottom of the sample chamber. After filling, the sample chamber was
placed in a vacuum desiccator, under 600 mg Hg of vacuum, for a few minutes to remove
entrapped air, if present. It was assumed that this procedure gave negligible moisture loss
from the samples. Four different products were tested, designated at "TC1", "TC2", "TC3"
and "TC4". Each product was supplied in various solids contents. The products were tested at
two temperatures, 95 and 150 °C. Duplicate samples were used for all but product "TC2".
The combinations of product, solids content and temperature that were tested were specified
by the supplier and can be found in Table (5.23) in the results. The samples were submerged
in the bath, already at measurement temperature. Doing so minimized the heat-up time.
Approximately 20 minutes were required, after submersion, for temperature equilibration to
take place. Property measurement was performed with the protocol as described in "4.6
Effective Probe Properties".

4.10 THERMAL STANDARDS
Thermistor probes require calibration to obtain their effective properties. Calibration takes
place with thermal standards, media of known thermal conductivity and diffusivity. The
accuracy of the thermistors in subsequent use is dependent on the accuracy of property data
of the calibration media. One of the conclusions reached by Kravets (1988) was the need for
well defined thermal reference materials in the range of food products. Dougherty (1987)
observed that probe parameters found from a calibration with water and ethylene glycol were
different from those found for the same probe with water, glycerol and toluene. This
dependency of probe properties on thermal standards was partly attributed to inaccurate
literature data.
A potential calibration material should meet a number of criteria:
1. Its thermal conductivity and diffusivity should be similar to the values found in moist
food materials,
2. Its thermophysical properties must be well documented over 20 - 150 °C,
3. The transfer of energy in the fluid by means other then conduction (convection,
radiation) should be minimal.
Literature was searched for liquid thermal standards that would meet the criteria. Finding
materials that are viscous enough to be used at high temperatures and have well documented
properties proved difficult. The starting point were the media used by prior researchers:
water, glycerol, castor oil, toluene and ethylene glycol.
The thermal conductivity and diffusivity of water is well documented at high temperatures.
Convection is a potential problem with water, and agar, an effective agent at low
temperatures, breaks down at the high temperatures encountered in the present study. An
alternative was found in Kelset, a sodium alginate. This material proved to control
convection satisfactory in low enough concentrations, 1%, as not to alter the thermal
properties of water. Glycerol is a high viscosity fluid at room temperature but at 150 °C its

viscosity is only slightly higher than water at room temperature. However, with glass fiber
material to suppress convection, it tested well at 150 °C. Castor oil is not documented at high
temperatures. Toluene is a recommended thermal conductivity reference material (Marsh,
1987). Its thermal conductivity is low, around 0.13 W/m°C, making it a good complement to
water (0.61) and glycerol (0.30) in covering the range of values found in food materials. Its
viscosity however is very low. Even with aid of fiber glass material, it failed to perform well.
Ethylene glycol has a thermal conductivity comparable to that of glycerol. Its thermal
conductivity is documented at high temperatures, but thermal diffusivity data were neither
found nor could be calculated from thermal conductivity, specific heat and density for lack of
data for the latter two properties. Of the four (liquid) recommended thermal conductivity
reference materials (Marsh, 1987), only water was considered to meet the criteria formulated
earlier.
A heat transfer fluid was chosen as third calibration fluid. HTF 500 (Union Carbide, 1989 )
has a thermal conductivity below 0.2 W/m°C. Glass fiber material sufficiently suppressed
convection to allow HTF 500 to be used as a calibration material. Thermophysical properties
were obtained from the manufacturer. Their accuracy were not verified. Tables 4.1 through
4.3 list the properties of the calibration fluids used in this research. In all cases, thermal
diffusivity was calculated from thermal conductivity, density and specific heat. Thermal
conductivity and specific heat values for glycerol were taken from CINDAS (1988) and its
density was measured (Appendix A).

Table 4.1 Thermophysical Properties of Water
Temperature Thermal Density2 Specific Heat2 Thermal

Conductivity1 Diffusivity3
[°C] [W/m°C] [kg/m3] [J/kg°C] [10-6m2/s]
20 0.5981 998.5 4181 0.1432
25 0.6067 997.1 4180 0.1456
30 0.6149 995.5 4179 0.1478
40 0.6297 992.0 4179 0.1519
50 0.6426 987.8 4181 0.1556
60 0.6535 983.1 4184 0.1589
70 0.6624 977.7 4190 0.1617
80 0.6694 971.7 4197 0.1641
90 0.6744 965.1 4207 0.1661
100 0.6790 958.2 4218 0.1680
110 0.6816 950.8 4229 0.1695
120 0.6832 942.8 4244 0.1707
130 0.6838 934.5 4263 0.1717
140 0.6834 925.6 4285 0.1723
150 0.6820 916.4 4311 0.1726
1
Nieto de Castro (1986), 20 - 90 °C; adapted from Sengers et. al (1984), 100 -
150 °C.
2
Adapted from Incropera and DeWitt (1985).
3
Calculated as α = k/ρ Cp, α = thermal diffusivity, k = thermal conductivity, ρ =
density, Cp = specific heat.

Table 4.2 Thermophysical Properties of Glycerol.

[°C] [W/m°C] [kg/m3] [J/kg°C] [10-6m2/s]
20 0.2843 1260.9 2354 0.0958
25 0.2850 1257.7 2381 0.0952
30 0.2856 1254.6 2407 0.0946
40 0.2869 1248.1 2458 0.0935
50 0.2882 1241.6 2505 0.0927
60 0.2895 1235.0 2549 0.0920
70 0.2908 1228.3 2588 0.0915
80 0.2921 1221.5 2625 0.0911
90 0.2934 1214.6 2658 0.0909
100 0.2947 1207.6 2687 0.0908
110 0.2960 1200.5 2712 0.0909
120 0.2973 1193.3 2733 0.0911
130 0.2986 1186.0 2751 0.0915
140 0.2998 1178.5 2764 0.0920
150 0.3011 1171.0 2774 0.0927
1
CINDAS (1988).
2
Measured (Appendix A).
3

Table 4.3 Thermophysical Properties of HTF 500

[°C] [W/m°C] [kg/m3] [J/kg°C] [10-6m2/s]
20 0.1679 1038.9 1940 0.0833
25 0.1671 1035.3 1958 0.0824
30 0.1663 1031.8 1976 0.0816
40 0.1647 1024.6 2010 0.0800
50 0.1632 1017.5 2043 0.0785
60 0.1616 1010.4 2074 0.0771
70 0.1600 1003.2 2104 0.0758
80 0.1584 996.1 2132 0.0746
90 0.1569 988.9 2158 0.0735
100 0.1553 981.8 2182 0.0725
110 0.1537 974.7 2205 0.0715
120 0.1522 967.5 2226 0.0706
130 0.1506 960.4 2246 0.0698
140 0.1490 953.2 2264 0.0691
150 0.1475 946.1 2280 0.0684
1
Adapted from manufacturer supplied graphs (Union Carbide, 1989).
2

CHAPTER 5
RESULTS AND DISCUSSION
The first part of this chapter discusses the calibration of the thermistors. Subsequently, the
results of the comparative study are given. These are followed by those for the study into
natural convection. In the last part of the chapter, the high temperature measurement of two
types of food material are discussed.
5.1 TEMPERATURE-RESISTANCE CALIBRATION
The results of temperature-resistance calibrations are given in Table 5.1. The table shows the
probes that were calibrated, the temperature range over which a probe was calibrated, the
value of the coefficients of equation (2.4), and the value of the χ2 statistic. The values of χ2
for the low temperature probes were far lower than those for the high temperature probes.
This indicated a much better fit of the equation to the data for these probes. The agreement
between the equations and the data was better than 0.006 °C for the low temperature probes,
and better than 0.05 °C for the high temperature probes. This excellent agreement between
the equations and the data attest to the correctness of equation (2.4), and to the accuracy of
both the temperature and resistance measurement.
1 = a + a ln(R) + a ln(R) 3 (2.4)

0 1 2
T
Not all the probes of Table 5.1 were used after they were calibrated. The
temperature-resistance calibration was the first step of the two-step process of calibrating a
thermistor probe for thermal property measurements. This first step was the least
time-consuming part, and could be done for four probes simultaneously. The second step, in
which the effective probe properties were determined, was a fairly lengthy endeavor. It had to
be done for each probe separately. Therefore, just one low and one high temperature probe
Results and Discussion 66

were completely calibrated at a given time. Only in case of probe failure would another probe
be taken through the second step.
Table 5.1 Calibration coefficients and value of the χ2 statistic for the temperature-resistance
function of the probes
Range of a0 a1 a2 χ2
Calibration and
Probe
20 - 62 °C
#L1 1.30486 E-03 2.58926 E-04 1.33393 E-07 6.0 E-05
#L2 1.27912 E-03 2.58769 E-04 1.33874 E-07 4.3 E-05
#L3 1.30103 E-03 2.60179 E-04 1.32498 E-07 1.1 E-04
#L4 1.28409 E-03 2.59543 E-04 1.37882 E-07 6.1 E-05
95 - 155 °C
#H1 8.63309 E-04 2.11965 E-04 1.52697 E-07 6.5 E-03
#H2 8.46124 E-04 2.14167 E-04 1.52067 E-07 6.8 E-03
#H3 8.80808 E-04 2.09922 E-04 1.60020 E-07 3.3 E-03
#H4 8.81477 E-04 2.10426 E-04 1.59332 E-07 3.9 E-03
#TT1 8.31390 E-04 2.19282 E-04 1.17724 E-07 1.2 E-02
#TT2 8.42144 E-04 2.20328 E-04 1.14475 E-07 1.1 E-02
#TT3 8.71597 E-04 2.16516 E-04 1.25304 E-07 1.6 E-02
#TT4 8.34315 E-04 2.18145 E-04 1.11302 E-07 1.3 E-02
5.2 EFFECTIVE PROBE PROPERTIES
5.2.1 Low Temperature Probes
The effective probe properties were determined in the second step of the calibration for
thermal property measurement. The probe parameters (a1, kb1, a2, kb2) as required by equations
(3.14) and (3.16) were estimated from measurements in reference media.
1 = 4π a 1 ∆T R f − 0.2 (3.14)
km V 2ss k b1

2
 
 a2 
αm =   (3.16)
 π (S/I)  1 + km  
  5k b2  
Of the low temperature probes, only probe #L4 was calibrated for use in thermal property
measurements. It was calibrated with water and glycerol. Regression analysis yielded values
for Vss2 (replaced by I in the equation 3.16) and S. I and S represent the two parameters of the
model that described the voltage drop across the thermistor, V(t)2 = I + S t-0.5. The values of
Rf and ∆T were known constants for each measurement. Linear regression used the data of
the interval 3 to 9 seconds. The regression results were used to generate values for the terms
Rf∆T/Vss2, and S/I. The sample average at each temperature was found by averaging these
terms over the three measurements that were done on each sample at that temperature. The
medium average was found as the average of the three sample averages. The replication was
good. The largest difference between sample averages was 0.6%, which was for S/I of water,
at 25 and 50 °C. The replication was marginally better for the term Rf∆T/Vss2, with 0.5%
being the largest difference. The regression results are given in Appendix B. The medium
averages, and the reference data for the calibration media, were substituted into the two
equations to find the properties of the probe. The properties of probe #L4 at 25 and 50 °C are
given in Table 5.2.
Table 5.2 Properties for low temperature probe #L4, corresponding to the regression
interval 3 to 9 seconds.
Temperature a1 kb1 a2 kb2

[°C] [mm] [W/m°C] [mm] [W/m°C]
25 °C 0.9874 0.1032 0.7467 0.1185
50 °C 0.9707 0.1080 0.7602 0.1168

The radius as obtained for estimation of thermal conductivity is larger than the physical
radius of 0.76 mm. This is in agreement with findings of other researchers
(Balasubramaniam, 1975, Valvano, 1981, Dougherty, 1987). The radii found for estimation
of thermal conductivity, a1, were also larger than those for estimation of thermal diffusivity,
a2. Valvano (1981), the only other researcher who calibrated probes for measurement of
thermal diffusivity found that as well. Inaccuracies in the reported reference data as well as
stem losses were given as likely causes for the differences in probe properties. The values for
the calibrated thermal conductivity of the probe fall within reported values.
5.2.2 High Temperature Probes
High temperature probes were calibrated with three calibration media. During the calibration
of the low temperature probes it was noticed that the response of the thermistor in water and
glycerol deviated somewhat from the model, in particular when data early in the
measurement were used in regression. At longer times, the response more closely followed
the model. This suggested the use of long sample times. However, the duration of the test in
low viscosity fluids was limited by the onset of natural convection, despite efforts to delay
that as much as possible. Thus, a regression interval had to be found that excluded initial
effects and also ended before convection effects disrupted the measurement.
Software was written in Turbo Pascal (Borland International, Scotts Valley, CA) to perform
the linear regression. The software performed linear regression on successive intervals. Input
to the program consisted of the length of the regression interval, the start time of the first
interval and the magnitude of the step with which the start time was incremented. The length
of the regression interval was chosen as 10 seconds. The first interval started at 1 second. For
each successive interval the start time was incremented by 1 second. Linear regression as
described above was done on the data files that were collected for calibration. For each data
file, the program would generate a table with regression results. In those, regression statistics
were tabulated for the successive regression intervals. Included for each regression interval
were the secondary statistics Rf∆T/Vss2 and S/I. The probe properties were obtained much like

that done for the low temperature probe. Software was used to calculate the probe properties
for each of the three two-media calibrations, (water-glycerol, water-HTF 500, and
glycerol-HT 500). Those were averaged to find the properties for the three media calibration.
They were calculated for each regression interval. The parameters of probe #H3 at 100 °C for
estimation of thermal conductivity are given in Table 5.3, those for estimation of thermal
diffusivity in Table 5.4.
The calibrated probes were used to estimate the thermal properties of the calibration media.
The difference between the estimated medium properties and their reference values is shown
in Tables 5.3 and 5.4 as estimation errors. Ideally, each of the three combinations of two
calibration media would have yielded identical probe properties, and all estimation errors
would have been zero. This was not the case, and the calibration accuracy could be assessed
from the estimation errors. These errors are also shown in Figure 5.1 for estimation of
thermal conductivity and Figure 5.2 for estimation of thermal diffusivity.

Table 5.3 Radius and thermal conductivity of the #H3-probe, calibrated for the
measurement of thermal conductivity at 100 °C. The error in estimation of
medium thermal conductivity for the calibration media is given. Regression
interval was 10 seconds.
Start Time of Probe Probe Errora in estimation of

Regression Radius Thermal
Interval Conductivity Water Glycerol HTF 500
[sec] [mm] [W/m°C] [%] [%] [%]
1 0.9528 0.0968 -1.93 0.35 -0.70
2 1.0106 0.0844 2.18 -0.95 0.80
3 1.0374 0.0801 3.18 -1.94 1.19
4 1.0514 0.0780 3.50 -2.55 1.33
5 1.0596 0.0768 3.62 -2.93 1.39
6 1.0647 0.0761 3.68 -3.20 1.42
7 1.0684 0.0755 3.70 -3.39 1.44
8 1.0711 0.0751 3.72 -3.57 1.45
9 1.0731 0.0748 3.72 -3.72 1.46
10 1.0746 0.0746 3.72 -3.86 1.47
11 1.0759 0.0744 3.72 -4.00 1.47
12 1.0777 0.0742 3.71 -4.09 1.47
13 1.0784 0.0741 3.70 -4.16 1.48
14 1.0785 0.0740 3.70 -4.18 1.48
15 1.0786 0.0740 3.70 -4.17 1.48
16 1.0783 0.0740 3.71 -4.15 1.48
17 1.0778 0.0740 3.71 -4.07 1.47
18 1.0780 0.0740 3.71 -4.08 1.47
19 1.0778 0.0739 3.71 -4.08 1.47
a
A positive value indicates overestimation by the thermistor based method

Table 5.4 Radius and thermal conductivity of the #H3-probe, calibrated for the
measurement of thermal diffusivity at 100 °C. The error in estimation of medium
thermal conductivity for the calibration media is given. Regression interval was
10 seconds.
Start Time of Probe Probe Errora in estimation of

Regression Radius Thermal
Interval Conductivity Water Glycerol HTF 500
[sec] [mm] [W/m°C] [%] [%] [%]
1 1.2025 0.0505 13.47 29.77 7.21
2 0.9326 0.0714 -4.60 9.15 -2.25
3 0.8474 0.0858 -6.28 4.93 -2.54
4 0.8073 0.0962 -5.86 3.48 -1.98
5 0.7844 0.1041 -5.20 2.88 -1.47
6 0.7701 0.1102 -4.52 2.55 -1.00
7 0.7598 0.1153 -3.91 2.34 -0.62
8 0.7522 0.1196 -3.22 2.19 -0.21
9 0.7466 0.1236 -2.57 2.09 0.17
10 0.7424 0.1269 -1.86 2.02 0.57
11 0.7395 0.1296 -0.90 2.00 1.11
12 0.7342 0.1333 -0.41 1.99 1.36
13 0.7322 0.1354 0.10 2.00 1.63
14 0.7311 0.1370 0.12 1.99 1.63
15 0.7300 0.1380 -0.06 1.97 1.51
16 0.7301 0.1384 -0.24 1.96 1.41
17 0.7304 0.1378 -0.97 1.95 1.00
18 0.7290 0.1392 -1.06 1.94 0.94
19 0.7288 0.1403 -1.13 1.94 0.89
a
A positive value indicates overestimation by the thermistor based method

5
4
3
Water
2
Glycerol
Error [%]
1
HTF 500
0
-1 0 2 4 6 8 10 12 14 16 18 20
-2
-3
-4
-5
Start Time of Regression Interval [s]
Figure 5.1 Error in estimation of thermal conductivity of the calibration media for Probe
#H3 at 100 °C.
10
8
6
Water
4
Glycerol
Error [%]
2
HTF 500
0
-2 0 2 4 6 8 10 12 14 16 18 20
-4
-6
-8
-10
Figure 5.2 Error in estimation of thermal diffusivity of the calibration media for Probe
#H3 at 100 °C.

