Accurate Infrared Temperature

Measurement in Reformers
An overview of the methodology for measuring tube skin temperatures in reformers using infrared
techniques is presented. The major sources of error are examined and advice is given on how best to
avoid or correct for these errors. An uncertainty analysis is also given, providing the means for
reformer operators to assess the risks associated with decisions based on the results of temperature
measurements. Maintenance and calibration of infrared thermometers is discussed.

Peter Saunders
Measurement Standards Laboratory
Industrial Research Limited
PO Box 31-310, Lower Hutt, New Zealand

Introduction measurements. There are many physical effects that

lead to errors in infrared thermometry, so that typically

C
atalyst tube skin temperature is a critical pa- a thermometer reading will be too high by between
rameter in the operation of primary reform- 20 °C and 100 °C depending on the conditions within
ers, governing both the plant efficiency and the reformer and the geometry of the reformer. A good
the remnant life of the tubes. Continual op- understanding of the effects leading to these errors is
eration at a temperature just 20 °C higher than nominal essential for obtaining accurate and reliable temperature
will halve the lifetime of a tube. Thus, accurate tem- measurements.
perature determinations are required in order to manage This paper describes the effects that lead to the
the risks associated with the operation of a reformer. largest errors, namely reflection of radiation from ob-
In the harsh environment created by flame exhausts, jects surrounding the target tube, unknown emissivity
contact thermometers such as thermocouples are unreli- of the tube, and absorption and emission effects due to
able and have limited lifetimes. Establishing good ther- the flue gas inside the reformer. Advice is given on the
mal contact with the tube without perturbing its tem- best measurement practice and how to minimize the er-
perature is also difficult. In most cases infrared rors. This includes choosing the most appropriate val-
thermometry (also known as radiation thermometry or ues for the operating wavelength and instrumental
pyrometry) is the only viable temperature measurement emissivity setting on the thermometer for given operat-
technique. Infrared thermometers operate remotely by ing conditions and reformer geometry. In most cases,
sensing the infrared radiation emitted by an object and corrections must be applied to the thermometer read-
inferring its temperature via the well-known Planck ings. The paper shows how these corrections are ap-
blackbody radiation law. plied and emphasizes the importance of uncertainty
Despite the widespread use of infrared thermometry analysis, which provides a measure of risk associated
in reformers, there is a general lack of good advice in with decisions based on the values of the corrected
the industry on how to obtain accurate temperature

2005 343 AMMONIA TECHNICAL MANUAL

temperatures. Advice on thermometer maintenance and providing the greatest sensitivity. Most industrial ther-
calibration is also given. mometers operate in the infrared part of the spectrum.
However, as discussed in subsequent sections, there are
other considerations that determine the best choice of
Infrared Thermometry Principles wavelength.
108
To understand the nature of the errors associated
with infrared temperature measurement, it is necessary 107
5800 K

to have a basic knowledge of the way in which infrared

thermometers operate. An understanding of some of the 106

fundamental optical properties of materials and gases

105
and their interaction with electromagnetic radiation is 2800 K
1300 K
also required. 104
Underlying infrared thermometry is the concept of
blackbody radiation. This is the spectrum of electro- 103
700 K
magnetic radiation emitted by a perfect emitter (called a 2
10
blackbody), and was described mathematically in 1900
by the German physicist Max Planck. Planck’s law [1] 101
states that the spectral radiance of a blackbody, 300 K

Lb(λ, T), is a function only of wavelength, λ, and tem- 100

perature, T:
10-1
0.1 1 10 100
−1 Wavelength / µm
c1 ⎡ ⎛ c ⎞ ⎤
Lb (λ , T ) = 5 ⎢
exp ⎜ 2 ⎟ − 1⎥ , (1) Figure 1. Blackbody spectral radiance curves, as given by
λ ⎣ ⎝ λT ⎠ ⎦ Planck’s law.

