A new experimental apparatus for emissivity measurements of steel and the

application of multi-wavelength thermometry to continuous casting billets

Peng Wang, Zhenwei Hu, Zhi Xie, and Ming Yan

Citation: Review of Scientific Instruments 89, 054903 (2018); doi: 10.1063/1.5007225

View online: https://doi.org/10.1063/1.5007225
View Table of Contents: http://aip.scitation.org/toc/rsi/89/5
Published by the American Institute of Physics

Articles you may be interested in

Design techniques for a stable operation of cryogenic field-programmable gate arrays
Review of Scientific Instruments 89, 014703 (2018); 10.1063/1.5004484

Development of ball surface acoustic wave trace moisture analyzer using burst waveform undersampling
circuit
Review of Scientific Instruments 89, 055006 (2018); 10.1063/1.4993928

Polydyne displacement interferometer using frequency-modulated light

Review of Scientific Instruments 89, 055007 (2018); 10.1063/1.5026569

Inertial piezoelectric linear motor driven by a single-phase harmonic wave with automatic clamping
mechanism
Review of Scientific Instruments 89, 055008 (2018); 10.1063/1.5025732

Immersion-scanning-tunneling-microscope for long-term variable-temperature experiments at liquid-solid

interfaces
Review of Scientific Instruments 89, 053707 (2018); 10.1063/1.5030407

Simultaneous reflectometry and interferometry for measuring thin-film thickness and curvature
Review of Scientific Instruments 89, 055117 (2018); 10.1063/1.5021704
 REVIEW OF SCIENTIFIC INSTRUMENTS 89, 054903 (2018)

A new experimental apparatus for emissivity measurements

of steel and the application of multi-wavelength thermometry
to continuous casting billets
Peng Wang,1,a) Zhenwei Hu,2 Zhi Xie,2 and Ming Yan1
1 College of Mechanical Engineering, Shenyang University of Technology, Shenyang 110870, China
2 College of Information Science and Engineering, Northeastern University, Shenyang 110819, China
(Received 1 October 2017; accepted 30 April 2018; published online 22 May 2018)
An experimental apparatus has been designed for measuring the emissivity of a steel surface in both
vacuum and oxidation atmosphere. The sample is heated with the method of electromagnetic induction
in order to ensure the temperature uniformity. The radiance emitted from a sample is measured using
a fiber-optic Fourier transform infrared spectrometer. Using this unique apparatus, we investigated the
spectral (2-6 µm) and directional (0◦ -86◦ ) emissivity of stainless steel 304 with different degrees of
surface oxidation at temperatures ranging from 800 to 1100 ◦ C. The experimental results show that
the emissivity increases slightly with increasing temperature, which accords with the Hagen-Rubens
relation. The emissivity increases rapidly at the initial stage of oxidation, but gradually reaches to
a constant value after 20 min. In addition, the directional emissivity has a maximum value at the
measuring angle of about 75◦ . The maximum uncertainty of emissivity is only 3.0% over all the
measuring ranges, indicating that this experimental apparatus has a high reliability. In order to mea-
sure the surface temperature of casting billets based on multi-wavelength thermometry, the bivariate
emissivity function with the two variables, wavelength and temperature, is determined. Temperature
measurement results based on our technique are compared with those from common dual-wavelength
radiation thermometry. Our approach reduces the measured temperature fluctuation from ±20.7 ◦ C
to ±2.8 ◦ C and reflects the temperature variation with the changes of production parameters in real
time. Published by AIP Publishing. https://doi.org/10.1063/1.5007225

