You are on page 1of 8

NDT&E International 51 (2012) 127134

Contents lists available at SciVerse ScienceDirect

NDT&E International
journal homepage: www.elsevier.com/locate/ndteint

Emissivity estimation for accurate quantitative thermography


Sergio Marinetti, Pier Giorgio Cesaratto n
ITC-CNR Istituto per le Tecnologie della Costruzione, sede di Padova, Consiglio Nazionale delle Ricerche, 4 Corso Stati Uniti, I-35127 Padova, Italy

a r t i c l e i n f o

abstract

Article history:
Received 3 November 2011
Received in revised form
16 April 2012
Accepted 1 June 2012
Available online 12 June 2012

In this paper an emissivity measurement method is proposed; it relies on individual calibration


functions, and is based on the spectral response of a specic IR sensor. The reected temperature is kept
constant during the test, and its knowledge is not required. Results of measurements on materials
commonly used as emissivity references in quantitative thermography are reported. Computer
simulations of indoor and outdoor thermographic inspections show the differentiated inuence of
emissivity and surroundings temperature on the measurement accuracy. Finally experimental results of
absolute and differential temperature measurements are discussed.
& 2012 Elsevier Ltd. All rights reserved.

Keywords:
Thermography
Emissivity measurement
Temperature measurements

1. Introduction
Infrared thermography is a technique that allows one to map
the radiative thermal eld coming from a scene. By this technique, it is possible to evaluate the two-dimensional temperature
eld on the target surface in a non-destructive manner. For this
purpose, target emissivity and surroundings radiation have to be
considered. Emissivity depends on many factors such as temperature, surface roughness, wavelength, and viewing angle.
Knowledge of total emissivity (hemispherical or directional) is
necessary to characterize the radiative behavior of the target,
while for temperature measurements, the equivalent target
emissivity in the IR detector spectrum must be known. Reected
temperature, depending mainly on the surrounding radiation, is
another factor which affects the thermographic temperature
measurement accuracy.
In the last years, both experimental and analytical studies on
the emissivity measurement have been done, and nowadays
different approaches are possible. Especel and Matte [1]
proposed an absolute method for emissivity measurement via
periodic radiometry. They rst implemented the direct and
indirect ordinary methods by separating a modulated emitted
and a non-modulated reected uxes for the direct one, and vice
versa for the indirect one. It is worth noting that both of these
methodologies require the use of an emissivity reference, while
the proposed absolute method did not require any reference, but
involved modulating the temperature of both a sample and a
hemispherical IR source. In the same work, measurements of the
total directional emissivity have been carried out and validated

Corresponding author. Tel.: 39 049 829 5882; fax: 39 049 829 5728.
E-mail addresses: sergio.marinetti@itc.cnr.it (S. Marinetti),
piergiorgio.cesaratto@gmail.com (P.G. Cesaratto).
0963-8695/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ndteint.2012.06.001

on different samples. Results for low emissivities near ambient


temperature were also reported. Although periodic radiometry
methods have proved to be reliable, the drawback is the complexity of the experimental setup. Schurer [2] elaborated a technique
for the assessment of IR emissivity of surfaces at approximately
room temperature. Measurements were done in two wavelength
bands, providing the surface reectivity of different samples with
the use of a black painted hood with known temperature.
Walach [3] dealt with the importance of infrared thermal
analysis for monitoring thermal conditions of electronic microcircuits. He discussed different methods for measuring emissivity,
and presented measurements for different samples with various
emissivities. The author remarked the importance of minimizing
changes of the radiation incident on the sample itself, and reaching the condition of a high uniformity of surface temperature of
the heated microcircuit during measurements.
Mazikowski and Chrzanowski [4] developed a model of
passive multiband system for emissivity measurement through
simulations. They highlighted that non-ideal approximation of
object emissivity by assumed mathematical function, and uctuation of object temperature during measurement process are the
most important sources of error. Furthermore, they remarked that
errors caused by background radiation reected by the object
strongly depend on background and object temperatures; in
particular, if the object temperature is sufciently higher than
the background one, the inuence of these sources of error can be
considered as non-signicant.
Chrzanowski [5] discussed different formulas for determining
the effective emissivity of some materials. He highlighted that
commonly used formulas could produce inaccurate results in the
case of systems of spectrally variable sensitivity, and proposed a
more accurate formula enabling the correct determination of
emissivity.

