Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002

THE NEXT GENERATION OF THERMOCOUPLES

FOR THE TURBINE ENGINE INDUSTRY

Daniel A. Barberree, PhD

Director of Technology
AccuTru International Corporation
817 Russell Palmer Road
Kingwood, Texas 77339
Phone: 281-358-5600 x516
Fax: 281-358-5605
Email: dbarberree@accutru.com

KEYWORDS

Thermocouple, RTD, Mineral Insulation, MIMS, Temperature, Temperature Measurement,

Sensor, Type K, Instrumentation, Calibration

ABSTRACT

Thermocouples are extremely important to the operation of turbine engines. While they are
the most rugged and reliable temperature sensors available, it is well known that
thermocouples experience de-calibration or “drift” while in service. There is no reliable
way to tell when thermocouple drift begins to occur or to determine its magnitude or
direction. De-calibration is not only a problem for turbine engine operators. The
uncertainties resulting from thermocouple “drift” cost industry millions of dollars each
year in off-quality product, less than optimum yields, underutilized capacity, unnecessary
emergency shutdowns, reduced equipment life and safety and emissions problems. Two
major improvements to thermocouple technology have been developed. A new Metal
Sheathed Mineral Insulated (MIMS) cable has been developed that is superior to that
commonly used to make the thermocouples used in turbine engines. Tests show that
thermocouples made with this new cable have greater signal stability and can give 3 – 4
times the life of standard thermocouples. Building on this technology, Dynamically Self-
Validating Sensors have been invented that eliminate unreliable readings altogether and
warn in advance of the onset of drift. This paper will focus on the development of the new
MIMS cable and also introduce the concept of Self-Validation. An example of the benefits
of applying these new technologies to a turbine engine powered electrical generation unit
will be discussed.

1
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002

INTRODUCTION

Temperature is one of the most important variables measured in industry. Often

temperature measurements are critical to the safe operation of a device or process.
Frequently they are crucial to the output of the process – the throughput or capacity,
quality, yield, energy efficiency, emissions, etc. often depend on reliable temperature
measurements.

Thermocouples and RTD’s have been developed that cover a wide range of temperatures
and provide different degrees of reliability and accuracy. Good correlations have been
developed between temperature and voltage output or resistance change of several standard
thermocouple and RTD designs. Using carefully formulated standard materials and
sophisticated polynomial relationships imbedded in electronic signal conditioners quite
accurate estimates of temperature can be obtained with new sensors.

However, it is well known that the sensors themselves are subject to change or de-
calibration when placed in service. This is sometimes called “drift”. While some causes of
drift have been identified and are somewhat predictable – such as the positive shift of type
K (or E) thermocouples (1) between 700F and 1000F - most are not. Attempts to predict
the onset, magnitude and even direction of de-calibration in thermocouples have been
largely unsuccessful.

Figure 1. depicts data illustrating drift of various types of thermocouples under conditions
of stress. Dr. Richard Anderson and coworkers (2) of Oak Ridge National Labs conducted
landmark experiments that demonstrated that different types of materials could interact
with each other to produce drift. He also demonstrated that different kinds of drift errors
contribute to the overall errors in thermocouple signals. Commonly reported thermocouple
drift performance data reflects only errors of the “first kind”, where tests are conducted
with the thermocouple in a fixed position. Drift measured in this way gives a much too
optimistic view of thermocouple stability. Thermocouple errors of the “second kind”
illustrated in Figure 1, where the thermocouple may be moved with respect to the
temperature gradient can produce large errors even when the thermocouple junction is
maintained at a constant temperature. Basically Dr. Anderson concluded that the sheathed
thermocouple is a complex system at elevated temperatures. It can begin drifting anytime
after being placed in service and the direction and magnitude of the drift depends on many
factors that cannot be quantified or predicted.