I. INTRODUCTION Multi-wavelength thermometry utilizes the measured infor-

mation at three or more wavelengths to obtain the target tem-
The surface temperature of casting billets is an important
perature that can reduce the influence of emissivity. However,
parameter in the process of continuous casting. For instance,
in order to obtain the real temperature, this method needs to
the temperature is a key feedback parameter for adjusting the
assume a mathematical model between the wavelength and
casting speed and the secondary cooling water distribution.1
emissivity. If the target spectral emissivity model and the hypo-
Stainless steel 304 (SS 304) is now one of the most widely
thetical model do not match well, the error of the measured
used stainless steel in the world by right of its corrosion resis-
temperature will be even larger.6 For these reasons, investiga-
tance, heat resistance, and excellent processing performance.
tions about the emissivity of SS 304 under different conditions
The annual output of SS 304 exceeds 30 × 106 tons. Therefore,
and applying the emissivity results to measure the temperature
accurate measurement of the temperature of casting billets
of casting billets are very necessary.
in real time can greatly improve product quality and offer
At present, several methods for measuring steel emissiv-
significant cost savings for the corporations.
ity such as the calorimetry method, reflection method, and
Nowadays, noncontact measurement methods are used for
radiance comparison method have been put forward. For exam-
measuring the surface temperature of casting billets, such as
ple, Wen7 measured the emissivity of some steel samples at
single-wavelength and dual-wavelength radiation thermome-
700, 800, and 900 K. The effects of temperature, wavelength,
try.2–4 For these technologies, we need to know the accurate
alloy composition, and heating time on emissivity were also
relationship between the emissivity and the temperature at
investigated. Furukawa and Iuchi8,9 studied the emissivity
the measuring wavelength. However, the emissivity of the
modeling of steel during the growth of the oxide film and devel-
measured billet varies with the wavelength, angle, and rough-
oped an experimental apparatus for measuring the emissivity
ness.5 More importantly, it also changes significantly with
of steel in oxidizing and reducing surroundings. del Campo
the surface oxidization film and surface temperature. Dur-
et al.10,11 studied the normal spectral emissivity of Armco
ing the production, casting billets are exposed to air at a high
iron as a function of temperature under a controlled atmo-
temperature and covered with oxide films. Therefore, it may
sphere. For the above research, however, most sample heating
cause large temperature measurement errors for the single-
methods employed heat conduction through the heating plate
wavelength and dual-wavelength radiation thermometry.
or the resistive wire below the sample. By this way, there are
obvious temperature gradients along the axis of the sample and
a) Authorto whom correspondence should be addressed: wangp816@ the heating time is long. Besides, few people established suit-
gmail.com. Tel.: +86 024 25496035. able emissivity models and applied them to the temperature

0034-6748/2018/89(5)/054903/9/$30.00 89, 054903-1 Published by AIP Publishing.

054903-2 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

measurement of continuous casting billets directly. In the field Lsur,ref (λ, Tsur ) = ε sur Lsur,b (λ, Tsur )ρ, (5)
of multi-wavelength thermometry, some people indeed stud-
ied and assumed some relationships between emissivity and where T sur is the temperature of the surroundings; ε sur is the
wavelength.12 However, in the process of continuous casting, emissivity of the surroundings which are treated as a gray
the temperature span of the billet surface can reach 250 ◦ C. and Lambertian emitter; and ρ is the directional-hemispherical
Therefore, the influence of temperature on emissivity must reflectivity of sample; according to Kirchhoff’s law, ρ = 1 ε.
also be considered. In this work, the emissivity of SS 304 L 0 is the background signal of the FTIR itself, and L 0 can be
at different wavelengths, temperatures, and oxidations were neglected because the experiment and calibration are carried
measured using the unique experimental apparatus, and then out in the same indoor environment. Therefore, Eq. (3) can be
the bivariate emissivity function with the two variables, wave- expressed as
length and temperature, was established. Moreover, we also Lmeas (λ, T ) = τ εLb (λ, T ) + ε sur Lsur,b (λ, Tsur )(1 − ε) . (6)
 