where T has the units of kelvin, and c1 and c2 are con-

Emissivity
stants given by c1 = 1.191044 × 10–16 W m2 and
c2 = 0.014388 m K. A blackbody is an idealized object. At a given tem-
Planck’s law, which paved the way to the develop- perature all real objects emit only a fraction of the ra-
ment of quantum mechanics, is plotted in Figure 1 for a diation of a blackbody, where this fraction is given the
range of temperatures. The curves demonstrate that, for name emissivity, with the symbol ε. By definition, the
a given temperature, there is a maximum value of the emissivity of an object is a number between 0 and 1,
spectral radiance and that this maximum moves to with 1 indicating a perfect emitter or blackbody, and 0 a
shorter wavelengths as the temperature increases. For perfect reflector.
temperatures below about 800 K almost all of the radia- The emissivity of a surface must be known in order
tion emitted by a blackbody is in the infrared part of the to determine the correct temperature from Planck’s law,
spectrum, but as the temperature increases this radiation since the signal detected by the thermometer will be
shifts into the visible and is perceived by the human eye less than that given by equation (1) by the factor ε. Any
as a glowing object. error in the value of the emissivity leads to an error in
Infrared thermometers are designed to measure the measured temperature. Lack of knowledge of the
spectral radiance and infer temperature using Planck’s correct emissivity is one of the largest sources of error
law. This is usually done electronically with a micro- in infrared thermometry.
processor or some sort of linearisation circuitry. The Most industrial infrared thermometers have an ad-
majority of infrared thermometers measure spectral ra- justment or knob for entering the value of the target’s
diance over a small wavelength range and are referred emissivity. This is called the instrumental emissivity, εI,
to as spectral-band thermometers. Figure 1 shows that and setting it causes the thermometer’s circuitry to di-
the choice of operating wavelength is primarily dictated vide the measured signal by εI before conversion to
by the temperature of the object. Generally, a wave- temperature. If the instrumental emissivity is set cor-
length on the short-wavelength side of the spectral radi- rectly then the measured signal becomes equivalent to
ance peak is optimal since it is here that the spectral ra- that from a blackbody and the calculated temperature is
diance increases most rapidly with temperature,

AMMONIA TECHNICAL MANUAL 344 2005

equal to the true temperature. We shall see below, that 300

in reformer applications the emissivity of the target is 250

almost never the correct value to use for the instrumen-
tal emissivity setting. 200

150
The emissivity of a material is usually wavelength
dependent [2]. Reformer tubes have an emissivity 100
around 0.85 at 1 μm and 0.82 at 3.9 μm (two operating
50
wavelengths commonly used in reformers). Emissivity
also depends quite strongly on the nature of the surface. 0
In general, rough and oxidized surfaces tend to have a 0.65 0.70 0.75 0.80 0.85 0.90

higher emissivity than smooth or polished surfaces. Emissivity

New reformer tubes, bearing the impressions of sand Figure 3. As for Figure 2 but the measurements where taken at
casting, can have emissivities as high as 0.9. As they tube locations where some direct flame impingement occurred.
age, they tend to shed oxide and gain a smooth surface The mean value is 0.82 and the standard deviation is 0.04.
finish resulting in a lower emissivity. Direct flame im- high values corresponding to new tubes. The measure-
pingement on tubes can accelerate this process. ments in Figure 3 were taken about 0.5 m below the
Figures 2 and 3 show in situ measurements [3] of natural gas burners, and demonstrate a much wider
the emissivity at 1 μm of all the tubes in a top-fired emissivity variation with a lower mean value.
methanol reformer, taken during a scheduled shutdown.
In Figure 2 the measurement position was well down Reflected Radiation
the tubes away from any direct flame impingement. The
distribution of emissivity is relatively narrow, with the Objects with an emissivity less than 1 are partial re-
few flectors of radiation. For any opaque object the sum of
the emissivity and reflectivity equals 1, so a reformer
tube, for example, with an emissivity of 0.85 has a re-
300
flectivity of 0.15. Thus, 15% of the radiation impinging
250 on the tube from its surroundings will be reflected and
will contribute to the signal recorded by an infrared
200 thermometer aimed at the tube. This reflected radiation
causes an error in the temperature determined using
150
equation (1).
100 In reformers the surroundings are usually not uni-
form in temperature, so the reflected radiation consists
50
of many different components. There will generally be
0
components from refractory walls, floor, ceiling,
0.83 0.84 0.85 0.86 0.87 0.88 0.89 0.90 flames, and other tubes. No matter what the surround-
Emissivity ings consist of, it is always possible to define an effec-
Figure 2. Histogram of measurements of tube emissivity at 1 μm tive background temperature, Tw, such that it appears
in a methanol reformer taken at tube locations well away from that the target tube is completely surrounded by a
direct flame impingement. The mean value is 0.849 and the stan- blackbody at temperature Tw [4].
dard deviation is 0.008.
Recognizing this fact, we can write down a meas-
urement equation for the infrared thermometer:

Lm = ε Lb (λ , T ) + (1 − ε ) Lb (λ , Tw ) , (2)

where Lm is the measured radiance and ε is the emissiv-

ity of the target. The first term on the right-hand side
corresponds to the radiation emitted from the target and
the second to the radiation reflected from the target. (It
is assumed that the reflectivity of the tube surface is dif-

2005 345 AMMONIA TECHNICAL MANUAL

fuse in nature.) The thermometer will determine its
measured temperature, Tm, using equation (1) based on 100
90
the measured value of Lm and the instrumental emissiv-
80
ity setting, εI:
70
60 3.9 μm
Lb (λ , Tm ) = Lm ε I . (3) 50
40
A reflection error occurs when Tm differs from the true 30

temperature, T. 20 1 μm

There are three basic measurement strategies [5] for 10

dealing with reflection errors, each corresponding to a 0

100 200 300 400 500 600 700 800 900 1000 1100
different regime of relative target and background tem- Effective background temperature / °C
peratures. These are discussed below. Figure 4. Reflection error for strategy 1 with εI = ε, for a re-
former tube at a true temperature of 900 °C. The shaded region
Measurement Strategy 1 (Tw << T) corresponds to typical values of Tw in a reformer.
Measurement Strategy 2 (Tw ≈ T)
The first strategy should be used when the effective
background temperature is much lower than the target The second measurement strategy should be used
temperature. In this regime the second term on the when the effective background temperature is approxi-
right-hand side of equation (2) (the reflection term) is mately equal to the target temperature. Indeed, if we set
negligible compared to the first term (the emission Tw = T in equation (2) then the measured radiance be-
term), and in this case the target is said to be freely ra- comes simply Lm = Lb(λ, T). Thus, the measured radi-
diating. Setting the instrumental emissivity on the ther- ance is equal to the radiance of a blackbody at the target
mometer to the emissivity of the target then gives the temperature, so the target appears to be a blackbody.
true temperature reading. Accordingly, the reflection error will be zero if the in-
This strategy is often the advice given by infrared strumental emissivity on the thermometer is set to
thermometer manufacturers, but it should be used with εI = 1. This strategy can be usefully applied in some re-
care. If there are any hot objects surrounding the target, formers or parts of some reformers.
as are always found in reformers, a reflection error will
occur. This is illustrated in Figure 4, where the reflec-
tion error is plotted against the effective background 60

temperature for a reformer tube at 900 °C. (Note that 40

the error is the difference between the thermometer
reading and the true temperature, so a positive error in- 20

dicates that the reading is too high.) The error is plotted

0
for two thermometers operating at typical wavelengths 1 μm
used for targets at this temperature, namely 1 μm and -20

