|
Thermocouples
are widely used as temperature sensors, but they can be strongly affected
by the environment in which they are employed
The
See beck emf (the thermocouple output) is generated by the product of
the temperature gradient and the specific thermoelectric power of the
wire ŋ, summed over the entire length of all materials in the thermocouple.
This can be expressed as in Eq. 1, where T is the temperature
(Kelvin) and x is the length coordinate. According to this
equation the emf is generated only in lengths containing temperature
gradients. Non contribution is made from isothermal portions.
see
back emf = ∫ŋ∆T•dx
(1)
If
the metals were perfectly homogeneous, the See beck emf would simply
be the difference between the junction temperatures times the difference
in the specific thermoelectric powers of the materials. But given
a non uniform ŋ, there are two ways a change in emf can occur while
maintaining the junction temperatures: either the temperature distribution
or ŋ could change throughout the wire.
As
chemical reactions progress in the thermoelements, their temperature
indication becomes increasingly erroneous. The constantly changing
spectrum of inhomogeneities in the wires causes these errors to develop.
Unless special precautions are taken to protect the thermoelements from
the environment, their calibration may valid only for a matter of hours.
Several
studies have attempted to discover the nature of thermocouple drift
and aging effects. One area of activity has been directed toward
oxidation and other redox reactions and their relation to thermocouple
outputs.
Exhaustive
3000-hour exposure tests were conducted with Type K (chromel-alumel)
thermocouples seleted from major manufacturers in Great Britain, Sweden,
Germany and the U. S. (Ref. 1). Some of the compositions of the
newly-developed Type Ks differed substantially from the trademarked
chromel (Ni-10% Cr) and alumel (Ni 5%, Al, Si, Mn). The couples
were exposed to air at 1223 K and were calibrated periodically at 473,
673, 873, 1073 and 1273 K. Two stages of chromel drift were discovered
- a fairly rapid drift of appreciable magnitude within the first 30
hours, and then a slow, almost linear drift after that.
Each
alloy was calibrated against platinum. In general, the drift of
the chromels was greater at 873 and 1073 K than at the higher temperatures.
The alumels had even larger drifts (around 200µ), the maximum occurring
at the highest temperature. The combined drifts amounted to an
indicated error of 11 K.
Two
of the alumels with high Si and low Al and Mn contents were found to
be more oxidation-resistant (Ref. 2). Hypothetically, the silicon
forms an impervious silical or silicate film at the metal/oxide interface.
This barrier and virtual absence of Al and Mn (Which are easily oxidized)
would account for the very small drifts found in these samples. See
Note 1 (Nicrosil-Nisil).
Pursuing
the concept of an oxidation-resistant couples further, a chromel alloy
containing 20% Cr has been proposed, since the oxidation resistance
of such alloys is far greater than that of the Ni-10% Cr. Controlled
small additions of Si would probably result in further improvement.
On the other hand, abnormally low Cr content could result in a negative
drift from the sart.
Exploration
into the the use of heat treatment processes for stabilizing couples
has led to the partial conclusion that heat treatment of the whole length
of a Type K thermocouple at 700 K for one hour would reduce considerably
the drift and gradient-shift errors for use up to 675 K. However,
one-hour treatment at 1273 K did not bring the necessary improvement.
A thermal cycling treatment was then attempted where one cycle consisted
of two hours of heating to 973 K and six hours of cooling to 323 K (Ref.
3). Every ten cycles the couples was held at 773 K for hours,
and then calibrated in intervals of 100 K over the cycling range.
A Total of eighty cycles were run, and the Type K couples were found
to stabilize after twenty cycles within 0.1 K.
Aging
was conducted in air at 1050 K for 1000 hours, and drifts in emf were
measured at the zinc point only. Immersion depth was closely controlled.
Chromel-Pt and Alumel-Pt thermocouples both changed severely in the
first 30 to 50 hours, resulting in an indicated error of 5 K.
Thereafter the drift was quite linear in time with no change in the
rate of drift in sight, After 1000 hours the error had increased another
1 K. At that time the effect of immersion depth was measured.
A Change of 15 cm caused an indicated error of 25 K, dramatically underscoring
the need for precise replication between calibration and use conditions.
Large
errors have been observed due to a small residual of sulphur-bearing
oil in the protection tubes that becomes corrosive when heated (Ref.
4) all such tubes should be burned out prior to assembly ant Ti included
as an "oxygen getter" in the hermetically-sealed tube.
The highest errors due to selective oxidation in rather short exposures
(23 or 48 hours) to both hydrogen and air at -47 K error at 1266 K,
and 89 K error at 1450 K. All of these tubes were open at the
cold end. Using identical wires in hermetically sealed tubes including
Ti, no drift of the See beck emf was observed under identical test conditions.
The
error due to diffusion of alloying elements across the junction has
been investigated in Type R (Pt-Pt-13% Rh) thermocouples (Ref. 5).