employed the multi-wavelength thermometry to measure the
temperature of continuous casting billets based on the experi- And then, the emissivity of sample can be obtained by the
mental results. The results of field application showed that the following equation:
measured temperature fluctuation was reduced obviously. Lmeas (λ, T )/τ − ε sur Lsur,b (λ, Tsur )
Hereinafter, the principle of emissivity measurement is ε= . (7)
Lb (λ, T ) − ε sur Lsur,b (λ, Tsur )
described. Then, the experimental apparatus is detailed and the
measurement results of emissivity of SS 304 steel are presented
and discussed in Secs. III and IV, respectively. In Sec. V, the
method of multi-wavelength thermometry and the application III. STRUCTURE OF APPARATUS
results for continuous casting billets are given. At last, the
concluding remarks are made in Sec. VI. The required ranges of the measuring parameters are listed
in Table I. The wavelength range takes into account of most
of the radiance emitted from the measured sample at the mea-
II. PRINCIPLE OF EMISSIVITY MEASUREMENT suring temperatures. The temperature range covers all of the
billet temperatures at the measuring position of the continuous
Directional spectral emissivity is defined as the ratio of the casting production line. The oxidation time ensures a com-
radiance emitted from a sample surface to that emitted from plete process of the sample from non-oxidation to complete
a blackbody at the same wavelength λ and temperature T. oxidation.
That is, Based on the measuring principle and the parameter
Ls (λ, T ) ranges described earlier, an experimental apparatus for mea-
ε(λ, T ) = , (1)
Lb (λ, T ) suring the emissivity of SS 304 is designed and consists of four
where L s (λ, T ) and L b (λ, T ) are the spectral radiances emitted parts: a sample chamber, atmosphere control system, heating
from the sample surface and blackbody, respectively. L b (λ, T ) system, and data acquisition system. The schematic diagram
can be obtained from the following equation: and photograph of apparatus are shown in Fig. 1.
The sample chamber, made of stainless steel, can remain
c1L
Lb = , (2) sealed during the experimental process, to complete the emis-
λ5 exp(c2 /λT ) − 1

sivity measurement under different environmental conditions.
where c1L is the first radiation constant for spectral radiance, In order to reduce background radiance, the chamber wall is
c1L = 2hc2 = 1.191 × 108 (W µm4 /m2 sr). c2 is the second designed with a hollow structure that can be full of circulat-
radiation constant, c2 = hc/K B = 1.439 × 104 (µm K). K B is ing cooling water. The temperature of the inner wall of the
the Boltzmann constant, h is the Planck constant, and c is the chamber, which is measured by two platinum resistance sen-
speed of light in a vacuum. T (K) is the real temperature of the sors (Pt 100), is 37.2-40.8 ◦ C when the sample temperature is
sample surface. heated from 800 to 1100 ◦ C. In addition, the radiance emitted
The spectral radiance L s (λ, T ) emitted from the sam- from the sample, after being reflected by the chamber wall and
ple can be obtained by a spectrometer. However, the spectral again by the sample, can be also received by the spectrome-
radiance measured by the spectrometer is expressed by the ter. In order to reduce this effect, the inner walls are painted
following equation: with aluminosilicate coating which has a high absorptance
f g (0.95 ± 0.004) within the spectral range of the measurements,
Lmeas (λ, T ) = τ Ls (λ, T ) + Lsur,ref (λ, Tsur ) + L0 , (3) and coating also has a high temperature resistance.
where τ is the transmittance of the observation window.
L meas (λ, T ) is the spectral radiance measured by the spec- TABLE I. Measuring parameter ranges of the apparatus.
trometer, L s (λ, T ) is the spectral radiance emitted from the
sample, Parameter Range

Ls (λ, T ) = εLb (λ, T ), (4) Wavelength (µm) 2-6

Temperature (◦ C) 800-1100
L sur,ref (λ, T sur ) is the spectral radiance emitted from the sur- Angle (deg) 0-86
roundings and then reflected by the sample. L sur,ref (λ, T sur ) Oxidation time (min) 0-30
can be expressed by the following equation:
054903-3 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

FIG. 1. (a) Schematic diagram. (b)

Photograph of apparatus.