3.9 μm. 3.9 μm

-40
For this strategy, the reflection error is smaller for
Equation (4)
shorter-wavelength thermometers and, for the target at -60
900 °C, is negligible when the effective background 500 600 700 800 900 1000 1100
Effective background temperature / °C
temperature is below 600 °C at 1 μm and below 200 °C Figure 5. Reflection error for strategy 2 with εI = 1, for a re-
at 3.9 μm. However, for typical conditions in a re- former tube at a true temperature of 900 °C.
former, where the effective background temperature is
around 950 °C to 1050 °C (the shaded region in Fig-
ure 4), the reflection error is unacceptably large. It is The reflection error for the εI = 1 strategy is shown
interesting to note that with Tw = 1000 °C, for example, in Figure 5 under the same conditions as Figure 4 ex-
the difference in readings between the two thermome- cept that the instrumental emissivity has been set to 1.
ters is 40 °C and both readings are in significant error. When Tw = T the reflection error is zero for all operat-
This very situation was the cause of much confusion ing wavelengths. When Tw is within 50 °C or so of T,
one known conscientious reformer operator!

AMMONIA TECHNICAL MANUAL 346 2005

the error is almost independent of wavelength and is inside the reformer and the geometry of the reformer
given approximately by itself.

Tm − T = (1 − ε )(Tw − T ) . (4)
1050°C
This equation is also plotted on Figure 5.
When Tw > T, longer-wavelength thermometers 900°C
have the smaller reflection error using this strategy, and
as the wavelength increases the error approaches the
limit given by equation (4). However, the advantage of
using longer and longer wavelengths does not continue
indefinitely due to the fact that the uncertainty in the
measured temperature due to the uncertainty in the
measured radiance increases proportionally with wave- Figure 6. Each tube in a reformer has a different effective back-
length: ground temperature according to its position in the reformer and the
geometry of the reformer itself.
λT 2 uLm
uTm = , (5) 20
c2 Lm 18
16

where the u quantities represent uncertainties. For ex- 14

ample, if the spectral radiance, Lm, can be measured to 12

an accuracy of 5% then, for a target at 900 °C, the un- 10

certainty in the measured temperature at 1 μm is about 8

6
5 °C and at 3.9 μm it is almost 20 °C. Thus, there is a
4
compromise to be made between reflection error and
2
uncertainty, leading to the existence of an optimum
0
wavelength for a set of reformer conditions. 810 820 830 840 850 860 870
Thermometer reading / °C
Measurement Strategy 3 (Tw >> T) Figure 7. Reflection errors for a set of tube measurements in a
reformer furnace. The effective background temperature varied
As Figure 5 shows, when using the εI = 1 strategy, between 867 °C and 892 °C.
if the effective background temperature is much higher
than the target temperature, then the reflection error
again becomes unacceptably large. This is most often This can be readily seen by comparing the two meas-
the case in reformers. The reflection error can only be urement locations for a multi-row reformer shown in
made lower by raising the instrumental emissivity to a Figure 6.
value greater than 1 [4]. Most infrared thermometers do For the measurement in the upper location, the lar-
not have the facility for doing this. ger part of the reflected radiation originates from the
Thus, the third measurement strategy is to apply hot refractory wall adjacent to the tube row, whereas at
corrections for the reflection error. As well as measur- lower location the reflected radiation comes predomi-
ing the target tube, this strategy requires carrying out nantly from the adjacent tube row, which is colder than
sufficient measurements of the background temperature the refractory wall. Thus, the value of Tw is higher in
distribution in order to calculate a value for the effec- the upper location than the lower. In other words, the
tive background temperature, Tw. Then, with an esti- upper measurement is subject to a larger reflection error
mate of the emissivity of the tube, the true temperature than the lower. Even for tubes in a given row, the value
can be calculated directly by rearranging equation (2). of Tw changes from tube to tube due to the different
It is important that all measurements are carried with furnace geometry as seen by each tube.
the instrumental emissivity set to 1. The value of the effective background temperature
The determination of Tw is usually the most diffi- can be thought of as a weighted average radiance tem-
cult part of this strategy. Each target tube will be ex- perature of all the surrounding objects [4] (the radiance
posed to a different value of Tw according to its position temperature is simply the thermometer reading with
εI = 1). The weighting factors are equal to the geometri-

