Nuclear Engineering and Design 239 (2009) 3132–3141

Contents lists available at ScienceDirect

Nuclear Engineering and Design

journal homepage: www.elsevier.com/locate/nucengdes

Nuclear plant temperature instrumentation

H.M. Hashemian a,∗ , Jin Jiang b,1
a
AMS Corporation, AMS Technology Center, 9111 Cross Park Drive, Knoxville, TN 37923, USA
b
Electrical & Computer Engineering Faculty of Engineering, The University of Western Ontario, London, Ontario, Canada N6A 5B9

a r t i c l e i n f o a b s t r a c t

Article history: Most critical process temperatures in nuclear power plants are measured using resistance tempera-
Received 26 March 2009 ture detectors (RTDs) and thermocouples. In addition to excellent reliability and accident survivability,
Received in revised form 14 July 2009 nuclear safety-related RTDs are expected to have good calibration and fast dynamic response time, as
Accepted 21 August 2009
these characteristics are important to plant safety and economy. In plants where RTDs are installed in
thermowells in the primary coolant pipes, response-time requirements have a range of 4.0–8.0 s versus
the direct-immersion RTDs installed in bypass loops which have a required response range of 1.0–3.0 s.
The variety of problems that can affect the accuracy and response time of RTDs is extensive: dynamic
response problems, failure of extension leads, low-insulation resistance, premature failure, wrong cali-
bration tables, loose or bad connections, large EMF effects, open elements, thinning of the platinum wire,
lead-wire imbalance, seeping of chemicals from the connection head into the thermowell, cracking of
the thermowell, and erroneous indication. The causes of core-exit thermocouples failure can take the
form of large calibration shifts, erratic and noisy output, saturated output, accidental reverse connec-
tions, and response-time degradation. Several effective methods for detecting RTD and thermocouple
performance failure while the plant is operating are available. To detect accuracy problems, the cross-
calibration technique is effective for both RTDs and core-exit thermocouples. It involves recording the
readings of redundant online RTDs, averaging these readings, and calculating the deviation of each RTD
from the average, less any outliers. To detect response time degradation online, the loop current step
response (LCSR) test is the most accurate method. However, the noise analysis technique remains the
most popular for detecting response time degradation in core-exit thermocouples.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction resistance is proportional to temperature. Today, the sensing ele-

ment of industrial RTDs is almost always made from platinum wire.
Most critical process temperatures in nuclear power plants are Early RTDs were often fragile and unstable because the platinum-
measured using resistance temperature detectors (RTDs) and ther- sensing element became contaminated. Today, industrial RTDs are
mocouples. In pressurized water reactor (PWR) plants, for example, very rugged and reliable, and can be used in applications as extreme
the primary coolant temperature and feedwater temperature are as: (1) measuring brake temperatures of over 1000 ◦ C in high-speed
measured using RTDs, and the temperature of the water that exits aircraft; (2) measuring oceanographic temperature under very high
the reactor core is measured using thermocouples. These ther- hydraulic pressures where high accuracy and quick response (less
mocouples, called core-exit thermocouples, are mainly used for than 0.5 s) are important; and (3) measuring primary coolant tem-
temperature monitoring purposes and are therefore not generally perature in PWRs at temperatures up to about 350 ◦ C, flow rates of
subject to very stringent requirements for accuracy and response- over 10 m s−1 , and pressures of about 150 bar (Plumb and Schooley,
time performance. In contrast, primary coolant RTDs typically feed 1982; NRC, 1990).
plant control and safety systems and must, therefore, be very accu-
rate and have good dynamic performance. 2. Nuclear-grade RTDs
RTDs have historically had sensing elements made out of plat-
inum, copper, nickel, and other metals or alloys in which electrical Nearly 100 suppliers worldwide manufacture RTDs today, but
fewer than 10 of them manufacture RTDs for safety-related applica-
tions in nuclear power plants. This is because the size of the market
∗ Corresponding author. Tel.: +1 865 691 1756; fax: +1 865 691 9344.
for this application is relatively small and the performance and
E-mail addresses: hash@ams-corp.com (H.M. Hashemian), jjiang@uwo.ca
reliability required of these RTDs are very stringent. Nuclear safety-
(J. Jiang). related RTDs must pass environmental and seismic testing under
1
Tel.: +1 519 661 2111x88320; fax: +1 519 850 2436. Institute of Electrical and Electronics Engineers (IEEE) standards

0029-5493/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2009.08.030
 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141 3133

Table 1
Partial listing of suppliers of nuclear-grade RTDs.

RTD manufacturer Model number RTD type

Conax 7N10 Thermowell-mounted

7RB4 Direct-immersion
7N13 Thermowell-mounted

RdF 21204 Direct-immersion

21297 Direct-immersion
21232 Thermowell-mounted
21458 Thermowell-mounted
21459 Thermowell-mounted
21465 Thermowell-mounted