In order to measure the directional emissivity at any angle In practice, industrial billets are always covered by oxide
from 0◦ to nearly 90◦ , a unique circular track is designed. As films. In order to make the oxide film on the sample surface
can be seen in Figs. 1 and 2, the sample is placed on a graphite grow, oxygen can be injected into the chamber using the oxy-
pallet that is in the center of the circular track, and the measur- gen cylinder. In this way, the changes of the emissivity with
ing radius is 350 mm. The focusing probe of the spectrometer different oxidation time can be also studied.
is installed in a special fixture that can be fixed on any position In general, most researchers adopt resistive wires or heat-
of the track. The observation window is made of CaF2 , and its ing electrodes to heat samples. Those heaters are usually
transmission wavelength range is between 1 and 9 µm. The placed at the bottom of the samples, and there are three prob-
wide band allows the radiance emitted from the sample pass lems. The first one is causing obvious temperature gradient
through the sample chamber. The window is coated to ensure along the axis of the sample when the sample temperature is
that the transmittance is 96% within the measured spectral high. The second one is that the heaters will generate much
range. In addition, moving the detector out of the chamber can radiance which can be received by the spectrometer. The third
ensure better tightness compared with rotating the sample in one is that the heating time will be long, and the sample tem-
the chamber. perature is difficult to control. Therefore, for this work, we
The atmosphere control system comprises a mechanical adopt the electromagnetic induction method to heat the sam-
vacuum pump, a molecular vacuum pump, and an oxygen ple. The schematic diagram of the heater is shown in Fig. 3.
cylinder. The emissivity of SS 304 under the vacuum con- Alternating current flowing through the induction coil will gen-
dition can be studied using the vacuum pumps. The minimum erate a high frequency electromagnetic field. Thus, the metal
pressure of the chamber can reach less than 3.2 × 10 2 Pa. sample in the magnetic field will generate a great vortex and
then generate much radiance. With this heater, the sample is
heated uniformly with all parts and the heating time is only
3 min from room temperature to 800 ◦ C. Besides, the induc-
tion coil is with a hollow structure and cooling water flows

FIG. 2. Schematic diagram of the circular track. FIG. 3. Schematic diagram of an induction heater.
054903-4 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

roughness of the finishing sample surface can be less than

30 nm. After the polishing, the sample surface was cleaned
with alcohol to remove grease and dirt. After installing the
sample and adjusting the position of the focusing probe of the
spectrometer, the pressure of the sample chamber was kept
less than 3.2 × 10 2 Pa, and then the samples were heated by
the heating system. When the sample temperature was sta-
ble at the measurement temperatures, the radiance and the
temperature of the sample were then measured by the spec-
trometer and thermocouple, respectively, and then the emis-
sivity was calculated. To study the effects of oxidation on
the emissivity, oxygen was injected into the sample cham-
ber at the measurement temperature and angle. The injection
flow rate of oxygen was 1.0 × 10 6 m3 /s. To improve the
accuracy of the measurements, each experiment was repli-
cated four times and the averages of these values are present
hereinafter.

A. Effects of temperature
FIG. 4. Schematic of the sample showing the position of the thermocouple. Emissivity measurements for SS 304 at different tempera-
tures (800-1100 ◦ C) over the wavelength range from 2 to 6 µm
inside so that it will not generate much radiance. The temper- were carried out. Figure 5 shows the difference of the spectral
ature range of the sample meets the measurement requirement normal emissivity between the four temperatures under a vac-
(800-1100 ◦ C). In addition, the sample temperature is con- uum environment. As shown in Fig. 5, the spectral emissivity
trolled using a proportional-integral-differential (PID) device of sample gradually increases with the increasing temperature.
and the temperature instability is limited to less than ±1 ◦ C. According to the Hagen–Rubens relation, emissivity increases
The data acquisition system is composed of two parts: a with the increasing resistivity, and for SS 304, the metal resis-
platinum rhodium 10-platinum thermocouple and a fiber-optic tivity indeed increases with increasing temperature. Therefore,
Fourier transform infrared spectrometer. The thermocouple is the emissivity measurements for temperatures agree with the
used to measure the real temperature of the sample. One pur- metal radiation theory.
pose is to calculate emissivity, and the other one is to serve In addition, emissivity fluctuates at two wavelength bands
as the input signal of the PID device. In order to reduce the around 2.6 and 4.3 µm, as shown in Fig. 5. The former is caused
temperature difference between the sample surface and the by the radiation absorption of water vapor and carbon dioxide
measuring point of the thermocouple, the thermocouple is in the air, and the latter is only caused by carbon dioxide.
inserted into the underside of the sample at a distance of 3 mm The measured data of the two wavelength bands should be
from the upside surface, as shown in Fig. 4. The thermocouple excluded when fitting the emissivity function because they
is a sheathed-type one, and the thermocouple wire is sealed will affect the fitting accuracy and cause measurement errors
in a metal sleeve which can provide an electromagnetic shield in multi-wavelength thermometry.
environment. Therefore, the thermocouple wire will not be In order to show the effect of the temperature more visu-
heated by the alternating EM (electromagnetic)-field. ally, the variations of emissivity with the temperature for dif-
The radiance of the sample is measured by a fiber-optic ferent wavelengths (2, 3, 4, 5, and 6 µm) are shown in Fig. 6.
Fourier transform infrared spectrometer (Arcoptix, model FT- The emissivity increases almost linearly with the increasing
MIR Rocket, Switzerland); the effective experimental wave-
lengths range from 2 to 6 µm. The spectrometer has been
calibrated by a blackbody furnace prior to conducting the
experiments to ensure correspondence between the radiance
and the output values of the spectrometer. The diameter of the
viewable area of the spectrometer can be adjusted to 2 mm
from the normal direction at a distance of 350 mm. However,
as the angle tilts, the viewable area becomes an ellipse and its
major axis is proportioned to 1/cos θ. Therefore, the maximum
measuring angle is 86◦ that depends on the size of the sample
surface (30 mm × 30 mm).