2005 347 AMMONIA TECHNICAL MANUAL

cal view factors between the surrounding object and the which, when applied to equation (2), gives for the first
target tube [6]. The larger and closer the object is to the uncertainty component:
tube, the larger the weighting factor.
Figure 7 shows the calculated reflection errors for a T 2 Lb (λ , Tm ) 1
set of reformer tube temperature measurements. The ef- u1 = uT . (6)
Tm2 Lb (λ , T ) ε m
fective background temperature varied throughout the
reformer from 867 °C to 892 °C. The true tube tempera-
tures are obtained by subtracting the reflection error The value of uTm is typically a few degrees Celsius.
from the thermometer reading. The second component arises from an uncertainty
A trend can be observed in Figure 7: namely, the in the value of the emissivity used to perform the calcu-
higher the thermometer reading the lower the reflection lation to determine T from equation (2):
error. This is a general conclusion when the effective
background temperature is higher than the tube tem- λT 2 ⎡ Lb (λ , Tw ) ⎤ uε
u2 = − 1⎥ . (7)
c2 ⎢⎣ Lb (λ , T )
perature. A direct consequence of this is that the tubes
in a furnace always appear more uniform in temperature ⎦ε
12
than they actually are [4]. In other words, the set of cor-
rected temperatures always has a wider spread of values 10
than the set of thermometer readings.
8

6
Measurement Uncertainty
4
Uncertainty analysis is an important aspect of tem-
perature measurement and has often been overlooked in 2
the literature. Without a determination of the uncer-
0
tainty in a measured or calculated temperature, a re- 810 820 830 840 850 860 870
former operator has no way to assess the risks associ- Thermometer reading / °C
ated with a decision based on the result of that Figure 8. Uncertainty components in the corrected temperature, us-
measurement. Such decisions could involve increasing ing strategy 3, for an effective background temperature of 880 °C.
or decreasing the firing in the reformer or leaving it the
way it is, or they may be whether or not to replace tubes
based on the outcome of a remnant life analysis, which With only limited knowledge of the surface condition
uses temperature as an input quantity. of the tubes in a reformer, the value of uε is often no
Equation (5) gave a formula for the temperature un- better than about 0.05. However, this can be reduced
certainty when measuring an effective blackbody target significantly if the emissivity of a good set of tube sam-
(strategy 2), and was derived directly from Planck’s ples has been measured, as for the reformer whose data
law. The uncertainty analysis for strategy 3, given in is shown in Figures 2 and 3.
this section, is more complicated because in order to The third uncertainty component arises from an un-
apply corrections for reflection errors using equa- certainty in the measurement of the effective back-
tion (2), we additionally need to know the values of the ground temperature:
emissivity and the effective background temperature.
Any uncertainty in the knowledge or measurement of T 2 Lb (λ , Tw ) (1 − ε )
u3 = uTw . (8)
these values will lead to additional uncertainty in the Tw2 Lb (λ , T ) ε
estimation of the true temperature.
Thus, there are three components that contribute to
The value of uTw is the most difficult to assess since it is
the total uncertainty. The first arises from an uncer-
tainty in the measured target temperature due to the in- a function of the furnace geometry as well as the radi-
trinsic accuracy of the thermometer, stray radiation ance temperatures of all the surrounding objects. How-
scattering off imperfections or dust in or on the ther- ever, it is possible to obtain a reasonable upper limit by
mometer’s optical system, and fluctuation in the fur- carrying out a sensitivity analysis on the geometric
nace conditions. Reference [7] outlines the internation- view factors.
ally accepted method for determining uncertainties,