Rosemount 104AFC Thermowell-mounted

176KF Direct-immersion
177HW Thermowell-mounted
177GY Direct-immersion

Sensycon 1703 Thermowell-mounted

1717 Thermowell-mounted
Fig. 1. Simplified diagram of a primary coolant loop of a PWR.
Weed N9004 Thermowell-mounted
N9007 Direct-immersion
N9019 Direct-immersion
tor power (P) is the product of the temperature difference (T)
across the core and the mass flow rate (ṁ) in the primary coolant
to demonstrate that they can survive a loss-of-coolant accident system (i.e., P ≈ ṁT ). The T is typically about 30 ◦ C; thus, a one-
(LOCA), withstand a seismic event, and continue to provide reliable degree error in measuring the T corresponds to 3.33 percent in
service under postaccident conditions. The relevant IEEE standards power output. Therefore, calibrating RTDs accurately is very impor-
are (1) IEEE Standard 323, “IEEE Standard for Qualifying Class 1E tant to plant economy. For that reason, the primary coolant RTDs
Equipment for Nuclear Power Generating Stations” (2004), and (2) in PWR plants are typically calibrated to an accuracy of 0.3 ◦ C or
IEEE Standard 338, “IEEE Standard Criteria for the Periodic Surveil- better before installation.
lance Testing of Nuclear Power Generating Station Safety Systems” Fig. 2 illustrates the importance of response time for nuclear-
(1988). grade RTDs. Fig. 2 shows the primary coolant temperature in a
RTDs qualified for nuclear safety-related applications are mostly PWR plant experiencing a step change. To ensure safety, the RTDs
used in the primary coolant system of PWR plants, normally of the plant are expected to promptly trigger a mitigating action,
between 16 and 32 RTD elements per plant. Some nuclear power including a reactor scram if needed. This is why the response time
plants have more RTDs. For example, the Arkansas Nuclear One, requirements of primary coolant RTDs in PWR plants are so strin-
Unit 2 plant (ANO-2) in the United States has more RTDs to off- gent. These requirements differ from plant to plant, however. For
set the impact of a temperature streaming problem in its hot-leg example, in plants where RTDs are installed in thermowells in the
pipes that, while severe, does not pose a risk to the operation or primary coolant pipes, typical response-time requirements have a
safety of the plant. This temperature streaming, known as “hot-leg range of 4.0–8.0 s. This contrasts with the 1.0–3.0 s that are required
anomaly,” is inherent in the hot-leg pipes of PWR plants, and its of the direct-immersion RTDs that are installed in bypass loops.
severity depends on the specific design of each PWR plant. Sim- Some plants use bypass lines to help sample the reactor water from
ilarly the Sizewell B nuclear power plant in Great Britain uses 60 all coolant loops and mix it before it is used to measure the primary
primary coolant RTDs simply because this plant has a digital instru- coolant temperature. The RTDs in bypass loops therefore must be
mentation and control (I&C) system with a complete analog backup fast to compensate for the time delay caused when the water is
system and therefore needs twice as many sensors as most other diverted from the primary coolant pipe to the location where its
PWR plants. temperature is measured.
The number of RTDs used per PWR plant and the total number of In some PWR plants, direct-immersion RTDs are used in primary
PWR plants in operation worldwide allows us to estimate an inven- coolant pipes as opposed to bypass loops. Fig. 3 shows photographs
tory of fewer than 10,000 nuclear safety-related RTDs, including
spares, worldwide. Considering that the average life of these RTDs
is about 20 years, fewer than 1000 new RTDs per year are needed
in any given year, underscoring the small size of the nuclear-grade
RTD market. For that reason, the cost of nuclear safety-related RTDs
is generally very high compared with their commercial counter-
parts, and only a handful of (mostly smaller) manufacturers are in
the business of producing them. Large companies cannot normally
justify the overhead, liability, and stringent quality assurance (QA)
required to supply equipment to nuclear power plants. Table 1 lists
some of the suppliers of nuclear-grade RTDs.

3. Calibration and response time

In addition to excellent reliability and accident survivability,

nuclear safety-related RTDs are expected to have good calibration
and fast dynamic response time, as these characteristics are impor-
tant to plant safety and economy. Fig. 1 shows a simplified diagram
of a primary coolant loop for a PWR plant. In principle, the reac- Fig. 2. Illustration of RTD response to a step change in temperature in the reactor.
3134 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141

Fig. 4. Photograph and x-ray of Rosemount Model 177HW RTD.

manufacturers have entered the market, and Rosemount has grad-

ually reduced its presence. Rosemount RTDs are nevertheless still
used in some PWR plants and have provided excellent service.
The experience of the nuclear industry with nuclear-grade RTDs
Fig. 3. Photograph and x-rays of direct-immersion Rosemount Model 177GY RTDs.
from other manufacturers is also very good. However, as Rose-
mount reduced its market presence, problems arose in the 1980s
and X-rays of two direct-immersion Rosemount Model 177GY RTDs as new manufacturers of nuclear-grade RTDs entered the market
installed directly in the primary coolant pipes of PWR plants by (Hashemian, 2005). Some examples of these problems are:
Babcock and Wilcox (B&W). Fig. 4 shows the thermowell-mounted
counterpart of this RTD, which is the Rosemount Model 177HW,
also used predominately in B&W plants. 1. Dynamic response problems
2. Failure of extension leads
3. Low-insulation resistance
4. Problems with nuclear-grade RTDs
4. Premature failure
5. Wrong calibration tables
Nuclear-grade RTDs, like their commercial-grade counterparts,
6. Loose or bad connections
can suffer from a wide range of problems affecting both their
7. Large EMF effects
accuracy and their response time. These include calibration drift,
8. Open element
response-time degradation, reduced insulation resistance, erratic
9. Thinning of platinum wire
output, and wiring problems, among others. These problems occur
10. Lead-wire imbalance
less often in nuclear-grade RTDs than in commercial-grade RTDs
11. Seeping of chemicals from connection head into thermowell
because the former are of much higher quality. A test in the
12. Cracking of the thermowell
late 1980s of nearly 100 nuclear-grade RTDs against comparable
13. Erroneous indication
commercial-grade RTDs under both normal and harsh conditions
showed that nuclear-grade RTDs are generally twice as resilient and
immune from performance problems as commercial-grade RTDs Although these problems eventually became less frequent as
(NRC, 1990). new manufacturers gained experience with design, development,
In the early 1970s, at the height of nuclear power plant devel- and testing of nuclear-grade RTDs, they still occur. Nuclear plant
opment in the United States, almost all nuclear safety-related RTDs personnel must have the training and tools to detect and mitigate
for PWR plants were supplied by Rosemount. Over the years, other them.
 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141 3135

Table 2 4.4. Premature failure

Examples of problems encountered with response time of nuclear plant RTDs.