128

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

Nomenclature
Ll,bb
Ll,e

l
T
Tsurr
Tt
TIR
TPt100
Tair
Tapp
DT
DTt

DTIR
DTPt100
R(l)

Spectral blackbody emissive power (W m  2 mm  1)


Spectral object emissive power (W m  2 mm  1)
Wavelength (mm)
Temperature (1C)
Equivalent blackbody surroundings temperature (1C)
Temperature of a target (1C)
Temperature
obtained
via
IR
thermography
analysis (1C)
Temperature obtained by means of measurements
with a Pt100 probe (1C)
Air temperature (1C)
Apparent temperature (1C)
Temperature difference (1C)
Absolute error on surface temperature evaluation via
numerical analysis (1C)
Temperature difference measured by the IR camera (1C)
Temperature difference measured by Pt100 probes (1C)
Detector spectral response

Concerning the reected temperature, it has been dened in


the ASTM E 1862-97 [6] as the temperature of the energy
incident upon and reected from the measurement surface of a
specimen. Two methods for measuring reected temperature
have been there dened. The rst one consists in pointing the
infrared imaging radiometer at the specimen and measuring the
apparent surface temperature of an infrared reector positioned
on the specimen, which represents the reected temperature of
the specimen itself. However, this method is not suitable if the
specimen has a diffuse reection. The second approach consists in
making the radiometer pointing away from the specimen in the
same direction of the angle of reection, so that the recorded
apparent temperature represents the reected temperature of the
specimen. In case of a diffuse surface reector, pointing the
radiometer at a variety of locations within 451 of both sides of
the angle of incidence and averaging the reported temperatures
leads to the denition of the reected temperature.
Datcu et al. [7] described a study to improve the accuracy of
surface temperature measurements under near-ambient conditions through infrared thermography, by using a highly reective
and diffusive aluminum mirror to quantify the surroundings
mean radiation. They validated the proposed method via measurements both in an experimental room with homogeneous
radiative surroundings, and for an external building wall under
in situ surrounding conditions. They found a good agreement in
comparing the temperature eld measured by infrared thermography with thermocouple recordings.
Concerning the effect of different factors on emissivity surface,
Del Campo et al. [8] discussed how radiation wavelength, sample
temperature, emission angle, surface nish and oxidation affected
the emissivity of three Ni and Co based aeronautical alloys. They
found that spectral emissivity decreased with increasing wavelength, and was nearly independent on temperature in the analyzed
temperature range, then the lowest emissivity was observed for
brushed samples. Moreover, they highlighted that the angular
emissivity dependence was different for short and long wavelengths,
and oxidized samples showed a higher emissivity.
Wen and Mudawar [9,10] presented an in-depth study on the
inuence of different factors, such as temperature and surface
roughness, on spectral emissivity of aluminum alloys via measurements of intensity of radiation at several wavelengths, in

Ibb
Iatm

Smeas
Se
Sr
Sa
Stot
t

tatm
e
el
s
OS

Sensed blackbody emissive power received by a


sensor with spectral response R(l) (W m  2)
Sensed equivalent blackbody emissive power coming
from the atmosphere received by a sensor with
spectral response R(l) (W m  2)
Signal captured by the camera detector (W m  2)
Signal captured by the camera corresponding to the
emitted radiation (W m  2)
Signal captured by the camera corresponding to the
reected radiation (W m  2)
Signal captured by the camera coming from the
atmosphere (W m  2)
Signal captured by the camera detector composed by
emitted and reected components (W m  2)
Time (s)
Atmospheric transmission coefcient
Emissivity in the camera sensitivity band
Spectral emissivity
Standard deviation
Object signal (a.u.)