There are many examples where sensors have provided misleading information that led to
costly misfortunes – explosions, fires, releases etc. Misleading sensors don’t always result
in catastrophes. But inaccurate sensor data causes process inefficiencies costing industry
millions of dollars every year in less than optimum performance. Several techniques have
been employed to try to detect failing or incorrect sensors to improve the effectiveness of
automated process controls. Further advancements in sophisticated model based controls
and a move to “lights out plant management” will require highly reliable sensor data as
companies try to squeeze the most out of their investment with the fewest people and
operate as close to equipment constraints as possible.

2
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002

STRENGTHS AND WEAKNESSES OF THERMOCOUPLES AND

RTD’S

Current thermocouple technology has a well-promoted set of strengths. Thermocouples are

rugged, have a wide range of configurations, cover wide temperature ranges and are
relatively inexpensive. Significant advancements in electronic signal conditioners used to
process the measurements into the estimate of the temperature have been made.

However, current thermocouple technology has a not-so-well understood set of problems.

One problem is installation errors. If care is not taken with hookups, connections, insertion
depths and eliminating outside interference, significant errors can be unknowingly
introduced into the system that can go undetected. Even diligent users, who check and re-
check calibrations, can be fooled by inhomogeneities and the effects of insertion depths as
illustrated by Dr. Anderson in Figure 1. The major problem, however, is the gradual de-
calibration of the sensor elements causing signals to drift at unknown time, rate and
direction. The uncertainties surrounding de-calibration cannot be quantified or removed by
even the most sophisticated electronic signal conditioning units and smart transmitters.

RTD’s are promoted as having very high accuracies in the ranges in which they can be
employed but they are also subject to de-calibration. While the causes of de-calibration of
RTD’s are different, they too are unpredictable and cannot be detected. The range of
temperatures in which RTD’s can be used is much more restricted than thermocouples and
in some applications they are too fragile to withstand vibration from the process. Except by
gross comparisons and reasonableness checks vs. other sensors, drift is not detectable.

THE NEXT GENERATION OF SENSORS

A 1997 survey of industry process control specialists and managers concluded that the two
most important improvements being looked for in temperature sensors were:

1. “Get rid of the drifting signals while the probe is in service”.

2. “Give us longer useful life”.

Recently several industry leaders have declared that sensors must become more robust,
reliable, maintainable, self-checking and automatically self-calibrating. Some have
predicted that self-diagnosing field devices will be able to validate data within their loops
and manage communications to others.

Progress has now been made toward both of these objectives:

• Efforts to eliminate drift focused on implanting a credible calibration reference

inside a sensor probe capable of continuous calibration while the probe is in
service. This has led to the development of the Self-Validating Sensor.

3
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002
• Efforts to achieve longer life focused on materials. A new mineral insulation
material has been developed that has demonstrated 3-4 times the life and greater
signal stability in thermocouple sensors vs. those made with Magnesium Oxide
(MgO).

The primary goal of this paper is to present results of the work on the new mineral
insulation material. It will have immediate application for many situations. Another goal is
to introduce the concepts of Self-Validation. This will be done first. Self-Validation will be
the main subject of another paper to be presented at ISA later this year.

SELF-VALIDATION

Over the years a number of improvements have been made in thermocouple and RTD
technology. The materials have been improved and industry standards (ASTM, DIN) have
been developed for performance of materials and also for the construction of sensors.
Advances in electronics have led to better signal conditioners that can check themselves to
be sure they are operating properly. Despite these refinements, we have not been able to
reliably predict nor monitor the performance of these sensors when in service - especially
when they are used under conditions that stress the materials.

The challenge then is how to construct a sensor so that the performance and health of its
internal components can be monitored continuously while it is in service. Self-Validation,
then, is the ability to measure the process variable, in this case temperature, with a high
degree of confidence and at the same time determine whether the materials producing the
signal representing that variable are stable or showing signs of impairment. This is
sometimes called Self-Calibrating or Self-Correcting or Self-Diagnosing or Dynamically
Self-Validating, or simply a Self-Validating Sensor (SVS).