Plant Date of problem Response time (s) Another consequence of the emergence of new RTD manu-
facturers in the early 1980s was an unusually high failure rate
Expected Measured Manufacturer
after installation. In fact, a batch of nuclear-grade RTDs from
A 1978 5.4 21 X
one of the recognized manufacturers experienced a failure rate
B 1984 4.5 37 Y
C 1988 3.6 12 Z of about 50 percent early in their life. Although the failure
rate of new nuclear-grade RTDs has been low; nuclear plant
The measured values are from in situ response-time testing performed using the
LCSR method while the plant was operating.
personnel need to be aware that “infant mortality” can be a
source of component failure for RTDs and thermocouples (NRC,
1990).
4.1. Dynamic response problems
4.5. Wrong calibration tables
The dynamic response of an RTD is obtained by measuring
its reaction to a step change in temperature outside the sensing Nuclear plant RTDs are supplied with calibration tables gener-
element (Wu et al., 1979). Table 2 provides three examples of RTD ated by the manufacturer. A few cases of significant temperature
response-time failures in nuclear power plants. The RTDs involved errors have been traced to improper factory calibration. In one case,
were from three different manufacturers, and these problems a computer-aided calibration of several RTDs had not allowed the
occurred in three different plants. Almost all cases involving RTDs to come to equilibrium with the bath before measurements
dynamic response problems are caused by problems at the point were made. As a result, the RTDs indicated lower than the actual
where the RTD and its thermowell interface, at the sensing tip of temperatures. Cases have also occurred in which the calibration
the assembly, specifically, dirty RTDs, dirty thermowells, residue charts of different batches of RTDs were interchanged, leaving a
left from using thermal coupling compounds in the thermowell, nuclear plant with a batch of RTDs but calibration charts belonging
and dimensional tolerance issues involving the RTD and/or to a different batch.
thermowell.
The best method for detecting dynamic response problems is 4.6. Loose or bad connections
to use an in situ response-time measurement technique called the
loop current step response test (see Section 6.2.1). There are a number of transition points in an RTD circuit from
the field to the instrument cabinets in the control room area. Along
this path are terminal blocks, weld/solder joints, or splices where
4.2. Failure of extension leads
loose or bad connections have frequently been found. Licensee
Event Reports (LERs) of the Nuclear Regulatory Commission and
In a typical RTD, an outer sheath and internal insulation hold in
the Nuclear Plant Reliability Data System (NPRDS) of the nuclear
place both the sensing element and extension lead wires that carry
industry contain multiple examples of loose or bad connections
the signal of the RTD from the sensing element to the transmitter.
relating to RTDs. Of the 318 RTD failures reported in NPRDS over
As new manufacturers of nuclear-grade RTDs were emerging in
a period of about 10 years, for example, loose or bad connec-
the early 1980s, a number of RTD failures were caused by failure of
tions were the sixth-leading cause of RTD failure (accounting
the extension leads. Typically, the reason was that defective silver
for almost 10% of all failures). (Defective circuits, open circuits,
soldering had been used to attach the RTD leads to the extension
normal and abnormal wear, miscalibration, and short/grounded
wires protruding from the sensor.
failures were, in descending order, the more common causes of RTD
failure.)
4.3. Low-insulation resistance
4.7. Large EMF errors
Insulation resistance (IR) is the electrical resistance between
any extension lead that exits the RTD and earth ground. As the EMF, which stands for electromotive force, is a voltage sig-
temperature increases, the insulation resistance decreases. A high nal that may develop in an RTD circuit if an RTD is designed
and stable insulation resistance is important for the accuracy of an poorly or contains dissimilar metals. EMF interferes with resistance
RTD. Most manufacturers accept that industrial RTDs must have measurement and can cause as much as 0.5 ◦ C error depend-
an insulation resistance of at least 100 M at room temperature ing on the RTD, the application, and the temperature in which
(20 ◦ C) when measured with an applied voltage of 100 VDC. Most it is used. If EMF occurs, the resistance of the RTD will depend
nuclear-grade RTDs readily meet this requirement, and their IR on the measurement polarity. That is, if the resistance is mea-
often reaches the giga-ohm range or higher. However, if moisture sured with one polarity, then the result will be slightly different
enters the RTD, the IR value can drop to as low as a few kilo-ohms. than when the resistance measurement is repeated with reverse
Often, even a very large drop in the IR is not apparent unless IR is polarity.
measured. Table 3 shows the results of a laboratory experiment that
To avoid this problem, before RTDs are installed in a plant involved six RTDs from two manufacturers of nuclear-grade RTDs.
they should be tested to ensure sufficient IR. Insulation resistance These RTDs were placed in an oil bath along with a standard RTD
is measured using an instrument called a megohm meter that that was used to measure the bath temperature. The output of each
applies 50–100 V DC across the insulation between any RTD wire RTD was measured with normal and reversed polarity. At the same
and the sheath. Insulation resistance measurements are often dif- time, the open-circuit voltage at the RTD output was measured. For
ficult to make if there is much moisture in the RTD because the Manufacturer A, the three RTDs showed 80 ␮V of EMF voltage, and
megohm meter will not remain stable enough to make a reliable the temperature indication of the RTDs depended on measurement
measurement. To overcome this problem, most procedures give polarity. For Manufacturer B, there was no EMF effect and almost
a specific duration for the measurement. Minor insulation resis- no difference between the temperatures indicated by each RTD in
tance problems can be corrected by heating to drive the moisture the normal or reverse polarity. In other words, the EMF effect, if
out. present, will cause temperature error. The error changes with time
3136 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141

Table 3
Example of EMF problems with nuclear plant RTDs.