The tables and the figures show the trends in the probe parameters and estimation errors with
the successive regression intervals. The tables show that the properties of the probe changed
rapidly for successive intervals for small values of the start time of the regression interval and
more or less stabilized for larger values of the start time. The estimation errors reflected that
pattern. The estimation errors for the estimation of thermal conductivity were the smallest for
the interval 1-11 seconds. For later intervals, the estimation errors increased rapidly first, then
more or less stabilized. The estimation errors for glycerol were the largest, up to 4.2%, water
followed with -3.7%, and HTF 500 was best at 1.5%. As was the case with the low
temperature probe #L4, the radii for estimation of thermal conductivity were larger than the
physical radius and larger than the radii for estimation of thermal diffusivity. The finding that
the lowest estimation errors were found for the interval 1-11 seconds was not expected from
the theory. Equation 3.14 requires a value for the steady state power dissipation in the probe.
The power dissipation in the probe towards the end of a measurement should give a better
estimate of steady state power dissipation than early, yet was associated with larger errors in
the estimated thermal conductivity of the calibration media. It is possible that for longer
times heat flow along the stem of the probe exerted a stronger effect. Inaccuracies in the
reference data could have been another factor. When the probe properties were calculated
with a 8% higher thermal conductivity of HTF 500, the result was quite different from that
displayed in Figure 5.1. The lines for water and HTF 500 shifted downward whereas the line
for glycerol became nearly horizontal. The interval 1-11 seconds had far from optimal
estimation errors as the thermal conductivity of water was under predicted by almost 15%.
However, the estimation errors for all three media were below 1% for all intervals beyond
5-15 seconds. This illustrates the importance of accurate reference data for the calibration.
Estimation of medium thermal diffusivity tended to improve by using data later in the
experiment. The estimation improved up to a start time of 11 seconds. Estimation errors
corresponding to the interval 11-21 seconds were equal to or below 2% for each material.
Medium thermal diffusivity was however calculated with an estimated value for thermal
conductivity for that interval, which could be deduced from Figure 5.1, was different from

the reference value. When the literature value for thermal conductivity was used, optimal
estimation took place for interval 7-17 seconds. The estimation errors increased for later
intervals. Compared to Figure 5.2, the lines for water and HTF 500 were shifted up and the
line for glycerol shifted down by an amount reflecting their estimation errors for thermal
conductivity (Figure 5.1). Unfortunately, the intervals giving the smallest estimation errors
for thermal conductivity and diffusivity didn’t coincide. The best interval for thermal
conductivity gave poor estimation of thermal diffusivity.
In the foregoing, the results for probe #H3 at 100 °C, were discussed. Results at the higher
temperatures, 110 to 150 °C, were similar to those at 100 °C. The estimation errors for
thermal conductivity showed hardly a trend with temperature. For all temperatures, the
interval 1-11 seconds gave the best estimation of thermal conductivity. For later intervals, the
trend in the estimation errors replicated that seen at 100 °C. The magnitude of the errors
changed marginally. The maximum estimation errors for glycerol gradually worsened with
increasing temperature to about -6% at 150 °C (Figure 5.3). Those in water for long times
decreased with temperature to about 3.1% at 150 °C, whereas the maximum errors in HTF
500 gradually increased to about 1.8% at 150 °C.
Estimation of thermal diffusivity showed a larger change with temperature. The interval for
which the estimation errors were the smallest, moved towards earlier intervals (Figure 5.4).
At 150 °C, the interval with a start time of either 3 or 4 seconds yielded the smallest
estimation errors. When estimation used the reference values for thermal conductivity, the
interval 2-12 seconds gave optimal results. The errors increased rapidly for intervals beyond
that of optimal estimation. This was probably caused by convection. The deterioration in
estimation for intervals beyond that of optimal estimation, became noticeable for
temperatures above 120 °C. The viscosity of the calibration materials was strongly reduced at
these high temperatures. The small amount of pipe insulation may not have been sufficient to
restrict convection for a long time. If convection was the main cause for the deterioration,
then the term S/I was more sensitive to its effects than the term Rf∆T/Vss2. Convection will

lead to a decrease in the slope, S, and an increase in the intercept, I (= Vss2). The relative
change in S/I will thus be larger than in Rf∆T/Vss2.
5
4
3
Water
2
Glycerol
1
Error [%]
0 HTF 500
-1 0 2 4 6 8 10 12 14 16 18 20
-2
-3
-4
-5
-6
#H3 at 150 °C.
10
8
6
Water
4
Glycerol
Error [%]
2
HTF 500
0
-2 0 2 4 6 8 10 12 14 16 18 20
-4
-6
-8
-10
#H3 at 150 °C.

The optimal intervals were dependent on the data for the calibration media as was shown
with the hypothetical example under the discussion of the estimation errors for thermal
conductivity at 100 °C. Reference data for thermal properties can vary with the literature
source. While this research was in progress, property data for glycerol were found that were
more recent, and supposedly more accurate. Initially property data for the thermal
conductivity and specific heat were taken from Touloukian (1970a, 1970b). When a more
recent source of reference for these properties was found (CINDAS, 1988), the probe
properties were re-estimated. These latter properties were shown in Table 5.3 and 5.4. The
difference in thermal conductivity data between the two sources was minor. The CINDAS
data were 1% lower than those given by Touloukian over the temperature range 25 to 150 °C.
The difference in specific heat data however was small at low temperatures but substantial at
high temperatures. At 25 and 50 °C, the difference was well below 1%. At 100 °C, the
CINDAS data were 4% lower than those given by Touloukian. At 150 °C, this difference had
grown to 10%. The effect of this change in reference data on the estimation errors is best
illustrated by Figures 5.5 and 5.6 that show the estimation errors for probe #H3 at 150 °C,
when calibrated using the glycerol data that were taken from Touloukian. Comparison of
Figure 5.5 to Figure 5.3 showed that the more recent relationship given by CINDAS yielded
smaller maximum estimation errors for thermal conductivity. On the other hand, comparison
of Figures 5.6 and 5.4 indicated a better calibration result for estimation of thermal
diffusivity when the data from Touloukian were used. The use of more recent and supposedly
more accurate reference data had led to an improvement in the calibration for thermal
conductivity but a deterioration in the calibration for thermal diffusivity.

5
4
3
2 Water
1 Glycerol
0
Error [%]
-1 0 HTF 500
2 4 6 8 10 12 14 16 18 20
-2
-3
-4
-5
-6
-7
-8
#H3 at 150 °C, when using reference data for the thermal conductivity of
glycerol from Touloukian (1970).
10
8
6
Water
4
Glycerol
Error [%]
2
HTF 500
0
-2 0 2 4 6 8 10 12 14 16 18 20
-4
-6
-8
-10
#H3 at 150 °C, when using reference data for the thermal conductivity and
specific heat of glycerol from Touloukian (1970).

5.3 ADEQUACY OF THERMAL CONTACT
5.3.1 Introduction
The thermal conductivity of potato and lean beef was measured with a miniature line heat
source probe and the bead thermistor probe. The objective was to assess the adequacy of
thermal contact for the bead thermistor during measurements in materials of different textural
properties. The line heat source probe was used as a reference method. Thermal contact was
evaluated by a comparison of the test results for both methods.
The accuracy of the line heat source probe was assessed from measurements in thickened
water and glycerol at temperatures from 23 to 93 °C. Agreement with published data was
better than 3% for temperatures up to 88 °C. Based on these results, the probe was used
without a calibration factor (Wolters, 1992). The data for thickened water at 93 °C were
inconsistent, and largely overestimated the thermal conductivity of water. The over
estimation was ascribed to convection as the sample chamber was not pressureable. Line heat
source measurements were standardized to 30 seconds sample time. Thermal conductivity
was calculated over the data of the interval 10 to 30 seconds.
5.3.2 Potato Measurements
During the early potato measurements with the bead thermistor probe, all at 25 °C, a slight
decrease in thermal conductivity was observed with successive measurements on the same
sample. The decrease was small, but seemed systematic. It was decided to change the
experimental design somewhat. Instead of using one sample of each potato and measure that
sample three times, two samples were taken of each potato, and measured twice. This
approach was followed for the remaining potatoes at 25 °C and for all potatoes at 50 °C.
These later measurements did not consistently confirm the suspected effect of repeat
measurements on the apparent thermal conductivity. At 100 °C, the original approach of one
sample per potato with three measurements per sample was again used. This inconsistency in

measurement protocol is mentioned for completeness, and as clarification as to why in
Appendix D (Tables D.2 and D.3) three values are given for some potatoes, and four for
others. The effect on the overall results of this protocol switch was minute. Analysis of the
data showed that regardless of protocol, the standard deviation within samples was much
smaller than between samples.
Figure 5.7 shows the properties for potato as measured with both methods at the three
temperatures. The data are given in Appendix D (Table D.1). Included in the figure is thermal
conductivity of water, taken from literature (Nieto de Castro, 1986). The thermal
conductivity of moist foods is strongly dependent on moisture content and, barring
significant changes in product composition, is expected to follow the trend for water with
temperature. The measurement data for both methods followed this trend, as can be seen
from Figure 5.7. An exception to this were the results for the line heat source probe at 50 °C.
At that temperature the data appeared to follow a dichotomous distribution, whereby about
half of the samples were centered around a value that was lower than measured at 25 °C.
The agreement between both methods was good at 25 and 50 °C, and fair at 100 °C. At all
temperatures, the bead thermistor probe measured lower values than the line heat source
probe. The average thermal conductivity at 25 °C with the bead thermistor was 0.5517
W/m°C, 7.4% lower than the value of 0.5957, measured with the line heat source probe. At
50 °C, the average thermal conductivity for the bead thermistor was 0.5842 W/m°C. This
was 4.5% lower than the average for the line heat source probe, 0.6117 W/m°C. At 100 °C,
the difference between the methods was 15%. The bead thermistor probe yielded an average
value of 0.6015 W/m°C, and the line heat source probe an average of 0.7075 W/m°C. The
latter value seems rather high as it exceeds the thermal conductivity of water at 100 °C
(0.6779). The cause for the high value was sought in evaporative heat loss. The sample
chamber for the line heat source probe was not pressureable. In an initial test with the
circulating bath at 97 °C, an even larger apparent thermal conductivity was measured
(0.77 W/m°C). All subsequent measurements were done with the bath at 94 °C. This may

have only partly cured the problem. Both methods measured the thermal conductivity with
good replication.
0.75
2 points
0.70
2 points
Thermal Conductivity [W/m°C]
0.65
2 points
0.60
0.55 line heat source probe; Idaho Potatoes

line heat source probe; California White Potatoes
0.50 bead thermistor probe; Idaho Potatoes
bead thermistor probe: California Potatoes
water (Nieto de Castro, 1986)
0.45
Gratzek and Toledo (1993)
Rao et. al. (1975)
0.40
10 20 30 40 50 60 70 80 90 100 110
Temperature [°C]
Figure 5.7 Thermal Conductivity of Potato at 25, 50 and 100 °C, measured with the Line
Heat Source Probe and the Bead Thermistor Probe.
At 50 °C, the data for the line heat source probe fell in two distinct groups. The two groups
corresponded to the two brands of potato, California White and Idaho. The bead thermistor
probe however could not distinguish between the two brands. A brand effect is possible, but
clouded by the fact that at 25 °C, the line heat source probe did not distinguish the one
sample of California White from the samples of Idaho potato. Statistically, it would be more
likely that the one sample (at 25 °C) was an outlier than three samples (at 50 °C). That would
imply that there was a measurable difference between the two varieties. Rao et. al. (1975)

found significant differences between potato varieties, when measured at 25 °C, despite
similar moisture contents. However, the value measured for the California White potatoes at
50 °C, 0.5946 W/m°C, was lower than the average at 25 °C. One would expect a higher value
because of the positive effect of temperature on the thermal conductivity.
Literature data for potato are given for comparison in Figure 5.7. The values for the bead
thermistor probe are closer to the literature data than those of the line heat source probe. At
all three temperatures, the line heat source probe measured a thermal conductivity that was
higher than the literature data. Yamada (1970) reported values of 0.485 at 10 °C, to 0.556 at
75 °C, and mentioned that the thermal conductivity of potato followed the trend in water with
temperature. The data from Rao et. al. (1975) and Gratzek and Toledo (1993), both shown in
Figure 5.7 were also much lower than those found with the line heat source probe in this
study. Not shown in the figure are the results of two studies that report a thermal conductivity
of potato higher than that of water. Lamberg and Hallström (1986) measured the thermal
conductivity of potato (moisture content 80% w.b.) with the transient hot-strip method. They
report an equation for thermal conductivity: k = 0.624 + 1.19 10-3T W/m°C where T is in °C,
valid over the range 20 to 85 °C. The transient hot-strip method used had not been used with
food materials before (according to the authors). As an unestablished method for food
thermal property measurement, one can not give too much weight to the data it yielded.
Califano and Calvelo (1991) obtained the thermal conductivity of potato (moisture content
80% w.b.) from heat penetration data. The data were fitted to a 2nd order polynomial:
k = 1.05 - 1.96 10-2T + 1.90 10-4T2 W/m°C, valid for the range 50 to 100 °C, T in °C. The
equation from Califano and Calvelo is at odds with most of the published data for moist food
materials in that it has a positive coefficient for the quadratic term. In general, the thermal
conductivity is shown to curve downward with temperature, much like the thermal
conductivity of water.
Based on the comparison of both methods, it was concluded that thermal contact was
adequate for the bead thermistor during the measurements of potato. This conclusion was
supported by visual examination of the samples after the measurements, where the sample

was sliced in half at the insertion point. The samples showed evidence of a thight fit around
the thermistor.
5.3.3 Beef Measurements
Figure 5.8 shows the thermal conductivity of beef. The data can be found in Appendix D
(Tables D.5 to D.8). Six samples were measured at each temperature. The values in the figure
represent the average thermal conductivity for each sample, found from three measurements.
Because some samples gave nearly identical values, their associated data markers can be hard
to distinguish. The sample averages are listed in Appendix D (Table D.5). Two samples were
excluded from contributing to the average, one sample at 25 °C with the bead thermistor
probe, and one sample at 50 °C with the line heat source probe. Both samples had a
substantially lower apparent thermal conductivity than the other samples. Post-test
examination of all samples revealed the presence of fat near the probe only in the two
samples that had shown a much lower thermal conductivity than the others.
Both methods show a larger variation than when used with the potato samples. This was
expected as the beef samples were less homogeneous than the potato samples. Surprisingly,
both methods measured with comparable replication. It was expected that the line heat source
probe would show better replication because of its larger sample volume. As was the case
with the potato samples, the line heat source probe yielded values that were higher than those
from the bead thermistor probe. The differences between the methods, as judged by the
average at 25 and 50 °C, was similar to those seen with potato. At 25 °C, the average for the
bead thermistor probe was 0.4937 W/m°C, 7.3% lower than the value of 0.5326 W/m°C for
the line heat source probe. At 50 °C, the bead thermistor probe measured an average value of
0.5084 W/m°C, 7.3% lower than the value of 0.5482 W/m°C, found with the line heat source
probe. At 100 °C, agreement between both methods was the best (quite unlike with the potato
samples). The average value of 0.4656 W/m°C with the bead thermistor probe is only 2%
lower than measured with the line heat source probe (0.4750 W/m°C). A possible explanation
as to why the line heat source probe tests with beef at 100 °C (actual bath temperature, 94 °C;

temperature rise during test was about 6 °C) did not give the inflated values as it did with
potato may be found in the textural differences. The elasticity of beef might have provided a
tighter seal around the probe whereas insertion into the cooked potato might have created a
small gap between the probe and the tissue. Agreement between the bead thermistor probe
measurements and the data given by Baghe-Khandan (1981a) for whole round beef is good.
The difference was less than 4%.
0.75
0.70
0.65
0.60
2 points
0.55
2 points
0.50
0.45
0.40
Line Heat Source Probe

0.35
Bead Thermistor Probe
Water (Nieto de Castro, 1986)
0.30
Beef, Whole Round (Baghe-Khandan, 1982)
0.25
10 20 30 40 50 60 70 80 90 100 110
Temperature [°C]
Figure 5.8 Thermal Conductivity of Lean Beef at 25, 50 and 100 °C, measured with the
Line Heat Source Probe and the Bead Thermistor Probe.
The thermal conductivity of beef appeared to show no clear trend with temperature. The bead
thermistor measured a slight increase in thermal conductivity from 25 to 50 °C. The same can
be said for the line heat source probe. At 50 °C, the composition and texture of beef was still
unchanged from that at 25 °C, and any effect of heating induced changes was not expected.
Denaturation of meat proteins did not take place until higher temperatures. Baghe-Khandan
(1981a) reported denaturation of beef muscle protein between 60 and 70 °C, leading to

shrinkage and a decreased moisture content. In the present experiments it was observed that
the beef samples shrunk and lost moisture as a result of the cook at 100 °C. The decrease in
thermal conductivity at 100 °C was attributed to the moisture loss that occurred as a result of
the heating. Other heating induced effects might not have played a role. Perez and Calvelo
(1984) took room temperature thermal conductivity measurements of beef samples that were
either subjected to a heat treatment at temperatures between 80 and 180 °C, or vacuum dried
at low temperatures. They concluded that "both the apparent density and thermal conductivity
of cooked beef are dependent on the water content of the product, being independent of the
thermal history to which it has been submitted".
5.3.4 Conclusion
Good agreement was found between the bead thermistor probe and the miniature line heat
source probe for the thermal conductivity of potato and lean beef at 25 and 50 °C. The
agreement between both methods at 100 °C was very good for beef and fair for potato. The
latter was attributed to problematic measurements with the line heat source probe. Based on
the results of the comparative study, thermal contact between the thermistor probe and both
potato and lean beef was considered adequate. Examination of the samples after the tests
supported this conclusion. The bead thermistor probe measured the thermal conductivity of
both materials with replication that was comparable to that of the line heat source probe.

5.4.1 Viscosity Measurements
The intended viscosities were 1, 5, 25, 50, 100 and 150 cp. The consistency coefficients of
the solutions in this study came close to these values. In the following discussion the
intended values are used. It should be understood that these refer to the corresponding actual
solutions. Data on the solutions are given in Table 5.5. The solutions below 100 cp were
almost Newtonian in behavior as witnessed from their near unity flow behavior indices. The
replication was good.
Table 5.5 Viscosity of sodium alginate solutions
Intended Measured value, Measured value, Average

viscosity first sample1 second sample1
consistency flow consistency flow consistency
coefficient, K behavior coefficient, K behavior coefficient, K
index, n index, n
[cp] [10-3 N·sn/m2] [10-3 N·sn/m2] [10-3 N·sn/m2]
1 0.55 2 0.55 2 0.55
5 4.7 1.010 4.9 1.009 4.8
25 24.9 0.993 24.2 0.998 24.6
50 48.6 0.991 47.9 0.998 48.3
100 101.6 0.970 101.4 0.963 101.5
150 151.4 0.967 154.2 0.968 152.8
1
Average of triplicate measurement.
2
Literature value (Incropera and DeWitt,1985, Appendix 6).

