Krishpersad Manohar, I David W. Yarbrough,2 and James R.

Booth3

Measurement of Apparent Thermal Conductivity by the Thermal Probe Method

Authorized Reprint from Journal of Testing and Evaluation, Sept. 2000 @Copyright 2000

American Society for Testing and Materials, 100 Barr Harbor Drive, West Conshohocken, PA 19428-2959

REFERENCE: Manohar, K., Yarbrough, D. W., and Booth, J. R., "Measurement of Apparent Thermal Conductivity by the Thermal Probe Method," Journal of Testing and Evaluation, HEV A, Vol. 28, No.5, September 2000, pp. 345-351. ABSTRACT: Three thermal probes were constructed in accordance with ASTM D 5334 and calibrated using heat-flow metre data. The temperature-time response of the thermal probes for determining apparent thermal conductivity Aunder transient state conditions was logged at 1 s intervals. The instrumentation used reduced the determinate error associated with voltage and current measurements to a negligibly small value that made the uncertainty in A dependent on the uncertainty of the slope dTldlnt. A test method was run in 1000 s, in which a criterion of 2.5% spread among three consecutive slope values was used to determine the extent of the linear segment of the T-In t curve. The probes demonstrated repeatability within :!::3.5% but had definite individual bias indicating a need for individual calibration. Using individual probe calibration factors, the experimentally determined ASfor all-purpose sand, sifted sand, and soil were determined to be 0.520, 0.445, and 2.11 W1m'K, respectively. KEYWORDS: method thermal probe, thermal conductivity, line source

ture data from the probe are analyzed to obtain the mean apparent thermal conductivity iI.for the material over the temperature range introduced in the experiment. Apparent thermal conductivities are obtained from the data using Eq 1, which is an approximation of the solution for the boundary value problem describing the physical situation [4]. (Q/47T)
iI. = (dT/dlnt)

(1)

Equation 1 indicates an attractive method for thermal conductivity measurement with the advantages of low cost and short test time. Hence, this method would be ideal for conducting many tests per day and testing material with moisture. However, in the past the reliability of the results was poor [1-4] and hence the method was seldom used. This study concentrated on improving the reliability of the test method by reducing the associated equipment error and monitoring the time-temperature response of the thermal probe at I s intervals. This paper focuses on probe calibration, data acquisition, and apparent thermal conductivity data obtained for four materials. Thermal Probe The thermal probes in this work were constructed from seamless thin-walled stainless steel tubes, 100 mm long and 3 mm outer diameter. The main probe components are a manganin heater wire and a thermocouple as shown in Fig. 1. The probe assembly is similar to that described in ASTM D 5334 [5]. Insulated manganin heater wire, 0.358 mm in diameter, was looped and inserted into the full length of the tube to form the heating element; a 30 gage T-type thermocouple (wires supplied by Cole-Parmer) was inserted midway along the length of the tube. Both ends of the heating element and the thermocouple wires protruded from the same end of the tube and were passed through a nylon cap. Following the procedure described in ASTM D 5334 [5], the tube was filled with a clear plastic epoxy resin and a metal tip was placed on the open end of the tube. The epoxy was allowed 24 h setting time, and the extended heater wire ends were cut close to the thermal probe and power supply leads were soldered to them. Electrical Circuit The thermal probe was powered by a regulated doc power supply. The electrical circuit used is shown schematically in Fig. 2. The system employed a Hewlett Packard 6114A precision power supply capable of varying the voltage in steps of 0.001 V from 0 to 30 V and holding steady at the set value. The voltage across the thermal probe was monitored continuously with a 7081 Schlumberger precision voltmeter [6]. The voltmeter was calibrated by the manufacturer and checked with the Hewlett Packard precision 345

Nomenclature n Q R r T To t V K iI. Number of data points in chosen time interval Power emitted by probe per unit length, Wlm Electrical resistance, n Radial distance from probe, mm Temperature, C Initial temperature, C Time, s Voltage, V Thermal diffusivity, m2/s Apparent thermal conductivity, W/m.K

The use of small-diameter heated probes to approximate a linesource in a semi-infinite medium has been in the literature for many years [1-4]. The probe normally consists of an electrical heater and a temperature sensor, typically a thermocouple. The probe is inserted into an isothermal material and powered. The time-tempera-

Manuscript received 09/22/99; accepted for publication 6107100.