The diffusion was thermally generated by the equivalent of 100 days
at 1773 K. In a temperature gradient of 10 K/cm, an error of 1.3
K may be expected at all temperatures in the operating range.
In
a detailed investigation of Types K, Y, and J (Ref.
|
6),
thermocouples were cut into 7.6-cm lengths from the wire as received
from the manufacturer. Calibrating each section disclosed cyclic
variations in emf of wavelengths as short as 15 cm and up to one meter.
These variations caused errors of up to 1 K in the temperature range
from room ambient to 580 K.
An
investigation of the refractory metal thermocouples (Tungsten, Rhenium,
Tantalum, and Molybdenum), which are used for high temperatures in a
vacuum or inert environment, discovered that after 200 hours at 1900
K in argon errors of the magnitude of 5 K occur at calibration temperature,
1133 K (Ref. 7). Longer term tests (240 to 360 hours) resulted
in errors up to -66 K. A nonlinear positive correlation exists
between ductility and instability for any given material regardless
of the type of contamination. For example, contamination.
For example, contamination by oxygen and carbon happened in the range
of 880 K to 1500 K even in vacuum of 10-7 Torr. The
reaction of the ceramic protection tube also is illustrative of the
non -linear character of these effects. No effect was measured
up to 150 hours, bt after that the alumina-refractory reaction proceeds
rapidly (at 1900 K in argon). The conclusion was that pure Re
thermoelements have the most promise for stability. Sheathed in
alumina these remain very stable for 360 hours at 1900 K.
Nobel
metal thermocouples were tested in vacuum, argon and air (Ref. 8).
The chief cause of emf drift in these couples was analyzed as contamination
by Fe from the ceramic protection tubes. Accordingly, eschewing
ceramics with iron content is recommended. Fe had the greatest
effect on Pt, and much less for Iridium (Ir) and Rh. The instabilities
in vacuum were found to be comparable to the data obtained in argon,
and these instabilities were of much greater magnitude than those
resulting from aire environment. Exposures at 1873 K for 120 hours
resulted in -39K error for type S (Pt-Pt-10% Rh) calibrated at 1133
K. Other alloys, such as Pt-6% Rh-Pt-30% Rh, exhibited-29K errors,
under the same conditions. After these tests were completed, the
immersion depth was changed on each couple, and errors of -100 to -160
K were commonly witnessed.
The
effect of dilute solutions of iron group transition metals on the thermoelectric
power of pure copper have also been looked into (Ref. 9). Alloys
were made under carefully controlled conditions, melting various compositions
of transition metals into 99.9997% pure copper. These ingots were
drawn into 0.2 mm wires from which the thermocouples were fashioned.
The experiments were conducted in the temperature range from room ambient
to cryogenic values. The data shows rather dramatic changes in
µ, specific thermoelectric power, for dilute alloys up to 0.1 atomic
percent.
The
effect of shock waves was investigated by measuring a transient emf
due to shock corresponding to a 180 K rise (Ref. 10). the pressure
of the shock wave was measured at 0.3 Mbar. The emf of the Type
T Couples had a rise time of 2 µsec and a duration of 0.5 sec.
Strong
magnetic fields have a very great effect on the thermoelectric properties
of materials. A 10-KiloOersteds field was used to measure the
change in specific thermoelectric power of Bi (ref. 11). The transverse
field caused an increase of 24% in µ and an increase of 100% with the
field reversed compared to the condition of no magnetic field.
Pressure
also has a discernible effect on the See beck emf of Types Y, J and
K. The deviation in emf is close to linear with the applied pressure
(Ref. 12). When calibrations were performed at 873 K, ethe error
at 50 kbar for Type K was 10 K and the error for Type J at 40 kbar was
30 K.
Calibration
of any thermocouple must duplicate exactly its intended installation,
right down to the temperature distribution along the wire. The
reason for this is that there exists no homogeneous thermocouple, and
the calibration is of the length of wire containing the temperature
gradients. It would therefore be best to calibrate the thermocouples
in place.
Where
the accuracy need not be better than ±1K or 0.375% of reading, calibration
is not necessary since special thermocouples can be purchased with this
guaranteed accuracy. Indeed, even if the wire were calibrated
in the laboratory, this calibration should not be applied willy-nilly
to particular installations. A Thermocouple, homogeneous in the
laboratory, soon would become inhomogeneous from the chemical reactions
proceeding at different temperature-dependent rates along the wire.
Unless special precautions are taken to protect the thermocouple, its
calibration will be valid only for a matter of hours (roughly 30 to
100). Large drifts have been reported while maintaining constant
immersion, and drifts larger by an order of magnitude or more have been
witnessed due to changes in immersion.