IV. RESULTS AND DISCUSSION

Prior to the experiments, the sample was grinded using
a polishing wheel and then polished with sand paper; the FIG. 5. Spectral normal emissivity at four temperatures.
054903-5 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

FIG. 6. Emissivity variations with the temperature for five wavelengths.

FIG. 8. Emissivity variations with oxidation time at 4 µm and at four
temperatures.
temperature. Besides, the increasing rate at longer wavelengths
is greater than that at shorter wavelengths.
In addition, at higher temperatures, the increasing speed of
emissivity is greater and the oxidation time is less for the emis-
B. Effects of oxidation
sivity to become constant. This phenomenon can be explained
The emissivity behaviors during the growth of the oxide by the Arrhenius equation,13
film over the wavelength range from 2 to 6 µm were mea-
k = A exp(−Ea /RT ), (8)
sured. Some emissivity characteristics at four oxidation time
and 1000 ◦ C are shown in Fig. 7. The effect of the oxide films where k is the rate constant of oxidation reaction, A is the
on the emissivity is very obvious: in this wavelength range, the pre-exponential factor, E a is the activation energy, R is the
emissivity of SS 304 varies from 0.419 to 0.482 when the steel universal gas constant, and T is the temperature. This formula
is not oxidized; however, it increases substantially to 0.637- illustrates that the oxidation rate is greater at a higher temper-
0.726 when the oxidation time is 5 min. This phenomenon ature, and thus it requires less time for the surface of SS 304
reflects the interference that is caused by multiple radiation to become completely oxidized at a higher temperature. Con-
reflections between the sample surface and the oxide film.8 sequently, the emissivity of SS 304 increases faster at higher
Furthermore, the emissivity of non-oxidized steel decreases temperatures.
with increasing wavelength, while for the oxidized steel, the
opposite trend is observed. The emissivity properties change C. Effects of angle
dramatically, which are more similar to those of nonmetals,
when steel is oxidized. Therefore, the effect of oxidation on In Fig. 9, the directional emissivity for the completely oxi-
the emissivity of SS 304 must be considered since the billets dized sample at different temperatures is shown; the measuring
are always exposed to air in the process of continuous casting. wavelength is 4 µm. The emissivity data were measured every
Figure 8 shows the emissivity results versus the oxida- 5◦ less than 50◦ and every 2.5◦ from 50◦ to the maximum
tion time at four temperatures (800, 900, 1000, 1100 ◦ C) with measuring angle (86◦ ). As Fig. 9 shows, all the emissivity
the wavelength of 4 µm. As shown in Fig. 8, the emissiv- barely changes between 0◦ and 60◦ and then increases sud-
ity increases rapidly with the oxidation time within the first denly with the increasing angle. The maximum emissivity
10 min and remains approximately constant after 20 min.