]

Lecturer, Mechanical Engineering Department, The University of the West

Indies, St. Augustine, Trinidad, West Indies. 2 Professor and chairman, Chemical Engineering Department, Tennessee Technological University, Cookeville, TN. 3 Adjunct professor, Chemical Engineering Department, Tennessee Technological University, Cookeville, TN. @2000 by the American Society for Testing and Materials

346

JOURNAL OF TESTING AND EVALUATION

Steady state
Thennocouple wire

Nyloncap

g '"
2

I,~
Linear
Nonlinear

Steelprobe
Thermocouple

f-

~ '" 0. E '"

Epoxy Heaterwire
FIG. I-Schematic of thermal probe used.

10

100
Time (s)

1000

10000

FIG. 3-Typical

experimental test results (idealized curve) [2,5].

Thermocouple

wires

In practical situations, however, the finite radius ofthe probe has the effect of a time delay before the theoretical rate of radial heat flow through the surface of the probe is equal to the heat dissipated by the heater filament [2]. Therefore, for practical situations, the general shape of the T-In t plot includes an initial nonlinear region followed by a linear region. With the passage of time, the probe temperature will level off to a steady-state value because of an isothermal specimen boundary. The general shape of a typical T-In t plot is shown in Fig. 3 [2,5].
probe

Thenna1

Test specimen

Experimental

Procedure

FIG. 2-Schematic used.

of electrical circuit and data acquisition system

power supply. This unit displayed voltage values to 10-6 V with an uncertainty of :!:: .3 X 10-5 V within the 0 to 10 V range [6]. The 1 current in the circuit was obtained by monitoring the voltage drop across a standard 1.0 0 resistor. Data acquisition was done with an IBM-compatible PC via an Advantech Enhanced Multi-Lab cardPCL812PG with Advantech PCL789D amplifier and multiplexer board. The computer was programmed by the timer function in QuickBasic V4.5 to monitor the voltage drop ten times each second, calculate the current, and display the mean current value each second. Data Acquisition The temperature-time variation of the thermocouple in the thermal probe was monitored and recorded by the computer through the I/O board. Electronic compensation was provided for the thermocouple. The computer was programmed to sense ten temperature signals per second, compute the average and display it each second, store and display readings of current, time, and temperature for a total of 1000 entries, and at the end prompt for input of the voltage across the thermal probe. Theory The general solution for the temperature variation with time of a cylinder of a perfect conductor with small radius transferring heat at a constant rate to the surrounding isothermal infinite medium with constant properties indicates that in all cases a plot of temperature against In t has a linear asymptote of slope (Q/47TA) [4]. Therefore, if the value of Q is known, the apparent thermal conductivity Ais determined by Eq I.

The thermal probe can be used to test specimens in the laboratory or materials in situ. The analysis characterized by Eq I required that the specimen be isothermal, thus limiting in situ usage. The experimental procedure involved inserting the probe in the material to be tested after both were in thermal equilibrium with the surroundings. Since the friction associated with inserting the probe may cause a measurable temperature increase, the probe temperature was monitored to make certain that isothermal conditions were attained before the heater was powered. This thermal equilibration usually takes 10 to IS min. The data acquisition system and heater power were turned on together once thermal equilibrium had been achieved. The data acquisition system recorded probe current, temperature, and time for approximately 1000 s. Since manganin wire has a negligible change in electrical resistance with temperature [7], the power to the probe heater remained constant. A properly configured test could be completed with about 1000 s of data collection. Within this time limit, the T-In t variation of the probe showed three distinct segments, i.e., an initial nonlinear segment, a linear segment, and a final steady state. The T-In t curve leveling off too soon 800 s) is an indication that the power supplied is too low. This resulted in the linear section of the plot being too brief for analysis and the temperature increase too small for reliable results. A rapid increase in probe temperature is an indication that the power supplied to the heater is too high. This can cause heat buildup in the probe resulting in damage [2], and for materials with moisture a large rapid temperature gradient could result in moisture migration, causing a change in material properties [8]. The power can be adjusted up or down to satisfy the test design time of 1000 s. Experimental Uncertainty