5.4.2 Effect of the Viscosity
The discussion pertains to analyses performed at a given value of the temperature step, ∆T.
The overall effect of viscosity, not differentiated with respect to ∆T is a meaningless statistic
in the context of this research.
The results of the multiple-comparison procedure with SAS are shown in Tables 5.6 and 5.7.
They present the Tukey's W groupings based on Rf∆T/Vss2 and S/I respectively. The columns
in the tables show the groupings for a particular value of the starting time of the regression
interval. Each of the three blocks in a column corresponds to one value of the temperature
step. The viscosities within each block are ordered by descending value of their associated
value of Rf∆T/Vss2 and S/I respectively. Those values can be found in Tables 5.8 and 5.9.
Viscosities within a block that are marked with the same letter did not test different.
Statistical differences between viscosities in a block were most likely caused by natural
convection.
An idea of the onset time of convection for the different viscosities and temperature steps,
could be obtained from Tables 5.9 and 5.10. When a solution of certain viscosity tested
significantly different from the more viscous solutions for successive start times of the
regression interval, the earliest start time at which that occurred could be used to make an
estimate of the onset time of natural convection in the solution of that viscosity. The start
times of these earliest intervals at which the solutions tested different (from the more
viscous) are shown in Table 5.10.

88
Table 5.6 Tukey’s W grouping for viscosity for each of the three temperature steps and different values for the start time (ST) of the
regression interval; SAS analysis was performed on the means of RDT/Vss2 (Table 5.8) corresponding to viscosity (and
DT); the means of RDT/Vss2 originate from linear regression over a 2 second wide interval.
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
A A A A A A A A A A A A A A A A A A
1.5 [°C] 100 100 100 5 100 5 100 100 100 100 150 100 100 150 150 100 150 150 150 A
A A A
5 5 5 150 A 5 A
100 A 5 A
5 A
150 A 150 A 50 A
25 A
50 A
50 A
100 A 150 A 25 A
100 A 50 A
150 A B 150 A 150 A 100 A 25 A

25 A
150 A 25 A
25 A
50 A
25 A
150 A 25 A
25 A
50 A
25 A
100 A 50 A
25 A
100 A B 25 A
50 A
25 A
150 A 150 A 25 A
50 A
5 A
25 A
100 A 50 A
150 A B 100 A 25 AB
50 A
50 A
25 A
100 A
AB A A A A A A
50 50 25 50 50 50 50 150 A 50 A
5 A
5 A
5 A
5 B
5 5 B
5 5 5 5
B
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
2.5 [°C] 25 150 25 150 100 25 150 150 100 25 150 25 50 100 25 150 150 150 100 A
150 A 25 A
150 A 25 A
50 A
150 A 25 A
100 A 50 A
50 A
50 A
150 A 100 A 150 A 50 A
100 A 50 A
100 A 150 A
A A A
5 5 50 100 A 25 A
50 AB
100 A 25 A
150 A 150 A 100 A 100 A 150 A 50 A
150 A 50 A
100 A 50 A
50 A
100 A 50 A
5 A
50 A
150 A 100 A B 50 A
50 A
25 A
100 A 25 A
50 A
25 A
25 A
100 A 25 A
25 A
25 A
25 A
A
50 100 A 100 A 5 A
5 A
5 B
5 5 5 5 5 5 5 5 5 5 5 5 5
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
5.0 [°C] 5 150 150 150 150 100 150 50 150 150 100 100 100 50 150 100 150 150 150 A
A A A A A
50 25 25 25 50 150 A 50 A
150 A 100 A 50 A
150 A 150 A 150 A 150 A 100 A 150 A 100 A 100 A 100 A B
100 B 5 A
50 A
50 A
100 A 50 A
100 A 100 A 50 A
100 A 50 A
50 A
50 A
100 A 50 AB
50 A
50 A
50 A
50 B
150 B C 100 A 100 A 100 A 25 A

25 A
25 A
25 A
25 A
25 A
25 A
25 25 25 25 B
25 25 25 25
BC A A
25 50 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
C
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
89
Table 5.7 Tukey’s W grouping for viscosity for each of the three temperature steps and different values for the start time (ST) of the
regression interval; SAS analysis was performed on the means of S/I (Table 5.9) corresponding to viscosity (and ∆T); the
means of S/I originate from linear regression over a 2 second wide interval.
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
1.5 [°C] 5 25 100 5 25 50 100 50 25 100 150 25 50 150 150 150 150 150 150 A
150 A 5 A
5 A
150 A 5 A
25 A
150 A 100 A 50 A
50 A
50 A
100 A 100 A 50 AB
100 A 100 A 25 A
100 A 50 A
A
25 150 A 50 A
25 A
100 A 5 A
25 A
5 A
150 A 150 A 25 AB
50 A
25 A
25 AB
50 A
25 A
100 A 50 A
25 A
A
50 100 A 25 A
50 A
50 A
150 A 5 A
25 A
100 A 25 A
100 A B 150 A 150 A 100 B 25 AB
50 A
50 A
25 A
100 A B
100 A B 50 A
150 A 100 A 150 A 100 A 50 A
150 A 5 A
5 A
5 B
5 A
5 5 5 B
5 5 5 5 B
B
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
2.5 [°C] 25 150 50 150 50 25 50 150 100 25 50 25 50 100 25 150 50 150 100 A
150 A 50 A
25 A
50 A
100 A 150 A 25 A
100 A 50 A
50 A
150 A 100 A 100 A 150 A 50 A
100 A 150 A 100 A 150 A
100 A 25 A
150 A 100 A 25 AB
50 A
150 A 50 A
150 A 100 A 100 A 50 A
150 A 50 A
150 A 50 A
100 A 50 A
50 A
A
50 100 A 100 A 25 A
150 A B 100 A 100 A 25 A
25 A
150 A 25 A
150 A 25 A
25 A
100 A 25 A
25 A
25 A
25 A
A A A A B
5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 5 5
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
5.0 [°C] 5 25 25 25 150 100 150 50 100 150 100 100 100 50 150 100 150 100 150 A
A
50 150 A B 150 A 150 A 50 A
150 A 50 A
150 A 150 A 100 A B 150 A B 50 A
150 A 150 A 100 A 150 A 100 A 150 A 100 A B
100 B 100 A B 100 A 100 A 100 A 50 A
100 A 100 A 50 A
50 AB
50 AB
150 A 50 A
100 A 50 AB
50 A
50 A
50 A
50 B
150 B C 50 AB
50 A
50 A
25 A
25 A
25 A
25 A
25 A
25 B
25 B
25 B
25 25 25 B
25 25 25 B
25
BC B
25 5 5 5 5 5 5 5 5 5 5 5 1 1 1 1 1 1 1
C
1 1 1 1 1 1 1 1 1 1 1 1 5 5 5 5 5 5 5
90
Table 5.8 Values of RfDT/Vss2 as related to temperature step and start time of the regression interval. The regression interval was 2
seconds wide. Values are ordered, per temperature step, in decreasing value from top down. The corresponding values for
viscosity can be found in Table 5.6.
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
1.5 [°C] 274.4 280.0 283.7 284.0 285.4 286.3 287.6 288.0 287.6 289.1 288.9 289.5 290.0 291.8 292.1 290.1 292.0 292.3 293.9
274.3 279.9 283.1 283.9 285.1 286.3 286.9 287.4 287.2 286.8 288.1 288.6 289.4 290.1 290.5 289.9 291.3 291.2 291.7
274.1 279.6 282.5 283.8 284.6 285.8 286.8 286.8 287.1 286.6 287.8 288.1 288.3 289.5 288.4 289.0 289.8 289.5 289.7
273.5 279.4 282.2 283.3 284.5 285.8 286.3 286.7 287.0 286.3 287.4 287.9 288.1 288.5 287.3 287.7 287.7 286.8 287.7
273.4 278.7 282.2 283.1 283.9 285.7 286.1 286.6 286.5 285.7 285.6 285.5 283.5 281.9 281.5 275.9 279.0 275.7 277.6
272.0 270.5 263.1 252.0 241.4 231.9 226.1 220.5 220.2 219.3 220.3 220.7 223.1 225.1 225.4 229.3 224.0 226.7 227.8
msd 2.25 3.01 3.85 4.19 4.46 4.34 3.51 3.99 3.62 4.10 3.64 5.60 4.86 3.84 5.96 9.43 7.05 7.51 8.98
% of max 0.8 1.1 1.4 1.5 1.6 1.5 1.2 1.4 1.3 1.4 1.3 1.9 1.7 1.3 2.0 3.3 2.4 2.6 3.1
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
2.5 [°C] 269.6 274.7 276.9 278.8 279.8 280.2 280.5 281.3 281.7 282.7 282.6 281.9 281.8 283.9 284.4 284.8 282.4 283.9 284.2
269.6 274.5 276.6 278.2 279.8 279.8 280.4 280.9 281.3 281.4 282.5 281.5 281.7 283.3 282.9 283.1 282.2 283.2 283.8
269.6 274.4 276.5 278.1 279.2 279.0 280.1 280.9 281.3 281.3 281.8 281.2 281.3 282.4 282.6 282.1 280.7 282.5 281.1
269.2 274.2 276.3 278.0 279.0 278.8 280.1 280.8 280.0 281.1 281.1 281.1 280.3 282.1 282.0 281.7 280.3 278.6 280.1
269.0 273.9 276.1 277.5 278.4 277.5 276.9 275.3 274.9 273.8 271.3 268.9 264.4 258.8 253.9 253.9 249.9 249.1 246.2
264.9 257.8 242.9 227.3 216.3 210.0 209.0 209.0 211.5 213.1 214.8 214.4 216.5 216.2 217.4 217.7 218.5 219.7 216.4
msd 1.36 1.73 1.99 2.32 1.49 1.98 1.80 2.34 2.40 2.47 2.40 4.05 2.85 3.45 4.38 3.72 4.23 5.56 5.05
% of max 0.5 0.6 0.7 0.8 0.5 0.7 0.6 0.8 0.9 0.9 0.8 1.4 1.0 1.2 1.5 1.3 1.5 2.0 1.8
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
5.0 [°C] 283.6 276.1 274.6 274.1 274.7 275.0 275.2 275.8 275.9 276.6 275.9 275.6 276.5 276.0 276.7 276.4 278.5 276.9 276.9
279.6 276.0 274.5 274.0 274.5 274.9 275.1 275.5 275.8 275.8 275.8 275.2 275.9 275.6 275.4 276.2 276.5 276.8 274.8
272.4 274.9 273.7 273.8 274.1 274.6 274.7 274.8 275.3 275.7 275.6 275.2 275.7 275.0 272.8 274.8 275.1 274.1 272.4
268.3 274.8 273.7 273.7 274.0 274.3 274.3 273.6 274.7 274.2 273.4 272.5 272.3 270.8 269.7 269.1 267.5 265.8 263.9
267.8 274.1 272.9 270.5 266.7 262.7 257.8 252.6 247.1 242.0 238.4 234.4 229.9 226.9 226.1 226.8 226.2 227.5 228.2
263.1 242.1 214.8 201.2 199.5 201.8 205.0 206.5 206.8 207.1 207.5 207.1 208.3 207.8 208.3 208.1 207.7 206.5 208.5
msd 6.84 3.33 2.48 1.92 2.12 2.21 2.48 2.61 2.60 2.64 2.63 2.50 2.64 2.72 4.15 3.97 3.62 4.49 3.78
% of max 2.4 1.2 0.9 0.7 0.8 0.8 0.9 0.9 0.9 1.0 1.0 0.9 1.0 1.0 1.5 1.4 1.3 1.6 1.4
91
Table 5.9 Values of S/I as related to temperature step and start time of the regression interval. Regression interval was 2 seconds
wide. Values are ordered, per temperature step, in decreasing value from top down. The corresponding values for viscosity
can be found in Table 5.7.
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
1.5 [°C] 0.4272 0.4687 0.4950 0.5027 0.5157 0.5291 0.5365 0.5403 0.5425 0.5552 0.5612 0.5647 0.5811 0.6084 0.6150 0.5771 0.6165 0.6230 0.6579
0.4265 0.4677 0.4945 0.5025 0.5136 0.5278 0.5338 0.5403 0.5378 0.5389 0.5597 0.5614 0.5705 0.5931 0.5789 0.5720 0.6142 0.5911 0.6276
0.4264 0.4655 0.4931 0.5025 0.5117 0.5269 0.5337 0.5400 0.5374 0.5331 0.5527 0.5567 0.5610 0.5793 0.5652 0.5710 0.5658 0.5855 0.5850
0.4261 0.4648 0.4917 0.5010 0.5089 0.5216 0.5333 0.5397 0.5360 0.5319 0.5321 0.5505 0.5488 0.5459 0.5421 0.5512 0.5505 0.5285 0.5222
0.4244 0.4643 0.4901 0.4959 0.5087 0.5215 0.5329 0.5311 0.5341 0.5173 0.5162 0.5140 0.4833 0.4575 0.4503 0.3496 0.4065 0.3443 0.3819
0.4167 0.4059 0.3436 0.2399 0.1314 0.0260 -0.0422 -0.1136 -0.1175 -0.1293 -0.1143 -0.1088 -0.0685 -0.0349 -0.0288 0.0493 -0.0594 -0.0018 0.0226
msd 0.0082 0.0135 0.0202 0.0280 0.0278 0.0297 0.0247 0.0336 0.0334 0.0500 0.0421 0.0778 0.0560 0.0588 0.0980 0.1557 0.1294 0.1406 0.1718
% of max 1.9 2.9 4.1 5.6 5.4 5.6 4.6 6.2 6.2 9.0 7.5 13.8 9.6 9.7 15.9 27.0 21.0 22.6 26.1
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
2.5 [°C] 0.4429 0.4792 0.4989 0.5167 0.5314 0.5329 0.5360 0.5451 0.5550 0.5663 0.5684 0.5560 0.5585 0.5921 0.5997 0.6033 0.5635 0.5872 0.6021
0.4420 0.4788 0.4988 0.5136 0.5309 0.5266 0.5354 0.5450 0.5516 0.5524 0.5623 0.5480 0.5547 0.5743 0.5765 0.5791 0.5582 0.5805 0.5850
0.4411 0.4787 0.4955 0.5128 0.5217 0.5227 0.5352 0.5442 0.5451 0.5469 0.5565 0.5475 0.5437 0.5684 0.5622 0.5621 0.5342 0.5714 0.5413
0.4406 0.4753 0.4944 0.5109 0.5187 0.5192 0.5350 0.5417 0.5304 0.5450 0.5440 0.5472 0.5308 0.5595 0.5596 0.5534 0.5273 0.4934 0.5228
0.4381 0.4735 0.4892 0.5008 0.5095 0.4994 0.4935 0.4729 0.4681 0.4540 0.4191 0.3835 0.3150 0.2222 0.1370 0.1371 0.0596 0.0715 -0.0075
0.4185 0.3651 0.2368 0.0864 -0.0315 -0.1037 -0.1159 -0.1168 -0.0809 -0.0569 -0.0302 -0.0365 -0.0008 -0.0069 0.0166 0.0247 0.0425 0.0451 -0.0150
msd 0.0081 0.0104 0.0128 0.0175 0.0128 0.0178 0.0201 0.0242 0.0298 0.0307 0.0308 0.0547 0.0437 0.0560 0.0714 0.0689 0.0776 0.1110 0.1044
% of max 1.8 2.2 2.6 3.4 2.4 3.3 3.8 4.4 5.4 5.4 5.4 9.8 7.8 9.5 11.9 11.4 13.8 18.9 17.3
ST [s] 1 2 3 4 5 6 7 8 9 10 11 12 14 16 18 20 22 24 26
5.0 [°C] 0.5987 0.5532 0.5398 0.5359 0.5401 0.5480 0.5467 0.5544 0.5590 0.5645 0.5596 0.5552 0.5700 0.5585 0.5681 0.5682 0.6035 0.5773 0.5724
0.5764 0.5523 0.5392 0.5339 0.5386 0.5431 0.5459 0.5500 0.5555 0.5581 0.5534 0.5458 0.5553 0.5502 0.5519 0.5587 0.5713 0.5722 0.5354
0.5289 0.5431 0.5338 0.5336 0.5378 0.5394 0.5444 0.5462 0.5474 0.5538 0.5516 0.5456 0.5536 0.5447 0.5025 0.5371 0.5430 0.5239 0.4876
0.4968 0.5337 0.5302 0.5313 0.5353 0.5387 0.5387 0.5309 0.5445 0.5377 0.5268 0.5129 0.5108 0.4861 0.4671 0.4562 0.4275 0.3930 0.3531
0.4952 0.5322 0.5153 0.4920 0.4521 0.4066 0.3488 0.2836 0.2100 0.1393 0.0865 0.0255 0.0299 0.0208 0.0310 0.0255 0.0168 -0.0106 0.0375
0.4726 0.3070 0.0638 -0.0743 -0.0930 -0.0642 -0.0222 -0.0017 0.0036 0.0086 0.0153 0.0085 -0.0461 -0.0983 -0.1122 -0.0984 -0.1114 -0.0842 -0.0667
msd 0.0460 0.0208 0.0128 0.0115 0.0161 0.0171 0.0191 0.0272 0.0221 0.0258 0.0288 0.0298 0.0348 0.0401 0.0721 0.0702 0.0643 0.0835 0.0778
% of max 7.7 3.8 2.4 2.1 3.0 3.1 3.5 4.9 4.0 4.6 5.1 5.4 6.1 7.2 12.7 12.4 10.7 14.5 13.6
Table 5.10 Earliest start time of regression interval (of 2 seconds wide) at which a viscosity
would test different (at the 5% level) of the more viscous solutions. Times are
given for both the analysis of RfDT/Vss2 and S/I
Viscosity 1 5 25
on RfDT/V ss
2
on S/I on RfDT/V ss
2
on S/I on RfDT/V ss
2
on S/I
1.5 °C 2 2 16 14
2.5 °C 1 1 7 6
1 1
5 °C x x 4 3 12 12
1
Initial transients for the temperature step of 5 °C lasted longer than one second. These
transients served to overcome the heat capacity of the thermistor. The analysis at 1
second therefore is not given any weight.
Since Tukey’s W groupings are shown at the start time of the regression interval, significant
differences reflect effects occurring within the following two seconds. A non-significant
difference compared to a more viscous solution could be interpreted as the absence of natural
convection up to the time shown. The analysis on S/I shows effects of convection for earlier
start times than the analysis for RfDT/Vss2. Table 5.11 gives the estimated onset time of
convection in thickened water at 50 °C, based on the analysis of S/I.
Table 5.11 Estimated onset time of convection in thickened water at 50 °C, [s].
Viscosity, cp 1 5 25
1.5 °C <3 15
2.5 °C <3 7
5 °C <3 4 13
Table 5.11 shows the delaying effect of an increase in viscosity on the onset of convection,
and an accelerating effect of an increase in the temperature step. According to the statistical
analysis, thickened water samples with a viscosity of at least 50 cp can be measured with a

temperature step of up to 5 °C, without significant convection. For a temperature step of 1.5
or 2.5 °C, this threshold can be lowered to 25 cp.
5.4.3 Effect of the Temperature Step
Table 5.8 suggested that the temperature step had an effect on the value of R∆T/Vss2. This
was noticed when these values were compared for the 150 cp solution, in which convection
had not taken place. The model for the thermistor based method did not predict such an
effect. In order to determine if there was a statistical correlation between the magnitude of the
temperature step and the value of R∆T/Vss2, the data of the 150 cp solution were used in a
multi-comparisons test. The procedure was much like the one that was used for estimating
the onset of convection. A regression interval of 10 seconds was used to smooth the effect of
the noise in the data. The linear regression was done on successive intervals whose start time
was incremented by one second, until it reached the value of 19. The regression results
formed the input to Tukey's W test, with a type 1 error of 0.05.
According to the theory, the steady state power dissipation of the probe should be a linear
function of the temperature step, ∆T. If that were the case, and when convection was
negligible, the values for RfDT/Vss2 and S/I should be independent of the size of the
temperature step. The values of RfDT/Vss2 and S/I are shown in Figures 5.9 and 5.10
respectively. The data in the figures originated from the above described regression
procedure. If measurement followed theory, the lines for the three temperature steps would be
(nearly) identical. They were not and the difference between the lines was statistically
significant for times larger than 2 seconds (Table 5.12).