Now temperature sensors have been developed that are capable of Self-Validation. This
technology evolved from combining a set of known technologies in a new way and has
resulted in a uniquely designed thermocouple like probe. Externally the probe looks and
feels like a normal thermocouple or RTD and can be a direct replacement in most
processes. It is constructed using similar techniques. It is rugged and robust, it bends,
welds and configures just like a typical metal-sheathed thermocouple.

However, imbedded in the tip of this new probe is a combination of thermally sensitive
materials called a Calibration Reference Matrix (CRM). A CRM provides all the
information needed to develop a far more accurate temperature estimate than a
thermocouple and also continuously monitor the health of the probe and it’s components
while it is in service. An electronic signal conditioner multiplexes several measurements
taken from the probe and monitors the health of each individual element in the probe. In
addition to the temperature, the signal conditioner provides sensor health status to the
operator or control system and notifies of impending loss of measurement validation
before it occurs. No element in the sensor can de-calibrate without detection. This is a
major breakthrough in contact temperature measurement technology.

4
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002
To build such a probe and achieve satisfactory life and stability required better materials
than are currently used for thermocouples and RTD’s. A mineral insulation superior to
Magnesium Oxide (used in 90+% of MIMS Sensors) and Aluminum Oxide (the other
10%) was required. This led to the development of the new mineral insulation material
called MI-Dry.

THERMOCOUPLE DE-CALIBRATION

To discuss the advantages of MI-Dry requires some grounding in the causes of de-
calibration of thermocouple sensors. The temperature measurement in a thermocouple is
derived from the now famous observation by Tom Seebeck in 1821 that when two
dissimilar electrically conductive materials are joined at one end and that end is maintained
at a different temperature than the open end, a voltage or emf is generated across the open
end. Further Seebeck observed that that voltage could be reproducibly correlated with the
magnitude of the difference in temperatures of the two ends. We now know that the emf is
not generated at the junction of the two materials, but rather along the length of the two
materials as the temperature changes from one end to the other. This makes it very
important to have materials with consistent composition from one end to the other so that
the same signal is generated regardless of where the temperature profile is positioned. It is
also important that the electrical insulation surrounding the wires in a thermocouple be
stable to protect them from contamination and also to avoid shunting or short-circuiting the
wires. An excellent reference on thermocouple errors and inhomogeneities is “Traceable
Temperatures” by Nicholas and White (3).

The primary mechanism for thermocouple de-calibration and impairment is inhomogeneity

in the wires caused by a change in composition of the wires. As Dr Anderson and his
colleagues showed, this is frequently due to migration of impurities within the sensor from
wire to wire or sheath to wire as depicted in Figure 2. Small changes in composition cause
changes in the emf signal generated by the wire pair and cause errors in the temperature
estimate. Impurities may also come from the mineral insulation.

As stated earlier, Magnesium Oxide (MgO) is the mineral insulation used in 90+% of the
metal sheathed thermocouples made today. Pure MgO is a good insulator. A problem with
MgO is that it is hygroscopic. Any moisture absorbed by the MgO decreases it’s insulation
resistance and aids in the transport of ions within the sensor - from wire to wire or sheath
to wire. Moisture also contributes to corrosion of some materials used in thermocouples.
Moisture can be absorbed during the manufacturing process when the cable is opened up to
form the thermocouple junction and to expose the lead wires. Moisture can also be
absorbed if there is a breach in the sheath or thermocouple housing and the thermocouple
is stored for some time before use. Without great care in manufacturing techniques the
insulating properties of MgO can be severely compromised.

5
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002
A NEW MINERAL INSULATION MATERIAL

The new mineral insulation that has been developed is an extremely stable high
performance ceramic made specifically for use in mineral insulated metal-sheathed cables
used to make thermocouples and RTD’s. Unlike MgO, this material is non-hygroscopic. It
also has higher electrical resistance properties than MgO. And it can be used to fabricate
cable in the same manner as MgO.