FIG. 9. Normal spectral emissivity variations with angle at 4 µm for the

FIG. 7. Spectral normal emissivity during oxidation at 1000 ◦ C. completely oxidized sample at four temperatures.
054903-6 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

TABLE II. Emissivity uncertainty of the non-oxidized sample at T = 1000 ◦ C, D. Uncertainty analysis
λ = 3 µm, and θ = 0◦ . uBi is the component of the emissivity uncertainty due
to the error of variable xi (∆xi ). The uncertainty in emissivity measurement of this equip-
ment can be expressed by the following equation:
xi ∆xi uBi v
u
t 6
B1 2.2 ◦ C 1.2 × 10 2 X
2
uB = uBi 2 , (9)
B2 3.4 ◦ C 1.9 × 10
i=1
B3 5.376 W/(sr m2 µm) 1.8 × 10 2

B4 1.7 ◦ C 9.2 × 10 3 where uBi are the uncertainty components caused by the dif-
B5 0.5 ◦ C 2.2 × 10 3 ferent measurement uncertainty: uB1 is due to the measuring
4
B6 0.004 3.2 × 10 uncertainty of the thermocouple; uB2 is due to the tempera-
ture difference between the thermocouple and sample surface;
uB3 is due to the measuring uncertainty of the spectrometer;
uB4 is due to the non-uniform temperature of the sample sur-
appears mostly at the angle about 75◦ and decreases rapidly face; uB5 is due to the measuring uncertainty of the temperature
from the maximum point to 86◦ . This conclusion is consistent of surroundings; uB6 is due to the emissivity uncertainty of
with our previous theoretical derivation14 and other’s literature surroundings. The uncertainty of emissivity can therefore be
data.10 expressed by the following equation:

s !2 !2 !2
∂ε ∂Lb (λ, T ) ∆T ∂ε ∂Lb (λ, T ) ∆Tdiff ∂ε ∆Lmeas (λ, T )
uB = · +
· √ · · √ + · √
∂Lb (λ, T ) ∂T 3 ∂Lb (λ, T ) ∂T 3 ∂Lmeas (λ, T ) 3
s !2
∂Lsur,b (λ, Tsur ) ∆Tsur 2
!2
∂ε ∂Lb (λ, T ) ∆Tnonu ∂ε ∂ε
!
∆ε sur
+ + · · √ + · · √ + · √ . (10)
∂Lb (λ, T ) ∂T 3 ∂Lsur,b (λ, Tsur ) ∂Tsur 3 ∂ε sur 3

Substituting the measurement uncertainty into Eqs. (7) and researchers put forward a number of emissivity models and
(10), the uncertainty components uBi of the non-oxidized sam- obtained satisfactory measurement accuracies in their respec-
ple are shown in Table II, for values of real temperature tive application areas. Four most common emissivity models
1000 ◦ C and a wavelength of 3 µm, respectively. The combined are listed as follows:15
uncertainty u(ε λ ) is 0.0303, as calculated by Eq. (10), indicat- n
X
ing that the emissivity results measured by this experimental ε(λ) = ai λ i , (11)
apparatus have high accuracy. i
n
X
ε(λ) = exp* ai λ i +, (12)
V. APPLICATION OF MULTI-WAVELENGTH , i -
THERMOMETRY TO CONTINUOUS CASTING BILLETS
1
A. Method of multi-wavelength emissivity ε(λ) = , (13)
1 + a0 λ 2
Multi-wavelength thermometry, a development of single-
wavelength and dual-wavelength methods, measures the target
radiance at three or more wavelengths to deduce the real tem-
perature. It is well known that a hypothetical function relat-
ing emissivity is needed to calculate the temperature. Some

TABLE III. Bivariate emissivity function models.

Emissivity models Formula

  1/2  
Theoretical radiation model ε(λ, T ) = a0 Tλ + a1 Tλ
n m
ε(λ, T ) = ai λi + bj T j
P P
Polynomial model
i=0 j=1
n m
ε(λ, T ) = exp* ai λi + bj T j +
P P
Exponential polynomial model
,i = 0 j=0 -
Fractional polynomial model ε(λ, T ) = 1
1+a0 λ +a1 T
2
FIG. 10. Original emissivity data.
054903-7 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

FIG. 11. Fitted surfaces based on diffe-

rent emissivity models. (a) Theoretical
radiation model. (b) Polynomial model.
(c) Exponential polynomial model. (d)
Fractional polynomial model.

FIG. 12. The temperature measure-

ment system of casting billets. (a) Struc-
ture of the measurement system. (b)
Field photograph of the measurement
system.