292
1.5 °C
290
2.5 °C
288 5 °C
286
[°C/W]
284
282
2
R∆T/Vss
280
278
276
274
272
0 5 10 15 20
Start Tim e of Regression Interval [s]
Figure 5.9 Rf∆T/Vss2 for the 150 cp solution and temperature steps of 1.5, 2.5 and 5 °C.
Data originated from linear regression over a 10 second interval. They are the
average of 2 samples.
0.6000
1.5 °C
2.5 °C
0.5750
5 °C
0.5500
S/I [s ]
1/2
0.5250
0.5000
0.4750
0.4500
0 2 4 6 8 10 12 14 16 18 20
Figure 5.10 S/I for the 150 cp solution and temperature steps of 1.5, 2.5 and 5 °C. Data
were obtained from linear regression over 10 second intervals, with an
increment in its start time of 1 second. They are the average of two samples.

Figure 5.9 shows three lines that are almost equidistant from 4 seconds on. The lines for 1.5
and 5 °C are about 2% higher and lower respectively than the line for 2.5 °C. The differences
between these lines was significant according to the multiple-comparison procedure, Tukey's
W. This effect of the magnitude of the temperature step had not been reported by other
researchers. Valvano (1981) measured the thermal conductivity and diffusivity of water with
a temperature step ranging from 1 to 10 °C. He found no effect of the magnitude of the
temperature step on the observed thermal properties. Reproducibility, however, increased
with the magnitude of the temperature step. Dougherty (1984) found no difference in
measured thermal conductivity when using either 2 or 3 °C as temperature step. Based on this
study however, the probe should be used with the same temperature step as during
calibration. If the probe had been calibrated at 1.5 °C, and used with a step of 5 °C, the
thermal conductivity of water would be overestimated by about 10%. A probe calibrated at
2.5 °C can be used over the range 1.5 to 5 °C, with a temperature step induced error from
about -4-5 to +4-5% in estimated thermal conductivity of water.
The estimation of thermal diffusivity did not display the same sensitivity to the size of the
temperature step. Figure 5.10 shows how the lines - and thus the response of the thermistor -
displayed very much the same trend for the temperature step of 1.5 and 2.5 °C and converge
for longer times. The difference between the lines before convergence is only about 2%, but
statistically significant, at the 5% level, up to a starting time of 10 seconds (an exception are
the results for a starting time of 6 seconds). The line for the 5 °C temperature step doesn't
follow a similar trend as the others, and seemed to cross them after about 12 seconds. From
10 seconds on, all temperature steps tested the same at the 5% level.
The 100 cp solution was basically convection free as well. As such it could be used to test the
results of the 150 cp solution on representativeness. When the results for the 100 cp solution
were added to Figures 5.9 and 5.10, they confirmed those of the 150 cp solution.
The cause of the observed effects was unknown. It could be either that model inaccurately
predicted a linear relationship between the temperature step and the power dissipation in the

probe, or that the observed differences were caused by the instrumentation. Either way, it
appears advisable to use a probe with the same temperature step as in calibration. This
research did confirm Valvano’s conclusion that a higher value for the temperature step
improved the reproducibility, as indicated by the lower values of msd for higher values of the
temperature step (Table 5.8 and 5.9).

97
Table 5.12 Tukey’s W grouping for 150 cp solution for the effect of the magnitude of the temperature step on the value of R∆T/Vss2
and S/I
Effect of Temperature Step on Value of R∆T/Vss2 Effect of Temperature Step on Value of S/I
ST ∆T1 R∆T/Vss2 msd ST ∆T1 S/I msd
[s] [°C] [°C/W] [°C/W] [s] [°C] [s1/2] [s1/2]

1 1.5 2.5 5.0 281.8 276.3 274.5 0.92 1 5.0 2.5 1.5 0.5383 0.4873 0.4771 0.0080
2 1.5 2.5 5.0 284.3 278.5 275.0 0.50 2 5.0 2.5 1.5 0.5430 0.5111 0.5034 0.0032
3 1.5 2.5 5.0 285.6 279.7 275.0 0.52 3 5.0 2.5 1.5 0.5427 0.5244 0.5180 0.0032
4 1.5 2.5 5.0 286.4 280.5 275.2 0.58 4 5.0 2.5 1.5 0.5453 0.5339 0.5274 0.0032
5 1.5 2.5 5.0 287.0 281.0 275.5 0.56 5 5.0 2.5 1.5 0.5490 0.5402 0.5350 0.0038
A A
6 1.5 2.5 5.0 287.5 281.4 275.6 0.65 6 5.0 2.5 1.5 0.5516 0.5462 0.5417 0.0046
A A
7 1.5 2.5 5.0 287.8 281.8 275.8 0.60 7 5.0 2.5 1.5 0.5545 0.5512 0.5455 0.0042
A A
8 1.5 2.5 5.0 288.1 282.1 276.0 0.56 8 5.0 2.5 1.5 0.5567 0.5550 0.5495 0.0038
A A
9 1.5 2.5 5.0 288.4 282.3 276.1 0.61 9 2.5 5.0 1.5 0.5583 0.5581 0.5533 0.0042
A A A
10 1.5 2.5 5.0 288.6 282.5 276.2 0.60 10 2.5 5.0 1.5 0.5610 0.5592 0.5573 0.0046
A A A
12 1.5 2.5 5.0 289.2 282.8 276.5 0.62 12 2.5 1.5 5.0 0.5660 0.5655 0.5649 0.0063
A A A
15 1.5 2.5 5.0 289.7 283.3 276.8 0.84 15 2.5 1.5 5.0 0.5752 0.5747 0.5707 0.0110
A A A
18 1.5 2.5 5.0 290.1 283.6 277.0 1.06 18 1.5 2.5 5.0 0.5825 0.5802 0.5735 0.0150
A A A
19 1.5 2.5 5.0 290.2 283.7 277.0 0.87 19 1.5 2.5 5.0 0.5835 0.5823 0.5732 0.0113
1
Values that are marked with the same letter are not different at 5% level (Tukey)
Results and Discussion

5.4.4 Rayleigh Number Correlation
Rayleigh numbers were calculated for the viscosity-temperature step combinations that
appeared to exhibit natural convection. Figure 5.11 shows the Rayleigh numbers as function
of time. An attempt was made to find a critical Rayleigh number above which convection
would be likely. Table 5.11 was compared to Figure 5.11. The Rayleigh number
corresponding to estimated onset time of convection was, in increasing order, 43 (5 cp and
5 °C), 45 (25 cp and 5 °C), 48 (5 cp and 2.5 °C), and 84 (5 cp and 1.5 °C). In light of the
coarseness of the procedure with which the onset times were estimated, this range was
considered to be narrow. It appears then, that when the simulated Rayleigh number remained
under 43, convection had negligible effect.
90
80
70
60
Rayleigh Number
50
1.5 °C, 1 cps

40
1.5 °C, 5 cps
30 2.5 °C, 5 cps
2.5 °C, 25 cps
20
5 °C, 5 cps
5 °C, 25 cps
10
5 °C, 50 cps
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Time [s]
Figure 5.11 Simulated Rayleigh numbers for viscosity-DT combinations that appeared to
exhibit signs of convection during measurement of thickened water of various
viscosity.

5.4.5 Validation of Rayleigh Number Correlation
The Rayleigh number of 43, below which convection seemed unlikely, was derived from the
experiments in thickened water, and for a specific probe. Data gathered with another probe in
glycerol were scrutinized to investigate whether this Rayleigh number of 43 could be applied
to a different probe and a different test medium. A sample of glycerol was measured over the
range 100 to 150 °C in 10 °C increments, without any means of controlling convection in the
sample. The power response in the sample was compared to that in two other samples of
glycerol that were measured at the same temperatures, with the same probe and whereby
convection was inhibited (or limited) with 'angel hair' (in concentrations small enough as to
be considered of negligible impact on the thermal properties of interest). The onset time of
convection in sample #1 was estimated from comparison to samples #2 and #3. Linear
regression was performed over a two second wide interval with all the data files gathered.
The starting time of the regression interval was incremented by 1 second until it reached the
value of 27. The regression results were used in a t-test (Excel 97, Microsoft Corporation,
Seattle, WA) with type 1 error of 0.05. A significant difference between the mean of sample
#1 and that of samples #2 and #3 combined, was taken as indicative of convection in sample
#1. Onset time for convection in sample #1 was estimated from the regression results. Table
5.13 contains the estimated onset times for the 6 temperatures at which the glycerol was
tested.
Next the temperature distribution in the medium was simulated to find the Rayleigh numbers
corresponding to the onset times of Table 5.13. The results for S/I were slightly more
sensitive to convection. The estimates listed in that column were used. The relevant input
parameters and their values for simulation of the Rayleigh number are given in Table 5.14.
The probe parameters correspond to the regression interval 10-20 seconds. Rayleigh number
calculations were done at temperatures from 100 to 150 °C.

Table 5.13 Estimated onset time of convection in Glycerol sample #1, measured with probe
#H1.
Estimated onset time of convection

Temperature on RDT/Vss on S/I
[°C] [s] [s]
100 12 10
110 9 8
120 8 7
130 7 6
140 6 5
150 5 4
Table 5.14 Parameter values for simulation of temperature profiles in Glycerol with program
MEDTEMP.PAS
Parameter Simulation Temperature, °C

100 110 120 130 140 150
1
k , W/m°C 0.2947 0.2960 0.2973 0.2986 0.2998 0.3011
α , 10 m /s
2 -7 2
0.908 0.909 0.911 0.915 0.920 0.927
ρ , kg/m
3 3
1207.6 1200.5 1193.3 1186.0 1178.5 1171.0
β4 104, 1/°C 5.85 5.97 6.09 6.21 6.33 6.46
µ , cp
1
15 10.8 8.1 6.2 4.8 3.8
a1, mm 1.0687 1.0694 1.0778 1.1034 1.0991 1.1074
kb1, W/m°C 0.07332 0.07507 0.07566 0.07715 0.07964 0.07968
a2, mm 0.7609 0.7687 0.7718 0.7464 0.7661 0.7661
kb2, W/m°C 0.11081 0.11093 0.11066 0.13936 0.13957 0.14820
1
Adapted from CINDAS (1988).
2
Calculated as α = k/ρ Cp, α = thermal diffusivity, k = thermal conductivity, ρ = density, Cp =
specific heat.
3
Measured (Appendix A).
4
Appendix A.

Figure 5.12 shows the Rayleigh numbers in glycerol at temperatures from 100 to 150 °C.
Rayleigh numbers corresponding to the estimated onset times of Table 5.13, vary from a low
of about 46 at 110 °C to a high of 55 at 140 °C.
80
100 °C
70 110 °C
60 120 °C
Rayleigh Number
50 130 °C
140 °C
40
150 °C
30
20
10
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time [s]
Figure 5.12 Simulated Rayleigh numbers for measurements in glycerol at temperatures

from 100 to 150 °C, with a temperature step of 2.5 °C.
The Rayleigh numbers at the onset of convection in water varied from 43 to 82, those for the
sample of glycerol between 45 and 55. Considering the coarseness with which the onset time
of convection was estimated, this was considered a good result. A comparison of the numbers
to literature is of limited value. The Rayleigh numbers of this study are dependent on the
values of the input parameters. If for instance the penetration depth had been defined as
1/1000th of the surface temperature instead of 1/100th, much larger values of the critical
Rayleigh number would have resulted. Using different probe parameters will give a different
outcome. However when probes of the same series were used and their properties were taken
from the same interval, the variation in Rayleigh numbers was small. The properties of four

different probes were taken from the same regression interval. Rayleigh numbers were
generated for each probe in glycerol at three temperatures. Rayleigh numbers for different
probes were only 1 or 2% different. A larger effect was found when the properties of an
extreme interval were used. That resulted in a 10-15% change in the simulated Rayleigh
numbers. The net effect of that was a time shift of about 0.5-1 seconds before a Rayleigh
number of 40 was reached (or 30, 50, 60). Whether the correlation can be used for a larger
diameter probe is not sure. It is felt however, that similar values for the critical Rayleigh
number will be found.
5.5 EVALUATION AT HIGH TEMPERATURE
5.5.1 Tomato Concentrate
The solids content of tomato concentrate was found from refractometry to be 86% for the
tomato puree and 74% for the tomato paste. These values were used to compare the measured
values for the thermal properties of this study to those calculated with the equations from
Choi and Okos (1983).
The thermal properties of tomato concentrate were calculated from the results of linear
regression. Based on the results for the calibration of probe #H3, a regression interval of 1-11
seconds was used for estimation of thermal conductivity. The interval for estimation of
thermal diffusivity depended on the temperature as was discussed earlier in this Chapter
("5.2.2 High Temperature Probes"). At 100 °C, the interval 11-21 seconds was used, at
130 °C the interval 6-16 seconds, and at 150 °C the interval 3-13 seconds.
Thermal Conductivity and Thermal Diffusivity of Tomato Puree
The thermal conductivity and diffusivity of tomato puree are given in Tables 5.15 and 5.16.
The table lists the average for each sample over the three measurements that were taken with
each sample at each temperature. The coefficient of variation (COV) of the three

measurements is included in the table. A low value is an indication of unchanged conditions
during the test. A high value may suggest that the product was experiencing changes as a
result of the high temperature cook, or that the instrumentation malfunctioned.
Table 5.15 Average thermal conductivity and coefficient of variation (COV) of tomato puree
(solids content 14% w/w), at high temperature
100 °C 130 °C 150 °C

Sample Number Thermal COV Thermal COV Thermal COV
Conductivity Conductivity Conductivity
[W/m°C] [%] [W/m°C] [%] [W/m°C] [%]

1 0.6310 0.1 0.6476 0.6 0.6282 0.7
2 0.6293 0.3 0.6480 0.5 0.6455 1.5
3 0.6257 0.4 0.6409 0.7 0.6617 1.8
4 0.6256 0.4 0.6409 0.5 0.6623 0.6
Average 0.6279 0.6443 0.6494
StdDev. 0.0027 0.0040 0.0162
Choi and Okos (1983) 0.6176 0.6305 0.6335
Difference, % 1.7 2.2 2.5
The thermal conductivity is close to the values predicted by the equation for Choi and Okos.
The measured values are 1.7, 2.2 and 2.5% higher than the literature values at 100, 130 °C
and 150 °C respectively. The difference between the sample means at each temperature
increased from about 1% at 100 °C to about 4% at 150 °C. This is reflected in the increasing
value for the standard deviation.

Table 5.16 Average thermal diffusivity and coefficient of variation (COV) of tomato puree
100 °C 130 °C 150 °C

Diffusivity Diffusivity Diffusivity
[10-7m2/s] [%] [10-7m2/s] [%] [10-7m2/s] [%]

1 1.55 0.2 1.53 0.9 1.52 0.5
2 1.59 0.6 1.56 0.3 1.53 2.0
3 1.74 0.8 1.80 0.5 1.74 2.8
4 2.07 10.4 1.86 7.0 2.36 2.9
1
Average 1.63 1.63 1.60
1
StdDev. 0.10 0.15 0.12
Choi and Okos (1983) 1.52 1.58 1.60
Difference, % 7.2 3.2 0.0
1
The average and standard deviation excludes sample #4.
The measurement of thermal diffusivity was less repeatable than that of thermal conductivity
as shown by the standard deviation of the sample means. One sample (#4) was an outlier, and
excluded from the average and standard deviation at each temperature. With this sample
removed, the average values at each temperature compared favorably to the literature values.
The thermal diffusivity in this study was 7.2 and 3.2% higher at 100 and 130 °C respectively.
At 150 °C, the measured value and literature value were identical.
Thermal Conductivity and Thermal Diffusivity of Tomato Paste
A total of six samples were used in the tests. The main reason for exceeding the initially
planned total of four were the poor results obtained with two of the samples. These two
outliers were removed from the analysis. A cause for the poor results could have been the
inclusion of air pockets near the probe. The remaining four samples gave acceptable results at
100 °C. At higher temperatures a few gave acceptable results, others not. Product changes
due to prolonged heating at high temperatures were suspected as the cause for the large
variability in the results. Some samples were re-tested at 100 °C following the tests at 150 °C

to see if the heating had caused irreversible effects. Not surprisingly the results of these
measurements differed greatly from the earlier tests at 100 °C. The apparent thermal
conductivity of the samples had increased substantially. Inspection of the tested tomato paste
samples showed that the paste had hardened to a brittle material. Some samples had a fracture
near the location of the bead suggesting the existence of water air/gas pockets there (pressure
buildup inside the chamber was noticed on several occasions).
The procedure had been to measure a given sample at 100 °C, then at 130 °C, and finally at
150 °C. The samples were thus subjected to prolonged heating. In an alternative approach, a
few samples were tested by placing the room temperature samples in the bath at 150 °C.
Doing so reduced the time during which the product was subjected to the high temperature to
the 20 minutes required for thermal equilibration. This approach however did not improve
the measurements.
The thermal conductivity and diffusivity of tomato paste is given in Tables 5.17 and 5.18. At
100 and 130 °C, the thermal conductivity compared well against the literature data. At
100 °C, a mean value of 0.5828 W/m°C was measured, which was only 3.3% higher than that
calculated with the equation from Choi and Okos. The mean value of 0.5942 at 130 °C was
only 1.8% higher than the literature value. The coefficient of variation (COV) for the samples
and the standard deviation of the sample means at 100 and 130 °C were also good. The
overall mean at 150 °C, 0.5457 W/m°C, was 7.8% lower than that found from the equation of
Choi and Okos. The sample means at this temperature showed a large variation. Samples #2
and #3 had a low apparent thermal conductivity. Sample #4, on the other hand had a high
thermal conductivity. The low values for samples #2 and #3 may have been caused by
changes of the paste (hardening) which led to a localized drop in moisture content. The fairly
high value of sample #4 could have been caused by similar effects but with the formation of a
water filled cavity near the thermistor.