The experimentally determined A is calculated from Eq 1, in which A is proportional to the ratio of Q/(dT/dln f). The value of Q

MANOHAR ET AL. ON THE THERMAL PROBE METHOD

347

depends on the product of the voltage and current readings. The current reading is computed from a voltage taken across a standard resistor. (dT/dln t) is the slope b of the T-In t curve. From the theory of uncertainty analysis [9], the square of the uncertainty in the experimentally determined A.is given by Eq 2.

where Sy

r n(n

~2) ]
rnI(ln t)(T)
-

( ) ( )
2

~A.
A.

2 -

~V
V

+ ~V
V

Voltage

( )
2

+ ~R
R

Current

( )+( )
2 2

M
b

(2)

'"
LJ

In t) LJT )]

'"

. .. . M ~A. ~V ~R h d h were T' V' R' an bare t e uncertamtles m expenmentally determined A.,voltage, resistance, and slope, respectively [9]. The 7018 Schlumberger precision voltmeter displayed voltage readings with a limiting uncertainty of ::I:: X 10-5 V in the 0 to 1.3 10 V range [6]. The standard 1 D resistor showed a resistance of 1.00010 D, resulting in an uncertainty of 0.000 10 D. From Eq 2 the square of the uncertainty associated with the measured A.is:
( ~A.A. Y

X In~>'

(LT)' -

[nLOnt)'- (L Int)']
(6)

Calibration
Three thermal probes were constructed and calibrated using commercially available fine cryogenic perlite as the reference material. The A.of a 254 by 254 by 52 mm thick perlite specimen at a density of 51 kg/m3 was determined in a heat-flow meter apparatus built by LaserComp [11] at mean temperatures of9, 14, 19,24,29, 34, 39, 44, and 46e with a temperature difference of 22e between the hot and cold plates. This heat-flow meter was built and operated in accordance with ASTM e 518 [12]. The heat-flow meter was designed to provide A.measurements with an accuracy of
::I:: % with 1 ::1::0.2% repeatability and ::1::0.5% reproducibility.

(1.3 X 10-5)2

(3)

+ (1.3 X 10-5)2 + (0.0~01OY + ( ~: y

In comparison to the range of values associated with measured A., the uncertainty of the sum of the voltage and current readings is negligible. Hence, the uncertainty of A.is characterized by the uncertainty in the slope b of the straight line fit to the experimental data. Test Design Criteria and Data Analysis The theory associated with determining apparent thermal conductivity from the T-In t plot of the thermal probe (Eq 1) makes use of the slope associated with the linear segment of the heat-up curve. A least squares fit to a linear expression for T in terms of In twas used to describe this segment of the heat-up curve. The analysis involves determining the slope for candidate intervals of 0 to 1000 s, 50 to 950 s, 100 to 900 s, etc. The linear segment was taken to be such that three consecutive calculations of the slope differed from each other by no more than 2.5%. The slope, dT/dln t, of the bestfit line for the data points within the selected time intervals was calculated using Eq 4 [10].

The experimental A. ata for perlite were fit to a linear expression d in T using the method of least squares to obtain Eq 7.
A.

= 0.037493

+ (0.15727 X 1O-3)T

(7)

The thermal probes built for this project were checked in a 51 kg/m3 perlite specimen contained in a 305 by 305 by 305 mm wooden container. From the theory of the continuous line source heater [4], Eq 8 is the approximate solution for the temperature variation with time in an infinite solid, initially at temperature To, heated by a line source. T - To = (Q/4'ITK)ln(4Kt/r2)- 0.5772(Q/4'ITK) (8)

n ~dint -

(I In t)(I n I (In t)2 - (I In tY

(In t)(T)
-

T)

(4)

This equation gives the slope of the linear regression line through the selected data points. The experimentally determined value for A.is inversely proportional to dT/dln t, which is the result of an analysis of the (T, In t) data. A standard statistical technique [10] was used to estimate the 95% confidence interval of this slope. This is an important step since the uncertainty in the slope (dT/dln t) contributes to the overall uncertainty in A..Equations 5 and 6 were used for this purpose, where n is the number of data points within the chosen time interval for the analysis. 95% confidence limits of slope
= ::I:: 1.96