RTD I.D. Bath temp. (◦ C) EMF (␮V)

Standard RTD Normal polarity Reversed polarity

Manufacturer A
A-1 285.33 285.59 285.45 80
A-2 293.59 293.83 293.66 80
A-3 300.36 300.62 300.41 80

Manufacturer B
B-1 285.33 285.28 285.28 0
B-2 293.59 293.56 293.58 0
B-3 300.36 300.33 300.31 0

Fig. 7. Erratic behavior preceding the failure of a primary coolant RTD at a PWR
plant.

of a nuclear-grade RTD element, and Fig. 6 shows an electron

microscope photo of the platinum-sensing element of another
nuclear-grade RTD that has failed open. Usually, the weak points
where RTD elements fail are in weld points and places where the
element is bent.
RTD failures due to open elements are sometimes preceded
by erratic behavior whereby the RTD indication experiences large
swings, spikes, and random shift. Fig. 7 shows online monitoring
Fig. 5. Electron microscope photo of a platinum element in a nuclear-grade RTD. data of four hot-leg RTDs in a PWR plant (NRC, 1998). One of the four
RTDs exhibits erratic behavior. A month or so after this observation,
and temperature and is more pronounced at high temperatures and this RTD failed open and was replaced. Interestingly, this behavior
with shallow immersion depths. was not seen by the plant operators. In fact, during daily channel
To check for EMF, the RTD resistance is measured, the leads checks, the bad RTD continued to pass as its indication agreed well
are reversed, and the measurement repeated. If the results are dif- with the other three RTDs and met the acceptance criteria of the
ferent, thermoelectric effects are present in the RTD. When EMF plant.
occurs, the true temperature can still be obtained by averaging
the results of the two measurements. In precision thermometry, 4.9. Thinning of platinum wire
resistance measurements are made using AC bridges as opposed
to DC bridges. This is because an AC bridge cancels any EMF effect As noted, the sensing elements of industrial and nuclear-grade
in the circuit, yielding the true resistance of the RTD. In essence, RTDs are almost always made from fine platinum wire because plat-
an AC bridge works as if it measures the RTD resistance with inum has superior properties. Platinum wire is more linear than
forward and reverse polarities and displays the average of the metals like copper and nickel, and it has a much wider temperature
two. range. The sensing elements of nuclear-grade RTDs have experi-
To eliminate EMF errors, the extension wires in high quality enced corrosive thinning caused by the chemicals that were used
RTDs are made of platinum to minimize thermoelectric effects. to clean the elements when they or the RTD were manufactured.
This causes the cross section area of the sensing wire to decrease
4.8. Open element and its resistance to increase. Thinning of the RTD element can also
result from the chemical interaction between the element and the
The platinum element in RTDs is very fragile and can crack or RTD insulation material.
open as a result of vibration, stress, and interaction with other
material in the RTD. Fig. 5 shows an electron microscope photo 4.10. Lead-wire imbalance

Imbalanced lead wires are a potential problem in three-wire

RTDs that are connected to three-wire Wheatstone bridges for mea-
suring temperature. The two wires from across the RTD element
that run from the sensing element to the resistance measuring
equipment must have equal resistances.
Fig. 8 shows three-wire and four-wire RTD configurations. In a
four-wire arrangement, the lead-wire resistances are completely
compensated, while in a three-wire arrangement, the resistance of
wire 3 (R3) must be equal to the resistance of wire 1 (R1) or wire
2 (R2), depending on which wire is used as the common wire in
the three-wire bridge. If the resistance of wire 3 (R3) is not equal
to the resistance of one of the remaining two wires, then an error
will arise in temperature measurement. This lead-wire imbalance
Fig. 6. Electron microscope photo of an open platinum-sensing element in a error can cause about 0.10 ◦ C error, on average, at a temperature of
nuclear-grade RTD. about 300 ◦ C.
 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141 3137

4.13. Erroneous indication

Nuclear-grade RTDs have been found to suffer significant indi-

cation problems, the causes of which can be quite varied. Among
the worst RTD indication problems observed by the author in U.S.
nuclear power plants are a 4 ◦ C calibration shift within two years
for one RTD, a 2.7 ◦ C difference between two elements of a dual
RTD, a 3.3 ◦ C indication error caused by dirty RTD contacts, a 1.1 ◦ C
error caused by low-insulation resistance, and a 0.6 ◦ C indication
error resulting from a wire-shielding problem.