AMMONIA TECHNICAL MANUAL 348 2005

 The total uncertainty in the estimated true tempera- However, gas fired furnaces do not usually suffer from
ture is obtained by adding the three uncertainty compo- problems with scattering.
nents in quadrature [7]: Absorption and emission of radiation becomes a
problem when the spectral band of the infrared ther-
utotal = ( u12 + u22 + u32 ) . mometer overlaps the wavelengths corresponding to a-
12
(9)
tomic or molecular absorption lines or bands in the
components of the flue gas. For reformers, the most im-
Figure 8 shows the total uncertainty given by equa- portant components are carbon dioxide and water vapor
tion (9), along with each of the individual components, [8]. Most thermometers are designed to avoid the major
for data typical of that shown in Figure 7. The parame- absorption lines, but there is often some residual over-
ters used to generate these curves are: uTm = 3 °C, lap.
uε = 0.05, uTw = 10 °C, ε = 0.85, Tw = 880 °C, and As the radiation propagates through the gas, some
of it is absorbed and additional radiation is emitted
λ = 1 μm. The curves show that for low measured tem- from the gas itself. The error in the thermometer read-
peratures the total uncertainty is dominated by the un- ing depends on the net effect (i.e. the difference be-
certainty in the emissivity. This is because the lower- tween the amount of radiation absorbed and the amount
temperature tubes have larger reflection errors, and emitted), which in turn depends on the difference be-
knowledge of the emissivity is required to correct for tween the target tube radiance temperature and the tem-
the error. The higher-temperature tubes more closely perature of the flue gas. If the gas is hotter than the tube
resemble blackbody conditions (T ≈ Tw) so knowledge then there will be a net emission of radiation and the
of the tube emissivity is not so important. For these thermometer will read high. If the gas is at a lower tem-
tubes, the uncertainty in the measured temperature of perature absorption will dominate and the thermometer
the tube dominates the total uncertainty. Finally, despite will read low. When the two temperatures are the same
the relatively large uncertainty in the determination of the absorbed component will exactly match the emitted
the effective background temperature, its contribution component with no effect on the thermometer reading.
to the total uncertainty is smaller than that of the meas- The effect of absorption and emission also depends
ured tube temperature. on the concentrations of the absorbing gas species and
Analyzing each component of uncertainty in this the viewing path length through the flue gas. The error
manner allows us to identify which quantities need to becomes larger in magnitude as both of these quantities
be concentrated on if improved accuracy is required. increase. The nature of the error tends to be insidious
For a given set of operating conditions different com- and difficult to diagnose. Fortunately, however, the
ponents may dominate, so the analysis should be re- path length dependence provides the means to deter-
peated for each furnace design and temperature distri- mine whether or not the error exists. If it is possible to
bution. The uncertainties also differ with operating view the same tube over two different path lengths
wavelength. There is generally an optimum wavelength (from different sight doors), then obtaining different
that minimizes the total uncertainty [5]. At wavelengths temperature readings is indicative of an absorption and
shorter than the optimum, the uncertainties increase ex- emission error.
ponentially, while at longer wavelengths they increase It is even possible to obtain the correct tube reading
almost linearly. based on these two measurements and knowledge of the
two measurement distances [8]:
Flue Gas Absorption and Emission
d 2 L1 − d1 L2
Lcorrected = , (10)
The flue gas inside a reformer is potentially a large d 2 − d1
source of error in an infrared thermometer’s reading.
The radiation from the target must propagate some dis- where d1 and d2 are the distances corresponding to the
tance through the flue gas to the thermometer. Any in- two radiance measurements L1 and L2, respectively.
teraction between the gas and the radiation in the form Equation (1) can be used to convert radiance to tem-
of absorption and emission will alter the radiation de- perature and vice versa. The corrected radiance will still
tected by the thermometer. Scattering of radiation by contain the reflection error, so Lcorrected must then be
particulate matter, e.g. soot, can also be a problem. used as the value of Lm in equation (2) to correct for re-
flections.