Table 5.17 Average thermal conductivity and coefficient of variation (COV) of tomato paste
100 °C 130 °C 150 °C

Conductivity Conductivity Conductivity
[W/m°C] [%] [W/m°C] [%] [W/m°C] [%]

1 0.5817 0.4 0.5909 0.7 0.5820 1.6
2 0.5826 0.5 0.5646 0.3 0.4166 0.3
3 0.5835 0.1 0.6097 0.1 0.4987 1.3
6 0.5836 0.4 0.6114 0.7 0.6856 0.5
Average 0.5828 0.5942 0.5457
StdDev. 0.0009 0.0218 0.1151
Choi and Okos (1983) 0.5643 0.5836 0.5920
Difference, % 3.3 1.8 -7.8
Table 5.18 Average thermal diffusivity and coefficient of variation (COV) of tomato paste
100 °C 130 °C 150 °C

Diffusivity Diffusivity Diffusivity
[10-7m2/s] [%] [10-7m2/s] [%] [10-7m2/s] [%]

1 1.67 2.3 1.69 1.7 1.75 2.6
2 1.71 1.5 2.17 1.7 2.51 7.5
3 1.55 0.3 1.75 2.4 2.47 3.3
6 1.59 1.4 1.79 1.1 3.95 17.0
Average 1.64 1.85 2.67
StdDev. 0.08 0.22 0.92
Choi and Okos (1983) 1.38 1.46 1.50
Difference, % 18.8 26.7 78.0

The thermal diffusivity of tomato paste was very different from the literature data. At 100 °C,
the measured data exceeded the literature value by 18.8%. At 130 and 150 °C the difference
with the literature data was 26.7 and 78% respectively. Heating induced changes in the
tomato paste were the cause of the large discrepancies between the measured values and the
literature data. These changes were mentioned earlier. The effect of these changes on the
measurement of thermal diffusivity were much more pronounced than on the measurement of
thermal conductivity because of the way by which each property is calculated (Chapter 3).
Conclusion
The measurements of tomato puree gave better results than the measurements of tomato
paste. The higher moisture content of puree probably prevented formation of the hard lumps
that were noticed in the tomato paste samples upon examination after the measurements. The
puree thus remained more homogeneous. The measurement of thermal properties of the
tomato paste might have been more successful if the paste could have been heated faster. The
measurement of thermal diffusivity was more sensitive to (suspected) product variations and
other sources of variation (like noise) than the measurement of thermal conductivity.
5.5.2 Nutritional Supplement
Problems with the thermistor probes were encountered during the measurements. About
halfway into the research with the nutritional supplement, the calibrated probe (#TT2)
suffered catastrophic failure. No other calibrated probe was available. Time constraints,
dictated by the limited shelf life of the products, prohibited calibrating another probe. An
uncalibrated probe (#TT5) was used, with the intention to calibrate it after all the tests were
done. Unfortunately, this probe failed before it could be calibrated.
The properties of the uncalibrated probe at 150 °C were estimated using the results of the two
products that were measured with both the calibrated and uncalibrated probe ("TC3", 40%
solids, and "TC4", 30% solids). Property values of previously used probes were tried for the

uncalibrated probe. Those that yielded thermal conductivity values that best matched those
found with the calibrated probe were used to calculate the thermal conductivity at 150 °C for
all the samples that were measured with the uncalibrated probe. The closest match was found
with the properties of probe #H3. The probe whose calibrated properties at 150 °C were
substituted for those of the uncalibrated probe, also supplied the probe properties for the
uncalibrated probe at 95 °C. These properties were however obtained through calibration at
100 °C.
Thermal conductivity values for the fluid nutritional supplement as related to solids content
and temperature are given in Table 5.23. Reference values for water are given for
comparative purposes. The thermal conductivity of moist foods is strongly influenced by
their water content. Data gathered with the uncalibrated probe are indicated with a
superscript. Because of the assumed probe properties for the uncalibrated probe, the thermal
conductivity values obtained with it are less reliable than those found with the calibrated
probe.
For the nutritional supplement, no published thermal property data were available. The
accuracy of the values obtained thus can not be directly assessed. The values appear to be
reasonable. They follow an expected trend with respect to the effect of the solids content. An
increase in solids content appears to decrease the thermal conductivity of the products. This
finding is consistent with the commonly found positive relationship between thermal
conductivity and moisture content (Sweat, 1986). The replication is excellent (within 1%) for
the measurements at 95 °C at all solids levels and for those at 150 °C at solids levels of 25%
and below. The replication at 150 °C for the higher solids levels is worse than those at lower
solid levels and temperature. This is probably due to heat induced changes in the product
which are more pronounced at higher solids levels. A similar effect was observed with
tomato concentrate. Tomato paste yielded less accurate data (as compared to the literature
data) and displayed worse replication than tomato puree.

Table 5.19 Thermal Conductivity of "TC1", "TC2", "TC3" and "TC4" at High
Temperatures, [W/m°C]
TC1 TC2 TC3 TC4

Solids 95 °C 150 °C 95 °C 150 °C 95 °C 150 °C 95 °C 150 °C
content
[% wb]
15 0.6216 0.6152 0.6034
0.6170 0.6031
25 0.5594
0.5575
30 0.5527 0.5520 0.5501 0.4955 0.5230
0.5194b
35 0.5342
40 0.4864a 0.5047b 0.5048 0.5092 0.4575 0.4549a 0.4625b

0.4856a 0.4912b 0.4663b 0.4534a
50 0.4132a 0.4330b
0.4116a 0.4204b
water, 0.6762 0.6820 0.6762 0.6820 0.6762 0.6820 0.6762 0.6820
literature
value
a)
Measured with probe #TT5, properties taken from probe #H3 at 100 °C.
b)
Measured with probe #TT5, properties taken from probe #H3 at 150 °C.
5.5.3 Conclusions of High Temperature Measurements
The thermistor based method was found to be suitable for the measurement of thermal
conductivity and thermal diffusivity of tomato concentrate and of the nutritional supplement
at temperatures up to 150 °C. For the most part, the measurement data for tomato concentrate
agreed well with the published values. For those data, predominantly at 150 °C, were this
was not the case, the cause of disagreement was more likely related to heating induced
changes in the product, than to problems with the thermistor based method itself. The
nutritional supplement was measured with good replication. Measured values for thermal
conductivity followed the expected trend with moisture content.

CHAPTER 6
REMARKS ON CALIBRATION
The calibrated probe properties at high temperature, discussed in the previous Chapter ("5.2.2
High Temperature Probes"), depend on the part of the data (regression interval) that was used
for estimating the properties. The estimation errors of the calibration media also varied with
the regression interval. The dependency of these estimation errors on the regression interval
led to the selection of a certain intervals for the evaluation of thermal properties in
subsequent measurements of tomato concentrate and the nutritional supplement. These
intervals, dubbed optimal intervals, were selected based on the low estimation errors obtained
in the calibration. The optimal interval for the estimation of thermal conductivity was
different from the optimal interval for estimation of thermal diffusivity. Also, the optimal
interval for estimation of thermal diffusivity was different for the each of the six temperatures
of calibration. An evaluation interval that depends on the property of interest and on the
temperature is impractical. It was of interest to investigate the cause of the observed
dependency of estimation errors on the regression interval. A more practical calibration
procedure might follow out of that investigation.
The response characteristics of the thermistor in the three calibration media was analyzed. A
plot of the voltage response for different media doesn’t show the differences very well. The
derived properties R∆T/Vss2 and S/I much better illustrate the differences between the media.
Figures 6.1 and 6.2 show the values for these terms for the three calibration media. They
were used in estimating the probe properties of probe #H3 at 100 °C. Figures 6.3 and 6.4
present these data in a different format. The values for each interval are compared to those for
the interval 10-20 seconds.
Remarks on Calibration 110

750
700
650
600
Water
[°C/W]
550 Glycerol
HTF 500
500
2
R∆T/Vss
450
400
350
300
250
0 2 4 6 8 10 12 14 16 18 20
Figure 6.1 Value of R∆T/Vss2 for Probe #H3 at 100 °C for the three calibration materials
and a regression interval of 10 seconds. The average of three samples is shown.
1.6
1.5
Water
Glycerol
Ratio of Slope and Intercept [s ]
1.4
1/2
HTF 500
1.3
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0 2 4 6 8 10 12 14 16 18 20
Figure 6.2 Ratio of Slope and Intercept, S/I, for Probe #H3 at 100 °C for the three
calibration materials and a regression interval of 10 seconds. The average of
three samples is shown.

8
7 Water
Glycerol
6
HTF 500
Difference in Value [%]
5
0
0 2 4 6 8 10 12 14 16 18 20
-1
-2
-3
Figure 6.3 Difference in the value of R∆T/Vss2 from its value at ST = 10 seconds at 100
°C for the three calibration materials. The average of three samples is shown.
25
W ater
20 Glycerol
HTF 500
15
Difference in Value [%]
10
0
0 2 4 6 8 10 12 14 16 18 20
-5
-10
-15
Figure 6.4 Difference in the value of S/I from its value at ST = 10 seconds at 100 °C for
the three calibration materials. The average of three samples is shown.

Some conclusions can be drawn from these figures. According to the model, the power
dissipation in the probe should be linear in the inverse square root of time. In that case, the
linear regression results should have been independent of the regression interval; the lines in
Figures 6.1 through 6.4 should have been horizontal, with those of Figures 6.3 and 6.4
coinciding with the x-axes. The deviation from the model is most noticeable for early
intervals. At later times, the response in all fluids appeared to follow the model more closely.
The values of S/I showed a larger relative change with successive regression intervals than
the values of R∆T/Vss2. The response in the low thermal conductivity fluid, HTF 500, is
fundamentally different from that in either of the two fluids with a higher thermal
conductivity. The cause of this was not clear. The practical implication is that the calibration
of a probe for use with moist food materials might be better done with water and glycerol
only. The thermal properties of moist foods generally fall between those of water and
glycerol.
Tomato puree was measured in a study discussed earlier. The data from one sample were
used to compare the calibration of a probe (#H3) with water and glycerol only, to one where
HTF 500 was added as calibration material. Figures 6.5 to 6.8 show the apparent thermal
properties of tomato puree at 100 °C. The probe (#H3) was calibrated with water and glycerol
(Figures 6.5 and 6.6) or with water, glycerol and HTF 500 (Figures 6.7 and 6.8). Markers
indicating a plus or minus 1% deviation from the graphed values are shown in the figures to
aid in the interpretation.
Figures 6.5 and 6.6 clearly show that the response of the thermistor in the tomato puree
sample was quite similar to that in water or glycerol. The measured value of thermal
conductivity is virtually independent of the interval used to calculate the value. The thermal
diffusivity shows more variation with interval, but the variation is very small up to the
interval 18-28 seconds.

0.650
Thermal Conductivity - 1% + 1%
0.648
0.646
0.644
0.642
0.640
0.638
0.636
0.634
0.632
0 2 4 6 8 10 12 14 16 18 20
Figure 6.5 Thermal Conductivity of Tomato Puree at 100 °C, measured with a probe
(#H3) that was calibrated with water and glycerol.
1.61 E-07
Thermal Diffusivity - 1% + 1%
1.60 E-07
Thermal Diffusivity [m2/s]
1.59 E-07
1.58 E-07
1.57 E-07
1.56 E-07
1.55 E-07
1.54 E-07
1.53 E-07
0 2 4 6 8 10 12 14 16 18 20
Figure 6.6 Thermal Diffusivity of Tomato Puree at 100 °C, measured with a probe (#H3)
that was calibrated with water and glycerol.

0.670
0.665
0.660
0.655
0.650
0.645
0.640
0.635
0.630
0.625 Thermal Conductivity - 1% + 1%
0.620
0 2 4 6 8 10 12 14 16 18 20
Figure 6.7 Thermal Conductivity of Tomato Puree at 100 °C, measured with a probe
(#H3) that was calibrated with water, glycerol and HTF 500.
1.60 E-07
1.58 E-07
1.56 E-07
Thermal Diffusivity [m2/s]
1.54 E-07
1.52 E-07
1.50 E-07
1.48 E-07
1.46 E-07
1.44 E-07
1.42 E-07 Thermal Diffusivity - 1% + 1%
1.40 E-07
0 2 4 6 8 10 12 14 16 18 20
Figure 6.8 Thermal Diffusivity of Tomato Puree at 100 °C, measured with a probe (#H3)
that was calibrated with water, glycerol, and HTF 500.

Figures 6.7 and 6.8 demonstrate the effect of adding HTF 500 as calibration material. The
line in Figure 6.7 resembles that of the estimation errors for water that were shown in Figure
5.1. Except for the first two intervals, the calculated properties are only mildly dependent on
interval, and probably overestimated the actual values. The thermal conductivity of the
tomato puree sample found with the three-media calibration was about 3% higher than found
with the two-media calibration (excluding the values for starting times 1 and 2 seconds). The
value for thermal diffusivity was dependent on the interval from which it was obtained. As
was the case for the thermal conductivity, the graph resembles that shown for the estimation
errors for water (Figure 5.2). This reflects the high water content of tomato puree,
approximately 86%. The difference between the low value, at a starting time of 3 seconds,
and the high value, at a starting time of 14 seconds, was 6.9%. Figure 5.2 suggested the use
of an interval with a starting time in the range 10 to 19 seconds (this reduced all estimation
errors to below 2%). Doing so yielded thermal diffusivity values in the range 1.53 to 1.58
[10-7m2/s]. In comparison, the two-media calibration yielded values in the range 1.55 to 1.58
[10-7m2/s] (excluding the value for the last interval).
When the thermal properties of a tomato puree sample at 130 °C were calculated, similar
results were obtained as in Figures like 6.5 to 6.8. Again, the two media calibration showed
hardly any (estimation of thermal conductivity) to some (estimation of thermal diffusivity)
variation with the interval used for calculation. Estimation of thermal diffusivity with the
three-media calibration displayed a stronger relationship with regression interval than at
100 °C.
These findings suggest that a two media calibration with water and glycerol will give better
results when the thermistor probe is calibrated for thermal property measurements of moist
foods. With a two media calibration there was no need to pick an optimal regression interval
for each different temperature, as was the case with the three media calibration.
The accuracy with which the calibrated probe can determine the properties of a medium is
related to the accuracy of the data for the reference materials. The properties of water are well

documented. Those of glycerol were probably somewhat less accurate. The data for the third
material, HTF 500, were supplied by the manufacturer. The manufacturer could not give the
accuracy of the thermal property data. With the accuracy of the properties of HTF 500
uncertain, this too might suggest that the use of this material should be limited to the
calibration of a probe for low thermal conductivity applications.
An independent means of evaluating the accuracy of the probe for thermal property
measurement is to measure the properties of a material with well documented properties.
There were very few materials available for the evaluation. Ethylene glycol and dimethyl
phthalate were chosen. The thermal conductivity of dimethyl phthalate was taken from
literature (Marsh, 1987). The thermal diffusivity was calculated from the thermal
conductivity, the density and the specific heat. The last two properties were measured
(Appendix A and C). The thermal conductivity and specific heat for ethylene glycol were
taken from literature (CINDAS, 1988). The data covered temperatures from room
temperature to beyond 150 °C. Density data were found only up to 100 °C (Incropera, 1985).
Measurement of thermal diffusivity could thus only be evaluated at that temperature.
One sample of dimethyl phthalate and one sample of ethylene glycol were measured at 100,
130 and 150 °C. Convection in the samples was restricted with fiber glass pipe insulation in a
1.5% (w/w) concentration. Thermal properties were calculated and compared to the reference
data. Comparison was performed for results obtained with the three possible two-media
calibrations and the three-media calibration. The results are presented in Table 6.1 for
ethylene glycol and in Table 6.2 for dimethyl phthalate. Calibration of the probe with three
media used the interval 1-11 seconds for estimation of thermal conductivity at each
temperature, and for the estimation of thermal diffusivity the intervals 11-21 seconds, 6-16
seconds, 3-13 seconds at 100, 130 and 150 °C respectively. These intervals appeared to yield
optimal estimation errors for the calibration media. For the two-media calibration, the
interval 7-17 seconds was used. The choice was somewhat arbitrary. Calculation of the
thermal properties used the same interval as used to calibrate the probe.

Table 6.1 Thermal Properties of Ethylene Glycol at 100, 130 and 150 °C, measured with
probe #H3.
Thermal Conductivity
1
Temperature Literature WGH WG WH GH
[°C] [W/m°C] [W/m°C] [W/m°C] [W/m°C] [W/m°C]
100 0.2555 0.2640 0.2659 0.2509 0.2615
3
3.3 4.1 -1.8 2.
130 0.2560 0.2649 0.2646 0.2496 0.2625
3.5 3.4 -2.5 2.
150 0.2550 0.2639 0.2675 0.2472 0.2610
3.5 4.9 -3.1 2.
Thermal Diffusivity
Temperature Literature2 WGH WG WH GH
2 -8 2 -8 2 -8 2
[°C] [m /s] [10 m /s] [10 m /s] [10 m /s] [10-8m2/s]
100 8.67 9.04 8.89 9.14 8.90
4.3 2.5 5.4 2.
130 not available 8.95 8.76 8.77 8.76
150 not available 8.93 8.92 8.54 8.79
1
CINDAS (1988)
2
Calculated as α = k/ρ Cp
3
The values in italics give the difference from the literature value in %.
WGH Calibration with Water, Glycerol and HTF 500; estimation interval for thermal conductivity at
each temperature: 1-11 seconds; estimation interval for thermal diffusivity: 11-21 seconds at
100 °C, 6-16 seconds at 130 °C, and 3-13 seconds at 150 °C.
WG Calibration with Water and Glycerol; estimation interval 7-17 seconds.
WH Calibration with Water and HTF 500; id.
GH Calibration with Glycerol and HTF 500; id.