5. Problems with core-exit thermocouples

PWR plants typically have between 50 and 60 core-exit ther-

mocouples. Based on the author’s informal assessment of these
thermocouples in nearly 50 nuclear power plants, between 10 and
20 percent of PWR core-exit thermocouples fail within the first
20 years of plant operation. These failures take the form of large
calibration shifts (e.g., 10–30 ◦ C errors at 300 ◦ C), erratic and noisy
output, or saturated output. Some core-exit thermocouples develop
cable problems while the thermocouple is still intact. Plants have
been known to replace a core-exit thermocouple and later discover
that the problem was not in the thermocouple but in the thermo-
couple extension cables, connectors, or elsewhere in the circuit. To
avoid these problems, cable testing should be performed before a
thermocouple is replaced to distinguish cable/connector problems
from thermocouple problems.
Thermocouples can also be accidentally reverse-connected,
meaning that the positive and negative thermocouple wires may
be crossed during the installation or wiring of the thermocouple.
In such cases, at room temperature the thermocouple indication
could appear to be normal, but as the temperature is increased,
the thermocouple will show a negative reading. In one U.S. nuclear
fuel facility (see NRC Information Notice IN-96-33), engineers
attempted to correct for a reverse-connected thermocouple by
reversing its extension leads at the indicator. Originally, the ther-
mocouple appeared to read correctly at room temperature, but
its indication became negative when the process began to heat
up. Thus, the thermocouple leads were interchanged at the indi-
Fig. 8. HMH provide. cator, which made the reading positive and seemingly correct at
low temperatures. However, unbeknownst to the engineers, as the
temperature was increased, the thermocouple indication fell more
A similar problem can occur in RTDs that have dummy compen- and more below the actual process temperature to the point that at
sating leads. about 600 ◦ C, the thermocouple showed 450 ◦ C. This caused a fire
at the nuclear fuel fabrication facility.
Thermocouples, even those that are properly connected, can
4.11. Seeping of chemicals into thermowell
have good indication at room temperature but diverge from true
temperature as the plant heats up. Fig. 9 shows online monitoring
The thermowell is the protective jacket or tube that protects
results for a group of core-exit thermocouples in a PWR plant. The
well-type RTDs from the process fluid and allow them to be easily
thermocouple readings are comparable at cold shutdown, but one
replaced. The response time of well-type RTDs is extremely sensi-
deviates significantly from the others at plant operating tempera-
tive to installation in a thermowell. In some RTDs, the connection
ture.
head is filled with chemical foam to help it qualify for nuclear ser-
Thermocouples can also suffer response-time degradation as
vice. When the chemical seeps through into the RTD thermowell,
they age. Table 4 shows response-time results for core-exit ther-
as has occurred at some nuclear power plants, response time and
mocouples in four similar PWR plants with nearly identical
other problems have resulted.
thermocouples. Note that the average response time of the ther-

4.12. Cracking of thermowell Table 4

Results of trending the performance of core-exit thermocouples in PWR plants.
Another issue involving well-type RTDs is a cracked thermowell.
Plant Years in service Thermocouple response time (s)
RTDs should extend into the reactor coolant piping to a depth that
is consistent with the fluid forces on the RTD and the mechanical Average High Low
strength of the RTD/thermowell assembly. Instances have occurred A 30 2.01 2.9 0.6
where fluid forces have caused the assembly to bend and crack, B 30 1.96 2.8 0.6
increasing the potential for a Loss of Coolant Accident (LOCA) or C 10 0.97 1.5 0.5
D 10 1.10 1.5 0.6
shearing-off of the entire assembly.
3138 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141

6.1.1. Cross-calibration technique

The cross-calibration technique is a means for verifying online
that the calibration of a group of redundant sensors that measure
the same process parameters have not suffered a significant change.
Its basic principle of operation is to record the reading of redun-
dant sensors, average these readings, and calculate the deviation
of each sensor from the average of the redundant sensors, less any
outlier(s). The narrow-range RTDs are normally the most accurate
temperature sensors in a PWR plant and are therefore used to pro-
vide the reference temperature for the cross-calibration test. In
some plants, only the narrow-range RTDs are cross-calibrated; in
others, the wide-range RTDs and/or core-exit thermocouples are
also included in the cross-calibration test. Typically, the outputs
of plant sensors are sampled by the plant computer, converted
into temperature, and stored. The sampling period depends on the
plant and typically ranges from 1 to 10 s for RTDs and 10 to 60 s for
core-exit thermocouples.
Fig. 9. On-line monitoring results for a group of core-exit thermocouples. Table 6 shows results from a typical cross-calibration test in a
nuclear power plant. The data for this test was collected at isother-
Table 5 mal conditions during plant heatup at approximately 280 ◦ C. This
Potential sources of error and their estimated values in industrial temperature mea- printout, referred to as a cross-calibration run, typically includes
surements with thermocouples (for 50–500 ◦ C range). four data-collection passes for each sensor. After data for a run is
Source Abbreviation Range of error collected, the results of the four measurements for each sensor are
averaged and recorded under a column marked “Average Temp.” In
Inherent thermocouple errors (TC) ±0.5–5 ◦ C
Cold junction compensation error (CJ) ±0.1–0.5 ◦ C the example in Table 6, the average temperatures for all the narrow-
A/D errors (AD) ±0.1–0.2 ◦ C range RTDs are averaged, and the result is subtracted from each
Precision errors (Noise) ±0.1–0.5 ◦ C individual average to identify the “deviation” of each sensor from
Total error = (TC) + (CJ) the average of the narrow-range RTDs. The “Deviation” column is
+ (A/D) + (Noise) + (Other) the result of the cross-calibration run. These results are referred to
Other = extension wire tolerance or as the preliminary results of the cross-calibration test. Typically, the
mismatch, grounding, etc.
data is analyzed further to establish the final results of the cross-
calibration test and to quantify the uncertainty of the deviation
results. The final results must include the necessary corrections for
mocouples after 10 years of service is about 1 s and after 20 years
any significant instability and nonuniformity in the plant temper-
of service is about 2 s, a 100 percent increase. Although this increase
ature when the cross-calibration data was collected (NRC, 1997).
is not an operational or safety issue for the plant, it indicates that as
Also, test engineers must establish a means to verify that there is
thermocouples age, their response time increases. This and other
no common drift. This can be accomplished by, for example, cali-
similar observations have motivated some nuclear power plants
brating one of the RTDs that was involved in the cross-calibration
to perform response time testing on thermocouples as a means
test. Alternatively, test engineers can rely on historical drift data for
of aging management (Hashemian et al., 1990). In particular, it is
the type of RTDs that are cross-calibrated. An R&D effort in 1990
important to establish the baseline response time of thermocou-
found that most nuclear-grade RTDs showed random drift patterns
ples when the plant is new or the thermocouples are first installed
rather than systematic drift patterns (NRC, 1990).
and then repeat the measurements periodically (e.g., once every
Any narrow-range RTD that deviates from the average by more
five years) to determine if there is degradation from the nominal
than a predetermined criterion (e.g., 0.3 ◦ C) is excluded from the
performance.
average. The criteria are different in different plants and usually
In general, thermocouples are not as accurate as RTDs. This is
depend on plant accuracy requirements for the primary coolant
partly because thermocouples are not normally calibrated individ-
temperature. Table 7 provides the criteria for RTD cross-calibration
ually. Rather, thermocouple wires or a representative sample in a
used in seven nuclear plants and indicates both the deviation at
large batch of thermocouples are calibrated, and that calibration is
which a narrow-range RTD is excluded from the average and the
used for all thermocouples in the batch. Table 5 shows estimated
number of temperatures at which cross-calibration data is col-
temperature measurement accuracies of industrial thermocouples
lected.
as a result of four error source types.
The cross-calibration process results in a preliminary analysis of
the raw cross-calibration data and provides the preliminary results
6. Methods for detecting problems with RTDs and core-exit of the cross-calibration test. For more accurate results, the raw
thermocouples data are corrected for any significant temperature fluctuations that
might have been present when data was collected. This is referred
Appropriate online testing can avoid the problems that affect the to as instability correction. Also, the raw data should be corrected
accuracy and response time of nuclear-grade RTDs and core-exit for any significant temperature differences between the primary
thermocouples. coolant loops or the hot-leg and cold-leg temperatures. This is
referred to as nonuniformity correction.
6.1. Calibrating RTDs and core-exit thermocouples Although simple, the cross-calibration process involves numer-
ous calculations. As such, it is prudent to automate the process to
In most PWR plants, the accuracy of RTDs such as the primary facilitate the data-collection and data analysis tasks. Automated
coolant RTDs is verified periodically to ensure that any unaccept- software can read and plot the raw data, perform the analysis
able drift or deviation is identified and corrected. For this purpose, including corrections for plant temperature instability and nonuni-
the cross-calibration technique is used. formity, and print the results.
 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141 3139