1 experience complete oxidation. In addition, according to the

ε(λ) = [1 + sin(a0 + a1 λ)]. (14)
2 literature,16 the larger the emissivity is, the smaller the temper-
However, few people have studied the emissivity model ature error will be. Based on the experimental measurement
of billets for measuring its temperature, and the error of the
measured temperature will be larger if the target emissivity
model and hypothetical model do not match. More impor-
tantly, the temperature was rarely included in the previous
emissivity models. But the temperature span of billets at the
measuring position can reach 250 ◦ C from open-casting to
stop-casting. Therefore, for accurately measuring the temper-
ature of the billet surface, the bivariate emissivity function
with the two varies: wavelength and temperature should be
established based on the experimental data of the emissiv-
ity of SS 304. According to some previous emissivity models
and the radiation theory of the billet surface, we improved the
mathematical model and put forward some bivariate emissivity
functions, as shown in Table III.
After pulling the cast billets out of a tundish and spraying FIG. 13. Comparison between the temperatures measured by the spectrome-
them with water in the secondary cooling zone, most billets ter and dual-wavelength thermometer.
054903-8 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

FIG. 14. Temperature measurement re-

sults with changes of production param-
eters. (a) Temperature with changes
of the casting speed. (b) Tempera-
ture with changes of the cooling water
volume.

results, the emissivity of peroxide SS 304 becomes the largest secondary cooling zone and before the straightening area. The
when the measuring angle is 75◦ . For these reasons, we chose system optical axis was almost 75◦ to the central position of
the experimental data for the oxidation time of 30 min and the casting inner side direction in order to obtain the minimum
measuring angle of 75◦ as the representative of the billets in temperature measurement error. The field photograph of the
the casting field. Using the least-squares method, we fitted the measurement system is shown as Fig. 12(b).
bivariate functions with the emissivity models in Table III. The Based on the multi-wavelength thermometry theory and
original data are shown in Fig. 10, and the fitted surfaces are the above bivariate emissivity function, the temperature of
shown in Fig. 11, respectively. casting billets for a period of time was obtained, as shown
As shown in Figs. 10 and 11, the polynomial model is the in Fig. 13. The temperature fluctuation measured by the com-
most suitable model for peroxide SS 304 at 75◦ . The functional mon dual-wavelength thermometry method reaches ±20.7 ◦ C,
form is expressed by Eq. (15). The fitting accuracy of this but the temperature fluctuation measured by multi-wavelength
function is high: the correlation coefficient is 0.9867, thermometry with our acquired analytic model is only ±2.8 ◦ C.
Therefore, our multi-wavelength thermometry method can
ε(λ, T ) = −15.278 + 29.841λ − 23.179λ 2 + 9.662λ 3
reduce the temperature fluctuation effectively. Figure 14 shows
− 2.342λ 4 + 0.331λ 5 − 2.54 × 10−2 λ 6 the temperature measurement results with the change of cast-
ing speed (from 1.65 to 1.41 m/min) and with the change
+ 8.197 × 10−4 λ 7 + 2.177 × 10−4 T , (15) of the cooling water volume (from 140.9 to 120.1 l/min),
where λ (µm) is the wavelength and T (K) is the temperature. respectively. As can be seen from this figure, the measured
Replacing a universal hypothetical model by the certain func- temperature decreases with the decreasing casting speed and
tional given in Eq. (16), we can therefore effectively improve increases with the decreasing cooling water, which reflects the
the accuracy of multi-wavelength thermometry for SS 304 changes of production parameters in the continuous casting
under these conditions, line promptly. For these reasons, our study results can provide
much valuable and reliable information to dynamically adjust
ε(λ, T ) = a0 + a1 λ + a2 λ 2 + a3 λ 3 + a4 λ 4 the casting speed and cooling water volume.
+ a5 λ 5 + a6 λ 6 + a7 λ 7 + b1 T . (16)