Thermal conductivity of ethylene glycol was most accurately estimated from the calibration
with glycerol and HTF 500. The difference from the literature data was 2.3% at 100 °C, 2.5%
at 130 °C, and 2.4% at 150 °C. Compared to the other calibrations, the estimated value also
changed the least if another interval was used for evaluation of the thermal properties.
Regardless of interval (in the range 3-13 to 19-29 seconds), the calculated value deviated by
no more than 0.3% from the value for the interval 7-17 seconds (the value shown in Table 6.1
. Since the thermal conductivity of ethylene glycol fell between those of glycerol and HTF
500, it was expected that the calibration with these two media would give the best estimation.
The other calibrations gave acceptable results. The largest deviation was 4.9% for the
water-glycerol calibration at 150 °C. In terms of amount of variation in estimated thermal
conductivity with regression interval, the water-glycerol calibration was second to the
glycerol-HTF 500 calibration, followed by the three media calibration and the
water-HTF 500 calibration.
The thermal diffusivity of ethylene glycol at 100 °C was more accurately estimated with the
water-glycerol calibration (accurate to 2.5%) than with the glycerol-HTF 500 calibration
(accurate to 2.7%). The latter however showed a little less variation with regression interval.
The maximum inaccuracy was 5.4%, for the water-HTF 500 calibration. This calibration also
had the strongest dependence on regression interval.

Table 6.2 Thermal Properties of Dimethyl Phthalate at 100, 130 and 150 °C, measured with
probe #H3.
Thermal Conductivity
1
Temperature Literature WGH WG WH GH
[°C] [W/m°C] [W/m°C] [W/m°C] [W/m°C] [W/m°C]
100 0.1373 0.1373 0.1506 0.1399 0.1390
3
0.0 9.7 1.9 1.2
130 0.1326 0.1338 0.1494 0.1365 0.1355
0.9 12.7 2.9 2.2
150 0.1293 0.1307 0.1477 0.1336 0.1326
1.1 14.2 3.3 2.6
Thermal Diffusivity
Temperature Literature2 WGH WG WH GH
2 -8 2 -8 2 -8 2
[°C] [m /s] [10 m /s] [10 m /s] [10 m /s] [10-8m2/s]
100 6.98 6.67 6.40 6.60 6.62
-4.4 -8.3 -5.4 -5.2
130 6.68 6.71 6.77 6.60 6.58
0.4 1.3 -1.2 -1.5
150 6.47 6.44 7.16 6.49 6.45
-0.5 10.7 0.3 -0.3
1
CINDAS (1988)
2
Calculated as α = k/ρ Cp
3
The values in italics give the difference from the literature value in %.
WGH Calibration with Water, Glycerol and HTF 500; estimation interval for thermal conductivity at
each temperature: 1-11 seconds; estimation interval for thermal diffusivity: 11-21 seconds at
100 °C, 6-16 seconds at 130 °C, and 3-13 seconds at 150 °C.
WG Calibration with Water and Glycerol; estimation interval 7-17 seconds.
WH Calibration with Water and HTF 500; id.
GH Calibration with Glycerol and HTF 500; id.

The thermal properties of dimethyl phthalate fell outside the range for which the probe was
calibrated. The calibration material with the most similar thermal conductivity and diffusivity
was HTF 500. It was therefore expected that a calibration that used that material would yield
a more accurate estimate of the thermal properties of dimethyl phthalate, than the calibration
with only water and glycerol. Indeed the water-glycerol calibration gave the most inaccurate
results. The three media calibration gave the most accurate estimation of thermal
conductivity and diffusivity of dimethyl phthalate. The largest difference between the
measured values for thermal conductivity and the literature values was 1.1%, at 150 °C. For
thermal diffusivity, the worst estimation was at 100 °C (-4.4%). At 130 and 150 °C, the
difference from the literature was 0.4 and -0.5% respectively. The water-HTF 500 and
glycerol-HTF 500 calibration gave good results. The measured values for thermal
conductivity and the literature values were at most 3.3% different. The inaccuracy in thermal
diffusivity was the largest at 100 °C (5.4% maximally) and within 1.6% at 130 and 150 °C.
The variation in measured values with regression interval was small for the two-media
calibrations that used HTF 500 as one of the calibration values. Calculation of thermal
conductivity from a different interval than 7-17 seconds (in the range 3-13 to 19-29 seconds)
gave values that were at most 0.6% (100 and 130 °C) or 1% (150 °C) different from the
tabulated values. The estimation of thermal diffusivity was more sensitive to the choice of
regression interval, in particular at 150 °C. The apparent thermal properties increased as later
intervals were used. This could have been caused by convection. Convection will increase the
apparent thermal properties of the medium. The amount of convection a the lower
temperatures seemed limited as the apparent thermal diffusivity increased only little with
later intervals.

Conclusions
Since only one sample of ethylene glycol and dimethyl phthalate was used to evaluate the
different calibrations, conclusions have to be drawn with caution. The three-media calibration
gave good results for both ethylene glycol and dimethyl phthalate. Those good results were
obtained by using specific intervals for the evaluation of the properties. The use of
measurement data outside these intervals could result in significant inaccuracies in the
measured values.
The accuracy obtained with the two-media calibration was comparable to that for the
three-media calibration, provided the properties of the test material fell between those of the
calibration materials or were at least close to those of one of them. The advantage over the
three-media calibration was the low sensitivity of the measured property values to the choice
of regression interval. The results for ethylene glycol suggests that a two-media calibration
over a small thermal conductivity range around the test material gives the best results.

CHAPTER 7
SUMMARY AND CONCLUSIONS
The thermistor based method was investigated for its suitability to measure the thermal
conductivity and diffusivity of moist food materials at high temperatures. The following
summarizes the conclusions of this research. The conclusions are used to outline a
recommended calibration procedure for the thermistor.
Calibration of the thermistor with three standards, water, glycerol and HTF 500, gave
good results provided the optimal interval of 1-11 seconds was used for estimation of
thermal conductivity. The optimal interval for estimation of thermal diffusivity depended
on the temperature at which the probe was calibrated. The use of an interval other than the
optimal interval would result in marked loss of accuracy. The magnitude of this loss
increased as the temperature increased from 100 to 150 °C.
Thermal contact between the thermistor and both potato and beef was found to be
adequate for the measurements at temperatures of 25, 50 and 100 °C. A comparison of the
bead thermistor probe to a reference method, the line heat source probe, showed good
agreement between both methods, except for potato at 100 °C. The latter was attributed to
problematic measurements with the line heat source probe. Visual examination of the
samples after the tests showed evidence of good contact between the thermistor and the
potato and beef samples.
Natural convection occurred in samples of thickened water during the sample interval of
30 seconds. Statistical analysis of measurements in samples of thickened water
demonstrated convection in samples with a viscosity of 5 cp and less, when measured with
a temperature step of 1.5 and 2.5 °C, and in samples with a viscosity of 25 cp and less,
Summary and Conclusions 123

when measured with a temperature step of 5 °C. A Rayleigh number was defined and its
value at the onset of convection was numerically evaluated. For the four samples that
experienced convection, the Rayleigh number at the onset fell in the range 43 to 84. For
samples of glycerol, measured with a different probe, similar values for the Rayleigh
number were found at the onset of convection. Although extrapolation to probes of
slightly different size and to different test media is debatable, it is felt that onset for those
conditions will be characterized by a Rayleigh number of similar values as found in this
research.
The size of the temperature step had an effect on the outcome of a thermal property
measurement. The use of a different size temperature step then used during calibration led
to a marked loss of accuracy.
The high temperature thermal property measurement of tomato puree gave better results
than that of tomato paste. Product alterations and noticeable inhomogeneity were noticed
in the tomato paste samples after the measurements. The tomato puree experienced less
product alterations and remained almost homogeneous. The measurement of thermal
properties of tomato paste might have been more successful if the paste could have been
heated faster. The measurement of thermal diffusivity was more sensitive to (suspected)
product variations and other sources of variation (like noise) than the measurement of
thermal conductivity.
For the nutritional supplement at high temperature, the thermistor measured a decreased
thermal conductivity with increasing solids content. The data from the thermistor thus
followed the expected trend with respect to the effect of the solids content. The replication
was good for the measurements at 95 °C at all solids levels and for those at 150 °C at
solids levels of 25% and below. The replication at 150 °C for the higher solids levels was
worse than those at lower solid levels and temperature. This was probably due to heat
induced changes in the product which are more pronounced at higher solids levels.

The construction of a durable seal around the glass stem of the thermistor remained
unresolved at the end of this research. The epoxy, that was used initially, could not
withstand the repeated exposure to the test media at high temperatures. The final design
that used a rubber gasket to provide the seal led to loss of thermistors as well.
The power response of the thermistor probe deviated from the model function in a way
that was related to the fluid thermal properties. The response in materials with a thermal
conductivity below 0.2 W/m°C had a different character than that in materials with a
thermal conductivity of 0.26 W/m°C and higher.
Estimation of the thermal properties of a test medium was best accomplished when the
probe was calibrated with two thermal standards whose properties were close to that of the
test medium. The estimation deteriorated when the thermal properties of the standards
were more divergent from that of the test medium, or when three standards were used. A
calibration with two media did not show the sensitivity to regression interval that the three
media calibration did.
Recommended procedure for thermal property measurement with the thermistor probe.
1. Calibrate the probe with two materials whose thermal conductivity and diffusivity are
accurately known, and bracket the range of intended application.
2. The interval used for evaluation of the probe properties has to be used for the
evaluation of the thermal properties of a test medium. When calibrating to measure
the properties of a low viscosity medium in which convection is likely to occur,
standardization on a early interval is the preferred approach. If adequacy of thermal
contact is of concern, as can be with a solid test material, a later and possibly longer
interval is preferred.
3. It is recommended to use the thermistor probe with the same size temperature step as
used during its calibration.

4. One measurement per sample should suffice in most cases. Multiple measurements
can be done to either average over random instrument error or check for the stability
of the test conditions.
5. Perform the linear regression with successive intervals. Evaluate the power response
of the thermistor from the regression results for the successive intervals. The trend in
the results should be similar to that for the calibration materials. A strong upward
trend in the intercept is indicative of convection. A waving plot may indicate
inadequate surface contact and/or inhomogeneity.

RECOMMENDATIONS
The following recommendations were formulated based on the research of this study.
An alternative to the glass probes of this research should be developed. This alternative
should meet the following criteria. It uses a small spherical resistance heater as heat
source and temperature sensor. A uniform thin coating will insulate electrically the heat
source from the test medium. The coating as well as the heater should be good heat
conductors. A slender stem, both thermally and electrically an insulator, carries the lead
wires of the heater. While having the dimensions of the glass probes of this research, the
improved probe should be made of a material with superior strength and should be impact
resistant. Also, it should be manufactured as one piece so that it can be mounted into a
system that can withstand repeated exposure to high temperature and pressure.
A rapid means of heating food materials should be implemented in order to minimize the
heat up time. For fluid products a flow system with a tubular heat exchanger and valves
can be used. Thermal property measurement can take place after closing a valve to block
the flow. For solid food materials, a sample chamber with dimensions only slightly larger
than required to satisfy the infinite boundary condition can be used.
Recommendations 127
REFERENCES
Anderson, G.T., Valvano, J.W. and R.R. Santos. 1988. A self-heated thermistor technique to
measure perfusion. Proceedings of the IEEE Engineering in Medicine and Biology
Tenth Annual Meeting, New Orleans, pp. 776-777.
Anderson, G.T. and J.W. Valvano. 1989. An interlobular artery and vein based model for
self-heated thermistor measurements of perfusion in canine kidneys. ASME Winter
Annual Meeting, San Francisco, HTD Vol. 126, BED Vol. 12, pp. 29-35.
Arkin, H., Holmes, K.R., Chen, M.M. and W.G. Bottje. 1986. Thermal pulse decay method
for simultaneous measurement of local thermal conductivity and blood perfusion: a
theoretical analysis. J. of Biomechanical Engineering, Trans ASME, Vol. 108,
pp 208-214.
Baghe-Khandan, M.S. and M.R. Okos. 1981a. Effect of Cooking on the Thermal
Conductivity of Whole and Ground Lean Beef. Journal of Food Science, Vol. 46,
pp 1302-1305.
Baghe-Khandan, M.S. and M.R. Okos. 1981b. Improved Line Heat Source Thermal
Conductivity Probe. Journal of Food Science, Vol. 46, pp. 1430-1432.
Balasubramaniam, T.A. and H.F. Bowman. 1974. Temperature field due to a time dependent
heat source of spherical geometry in a infinite medium. Journal of Heat Transfer,
Transactions ASME Vol. 99, Series 3, No.3, pp. 296-299.
Balasubramaniam, T.A. 1975. Thermal conductivity and thermal diffusivity of biomaterials:

a simultaneous measurement technique. Ph.D. dissertation, Northeastern University,
Boston MA.
Balasubramaniam, T.A. and Bowman, H.F. 1977. Thermal conductivity and thermal
diffusivity of biomaterials: a simultaneous measurement technique. Journal
Biomechanical Engineering, Transactions ASME Vol. 99, pp. 148-154.
Bowman, H.F. and T.A. Balasubramaniam. 1976. A new technique utilizing thermistor
probes for the measurement of thermal properties of biomaterials. Cryobiology Vol.
13, pp. 572-580.
Califano, A.N. and A. Calvelo, 1991. Thermal conductivity of potato between 50 and
100 °C. Journal of Food Science, Vol. 56 (2), pp. 586-588.
Casada, M.E. and L.R. Walton, 1989. New model for determining thermal diffusivity with
the thermal probe. Transactions of the ASAE, Vol. 32 (3), pp. 973-976.
References 128
Chandarana, D.I., Gavin, A III and F.W Wheaton. 1989. Simulation of parameters for
modeling aseptic processing of foods containing particulates. Food Technol. Vol. 43,
pp. 137-143.
Chandrasekhar, S. 1961. Hydrodynamic and hydromagnetic stability. Oxford University

Press, London.
Chang, S.Y., and R.T. Toledo, 1990. Simultaneous determination of thermal diffusivity and
heat transfer coefficient during sterilization of carrot dices in a packed bed. Journal of
Food Science, Vol. 55 (1), pp. 199-205.
Chato, J.C. 1968. A method for the measurement of the thermal properties of biological
materials. In "Thermal Problems in Biotechnology", ASME, NY, pp. 16-25.
Chen, M.M. and V. Rupinskas. 1977. A simple method for measuring and monitoring
thermal properties in tissues. Advances in Bioengineering, ASME paper No.
77-WA/HT-42.
Chen, M.M and K.R. Holmes. 1980. The thermal pulse-decay method for simultaneous
measurement of thermal conductivity and local blood perfusion rate of living tissues.
Advances in Bioengineering, ASME, pp. 113-115.
Choi, Y. and M.R. Okos. 1983. The thermal properties of tomato juice concentrates.
Transactions of the ASAE Vol. , pp. 305-311.
CINDAS, 1988. Properties of inorganic and organic fluids. CINDAS Data Series on
Material Properties Volume V-1, C.Y. Ho (ed.), Hemisphere publishing corporation,
New York, London.
de Ruyter, P.W. de and R. Brunet. 1973. Estimation of process conditions for continuous
sterilization of foods containing particulates. Food Technology, Vol. 27, pp. 44-49.
Dougherty, B.P. 1987. An automated probe for thermal conductivity measurements. MSc
thesis in Mechanical Engineering, Virginia Polytechnic Institute and State University.
Drouzas, A.E., Maroulis, Z.B., Karathanos, V.T. and G.D. Saravacos. 1991. Direct and
indirect determination of the effective thermal diffusivity of granular starch. Journal
of Food Engineering, Vol. 13, pp. 91-101.
Genceli, O.F., 1980. The onset of manifest convection from suddenly heated horizontal
cylinders. Wärme und Stoffübertragung, Vol. 13, pp. 163-169.
Gratzek, J.P. and R.T. Toledo, 1993. Solid food thermal conductivity determination at high
temperatures. Journal of Food Science, Vol. 58 (4), pp. 908-913.
References 129
Hayes, L.J. and J.W. Valvano. 1985. Steady-State Analysis of Self-Heated Thermistors
Using Finite Elements. Journal of Biomechanical Engineering, Vol. 107, pp. 77-80.
Hooper, F.C. and F.R. Lepper. 1950. Transient heat flow apparatus for the determination of
thermal conductivities. ASHVE Transactions, Vol. 56, pp. 309.
Incropera, F.P. and D.P. DeWitt. 1985. Fundamentals of Heat and Mass Transfer, 2nd ed.,
John Wiley & Sons, New York.
Jia. H. and G. Gogos, 1996. Transient laminar natural convection heat transfer from
isothermal spheres. Numerical Heat Transfer, Part A, Vol. 29, pp. 83-101.
Kravets, R.R. 1988. Determination of thermal conductivity of food materials using a bead
thermistor. Ph.D. thesis in Food Science and Technology, Virginia Polytechnic
Institute and State University.
Kustermann, M.K. Scherer, R and H.D. Kutzbach. 1981. Thermal conductivity and
diffusivity of shelled corn and grain. Journal of Food Process Engineering, Vol. 4,
pp. 137-153.
Lamberg, I. and B. Hallström, 1986. Thermal properties of potatoes and a computer

simulation model of a blanching process. Journal of Food Technology, Vol. 21, pp.
577-585.
Lan, Y. and V.E. Sweat. 1992. Simultaneous determination of thermal conductivity and
thermal diffusivity with probe. ASAE Paper No. 92-6597, ASAE, St. Joseph, MI
49085.
Larkin, J.W. 1990. Mathematical analysis of critical parameters in aseptic particulate