2005 349 AMMONIA TECHNICAL MANUAL

 ured values highlights the insidious nature of absorption
950
and emission errors. The tuning of the burners was
930
based upon the measured temperatures only. Achieving
910
a fairly uniform set of measured temperatures unwit-
890
tingly had the undesirable result of creating a large cold
870
area in the middle of the reformer.
850
830
810 Measured Maintenance and Calibration
790 Corrected f or absorption and emission
770 Corrected f or ref lections In order to get continued good performance from an
750
0 5 10 15 20 25 30 35
infrared thermometer it is important that it is main-
Tube number tained with care and treated essentially as a very expen-
Figure 9. Measurements affected by absorption and emission by sive camera. Severe knocks can misalign the optical
flue gas and by reflected radiation. Measurements were taken system leading to an ill-defined field of view or, in the
from the end of a tube row so that tube number correlates with worst case, an effective change in the calibration con-
path length, up to about 12 m.
stants. Dust, grease, finger prints, or scratches on the
Figure 9 illustrates the effects of flue gas absorption
front lens cause scattering of radiation into and out of
and emission. The figure shows measurements of the
the field of view, resulting in reading errors. Scratching
temperatures (square symbols) of a row of tubes taken
tends to be irreversible, but dust can be removed from
from the end of a tube row (similar to the upper posi-
the lens by cleaning periodically with high-quality lens
tion in Figure 6). Tube number correlates with distance
tissue or an airbush. If absolutely necessary, grease can
so that Tube 1 was measured over a path length of
be removed using ethanol and lens tissue.
about 1 m and Tube 34 over about 12 m. The measure-
Over time, the responsivity of the detector and the
ments were taken with an infrared thermometer operat-
transmission characteristics of the wavelength-selecting
ing at 1 μm with a bandwidth of about 0.1 μm. This filter slowly change. These changes can be corrected
spectral region is known to contain water vapor absorp- for by recalibration of the thermometer.
tion lines, and since the reformer was fired with natural Calibration of an infrared thermometer is a techni-
gas, it is expected that the flue gas contains a relatively cally demanding exercise requiring specialized equip-
high concentration of water vapor. ment and procedures. In the first instance, a thermome-
Measurements of all the tubes shown in Figure 9 ter should be sent to the manufacturer or an accredited
were also possible from sight doors in the wall parallel calibration laboratory for recalibration. Preferably, the
to the tube row, allowing measurements to be made of thermometer should be calibrated on an annual basis, or
each tube over two different distances. The second every two years once confidence has been gained in its
curve in Figure 9 shows the corrected radiance tempera- long-term performance through observed consistency in
ture as determined by equation (10). The error in the its calibration history.
radiance temperatures increases almost linearly with A calibration laboratory should provide a set of cor-
path length, with errors of about 60 °C for the most dis- rections to readings over the temperature range of the
tant tube. This is most likely an extreme case, particular thermometer along with a statement of uncertainty in
to the high water vapor concentration combined with the corrected readings. In some cases, they (particularly
the relatively wide bandwidth of the 1 μm thermometer. the manufacturers) may modify the internal calibration
A thermometer with a narrower bandwidth or a differ- constants to correct for the reading errors, in which case
ent operating wavelength, such as 3.9 μm, would ex- it is important that they provide some before- and after-
hibit less error. In any case, it is important to establish calibration readings to enable you to track the history of
the presence or lack of absorption and emission errors performance.
when using a particular thermometer by measuring at Many reformer operators prefer to carry out in-
least one tube over sufficiently different path lengths. If house thermometer calibration. This requires the use of
the error is large then a different thermometer should be a blackbody cavity, which is a small temperature-
chosen. controlled furnace with an opening configured in such a
Finally, the third curve in Figure 9 shows the esti- way that the radiation emerging from the opening very
mated true temperatures after correcting for reflection closely follows Planck’s law (equation (1)). The cali-
errors. Comparison of this curve with the original meas- bration is performed [9] by setting the blackbody cavity