Table 6
Results of a typical cross-calibration run.

RTD Temperature (◦ C) Average temp. (◦ C) Dev. temp. (◦ C)

Pass 1 Pass 2 Pass 3 Pass 4

Narrow-range RTDs
1 280.3278 280.3274 280.3087 280.2956 280.315 −0.063
2 280.4091 280.3942 280.3853 280.3797 280.392 0.014
3 280.3616 280.3621 280.3426 280.3305 280.349 −0.029
4 280.3660 280.3655 280.344 280.3347 280.353 −0.026
5 280.4729 280.4599 280.4608 280.4571 280.463 0.084
6 280.3664 280.3329 280.3427 280.3274 280.342 −0.036
7 280.3392 280.3276 280.3230 280.3178 280.327 −0.051
8 280.4709 280.4574 280.4504 280.4453 280.456 0.078
9 280.3308 280.3312 280.3047 280.3029 280.317 −0.061
10 280.4369 280.4355 280.4081 280.4118 280.423 0.045
11 280.3765 280.3584 280.3477 280.3440 280.357 −0.022
12 280.4593 280.4584 280.4375 280.4296 280.446 0.068

Wide-range RTDs
13 280.0733 280.0612 280.0538 280.0352 280.056 −0.322
14 280.6964 280.6871 280.6741 280.6602 280.679 0.301
15 280.3290 280.3281 280.3067 280.3039 280.317 −0.061
16 280.4881 280.4899 280.4704 280.4686 280.479 0.101

Core-exit thermocouples
17 280.6723 280.6674 280.6261 280.6431 280.652 0.274
18 280.6301 280.6082 280.5928 280.6025 280.608 0.230
19 280.7786 280.7802 280.7640 280.7526 280.769 0.390
20 280.5482 280.5660 280.5474 280.5474 280.552 0.174
21 280.8232 280.8110 280.7940 280.7907 280.805 0.426
22 280.8978 280.8588 280.8483 280.8248 280.857 0.479
23 280.7680 280.7607 280.7445 280.7380 280.753 0.374
24 281.1411 281.1394 281.1394 281.1086 281.132 0.754
25 280.8037 280.7940 280.7510 280.7656 280.779 0.400

Average temperature indicated by the narrow-range RTDs = 280.378 ◦ C.

Note: The deviation column in this table is equal to the average temperature of each sensor minus the average temperature of the narrow-range RTDs.