Figure 3. illustrates these properties. This new material not only has superior electrical
resistance, but it is designed to block the diffusion of trace elements into thermo-element
wires. Because it is non-hygroscopic, the new material reduces the ingress of moisture into
the cable interior. This significantly increases resistance to corrosion and other processes
that promote thermocouple de-calibration. The new ceramic itself is non-corrosive to
metals up to 2000 oC, whereas MgO will react with most metals above 480 oC. It exhibits
negligible reaction with conducting wires or other materials up to 1300 C. The insulation
resistance of thermocouples fabricated from Ml-Dry is about 100 times higher than
conventional MgO insulated thermocouples or cable preventing virtual junction and
shunting errors.

Thermocouples made with this new mineral insulation material have demonstrated greater
signal stability and 3 to 4 times the life vs. similar sensors made with MgO.

COMPARISON TESTING

An accelerated comparison stress test of sensors made with the new mineral insulation
material and those made with MgO was done at 1200 C similar to the tests run by Dr.
Anderson. The test was conducted using a Lindberg box furnace with an upper temperature
limit of 1371 C. Figure 4. is a schematic diagram of the test apparatus. A 22” long ceramic
tube was inserted through a hole in the furnace door, with the tip of the tube located 5”
from the back interior wall of the furnace. Sensors were bundled together and wrapped
with platinum wire to keep them in thermal equilibrium. The sensor bundle was inserted
into the ceramic tube to a point where the tip of the bundle was ½” from the end of the
ceramic tube. The ceramic tube was held in place with two ring stands and appropriate
clamps to keep the ceramic tube at a fixed location during the test.

Five Type K sensors were bundled together with a Type S working reference standard and
held in a constant temperature zone of the furnace at 1200C until all had failed. Sensor
outputs were measured daily and compared to the Type S working reference standard. The
working standard had previously been calibrated vs. a NIST traceable Primary Reference
Type S sensor.

Figure 5. shows the results of the test. The vertical axis shows deviations in Deg C from
the working Type S reference standard after corrections for it’s deviation from “true”.

The test samples were all ¼” OD, Type K sensors in Inconel 600 sheath. Three of the
sensors were manufactured by three different well known companies using MgO mineral
6
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002
insulation and sold as within special limits of error tolerances. Two were manufactured
using Mi-Dry as the mineral insulation and were as close in design dimensions to the
three purchased sensors as possible. Special limits wires were also used in the manufacture
of the MI-Dry sensors.

The 1200C test temperature is near the upper limits of use for the materials used in a Type
K sensor, but below the use temperature of Inconel 600 so that this did not become a
sheath test.

The test protocol defined that any sensor with two consecutive daily readings outside
standard limits of error would be unacceptable for industrial use and therefore disqualified.
Recall that these sensors were all supposed to be manufactured to perform within special
limits of error.

The Mean Time to Failure of the two MI-Dry sensors was 4 times as long as the three
MgO sensors. The MI-Dry Sensors outlasted the MgO sensors by at least 3 to 1. In
addition, the MI-Dry sensors were still indicating within special limits of error at the
time of their failure. Mi-Dry insulating material demonstrated superior performance to
MgO.

In separate testing, a major turbine engine manufacturer has also performed cycling tests of
sensors made with MI-Dry. The tests simulated accelerated turbine engine operating
condition temperatures and cycles. The test covered 18,181 cycles between 300 deg F and
2150 deg F (2597 actual hours) that simulated about 150% of normal engine cycles and
500% of the time at max temperature There were at least 10 other Type K MIMS sensors
in the test. At 18,181 cycles the test was terminated. All the sensors had failed except the
sensor made with MI-Dry.

POTENTIAL VALUE TO A GAS TURBINE/GENERATOR

This research was conducted to develop superior sensors for the process industries. It has
been pointed out that with better temperature measurements there is a large potential for
improving many combustion processes and in particular the efficiency of turbine
generators. Type K MIMS sensors are commonly used to measure exhaust temperatures in
the turbine and control it’s operation.

A Houston based utility has assisted in generating an estimate of the value of applying
these technologies to a specific gas turbine/generation system. This is shown in Table 1.
Utility Savings. The areas of opportunities identified are:
• safely use previously off limits capacity
• improve fuel consumption efficiency
• reduce maintenance shut downs
• reduce or eliminate false trips
• improve equipment life
Additional benefits in emissions control are also possible in certain circumstances.