VI. CONCLUSIONS
B. Temperature measurements of casting billets
The emissivity characteristics of SS 304 at different tem-
To validate the reliability of the above research results, a
peratures, oxidation degrees, and angles were studied using
series of experiments for measuring the temperature of con-
a specifically designed experimental apparatus. The bivariate
tinuous casting billets have been done in Nanjing Iron and
emissivity function of two variables, wavelength and temper-
Steel Group Corp. The system consists of a spectrometer,
ature, was determined for the completely oxidized samples at
a dual-wavelength thermometer, a data processing unit, and
an angle of 75◦ . And then, the results were applied to the tem-
some other equipment, such as a dust-proof sleeve and cooling
perature measurement for continuous casting of billets. The
facility, as shown in Fig. 12(a).
key points of our study are as follows:
The spectrometer was installed in a securing position of
the dust-proof sleeve, and the frame rate of the spectrometer (1) The emissivity of non-oxidized SS 304 at the wavelength
was adjusted to 1 frame/s. The dual-wavelength thermome- range (2-6 µm) increases from 0.357-0.456 to 0.439-
ter was also installed in another dust-proof sleeve, and its 0.491 with increasing temperature from 800 to 1100 ◦ C.
operation wavelengths were 0.8 and 1.05 µm. The dust-proof (2) The emissivity increases significantly once the steel is
sleeve could reduce the dust influence on the spectrometer oxidized, but the increase trend becomes weaker with the
from the terrible field condition using a purged airflow. The oxidation time. Once the sample is completely oxidized,
cooling facility could ensure that the spectrometer and the the emissivity remains constant.
dual-wavelength thermometer operate normally at a high tem- (3) The emissivity of non-oxidized SS 304 decreases with
perature by the application of circulating cooling water. The increasing wavelength (2-6 µm); however, that of oxi-
measured position of the casting billet was located after the dized steel shows the opposite trend.
054903-9 Wang et al. Rev. Sci. Instrum. 89, 054903 (2018)

(4) The emissivity of oxidized SS 304 increases slowly 1 J. Spannar, P. Wide, and B. Sohlberg, IEEE Trans. Instrum. Meas. 51, 1240

from 0◦ to 60◦ and then increases more rapidly with the (2002).
2 M. Švantner, P. Vacı́ková, and M. Honner, Infrared Phys. Technol. 61, 20
increasing angle before suddenly dropping at angle near (2013).
75◦ . 3 Z. Xie and H. C. Bai, Rev. Sci. Instrum. 85, 024903 (2014).
4 T. Iuchi and T. Furukawa, Rev. Sci. Instrum. 75, 5326 (2004).
(5) Compared with the temperature measured by common
5 V. C. Raj and S. V. Prabhu, Rev. Sci. Instrum. 84, 124903 (2013).
dual-wavelength thermometry, the temperature fluctua- 6 F. Righini, A. Rosso, A. Cezairliyan, and A. P. Miiller, Metrologia 30, 197
tion reduces from ±20.7 ◦ C to ±2.8 ◦ C with the aid of our (1993).
multi-wavelength thermometry technique. The strange 7 C. D. Wen, Int. J. Heat Mass Transfer 53, 2035 (2010).
8 T. Iuchi, T. Furukawa, and S. Wada, Appl. Opt. 42, 2317 (2003).
noise in the dual-wavelength measurement is not present
9 T. Furukawa and T. Iuchi, Rev. Sci. Instrum. 71, 2843 (2000).
in the multi-wavelength measurement. 10 L. del Campo, R. B. Perez-Saez, X. Esquisabel, I. Fernandez, and M. J. Tello,

In addition, this apparatus is used for measuring the emis- Rev. Sci. Instrum. 77, 113111 (2006).
11 L. del Campo, R. B. Pérez-Sáez, M. J. Tello, X. Esquisabel, and I. Fernandez,
sivity of SS 304 in this work; moreover, it is designed for Int. J. Thermophys. 27, 1160 (2006).
measuring most metals and also for partly nonmetals and 12 T. Fu, M. Duan, J. Tang, and C. Shi, Int. J. Heat Mass Transfer 90, 1207

semiconductors. (2015).
13 L. D. Campo, R. B. Pérez-Sáez, and M. J. Tello, Corros. Sci. 50, 194

(2008).
14 P. Wang, X. Zhi, and Z. W. Hu, Acta Opt. Sin. 36, 0412001 (2016).
ACKNOWLEDGMENTS 15 J. Dai, X. G. Sun, X. D. Lu, and D. Cong, Theory and Practice of

Multi–Spectral Thermometry (High Education Press, Beijing, 2001),

This study was supported by the National Key R&D p. 10.
Program of China (Grant No. 2017YFC0805100). 16 C. T. Lu, Ph.D. thesis, National Cheng Kung University, Taiwan, 2007.