The cross-calibration technique has been validated for both 6.1.2. Cross-calibrating core-exit thermocouples
RTDs and thermocouples in a laboratory setting involving an oil Thermocouples are also cross-calibrated in some plants. In PWR
bath, an SPRT, and a set of cross-calibration test equipment (NRC, plants, at isothermal conditions, the primary coolant RTDs and
1990). core-exit thermocouples are at essentially the same temperature.
For that reason, the core-exit thermocouples are cross-calibrated
against the primary coolant RTDs because the latter are usually
Table 7
RTD cross-calibration criteria in various PWRs. more accurate than thermocouples. Cross-calibrating thermo-
couples involves subtracting the indication of each core-exit
Nuclear plant Temperature points Outlier criteria (◦ C) Remarks
thermocouple from the average temperature indicated by the
1 1 0.17 1 narrow-range RTDs. The results are then tabulated, as shown in
2 1 0.17 1 Table 8.
3 4 0.11 2
4 2 0.30 3
5 4 0.27 4 6.2. Response time testing of RTDs and core-exit thermocouples
6 1 0.17 5
7 4 0.11 6
Testing the response times of RTDs and thermocouples in situ
Footnotes: (1) Plant stability criteria for RTD cross-calibration are typically about (while the plant is operating) enables nuclear power plants to mea-
±0.15 to ±0.3 ◦ C. (2) Temperature points: number of temperatures at which cross-
sure the “in-service” response time of the sensor to meet technical
calibration data is collected.
Remarks:
specification requirements, regulatory regulations, or both. Plants
(1) Cross-calibration data is taken for any number of plateaus. However, only the can also verify that plant sensors bottom out in their thermowells
data for 292 ◦ C is used to meet acceptance criteria and adjust the temperature trans- and test for air gaps, dirt, and foreign objects in the thermowell.
mitters, as needed. In situ response time testing also provides for predictive main-
(2) Data can be taken at a constant heatup rate. On 16 RTDs, data is taken as follows:
tenance, incipient failure detection, and aging management and
RTD number 1–16, reverse current 16–1, reverse current 1–16, etc. This presumably
corrects for both the ramping temperature and for EMF effects (reversing the cur- enables plants to establish objective schedules for replacing sensors
rent). Data is taken around 95, 170, 230 and 292 ◦ C. (IAEA, 2000). Finally, response time testing enables plants to distin-
(3) Two plateaus: 170 and 292 ◦ C. For deviations greater than 0.17 ◦ C, the deviations guish between sensor problems and cable or connector problems
at 170 and 292 ◦ C are used to determine the error offset and the slope and to apply
and to diagnose sensor or process anomalies.
the corrections to the temperature transmitter.
(4) Data is taken at four temperatures on 16 RTDs sequentially, 1–16, 16–1, etc. The
In almost all U.S.-made PWR plants, testing RTD response times
heatup rate is also measured. is required and is performed on one or more RTD channels once
(5) The 16 RTDs in this plant are tested one channel (four RTDs) at a time. Data is every operating cycle (NRC, 1997). For thermocouples, response-
taken for 25 min at 5-min intervals. This is repeated for all four channels. The plant time testing is not mandatory, but some plants perform the tests
stability requirement for the tests is 0.17 ◦ C (i.e., the temperature cannot change by
for one or more of the reasons just noted.
more than 0.17 ◦ C from the beginning to the end of any test run).
(6) Tests are performed at 120, 180, 230, and 275 ◦ C. Data is taken on 16 RTDs, 1–16, Until 1977, testing the response times of temperature sensors
16–1, 1–16, and 16–1. The four calibration points are used to determine a zero and in nuclear power plants was almost always performed using the
a slope for the correction to temperature transmitters. plunge test (Kerlin et al., 1980). This is defined as the time it takes
3140 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141

Table 8
Results of thermocouple cross-calibration.

Fig. 10. Comparison of Raw and Transformed LCSR Data.

response-time standards as RTDs. We will therefore discuss the

LCSR method for response-time testing RTDs before discussing the
noise analysis technique for response-time testing of thermocou-
ples.

6.2.1. The LCSR response-time testing method for RTDs

The LCSR method is the most commonly used technique for test-
the sensor output to achieve 63.2 percent of its final value after a ing the response time of RTDs because it can yield the response time
step change in temperature is impressed on its surface. In nuclear of a sensor with such accuracy. The LCSR test involves injecting
reactors, however, plunge testing is inconvenient because the sen- the RTD with an electrical current applied at the end of the sensor
sor must be removed from the reactor coolant piping and taken to a extension leads. The current causes Joule heating in the sensor and
laboratory for testing. Nuclear reactor service conditions of 150 bar results in a temperature transient inside the sensor. The time plot,
(2,250 psi) and 300 ◦ C (572 ◦ F) are difficult to reproduce in the labo- of either the heating while the current is applied or the cooling
ratory. Therefore, all laboratory tests are performed at much milder after the current is discontinued, is recorded. From this plot, the
conditions, and the results are extrapolated to service conditions. response time of the RTD is obtained by means of the LCSR trans-
The combination of manipulating the sensor and extrapolating the formation, which is a way of converting the LCSR transient from
results to service conditions leads to significant errors in the mea- the internal heating of a sensor to yield the dynamic response of
surement of sensor response times, sometimes by as much as a the sensor to a step change outside the sensor. Fig. 10 compares a
factor of three. raw LCSR transient for an RTD, the converted LCSR transient, and
Because of these problems the nuclear industry therefore devel- the response of the RTD to a step change in temperature outside
oped the following three methods: the loop current step response the RTD, obtained using a plunge test. This data is derived from a
(LCSR) test, the self-heating test, and the noise analysis technique. laboratory test in which the RTD is placed in a rotating tank of water
In the LCSR method, the sensing element is heated by an electric at room temperature, ambient pressure, and a flow rate of 1 m s−1 .
current, and the temperature transient in the element is recorded. The LCSR test accounts for all the effects of installation and pro-
From this transient, the response time of the sensor to changes in cess conditions on response time and thereby provides the actual
external temperature is identified (Hashemian, 2005). In the self- “in-service” response time of the sensor. It has been validated for
heating test, as in the LCSR test, the sensing element is also heated several RTD models of the type used in nuclear power plants. Gen-
by an electric current. After the RTD output settles, the steady-state erally, the accuracy of in situ response-time measurements made
increase in RTD resistance is measured as a function of the elec- with the LCSR test is ±10 percent. This is provided that: (1) the LCSR
tric power applied to the sensor. The result is referred to as the method has been validated for the sensor under test; (2) the LCSR
self-heating index (SHI). Any significant change in SHI indicates a data is clean and smooth; and (3) optimum analysis parameters are
change in RTD response time. Therefore, the SHI can be tracked to used to process the data. Furthermore, the conditions of the plant
determine the degradation of RTD response time. In the noise anal- must be suitable for the LCSR test. This means that the plant tem-
ysis technique, the natural fluctuations (noise) that normally occur perature must be stable and undergo no significant fluctuations or
at the sensor output during plant operation are recorded and ana- drift during the LCSR tests.
lyzed to determine the response time of the sensor (Hashemian et
al., 1988). 6.2.2. The noise analysis response-time testing method for
The LCSR method is the most commonly used technique for thermocouples
RTDs. The self-heating test is applicable only to RTDs and is used Thermocouples are not routinely tested for response time in
to identify only changes in response time; it does not measure nuclear power plants because response time is not as important
response time itself. The noise analysis technique is useful in appli- for nuclear power plant thermocouples as it is for RTDs. When the
cations where accuracy is not critical, for example, where a large response time of thermocouples is tested, nuclear power plants
margin exists between the nominal response time of the sensor prefer to use the noise analysis technique, because the LCSR test
and the required response time of the plant. It is also the primary for thermocouples, although more accurate, requires high heating
method used in the nuclear power industry for testing the response currents (>500 mA), which could degrade the seal of the thermo-
time of thermocouples, which are not held to the same stringent couples and/or insulation material.
 H.M. Hashemian, J. Jiang / Nuclear Engineering and Design 239 (2009) 3132–3141 3141