processing systems. Journal of Food Process Engineering, Vol. 13, pp. 155-167.
Lee, J.H. and R.K. Singh. 1988. Determination of lethality in a continuous sterilization
system containing particulates. ASAE Paper No. 88-6600, ASAE, St. Joseph, MI
49085.
Manson, J.E. and J.F. Cullen 1974. Thermal process simulation for aseptic processing of
foods containing discrete particulates. Journal of Food Science, Vol. 39, pp.
1084-1089.
Marsh, K.N. (ed.) 1987. Recommended reference materials for the realization of
physicochemical properties. Blackwell Scientific Publications, Oxford.
Mohsenin, N.N., 1980. Thermal properties of foods and agricultural materials. Gordon and
Breach Science Publishers, NY.
References 130
Murakami, E.G., Sastry, S.K. and C.A. Zuritz, 1984. A modified Fitch apparatus for
measuring thermal conductivity of small food particles. ASAE Paper No. 84-6508,
ASAE, St. Joseph, MI 49085.
Nieto de Castro, C.A., Li, S.F.Y., Nagashima, A., Trengove, R.D. and W.A. Wakeham. 1986.
Standard reference data for the thermal conductivity of liquids. Journal Phys. Chem.
Ref. Data, Vol. 15 (3), pp. 1073-1086.
Pearson, D., 1973. Laboratory Techniques in Food Analysis. Butterworth & Co. (Publishers)
Ltd., London.
Ott, L. 1988. An Introduction to statistical methods and data analysis. Third ed., PWS-Kent
Publishing Company, Boston. Table 8, Appendix A-23.
Palaniappan, S. and C.E. Sizer, 1997. Aseptic process validated for foods containing
particulates. Food Technology, Vol. 51 (8), pp. 60-68.
Parsons, J.R. Jr., and J.C. Mulligan, 1980. Onset of natural convection from a suddenly
heated horizontal cylinder. Journal of Heat Transfer, Vol. 102, pp. 636-639.
Patel, P.A., Valvano, J.W., Pearce, J.A., Prahl, S.A. and C.R. Denham. 1987. A self-heated
thermistor technique to measure effective thermal properties from tissue surface.
Journal of Biomechanical Engineering, Transactions ASME, Vol. 109, pp. 330-335.
Perez, M.G.R. and A. Calvelo, 1984. Modeling the thermal conductivity of cooked meat.
Journal of Food Science, Vol. 49, pp. 152-156.
Press, W. H., Flannery, B.P., Teukolsky, S.A. and W.T. Vetterling, 1989. Numerical Recipes
in Pascal; The Art of Scientific Computing, Cambridge University Press, Cambridge.
Rahman, Md. S. 1991. Evaluation of the precision of the modified Fitch method for thermal
conductivity measurement of foods. Journal of Food Engineering, Vol. 14, pp. 71-82.
Rao, M.A., Barnard, J. and J.F. Kenny, 1975. Thermal conductivity and diffusivity of
process variety squash and white potatoes. Transactions of the ASAE, Vol. 18, pp.
1188-1192.
Sengers, J.V., Watson, J.T.R., Basu, R.S. and B. Kamgar-Parsi, 1984. Journal Phys. Chem.
Ref. Data, Vol. 13, table 1.6, p 925.
Sweat, V.E. and C.G. Haugh. 1974. A thermal conductivity probe for small food samples.
Transactions of ASAE, Vol. 17, No 1, pp. 56-58.
Sweat, V.E. 1986. Thermal properties of foods. in "Engineering properties of foods", Rhao,
M. and S.S.H. Rizvi (ed.), Marcel Dekker, New York.
References 131
Thermometrics. 1987. Thermistor sensor handbook. Thermometrics Inc., NJ.
Touloukian, Y.S., Liley, P.E. and S.C. Saxena. 1970a. Thermal Conductivity: Nonmetallic
Liquids and Gases. The TPRC data series, Vol. 3, IFI/Plenum Press, New York.
Touloukian, Y.S. 1970b. Specific Heat: Nonmetallic Liquids and Gases. The TPRC data
series, Vol. 6 Suppl., IFI/Plenum Press, New York.
Touloukian, Y.S., Kirby, R.K., Taylor, R.E. and P.D. Desai, 1975. Thermal Expansion:
Metallic Elements and Alloys. The TPRC data series, Vol. 12., IFI/Plenum Press,
New York.
UCON Heat Transfer Fluid 500. 1989. Union Carbide Chemicals and Plastics Company Inc.,
Specialty Chemicals Division, Danbury, CT.
Valvano, J.W. 1981. The use of thermal diffusivity to quantify tissue perfusion. Ph.D. thesis
in Medical Engineering, Harvard University - MIT Division of Health Sciences and
Technology.
Valvano, J.W., Patel, P.A. and L.J Hayes. 1983. A Finite Element Analysis of Self-Heated
Noninvasive Thermistors. Advances in Bioengineering, ASME, pp. 149-150.
Valvano, J.W., Allen, J.T. and H.F. Bowman. 1984. The simultaneous measurement of
thermal conductivity and perfusion in small volumes of tissue. Journal of
Biomechanical Engineering, Transactions ASME, Vol. 106, pp. 192-197.
Valvano, J.W., Cochran, J.R. and K.R. Diller. 1985. Thermal conductivity and diffusivity of
biomaterials measured with self-heated thermistors. Int. Journal of
Thermophysics,Vol. 6, pp. 301-311.
Valvano, J.W., Badeau, A.F. and J.A. Pearce. 1987. Simultaneous measurement of intrinsic
and effective thermal conductivity. Heat Transfer in Bioengineering and Medicine,
ASME Winter Annual Meeting, Boston, HTD Vol. 95, BED Vol. 7, ed. Chato, Diller,
Diller, Roemer, pp. 31-36.
Valvano, J.W., Nho, S. and G.T. Anderson. 1990. A Weinbaum-Jiji model of steady state
heated thermistors in the canine kidney cortex. Advances in measuring and
computing temperatures in biomedicine: thermal tomography techniques, bioheat
models. ASME Winter Annual Meeting., Dallas, TX, HTD Vol. 147, ed. Roemer,
Valvano, Hayes, Anderson, pp. 51-57.
van Gelder, M.F and K.C. Diehl. 1996. A thermistor based method for measuring thermal
conductivity and thermal diffusivity of moist food materials at high temperatures.
Thermal Conductivity 23: Proceedings of the Twenty-Third International Thermal
Conductivity Conference, Technomic Publishing Co. Inc., Lancaster, PA.
References 132
Vargaftik, N.B., 1975. Tables on the Thermophysical Properties of Liquids and Gases: in
Normal and Dissociated States, Hemisphere Publishing Corp., Washington.
Vest, C.M. and M.L. Lawson, 1972. Onset of convection near a suddenly heated horizontal
wire. Int. Journal of Heat and Mass Transfer, Vol. 15, pp. 1281-1283.
Wang, J. and K. Hayakawa, 1992. Thermal conductivity of foods at high temperatures as

influenced by moisture content. Act. Rep. R & D. Assoc., Research and Development
Associates for Military Food and Packaging Systems, San Antonio, TX.
Wang, P., Kahawita, R. and D.L. Nguyen, 1991. Transient laminar natural convection from
horizontal cylinders. Int. Journal of Heat and Mass Transfer, Vol. 34 (6), pp.
1429-1442.
Wolters, R.J.H., 1992. Development of a computer controlled method for measuring thermal
conductivity in small food samples. (unpublished), Virginia Polytechnic Institute and
State University .
Woodbury, K.A. 1984. An experimental and analytical investigation of liquid moisture

distribution in roof insulating systems. Ph.D. thesis in Mechanical Engineering,
Virginia Polytechnic Institute and State University.
Yamada, T. , 1970. The thermal properties of potato. (in Japanese) J. Agr. Chem. Soc. of
Japan 44:587, (Abstract in Food Science and Technology Abstracts, 1973, 5:Abstract
No. 12j2011).
Zuritz, C.A., Sastry, S.K., McCoy, S.C., Murakami, E.G. and J.L. Blaisdell. 1989. A
modified Fitch device for measuring the thermal conductivity of small food particles.
Transactions ASAE Vol. 32(2), pp. 711-718.
References 133
APPENDIX A
DENSITY OF GLYCEROL AND DIMETHYL PHTHALATE
Appendix A 134
A DENSITY OF GLYCEROL AND DIMETHYL PHTHALATE.
A.1 INTRODUCTION
In the initial stage of the research, dimethyl phthalate was considered as a calibration
material. Its thermal conductivity was well documented as it was included in a list of
recommended reference materials for thermal conductivity measurements (Marsh, 1987). Its
thermal diffusivity however was not documented. It could not be calculated from published
values for specific heat and density since those weren’t found for the temperature range of
interest (100 - 150 °C). A project was undertaken to measure the density of dimethyl
phthalate at temperatures up to 150 °C. A second objective was to obtain high temperature
density data for glycerol in order to calculate its thermal diffusivity. Thermal diffusivity of
glycerol was not available from literature at high temperature. Density data were limited and
old.
The density of fluids can be found by several methods. A simple approach is to weigh a
volume, measured with a graduated cylinder. Other methods use calibrated beads of known
density or an hydrometer. These methods however are limited to a narrow temperature range.
Graduated cylinders are calibrated for a certain temperature. Thermal expansion will
negatively impact their accuracy at the high temperatures intended in this study. The same
remarks could be made with regard to calibrated beads and the hydrometer. This research
looked to the specific gravity balance as a possible method.
A.2 METHOD
A specific gravity balance is generally used to measure the specific gravity, or the density of
a sample of solid material. In order to determine the density of a solid sample with the
specific gravity balance, the sample is first weighed suspended in air. Subsequently, the
sample is submerged in a liquid of known density and its weight is recorded again. The
difference between the two values (suspended in air versus suspended in the liquid) yields the
Appendix A 135
buoyancy of the liquid (the buoyancy of air is assumed negligible). The volume of the sample
can be calculated from the buoyancy and the known density of the liquid. Finally, the density
of the sample is found from its volume and dry weight.
For this research, the method was adapted. The unknown was the density of the fluid. An
object needed to be manufactured with an accurately known volume. The material for this
object had to be temperature and chemical resistant. Metal seemed a logical choice. Stainless
steel rod of known specifications was available. The object was manufactured from a
stainless steel (SS 304) rod (9.6 mm in diameter). The length was 51.5 mm. The bottom of
the object was rounded to prevent air bubbles from being trapped underneath. The top was
tapered and a hole was drilled to hang the object from a hook. The hole was machined such
that it would not retain any air. The temperature dependence of the volume of the object was
found by applying equations given by Touloukian (1975). The expansion of type SS 304
stainless steel could be described by the equation:
L1 − L0
= −0.00358 + 9.472 • 10 −6 T + 1.031 • 10 −8 T 2 − 2.978 • 10 −12 T 3 (A.1)
L0
where: L1 = Length at temperature T1, [m]

L0 = Length at T0, T0 = 20 °C, [m]
T = Temperature [K]
The equation gives the expansion as a fraction of the length at 20 °C. The result of equation
A.1 is used to find the mean coefficient of linear expansion, defined as:
L1 − L0
. α= (A.2)
L 0 (T 1 − T 0 )
where: L0, L1 = Length, at Temperature T0 and T1 respectively [m]

T0, T1 = Temperature [°C]
Appendix A 136
The volumetric thermal expansion coefficient, β, is approximately three times the value of α.
The value of α is referenced to 20 °C. The volume of the object at other temperatures can be
calculated with the expression:
V(T) = V 20 (1 + 3α∆T) (A.3)
where: V(T) = Volume of object at temperature T, [cm3]

V20 = Volume of object at 20 °C, [cm3]
α = mean coefficient of linear thermal expansion, [°C-1]
∆T = T - 20. [°C]
The volume of the object was determined by triplicate measurement in water. The density of
water taken from Incropera (1985).
A.3 EQUIPMENT
All weight measurements were taken with a digital analytical balance (Sartorius Model 1712
MP8 Silver edition, Brinkmann Instruments Inc., Div. of Sybron, Westbury, NY), with a
resolution of 0.1 mg. The balance had a hook underneath the weighing platform, allowing it
to be used as a specific gravity balance. The balance was placed on a stand, 29 cm above the
workbench. An opening in the stand gave access to the hook. A piece of thin wire with ends
shaped as hooks was hung from the hook under the balance. A similar shaped second piece of
wire was hung from the first one. The object was hung on this second piece of wire. The
volume of the submerged part of the wire was determined from its diameter and length.
The test fluid was contained in the sample cup borrowed from a viscometer (Brookfield LV,
Stoughton, MA). The cup had a diameter of 20.9 mm and a depth of 60.2 mm. The cup came
with a flow jacket for temperature control. The flow jacket was connected, through insulated
high temperature silicon tubing, to a circulating temperature bath (Exacal EX-251 HT,
NESLAB Instruments, Inc., Newington, NH). The bath contained a heat transfer fluid for
Appendix A 137
high temperature applications (HTF 500, Union Carbide, Danbury, CT). The viscosity of this
fluid at low temperatures was too high for the circulating bath to properly function. For this
reason, measurements were started at 35 °C. The sample cup was covered with a plastic top
(accessory to the small sample adapter) to minimize evaporation and heat loss from the
sample. The temperature of the test fluid was monitored with a thin thermocouple (24 ga.
copper-constantan). The thermocouple was connected to a cold junction compensator
(OMEGA CJ, Omega Engineering, Inc., Stamford, CT). The output of the cold junction
compensator was read with a voltmeter (Keithley 197 Autoranging Microvolt DMM,
Keithley Instruments Inc., Cleveland, OH). The thermocouple was calibrated against a
thermal standard with 0.001 °C resolution (Guildline 9540 Digital Platinum Resistance
Thermometer, Smith Falls, Ontario), traceable to the National Institute of Standards and
Technology (NIST).
The accuracy of the method was assessed with water. Distilled water was boiled and placed
under a vacuum of 600 mg Hg to expel air. The density was determined at 7.5 °C intervals
over the range 40 to 85 °C. Higher temperatures were not feasible due to excessive
evaporation. The measurement was replicated once.
Glycerol and dimethyl phthalate were measured over the temperature range 40 to 150 °C at
10 °C increments. Each material was measured in duplicate. Prior to filling the sample cup,
the glycerol was heated to approximately 85 °C to lower its viscosity, and so decrease the
likelihood that air bubbles would be incorporated into the fluid upon filling the cup, or when
the object was lowered into the fluid.
A.4 RESULTS
Because the circulating bath didn't allow measurements at 20 °C, the volume of the object
was determined at a higher temperature. The volume of the object at 50 °C was found from
triplicate measurements to be 3.3266 cm3. By using equations A.1-A.3, the volume at 20 °C
was found, 3.218 cm3. The volume of the submerged part of the wire was 0.0035 cm3.
Appendix A 138
The results of the density measurements in water are given in Table A.1. The replication was
excellent. The combined measurement data were used in quadratic linear regression to find a
prediction equation. The equation fitted the data with a mean absolute deviation of 0.04%
and a maximum absolute deviation of 0.14%. The equation was used to generate density
values for comparison to tabulated literature data. Compared to literature values for water
given by Incropera and DeWitt (1985) over the range 310 to 360 K (37 - 87 °C) the equation
underestimated the density of water by an average of 0.06% and a maximum of 0.08%. Based
on these results, the method was deemed of sufficient accuracy.
Table A.1 Density of water determined with a specific gravity balance.
First Sample Second Sample

Temperature Density Temperature Density
3
[°C] [kg/m ] [°C] [kg/m3]
34.5 993.6
39.4 991.8 39.4 991.4
46.9 988.3 46.8 988.2
54.2 985.5 54.1 984.8
61.6 983.0 61.7 981.3
69.1 977.7 68.9 977.3
76.5 973.6 76.2 973.4
83.8 969.5 83.4 970.4
Density of Glycerol
The data of the two measurements are given in Table A.2 and Figure A.1. The combined
measurement data were used in quadratic linear regression to find a prediction equation.
ρ = 1273.18 − 0.60577 T − 5.01 • 10 −4 T 2 (A.4)
where: ρ = Density [kg/m3]

T = Temperature [°C]
Appendix A 139
Equation A.4 fitted the data with a mean absolute deviation of 0.05% and a maximum
absolute deviation of 0.1%. The equation was compared to literature data given by Vargaftik
(1975) over the range 20 to 140 °C. The difference with these data is less than 0.25% up to
100 °C, -0.44% at 120 °C, and -1% at 140 °C. The measurement data and the reference
values from Vargaftik are given in Figure A.1. Equation A.4 was used the estimate the
coefficient of thermal expansion, b: b(T) = -1/r(T) (dr/dT)T. The estimated values for the
coefficient of thermal expansion deviated less that 5% from those given by Incropera (1985)
over the range 273 - 320 K.
Table A.2 Density of Glycerol with a specific gravity balance

3
[°C] [kg/m ] [°C] [kg/m3]
39.4 1247.5 49.5 1242.0
49.1 1242.6 59.2 1235.7
59.3 1235.6 69.0 1229.4
68.9 1229.6 78.9 1222.8
78.4 1222.7 88.6 1215.9
88.6 1215.0 99.3 1208.5
98.1 1207.9 109.1 1201.8
107.9 1201.2 119.2 1194.6
118.1 1193.8 129.2 1187.1
127.7 1186.7 139.1 1180.1
137.7 1179.1 149.1 1172.4
Appendix A 140
1280
First Sample
1260 Second Sample
Vargaftik (1975)
1240
Density [kg/m3] 1220
1200
1180
1160
1140
1120
0 20 40 60 80 100 120 140 160 180
Temperature [°C]
Figure A.1 Density of Glycerol with a specific gravity balance
density of dimethyl phthalate
The data of the two measurements are given in Table A.3 and Figure A.2. The combined
measurement data were used in linear regression to find a prediction equation.
ρ = 1210.00 − 0.92027 T (A.5)
where: ρ = Density [kg/m3]

The data fit the equation with an average absolute deviation of 0.02% and a maximum
deviation of 0.06% (0.22 and 0.62 [kg/m3] respectively). This excellent fit is testament of the
excellent replication.
Appendix A 141
Table A.3 Density of dimethyl phthalate with a specific gravity balance.