One
suitable procedure for checking the accuracy of thermocouples in services
is described in the American Standard for Temperature Measurement
Thermocouples (C96.1-1964) , which recommends the temporary
installation of a new or checking thermocouple alongside the service
thermocouple or in its place (in this case, it is essential that stable
temperature conditions be maintained) and comparing the readings.
if the installed thermocouple is used to measure a wide range of
temperatures, it should be checked at more than one temperature
within the range of its use. Testing of a thermocouple at a single
temperature yields some information, but it is not safe to assume that
the changes in the emf of the couple are proportional to the temperature
or to the emf. A separate checking instrument should be used with
the checking thermocouple, to permit checking of the service instrument,
as well as of the service thermocouple. In general, the higher
the temperature or more contaminating the atmosphere, the more frequently
checks should be made.
|
-
Burley, N.A., Ackland,
R. G., "The stability of the Thermo-emf/Temperature Characteristics
of Nickel Base Thermocouple," J. Australian Inst. of Metals,
Feb., 1967
-
Hughes, P.C., Burely,
N. A. "Metallurgical Factors Affecting Stability of Nickel
Base Thermocouple," J. Inst. of Metals, 1962-63, v. 91. pp.
373-6.
-
Pirro, P., "The
Use of Thermocoax Chromel-Alumel Thermocouples for the Measurement
of Small Temperature Differences, " J. Sci. Instruments, 1967,
p. 1055
-
Spooner, N. F. Thomas,
J. M., "Longer Life for Chromel-Alumel Thermocouples,"
Metal Progress, Nov., 1955, pp. 81-85
-
Mortlock, A. J.,
"Error in Temperature Measurement Due to Interdiffusion at
the Hot Junction of a Thermocouple," J. Sci. Instruments, Aug.,
1958, p. 283.
-
Moffat, R. J., "Understanding
Thermocouple Behavior: The Key to Precision," ISA Proceedings,
Oct., 1968. #68-628.
-
Walker, B.E., Ewing,
C.T., Miller, R. R., "Instability of Refractory Metal Thermocouples,"
Rev. Sci. Instruments, June, 1963, pp. 816-25.
-
Walker, B.E. Ewing,
C. T., Miller, R. R., "A Study of Nobel Metal Thermocouples
in Vacuum," Naval Research Lab. Interim Report NRL-6236, May
1965.
-
Brewig, E., Kierspe,
W., Schotte, U., Wangner, D., "Effects of Transition Metal
Solutes on the Thermocouple power of Copper" J. Phys. Chem.
Solids, 1969, pp. 483-90
- Palmer, E.P., "Response
of a Thermocouple Junction to Shock Waves in Copper," J. Applied
Physics, Oct. 1964, p. 3055.
- Wolfe, R., Smith, G. E., "Experimental
Verification of the Kelvin Relation of Thermoelectricity in a Magnetic
Field," Physical Review, Feb. 1963, pp. 1086-7
- Henneman, R. E., Strong, H.
M., "Pressure Dependence of the emf of Thermocouples to 1300
C and 50 kbar," J. Applied Phys., Fe., 1965, pp. 523-8
The preceding article
on Thermocouple Accuracy by David R. Keyser was originally titled "How
Accurate are Thermocouples, Anyway?" How Accurate are Thermocouples,
Anyway?" appears on pages 51,52,53 and 54 the March, 1974, edition
of INSTRUMENTS AND CONTROL SYSTEMS.
|
Note
1:
The NICROSIL-NISIL
thermocouples are nickel-based alloys and therefore, have melting points
similar to the Type K thermocouples. NICROSIL-NISIL is not a substitute
for platinum thermocouples, but can be used under the same parameters
as for those of Type K Thermocouples.
The NICROSIL-NISIL elements,
both of which contain m magnesium, have been shown to be more stable,
and last longer, than Type K thermocouples, I has been recommended to
the ASTM that the maximum operating temperature range for Type K Thermocouples
be adopted for NICROSIL-NISIL
Originally, the NICROSIL-NISIL
thermocouple was developed to improve oxidation resistance; however,
the current NICROSIL (NICROSIL II) with chromium content of 14.5% reduces
the drift in the EMF found in the positive Type K thermoelements operating
at approximately 500oC.
The increase of silicon
plus the complete elimination of aluminum in the negative thermoelements,
along with the addition of magnesium in both legs, reduces the drift
in the EMF caused by preferential oxidation, and produces a better oxide
film which slows down physical failure of the thermo element.
The magnesium addition enhances the formation of oxide film which is
resistant to spalling and progressive and/or selective oxidation.
Numerous papers have
been present at ISA seminars and various conferences showing the increase
in useful life of these alloys. These tests indicate longer life,
better accuracy for longer time periods and/or faster temperature response
because finer sizes of NICROSIL-NISIL will give equivalent life to heavier
Type K sizes( i.e. .064 NICROSIL-NISIL will last longer than .128 Type
K.
Limits of Error:
Extension Wire
: 0oC to 20oC ±2.2oC
Thermocouple Wire : 0oC to 1250oC
±2.2oC or 0.75%
(Whichever is greater)
|