The noise analysis technique is based on the fact that the cial, industrial-grade temperature sensors. Nevertheless, they are
output of all process sensors in nuclear power plants normally susceptible to a range of potential problems, affecting both their
contains fluctuations due to random flux, random heat transfer, accuracy and their response time. The industry has developed a
turbulence, vibration, and other mechanical and thermal hydraulic range of methods for detecting these problems while these sen-
phenomenon. These fluctuations (noise) can be extracted from the sors are online in operating nuclear plants. The cross-calibration
sensor output and analyzed to yield the response time of the sensor. method (for accuracy) and the loop current step response test
The test involves three steps – data acquisition, data qualification, (for response time) are among the most effective and the most
and data analysis. widely used RTDs. Potential sensor problems notwithstanding,
The normal output of a thermocouple is a DC signal on which the nuclear-grade RTDs and thermocouples remain durable instru-
process noise (AC signal) is superimposed. During the data acqui- ments. Unless specific problems arise that require more frequent
sition step, the noise is extracted from the thermocouple output testing or plants have reason to suspect that their sensors are defi-
by removing the DC component of the signal and amplifying the cient in design, fabrication, or installation, the current industry
AC component. This is accomplished simply by using commercial practice of verifying accuracy and response time at least once every
signal-conditioning equipment, including amplifiers, filters, and fuel cycle is reasonable. Barring these problems, the typical nuclear-
other components. The AC signal is then digitized using a high grade RTDs can last 10–40 years depending on the conditions of
sampling rate (e.g., 1 kHz) and stored for subsequent analysis. its use.
During data qualification, the raw data must first be thor-
oughly scanned and screened before any analysis can resume. This References
is normally accomplished by using data qualification algorithms
embedded in software, which check for the stationary and linearity Hashemian, H.M., 2005. Sensor Performance and Reliability. ISA – Instrumentation,
Systems, and Automation Society, Research Triangle Park, North Carolina.
of the raw data and look for other abnormalities. Hashemian, H.M., Petersen, K.M., Mitchell, D.W., Hashemian, M., Beverly, D.D., 1990.
During data analysis, noise data is analyzed in the frequency In Situ Response Time Testing of Thermocouples, 29. ISA Transactions, p. 97+,
domain and/or time domain. For frequency domain analysis, the no. 4.
Hashemian, H.M., Thie, J.A., Upadhyaya, B.R., Holbert, K.E., 1988. Sensor Response
Power Spectral Density (PSD) of the noise signal is first obtained Time Monitoring Using Noise Analysis. Prog. Nucl. Energy 21, 583–592.
through a Fast Fourier Transform algorithm or its equivalent. Next, International Atomic Energy Agency, 2000. Management of Ageing of I&C Equipment
a mathematical model of the thermocouple is fit to the PSD, from in Nuclear Power Plants. IAEA Publication TECDOC-II47, Vienna, Austria.
Kerlin, T.W., Miller, L.F., Hashemian, H.M., Poore, W.P., Shorska, M., Upadhyaya, B.R.,
which the response time of the system is calculated. The PSDs of
Cormault, P., Jacquot, J.P., 1980. Temperature sensor response and character-
nuclear plant thermocouples have various shapes, depending on ization. EPRI Report No. NP-I486, Electric Power Research Institute, Palo Alto,
the plant, the thermocouple installation and service, the process California.
conditions, and other effects. For time domain analysis, the noise Plumb, H.H., Schooley, J.F., 1982. Temperature, its Measurement and Control in Sci-
ence and Industry. Office of Ordnance Research, Instrument Society of America,
data is processed using a univariate Autoregression (AR) model- National Bureau of Standards, American Institute of Physics.
ing program. This provides the impulse response (i.e., response to U.S. Nuclear Regulatory Commission (NRC), 1990. Aging of Nuclear Plant Resistance
a narrow pressure pulse) and the step response, from which the Temperature Detectors. NUREG/CR-5560, Washington, DC.
U.S. Nuclear Regulatory Commission (NRC), 1997. Standard review plan for the
response time of the system is calculated. Typically, the noise data review of safety analysis reports for nuclear power plants. NUREG-0800, Wash-
is analyzed in both the frequency domain and time domain, and the ington, DC.
results are averaged to obtain the response time of the system. U.S. Nuclear Regulatory Commission (NRC), 1998. Advanced instrumentation and
maintenance technologies for nuclear power plants. NUREG/CR-5501, Wash-
ington, DC.
7. Conclusion Wu, S.M., Hsu, M.C., Chow, M.C., 1979. Determination of time constants of reactor
pressure and temperature sensors: the dynamic data system method. Nucl. Sci.
Eng. 72 (1), 84–96.
Nuclear-grade RTDs and core-exit thermocouples are made to
higher specifications from higher quality parts than are commer-