Using a specific heat estimate of 1005 J/kg e for 51 kg/m3 perlite at 31e, the radial distance r at which (T - To) is zero after 1000 s is 43 mm. Therefore, for all practical purposes, the 305 mm cubed specimen simulated an infinite medium with respect to the thermal probes over the test time of 1000 s. Eight calibration tests were conducted for each thermal probe. For each test the slope was determined using the procedure outlined earlier. The linear region of the slope was selected in accordance with the 2.5% criterion. Test data that did not satisfy the conditions of the test design were discarded. The 95% confidence interval, the mean value for the slope, and A.for data that satisfied the test conditions were calculated. Table 1 shows the calibration test results for probe 1. Of the eight calibration tests for probe 1, six satisfied the test design conditions (Table 1) and two were discarded. The results show that the 95% confidence interval for the slope that satisfied

the test conditions was

::I::

1.55%. The A.indicated a spread from

s;

(5)

0.0410 to 0.0422 W/m'K and the mean A.was calculated to be 0.0414 W/m.K at a mean temperature of 28.5e. From Eq 7, the e 518 value of A.for perlite at 28.5e is 0,0420 W/m'K. The difference between A.c518and A.probe is 0.0006 W/m'K. To account 1 for the difference in A.values, the assigned calibration factor cf for probe J is: cfI

(lnt)2 -

(I In ry

= (A.cm)I(A.probe

1)

= (0.0420)/(0.0414)= 1.01

348

JOURNALOF TESTING AND EVALUATION TABLE I-Calibration Test 1 results-Probe 1.

Voltage Time Period, s 0-1000 50-950 100-900 150-850 200-800 250-750 Slope, dTldln t 2.2007 2.4813 2.4918 2.4699 2.4464 2.3768

= 0.5609

Calibration Test 1 Results-Probe 1 CUlTent = 0.4102 A Mean Test Temperature = 28.5C

95% Confidence Mean Slope, dTldln t

% Spread

Interval

of Slope

A, W/m.K

::':0.0222 (::':0.89%)

0.46%

::':0.0285 (::': 1.14%) ::':0.0382 (::': 1.55%)

2.48

0.0410

Calibration Test 2 Results-Probe

Voltage Time Period, s

1 = 28.5C
Mean Slope, dTldln t

= 0.5609

Current

0.4102 A

Mean Test Temperature 95% Confidence

Slope, dTldln t 2.2261 2.4943 2.4876 2.4593 2.4324 2.4176

% Spread

Interval

of Slope

A, W/m.K

0-1 000 50-950 100-900 150-850 200-800 250-750

::':0.0128 (::':0.51%)

1.42%

::':0.0168 (::':0.65%) ::':0.0197 (::':0.80%)

2.48

0.0410

Voltage Time Period, s 0-1000 50-950 100-900 150-850 200-800 250-750 Slope, dTldln t 2.2368 2.5216 2.5323 2.5043 2.4888 2.4681

= 0.5606

Calibration Test 3 Results-Probe I Current = 0.4102 A Mean Test Temperature = 28.5C

95% Confidence Mean Slope, dT/dln t

% Spread

Interval

of Slope

A. W/m.K

::':0.0133 (::':0.53%)

1.12%

::':0.0164 (::':0.65%) ::':0.0198 (::':0.79%)

2.52

0.0404

Voltage Time Period, s 0-1000 50-950 100-900 150-850 200-800 250-750 Slope, dTldln t 1.2884 1.4569 1.4577 1.4372 1.4132 1.4086

= 0.4262

Calibration Test 4 Results-Probe 1 Current = 0.3252 A Mean Test Temperature = 27C

95% Confidence

% Spread

Interval

of Slope

Mean Slope, dTldln t

A, W/m.K

::':0.0094 (::':0.65%)

1.43%

::':0.0121 (::':0.83%) ::':0.0152 (::':1.06%)