7
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002
As experience is gained with the new more stable MI-Dry sensors, turbine operators will
gain confidence in operating closer to the constraints of these engines. When the switch to
Self-Validating Sensors is made it will be possible to further optimize the performance of
these engines with significant economic benefits.

SUMMARY AND CONCLUSIONS

A new mineral insulation has been developed for use in manufacturing extended life
mineral insulated metal sheathed (MIMS) thermocouples and RTD’s. It has the following
benefits:
• Direct replacement for MgO in many applications
• Long life in high temperature applications
• Extremely inert
• Non-hygroscopic
• High electrical resistance
• Superior resistance to neutron radiation compared to MgO
• Reduced de-calibration of thermo-elements due to
Lower corrosion
Blocking ion migration
This development will be of use to many users who need longer life and greater signal
stability from their thermocouples and RTD’s.

A new class of sensors that can validate themselves while in service and warn in advance
of their own de-calibration has been introduced. The 1st of these is the SVS Self-Validating
Temperature Sensor.

The implications of these technologies for process operators are significant. They will be
positioned to achieve a big step forward in control and optimization of their processes.

REFERENCES

1. Sibley, F.S., et al., “Aging in Type K Couples”, Instrumentation Technology, Vol.

15, p 107, 1968.
2. Anderson, R.L., et al., “Decalibration of sheathed thermocouples”, Temperature,
it’s Measurement and Control in Science and Industry, Vol. 5, pp. 977-1007,
American Institute of Physics, N.Y., 1982.
3. Nicholas, J.V., and White, D.R., “Thermocouple Thermometry”, Traceable
Temperatures, John Wiley & Sons Ltd., Chichester, England, 1994.

8
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002

25
Anderson, et al
Decalibration of Sheathed Thermocouples
DRIFT, DEG C

-50

0
200 400 600 800 1000 1200
ELAPSED TIME, HOURS
FIG. 1. Drift of 3-mm-diameter stainless steel-sheathed and Inconel 600-sheathed type K and Nicrosil versus Nisil
thermocouples at 1200 C in vacuum. The dips in the drift curve are the result of the "in-place inhomogeneity test"
where the samples were extracted from the furnace by 5 cm. The symbols used for the various samples are: type K
in Inconel - circles; Nicrosil versus Nisil in Inconel -triangles; Nicrosil versus Nisil in stainless steel - crosses; and
type K in stainless steel - squares.

SHEATH IMPURITIES MINERAL

THERMO-ELEMENTS INSULATION

FIG. 2 – FAILURE MECHANISM IN THERMOCOUPLES

9
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002

SHEATH IMPURITIES MINERAL

THERMO-ELEMENTS INSULATION

FIG. 3 – BLOCKING MECHANISM IN MI-DRY

LINDBERG BOX Sensors

FURNACE

22” Ceramic Tube 1-1/2’ OD

10”
Furnace 4” 15” 4” Holders/Clamps
Control
Sensor

FIG. 4 – TEST EQUIPMENT

10
 Reprint – ISA 48th Annual International Instrumentation Symposium – San Diego, CA – May, 2002

20

15

MgO TC #3
10
Standard Limits
ERROR FROM TRUE, DEG C

5
MI-Dry TC #1 Special Limits

0
0 500 1000 1500 2000 MI-Dry TC #2 2500
-5
Special Limits
MgO TC #2
-10 Standard Limits

-15

-20

-25 MgO TC #1

-30
TIME AT 1200 C, HR

FIG. 5 – MI-Dry VS. MgO TEST RESULTS

TABLE 1 – UTILITY SAVINGS

Gas Turbine - Houston Based Utility

(80 Mw Unit, $35/Mw)

Opportunity: Benefit:
Unused Capacity > $500 K/YR
Fuel Consumption > $100 K/YR
Maintenance S/D’s $300 K/YR
False Trips $100 K
Equipment Life $600 K

11