AMMONIA TECHNICAL MANUAL 350 2005

to a range of temperatures and comparing the infrared when measuring the temperature of real non-blackbody
thermometer’s readings with the true temperature of the targets such as reformer tubes.
blackbody as measured by either a reference contact
thermometer (a thermocouple or a platinum resistance
thermometer) or a separate reference infrared ther- References
mometer. Whichever type of reference thermometer is
[1] D P DeWitt, G D Nutter, Theory and Practice of
used, it must be independently calibrated by an accred-
Radiation Thermometry, (New York: John Wiley
ited calibration laboratory.
& Sons), 1988.
The best results are usually obtained using a second
[2] Y S Touloukian, D P DeWitt, Thermophysical
infrared thermometer as the reference because residual
Properties of Matter, Volume 7, Thermal Radia-
temperature gradients and non-uniformities in the
tive Properties. Metallic Elements and Alloys,
blackbody cavity lead to additional errors when a con-
(New York: IFI/Plenum), 1970.
tact thermometer is used. The reference infrared ther-
[3] P Saunders, A B Trotter, H Edgar, D M J Coch-
mometer should only be used as a laboratory reference,
rane, “In situ measurement of catalyst tube emis-
and with care its calibration interval can be extended
sivity by means of a portable solid integrating
out to a maximum of five years.
sphere reflectometer”, Measurement Science and
Technology, 12, 622–626, 2001.
Conclusion [4] P Saunders, “Reflection errors in industrial ra-
diation thermometry”, in Proceedings of
Infrared thermometers provide the most reliable TEMPMEKO ’99, Seventh International Sympo-
means of measuring tube skin temperatures in reform- sium on Temperature and Thermal Measure-
ers. However, their use is generally not straightforward ments in Industry and Science, edited by J F
due to the necessity in most cases of applying correc- Dubbeldam, M J de Groot, IMEKO/NMi Van
tions to the thermometer readings for reflected radiation Swinden Laboratorium, Delft, 631–636, 1999.
and flue gas absorption and emission. The methodology [5] P Saunders, D R White, “A theory of reflections
outlined in this paper shows how to apply these correc- for traceable radiation thermometry”, Metrolo-
tions and how to determine the accuracy in the calcu- gia, 32, 1–10, 1995.
lated true temperatures. [6] F P Incropera, D P DeWitt, Fundamentals of
The uncertainties in the corrected temperatures are Heat and Mass Transfer, Fourth Edition, (New
minimized when the reflection errors are minimized. In York: John Wiley & Sons), 1996.
all cases in reformer thermometry, this is achieved by [7] Guide to the Expression of Uncertainty in Meas-
setting the instrumental emissivity to 1 for all meas- urement, (Geneva: International Organization for
urements. Operating wavelength is also a consideration Standardization), 1993.
in minimizing the temperature uncertainty. An optimum [8] P Saunders, “Absorption and emission effects on
wavelength exists for a set of furnace conditions, which radiation thermometry measurements in reformer
is essentially a compromise between achieving a low furnaces”, in Temperature: Its Measurement and
reflection error at long wavelengths and a low meas- Control in Science and Industry, Vol. 7, edited
urement uncertainty at short wavelengths. Care must by D. C. Ripple et al., AIP Conference Proceed-
also be exercised in selecting the operating wavelength ings, Melville, New York, 825–830, 2003.
to avoid spectral regions where flue gases have strong [9] J V Nicholas, D R White, “Traceable Tempera-
absorption lines. Gas flames are mostly transparent at tures: An Introduction to Temperature Measure-
3.9 μm, so this wavelength should be selected if is nec- ment and Calibration”, second edition, John
essary to view directly through gas flames. Wiley & Sons, Chichester, 2001.
Calibration is an essential aspect of the mainte-
nance of an infrared thermometer. It should be stressed
that, because infrared thermometers are designed to
measure a variety of targets in a variety of surround-
ings, calibration can only establish the link between de-
tected signal and temperature for a blackbody target.
The methodology described in this paper must be used

2005 351 AMMONIA TECHNICAL MANUAL