1.45

0.0422

MANOHAR

ET AL. ON THE THERMAL

PROBE

METHOD

349

TABLE I-Continued Calibration Test 5 Results-Probe I Current = 0.3950 A Mean Test Temperature = 28.4C
95% Confidence Mean Slope, dT/dln t

Voltage = 0.5714 V Time Period, s 0-1000 50-950 100-900 150-850 200-800 250-750 Slope, dTldln t 1.9114 2.1526 2.1587 1.1237 2.0964 2.0774

% Spread

Interval

of Slope

A, W/m.K

::':0.0112 (::':0.52%)

1.65%

::':0.0141 (::':0.65%) ::':0.0165 (::':0.78%)

2.15

0.0421

Voltage Time Period, s 0-1000 50-950 100-900 150-850 200-800 250-750 Slope, dTldln t 1.9070 2.1562 2.1631 2.1396 2.0932 2.0501

= 0.5172

Calibration Test 6 Results-Probe 1 Current = 0.3926 A Mean Test Temperature = 28.4C

95% Confidence Mean Slope, dTldln t

% Spread

Interval

of Slope

A, W/m.K

::':0.0149 (::':0.69%)

1.11%

::':0.0194 (::':0.90%) ::':0.0243 (::':1.14%)

2.15

0.0417

The calibration test for probe 2 showed that the 95% confidence interval for the slope that satisfied the test conditions was::'::1.07%. The A indicated a spread from 0.0319 to 0.0340 W/m.K, and the mean A was calculated to be 0.0327 W/m' K at a mean temperature of 31c. From Eq 7, the C 518 value of A for perlite at 31C is 0.0424 W/m' K. The difference between AC518 and Aprobe is 2 0.00963 W/m' K. To account for the difference in A values, the assigned calibration factor for probe 2 is: cfz = (AC5I8)/(Aprobe = (0.0424)/(0.0327) = 1.30 2) The calibration test for probe 3 showed that the 95% confidence interval for the slope that satisfied the test conditions was::'::1.05%. The A indicated a spread from 0.0370 to 0.0385 W/m'K, and the mean Awas calculated to be 0.0378 W/m' K at a mean temperature of 31c. From Eq 7, the C 518 value of A for perlite at 31C is 0.0424 W/m'K. The difference between AC518 and Aprobe is 3 0.00459 W/m'K. To account for the difference in A values the assigned calibration factor for probe 3 is: cf3

TABLE 2-Afromprobes A, W/m.K Probe 1 (28.5C) Probe 2 (31C) Probe 3 (31C) 0.0414 0.0327 0.0378 A X cf, W/m.K 0.0418 0.0425 0.0423

and C 518. AC5IS, W/m.K 0.0420 0.0424 0.0424 Difference, ACSIS- A X cf 0.0002 -0.0001 0.0001

= (AC5I8)/(Aprobe3) = (0.0424)/(0.0378)= 1.12

The apparent thermal conductivity of the three probes with their respective calibration factors are compared with the C 518 measurements in Table 2.
A Measurements

Using the three calibrated probes, A measurements were conducted on samples of general all-purpose sand, sifted sand, and

soil. The general all-purpose sand is commercially available in prepackaged bags ("Quikrete" brand), and the material was allowed to attain equilibrium with laboratory conditions of 24C and 50% relative humidity for three days. The sifted sand was sized between 106 and 500 /Lmusing the appropriate size sieves, and the material was allowed to attain equilibrium with laboratory conditions of 24C and 50% relative humidity for three days. The soil sample used was 152 mm diameter, 200 mm long, and was obtained from the Cookeville, Tennessee area. The soil was of a brown, sandyclay nature with a density of 1893 kg/m3 and moisture content of 20.3% by weight. The all-purpose and sifted sand specimens were prepared using commercially available 2000 mL plastic containers. The sizes of the specimens were 108 mm diameter, 200 mm high, and the densities were 1730 and 1654 kglm3, respectively. Using thermal diffusivity values of K = 0.46 mm2/s and K = 0.20 mm2/s for soil and sand, respectively [4], the radial distances at which there is zero temperature change after 1000 s were calculated from Eq 8 to be 32.1 and 21.2 mm, respectively.