3
[°C] [kg/m ] [°C] [kg/m3]
39.2 1173.7 39.1 1174.1
49.2 1164.5 48.9 1164.4
59.0 1155.5 59.0 1155.6
69.3 1146.1 69.0 1146.6
79.6 1136.7 79.7 1137.0
90.1 1127.3 89.2 1128.5
99.9 1118.3 99.7 1118.5
110.1 1108.7 110.1 1109.1
119.7 1099.8 120.0 1099.9
129.6 1090.8 129.7 1090.7
139.7 1081.5 139.5 1081.4
149.6 1072.1 149.3 1072.0
1180
First Sample
1160
Second Sample
1140
Density [kg/m3]
1120
1100
1080
1060
30 40 50 60 70 80 90 100 110 120 130 140 150 160
Temperature [°C]
Figure A.2 Density of dimethyl phthalate measured with a specific gravity balance.
Appendix A 142
APPENDIX B
MEASUREMENT DATA FOR THE CALIBRATION OF THE

LOW TEMPERATURE PROBE
Appendix B 143
Table B.1 Regression Results for Water at 25 °C with Probe
#L4, from regression over the interval 3-9
seconds.
File R∆T/Vss2 Intercept Slope S/I
L4W25A.D1 288.3 17.7947 9.6738 0.5436
L4W25B.D1 289.5 17.7182 9.6828 0.5465
L4W25C.B1 289.1 17.7003 9.5990 0.5423
Sample Average 289.0 0.5441
L4W25X.D2 290.8 17.6358 9.6215 0.5456

L4W25Y.B2 289.4 17.7393 9.6562 0.5443
L4W25Z.D2 289.1 17.7505 9.6906 0.5459
L4W25A.D3 288.5 17.7340 9.6811 0.5459

L4W25B.D3 288.4 17.7017 9.6594 0.5457
L4W25C.D3 288.7 17.7559 9.7675 0.5501
Average at 25 °C 289.1 0.5456
Appendix B 144
Table B.2 Regression Results for Glycerol at 25 °C with
Probe #L4, from regression over the interval 3-9
seconds.
L4G25A.D1 439.2 11.7065 10.7951 0.9222
L4G25B.D1 438.4 11.6589 10.7683 0.9236
L4G25C.D1 438.5 11.7142 10.8208 0.9237
L4G25A.D2 439.8 11.6407 10.7545 0.9239

L4G25B.D2 439.9 11.6382 10.7290 0.9219
L4G25C.D2 440.3 11.6351 10.7597 0.9248
L4G25A.D3 438.6 11.6788 10.7604 0.9214

L4G25C.D3 438.3 11.6975 10.7579 0.9197
L4G25D.D3 438.4 11.6895 10.7187 0.9170
Average at 25 °C 439.0 0.9220
Appendix B 145
Table B.3 Regression Results for Water at 50 °C with Probe
#L4, from regression over the interval 3-9
seconds.
L4W50A.B1 279.7 7.3673 3.8184 0.5183
L4W50B.B1 279.1 7.3768 3.7977 0.5148
L4W50C.D1 279.5 7.3913 3.8188 0.5167
L4W50A.D2 279.2 7.4098 3.8260 0.5164

L4W50B.B2 279.7 7.3960 3.8229 0.5169
L4W50C.D2 279.8 7.4102 3.8300 0.5169
L4W50A.B3 279.2 7.3276 3.8115 0.5202

L4W50B.D3 280.0 7.3330 3.8121 0.5199
L4W50C.D3 279.1 7.3248 3.8009 0.5189
Average at 50 °C 279.5 0.5176
Appendix B 146
Table B.4 Regression Results for Glycerol at 50 °C with
Probe #L4, from regression over the interval 3-9
seconds.
L4G50A.B1 436.9 4.6861 4.4293 0.9452
L4G50B.B1 436.1 4.6718 4.4024 0.9423
L4G50C.B1 436.6 4.6624 4.3955 0.9428
L4G50A.D2 435.7 4.7915 4.5261 0.9446

L4G50B.D2 435.4 4.7841 4.5153 0.9438
L4G50C.D2 435.6 4.7742 4.5119 0.9451
L4G50A.D3 436.2 4.7546 4.4600 0.9381

L4G50B.D3 437.2 4.7477 4.4789 0.9434
L4G50C.D3 437.3 4.7457 4.4774 0.9435
Average at 50 °C 436.3 0.9432
Appendix B 147
APPENDIX C
SPECIFIC HEAT OF DIMETHYL PHTHALATE
Appendix C 148
C SPECIFIC HEAT OF DIMETHYL PHTHALATE
C.1 INTRODUCTION
In order to calculate the thermal diffusivity of dimethyl phthalate, values for thermal
conductivity, density, and specific heat had to be found. Thermal conductivity was found
from literature (Marsh, 1987). The density was determined with a specific gravity balance
(Appendix A). The determination of specific heat in the temperature range 100 to 150 °C was
the objective of this study.
C.2 METHODS AND MATERIALS
The specific heat of dimethyl phthalate was measured with a differential scanning calorimeter
(Metler Thermal Analysis Systems TA 4000, consisting of a TC11 TA processor and a
DSC 30 measuring cell). The DSC was fully calibrated prior to the experiments.
The DSC was run from 50 °C to 160 °C with a scan rate of 10 °C per minute. Four samples
were tested. The samples were weighed on a digital analytical balance (Sartorius model 1712
MP8 silver edition, Brinkmann Instruments Inc., Div. of Sybron, Westbury, NY) with a
resolution of 0.01 mg. Sample weight was approximately 30 mg.
C.3 RESULTS
The thermal analysis processor evaluated the specific heat at 5 °C increments over the range
80 to 150 °C. The data of the four samples are given in Table C.1. Replication was good as
demonstrated by the low values of the standard deviation. The data were to a linear equation
with a R2 value of 0.944. The data fitted the equation with an average absolute deviation of
0.52% and a maximum deviation of 1.33%.
C p = 1.5495 + 0.002091 T (C.1)
Appendix C 149
where: Cp = Specific Heat [kJ/kg°C]
A quadratic equation improved the fit marginally, but the parameter of the quadratic term had
a standard error that was of the same magnitude as the parameter itself.
Table C.1 Specific heat of Dimethyl Phthalate, kJ/kg°C
Temperature Specific Heat

[°C] Sample #1 Sample #2 Sample #3 Sample #4 Average StdDev
80 1.72 1.72 1.73 1.74 1.73 0.0096
85 1.72 1.71 1.73 1.74 1.73 0.0129
90 1.73 1.73 1.74 1.75 1.74 0.0096
95 1.74 1.73 1.76 1.76 1.75 0.0150
100 1.75 1.75 1.76 1.78 1.76 0.0141
105 1.75 1.75 1.76 1.78 1.76 0.0141
110 1.77 1.77 1.78 1.79 1.78 0.0096
115 1.78 1.78 1.79 1.80 1.79 0.0096
120 1.79 1.80 1.81 1.82 1.81 0.0129
125 1.79 1.80 1.81 1.81 1.80 0.0096
130 1.81 1.81 1.83 1.83 1.82 0.0115
135 1.82 1.83 1.84 1.84 1.83 0.0096
140 1.84 1.83 1.85 1.86 1.85 0.0129
145 1.85 1.85 1.86 1.87 1.86 0.0096
150 1.86 1.86 1.87 1.87 1.87 0.0058
Appendix C 150
APPENDIX D
THERMAL CONDUCTIVITY OF POTATO AND BEEF:

MEASUREMENT DATA
Appendix D 151
Table D.1 Average Thermal Conductivity of Potato at 25, 50 and 100 °C, measured with the
Line Heat Source Probe and the Bead Thermistor Probe
Measurement Potato Variety Thermal Conductivity

Temperature Line Heat Source Probe Bead Thermistor Probe
Sample Temperature Sample Temperature
Averagea Average Averagea Average
(°C) (W/m°C) (W/m°C) (W/m°C) (W/m°C)
25 1 Idaho 0.5846 0.5500
2 Idaho 0.5986 0.5598
3 California White 0.5873 0.5580
4 Idaho 0.5503
5 Idaho 0.5961 0.5452
6 Idaho 0.6004 0.5467
7 Idaho 0.6070 0.5957 0.5517
50 1 California White 0.5942 0.5918
5 Idaho 0.6355 0.5773
6 Idaho 0.6357 0.5878
7 Idaho 0.6325 0.6117 0.5710 0.5841
100 1 Idaho 0.7790
2 Idaho 0.7068 0.6135
3 Idaho 0.6970 0.5986
4 Idaho 0.7122 0.5892
5 Idaho 0.6968 0.6133
6 Idaho 0.7028 0.6068
b
7 Idaho 0.7296 0.7075 0.5877 0.6015
a
Average is calculated over 3 repeats.
b
Average excludes sample #1.
Appendix D 152
Table D.2 Thermal Conductivity of Potato at 25 °C, measured with the Line Heat Source Probe
and the Bead Thermistor Probe; results of individual measurements
Potato Variety Thermal Conductivity

Line Heat Source Probe Bead Thermistor Probe
Mean Value Mean Value
for Potato for Potato
[W/m°C] [W/m°C] [W/m°C] [W/m°C]
Idaho 0.5872 0.5495
0.5809 0.5505
0.5857 0.5846 0.5500 0.5500
Idaho 0.5985 0.5596
0.5981 0.5617
0.5993 0.5986 0.5582 0.5598
California White 0.5900 0.5588
0.5836 0.5570
0.5883 0.5873 0.5581 0.5580
Idaho 0.5961 0.5466
0.5951 0.5515
0.5970 0.5961 0.5529 0.5503
Idaho 0.6038 0.5425
0.6030 0.5411
0.5944 0.6004 0.5493
0.5479 0.5452
Idaho 0.6042 0.5426
0.6075 0.5419
0.6092 0.6070 0.5510
0.5510 0.5466
Average at 25 °C. 0.5957 0.5517
Standard Deviation of Means 0.0084 0.0060
Appendix D 153
Table D.3 Thermal Conductivity of Potato at 50 °C, measured with the Line Heat Source Probe
and the Bead Thermistor Probe; results of individual measurements

[W/m°C] [W/m°C] [W/m°C] [W/m°C]
0.5925 0.5894
0.5919 0.5942 0.5948
0.5911 0.5918
0.5947 0.5885
0.5952 0.5969 0.5894
0.5900 0.5891
0.5958 0.5837
0.5847 0.5932 0.5850
0.5866 0.5850
0.5967 0.5865
0.5928 0.5942 0.5871
0.5860 0.5870
Idaho 0.6388 0.5793
0.6360 0.5743
0.6316 0.6355 0.5773
0.5785 0.5774
Idaho 0.6371 0.5914
0.6342 0.5902
0.6359 0.6357 0.5853
0.5843 0.5878
0.6357 0.5634
0.6300 0.5610
0.6318 0.6325 0.5781
0.5815 0.5710
Average at 50 °C. 0.6117 0.5842
Appendix D 154
Table D.4 Thermal Conductivity of Potato at 100 °C, measured with the Line Heat Source
Probe and the Bead Thermistor Probe; results of individual measurements

[W/m°C] [W/m°C] [W/m°C] [W/m°C]
Idaho 0.7088 0.6134
0.6917 0.6138
0.7199 0.7068 0.6133 0.6135
Idaho 0.6938 0.5849
0.6949 0.6089
0.7022 0.6970 0.6020 0.5986
Idaho 0.7055 0.5930
0.7223 0.5911
0.7088 0.7122 0.5834 0.5892
Idaho 0.6958 0.6000
0.7018 0.6220
0.6927 0.6968 0.6180 0.6133
Idaho 0.7023 0.6095
0.6986 0.6040
0.7076 0.7028 0.6069 0.6068
Idaho 0.7226 0.5820
0.7371 0.5868
0.7291 0.7296 0.5943 0.5877
Average at 100 °C. 0.7075 0.6015
Appendix D 155
Table D.5 Average Thermal Conductivity of Lean Beef at 25, 50 and 100 °C, measured with
the Line Heat Source Probe and the Bead Thermistor Probe

Measurement Sample Thermal Conductivity Sample Thermal Conductivity
Temperature
Sample Temperature Sample Temperature
Average1 Average Average1 Average
(°C) (W/m°C) (W/m°C) (W/m°C) (W/m°C)
25 1 0.5465 1 0.5025
2 0.5458 2 0.4007
3 0.5331 3 0.4590
4 0.5435 4 0.5083
5 0.5085 5 0.5047
6 0.5184 0.5326 6 0.4941 0.4782
50 1 0.5321 1 0.5169
2 0.5504 2 0.5195
3 0.5532 3 0.5116
4 0.5488 4 0.5031
5 0.5562 5 0.5057
6 0.4600 0.5335 6 0.4938 0.5084
100 1 0.4669 1 0.4829
2 0.4145 2 0.4669
3 0.5234 3 0.4671
4 0.4726 4 0.4888
5 0.4977 0.4750 5 0.4117
6 6 0.4764 0.4656
1
Average is calculated over 3 repeats.
Appendix D 156
Table D.6 Thermal Conductivity of Beef (Whole Round) at 25 °C, measured with the Line
Heat Source Probe and the Bead Thermistor Probe; results of individual
measurements
Sample Thermal Conductivity

Number Line Heat Source Probe Bead Thermistor Probe
for Sample for Sample
[W/m°C] [W/m°C] [W/m°C] [W/m°C]
1 0.5482 0.5040
0.5446 0.5010
0.5467 0.5465 0.5024 0.5025
2 0.5468 0.4019
0.5461 0.3989
0.5444 0.5458 0.4015 0.4008
3 0.5315 0.4532
0.5351 0.4622
0.5327 0.5331 0.4616 0.4590
4 0.5403 0.5083
0.5468 0.5079
0.5433 0.5435 0.5087 0.5083
5 0.5138 0.5045
0.5046 0.5034
0.5070 0.5085 0.5063 0.5047
6 0.5174 0.4949
0.5173 0.4928
0.5205 0.5184 0.4946 0.4941
Average at 25 °C. 0.5326 0.47821
1
After the test a lump of fat was discovered near the probe. Excluding
sample number 2 from the analysis changes the average and standard
deviation to 0.4937and 0.0201 respectively.
Appendix D 157
measurements

[W/m°C] [W/m°C] [W/m°C] [W/m°C]
1 0.5309 0.5195
0.5308 0.5174
0.5347 0.5321 0.5138 0.5169
2 0.5509 0.5212
0.5521 0.5185
0.5482 0.5504 0.5188 0.5195
3 0.5566 0.5119
0.5510 0.5121
0.5521 0.5532 0.5106 0.5115
4 0.5528 0.5026
0.5479 0.5053
0.5456 0.5488 0.5013 0.5031
5 0.5543 0.5053
0.5625 0.5052
0.5519 0.5562 0.5065 0.5057
6 0.4624 0.4930
0.4586 0.4948
0.4589 0.4600 0.4936 0.4938
Average at 50 °C. 0.5335 0.5084
Appendix D 158
measurements

[W/m°C] [W/m°C] [W/m°C] [W/m°C]
1 0.4695 0.4885
0.4675 0.4778
0.4637 0.4669 0.4824 0.4829
2 0.4120 0.4670
0.4131 0.4696
0.4183 0.4145 0.4641 0.4669
3 0.5219 0.4655
0.5180 0.4745
0.5304 0.5234 0.4613 0.4671
4 0.4771 0.4856
0.4760 0.4875
0.4647 0.4726 0.4933 0.4888
5 0.4936 0.4147
0.5044 0.4095
0.4952 0.4977 0.4109 0.4117
6 0.4756
0.4738
0.4798 0.4764
Average at 100 °C. 0.4750 0.4656
Appendix D 159
VITAE
Maarten F. van Gelder was born on February, 1960 in The Hague, The Netherlands. He
studied at Wageningen Agricultural University and graduated in 1989 with a MSc. in
Agricultural Engineering. A practical internship as part of his Masters program brought him
to Virginia Polytechnic Institute and State University. For five months he worked for Dr. K.
C. Diehl, Jr. on a project involving the physical properties of oysters. After completion of his
Masters degree in Wageningen, he decided to come back to Virginia Tech to pursue a
doctoral degree in the field of Food Engineering under the guidance of Dr. K.C. Diehl, Jr.
‘ 160

Thermistor PDF

Uploaded by

Copyright:

Available Formats

Thermistor PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thermistor PDF

Uploaded by

Copyright:

Available Formats

A Thermistor Based Method for Measurement of Thermal Conductivity and

Thermal Diffusivity of Moist Food Materials at High Temperatures

Dissertation submitted to the Faculty of the

Kenneth C. Diehl, Jr., Chairman

Table of Contents vii

List of Figures viii

Table 4.1 Thermophysical Properties of Water 63

2.1 STEADY STATE METHODS FOR MEASUREMENT OF THERMAL

where: Q = heat flow [W]

2.1.2 Guarded Hot Plate

2.1.3 Radial Heat Flow Method

2.2 TRANSIENT STATE METHODS

2.2.2 Fitch Method

where: A = heat transfer area, [m]

2.2.3 Transient Hot Wire

where: k = thermal conductivity of sample [W/m°C]

Although often reported as an absolute method, it is recommended to calibrate the probe.

2.2.5 Thermistor Based Method

A thermistor is a temperature dependent resistor. Those used for thermal property

1 = a + a ln (R) + a [ln (R)] 3 (2.4)

where: T = temperature, [K]

The constants a0, a1 and a2 are obtained in a calibration procedure.

and the heat dissipation by the medium is given by:

where: a = radius of thermistor bead [m]

1 = 4πa ∆T − 0.2 (2.7)

Woodbury (1984) used self-heated thermistors to measure the thermal conductivity of

Alternate Modes of Operation of Thermistor Probes

where: g = gravitational acceleration [m/s2]

PRINCIPLES OF THE THERMISTOR BASED METHOD

3.1 HEAT TRANSFER MODEL

Principles of Thermistor Based Method 27

Principles of Thermistor Based Method 28

T(r, t) = T i 0 ≤ r ≤ a, t=0 (3.7)

Principles of Thermistor Based Method 29

 (a − r)f(t)  βa 3 f(t) Γa 4 f(t)

3.3.1 Measurement of Thermal Conductivity

Principles of Thermistor Based Method 30

The constant internal heat generation in the bead is:

Substituting this expression into equation (3.12) gives, after rearranging:

3.3.2 Measurement of Thermal Diffusivity

Principles of Thermistor Based Method 31

3.4 CALIBRATION: ESTIMATION OF PROBE PROPERTIES

Principles of Thermistor Based Method 32

3.5 PRACTICAL CONSIDERATIONS

3.5.1 Sample Dimension

Principles of Thermistor Based Method 33

Principles of Thermistor Based Method 34

3.5.2 Temperature Step

Principles of Thermistor Based Method 35

Principles of Thermistor Based Method 36

a = radius of thermistor bead [m]

Principles of Thermistor Based Method 37

METHODS AND MATERIALS

Methods and Materials 38

1 = a + a ln (R) + a [ln (R)] 3 (2.4)

4.2 THERMISTOR ASSEMBLY

Methods and Materials 39

Figure 4.1 Probe assembly for low temperature measurements.

4.2.2 High Temperature Probes