350

JOURNALOF TESTING AND EVALUATION

For each test material, four tests were conducted with each probe, and the voltage, current, and time-temperature data were logged. For each test the time regression analysis on the slope described earlier was conducted for the T-ln tdata. For the results that satisfied the 2.5% criterion, the 95% confidence intervals were calculated, the mean slope determined, and the A evaluated using Eq I. Table 3 shows the experimental test data for tests conducted with probe 1 on all-purpose sand (1730 kg/m3). Figure 4 shows the T-In t curve for probe I experimental data for all-purpose sand. This is a typical curve of T-In t variation for the probes. A summary of the results is shown in Table 4. Discussion The theory associated with A measurements using small diameter heated probes has been in existence for over four decades [1-4]. However, the stringent conditions under which the theory holds are difficult to meet in practical situations because of many compromising conditions. As such the reported AS determined by this method range in precision from 5 to 50% [3,5]. Another factor contributing to the poor precision of reported AS is the difficulty in monitoring the temperature-time response of the probe on which the test method hinges (Eq 1) [1,2,5]. In this work, an electrical circuit and data acquisition system was assembled to monitor precisely the temperature-time response. The circuit design and equipment brought the determinate error associated with the voltage and current measurements to a negligibly

G ~
G)

35
30

rG)

cri ::; 0. E

;:;

25 20

:~-~
,0

4 In t (s)

FIG. 4--T-In

t curve for thermal probe 1: Test I-All-Purpose

Sand.

TABLE 4--Experimentally

determined >.. y the thermal probes. b >"probe 2, W/m'K 0.5168 0.4432 1.9238 >"probe 3, W/m'K 0.5129 0.4288 2.1595 Mean>.., W/m'K 0.52 0.44 2.11

>"probe " W/m'K All-purpose sand, 0.5308 0.4614 2.2412

1730kg/m3

Sifted sand, 1654 kg/m3 Soil, 1893 kg/m3, 20.3% moisture by wt.

TABLE 3-Thermal

probe 1: test 1: all-purpose sand (1730 kg/m3). Test 2 Valid 1.4001 1.063480 30.0 1.3087 1.2526 1.2276 1.2122 1.2207 1.2186 1.27% Test 3 Valid 1.4009 1.0635 30.0 1.2594 1.2354 1.2133 1.1970 1.1989 1.2076 1.36 Test 4 Not-Valid 1.4010 1.0635 30.0 1.3279 1.2357 1.2039 1.1685 1.1260 1.1027 5.75%

Test 1 Test critena Voltage, V Current, A Mean test temperature, C Slope, time interval (0-1000) (50-950) (100-900) (150-850) (200-800) (250-750) Minimum % spread between three consecutive slopes 95% Confidence Interval of Slope Mean Slope aTtaint >.., /m.K W (>..)(1.01) (W/m'K) Mean corrected >.., W/m'K) ( (100-900) (150-850) (200-800) Valid 1.5228 1.1572 31.5 1.6883 1.6220 1.5999 1.5898 1.5858 1.5597 0.89% ::to.0I07 ::to.0146 ::to.0224 1.5918 0.4894 0.4943

::to.0094 ::to.0I25 ::'::0.0190 1.2202 0.5398 0.5452 0.5308

::'::0.0101 ::to.0141 ::to.0214 1.2031 0.5475 0.5530

MANOHAR ET AL. ON THE THERMAL PROBE METHOD

351

small value. Therefore, from Eq I, the precision of A.becomes dependent on the precision of determining the slope of the linear segment of the T-In t curve. In the past this was estimated from a physical plot of the T-In t curve [1,2,5]. In this work, the availability of a large database made it possible to use standard statistical methods for determining the slope. Specific test design conditions were set for determining the linear region of the T-In t curve as described in the text, and the probes were checked against a C 518 measurement. The 95% confidence interval of the slope, and hence A.,for all the calibration tests was within::!::1.55%. This indicates that the criteria set and the statistical analysis are effective in providing a precise determination of the slope. Probes I, 2, and 3 demonstrated repeatability of ::!::2.27%, :!::3.22%, nd::!::1.94%, respectively. However, a definite : a bias was observed between A.probe A.C5IS, and indicating the need for individual calibration. Calibration factors of 1.01, 1.30, and 1.12 were computed for the three probes in the project. Tests were conducted on samples of all-purpose sand, sifted sand, and soil. Using the respective calibration factors, the mean A. values listed in Table 4 demonstrate that the calibration and repeatability for the individual probes are consistent. The large variation in the measured soil samples (::!::8.7%)is indicative of the non homogeneity and moisture-associated complexity of soil samples [8]. The measured A.S were all within the range of published values [12-14]; however, these results were obtained with probes that were calibrated with a material of much lower A.. Conclusions I. A precision temperature-time data logging system for collecting probe data has been shown useful in automating the measurement of A.for particulates. 2. A statistical analysis and the 2.5% spread between three consecutive slope values for determining the linear segment of the T-In t curve can give results for homogenous material with a 95% confidence of about::!::1.5%. 3. Probes need to be individually calibrated with a reference material of known A..The A.of the reference material should be close to that of the materials to be tested. 4. The probe method is capable of providing reliable and repeatable A.measurements within ::!::3.5% homogeneous materials. for

[2]

[3]

[4] [5]

[6]

Probe Method," Journal of Geophysical Research, Vol. 64, No. 10, 1959, pp. 1557-1563. Winterkorn, H. F., "Suggested Method of Test for Thermal Resistivity of Soil by the Thermal Probe," Special Procedure for Testing Soil and Rock for Engineering Purposes, ASTM STP 479, ASTM, 1970, pp. 264-270. Tihen, S. S., Carpenter, H. C, and Sohns, H. W., "Thermal Conductivity and Thermal Diffusivity of Green River Oil Shale," Thermal Conductivity, Proceedings of the Seventh Conference, D. R. Flynn and B. A. Peavy, Jr., Eds., National Bureau of Standards Special Publication 302, September 1968, pp. 529-535. Carslaw, H. S. and Jaeger, J. C, Conduction of Heat in Solids, Oxford Press, 2nd ed., 1964, pp. 58-60, 344-345. ASTM D 5334 Standard Test Method for Determination of Thermal Conductivity of Soil and Soft Rock by Thermal Needle Probe Procedure, 1995 Annual Book of ASTM Standards, Vol. 04.09, pp. 225-229. Solartron Instruments, Solartron Schlumberger 7081 Precision Voltmeter, Instruction Manual, 2 Westchester Plaza, Elmsford, NY.

[7] Weast, R. C, Ed., CRCHandbookof Chemistryand Physics,

CRC Press Inc., 62nd ed., 1982, p. E-82. [8] Kersten, M. S., "Thermal Properties of Soils," Bulletin of the University of Minnesota Institute of Technology, Engineering Experiment Station, Vol. LII, No. 28, 1949. [9] Coleman, H. W. and Steller, W. G., Experimentation and Uncertainty Analysis for Engineers, 2nd ed., John Wiley and Sons, New York, 1999, pp. 47-64. [10] Natrella, M. G., Experimental Statistics, National Bureau of Standards Handbook 91, 1963. [11] LaserComp FOX 304, "Heat Flow Meter Thermal Conductivity Instrument," Information Sheet, LaserComp, Wakefield, MA, USA. [12] ASTM C 518 Standard Test Method for Steady-State Heat Flux Measurement and Thermal Transmission Properties by Means of the Heat Flow Meter Apparatus, 1998 Annual Book of ASTM Standards, Vol. 04.06, pp. 163-174. [13] Rohsenow, W. M., Hartnett, J. P., and Cho, Y. I., Eds., Handbook of Heat Transfer, 3rd ed., pp. 2.68-2.74. [14] American Society of Heating, Refrigerating and Air-Conditioning Engineers, ASHRAE Handbook of Fundamentals, 1993, pp. 22.13-22.22. [15] Meinel, A. B. and Meinel, M. P., Applied Solar Energy, Addison-Wesley Publishing Company, 1976, p. 619.

References [1] Herzen, R. V. and Maxwell, A. E., "The Measurement of Thermal Conductivity of Deep-Sea Sediments by a Needle-