conjunction with the denition of a multiwavelength emissivity


model to infer the surface temperature. They found that surfaces
of this kind of alloys bucked the general trend of increasing
emissivity with increasing temperature in the infrared range, then
underlined that the emissivity of aluminum samples generally
decreased appreciably between 2.05 and 3.5 mm, and increased
slightly between 3.5 and 4.72 mm. Moreover, in the same papers
[9,10] it was experimentally highlighted that surface roughness
affects spectral emissivity in two ways; the spectral emissivity
distribution was found to be very pronounced for the polished,
6- and 14-mm roughened surfaces, and far less pronounced for the
highly roughened extruded and saw-cut surfaces, furthermore,
roughness increased emissivity, especially for the saw-cut
surface. Concerning the emissivity models, Wen and Mudawar
[9,10] individuated two relatively simple models, providing the
best overall results, even if it was remarked that drastic changes
in the spectral emissivity distribution precluded the use of a
single function to accurately represent every band of the measured spectrum. Finally these papers underlined that signicant
improvements were needed in both instrumentation and emissivity models to achieve acceptable accuracy in the implementation of radiation thermometry in the aluminum industry.
Wen and Mudawar [11] continued their former studies by
exploring more in depth the relationship between the emissivity
of aluminum alloy surfaces and surface roughness. In particular,
they furthered an existent method for determining emissivity of
rough surfaces, nding a good agreement between measured
emissivity values and model predictions. Moreover, they
suggested a simple new model, providing the best overall predictions for different temperatures.
Inagaki and Okamoto [12] discussed emissivity measurements
for non-metal surfaces using infrared radiometers with different
wavelength bands, also in near ambient conditions. In particular
they developed an experimental apparatus to surround the testpiece by black-body surfaces which were maintained at a
monitored ambient temperature. Results described in this paper
highlighted that sample emissivity uctuates when the temperature difference between ambient and sample temperatures is
small (i.e. almost 20 K), while it seemed independent of temperature for high temperature differences (i.e. higher than 20 K).
Moreover, the authors remarked the dependence of sample

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

emissivity on the wavelength; in fact, experimental results


showed that diverse kinds of detectors operating in various
wavelengths led to different emissivity values for the same
sample under the same conditions.
Eto et al. [13] studied the infrared radiation properties and
surface characteristics of C/C composites and graphite for different temperatures with an apparatus similar to the one described
by Inagaki and Okamoto [12]. Independence of emissivity on
ambient temperature for high sample and ambient temperature
differences has been further investigated.
Papers discussed so far developed the characterization of
emissivity for materials in different spectra with various methods,
while the reected temperature has been neglected, in case of
sample emission sufciently high, approximated to ambient
temperature, or evaluated by means of a reector. Furthermore,
no paper highlights the possibility of characterizing the twodimensional thermal eld of a surface with a high accuracy, and
further developments are requested. In the present work, a
transient method for measuring emissivity without the use of
reference materials is proposed. The surroundings radiation is
unknown but kept approximately constant, thus freeing from the
need of accurately estimating the reected temperature. Results
of measurements on common materials are shown here. The
same materials were tested with different IR cameras and
obtained emissivity values are compared with those available in
literature. Then, through the use of calibration function, the
effects of emissivity and reected temperature uncertainties on
absolute and differential temperature measurements are theoretically discussed. Finally, experimental results are reported and
discussed. The proposed approach could be an important step for
developing existent IR thermography approach for estimating the
2-D temperature eld on a surface. This work is directed to
thermography users who carry out quantitative analysis, to
provide them with a simple and accurate method to characterize
their own set of emissivity references, tailored on their equipment and experimental needs. Such references could be used to
characterize other materials to improve the accuracy of thermographic temperature measurement.

129

where tatm is the atmospheric transmission coefcient, e is the


target emissivity in the camera sensitivity band, Tt is the target
temperature, and Tsurr is the equivalent blackbody surroundings
temperature that corresponds to the emissive power the sensor
would receive, if the target surface were a perfect reector. The last
term (1 tatm)  Iatm represents the contribution due to the atmospheric emission. Calibrated thermographic systems compensate for
the effect of atmosphere, taking into account environmental parameters like air temperature and relative humidity, thus providing
(when emissivity is set to 1) information on the overall signal Stot
(the term between square brackets in Eq. (4)) generated by the
target surface, that can be thought as the sensed emissive power
Ibb(Tapp) produced by a blackbody at temperature Tapp.
A key parameter determining the weights of the emitted and
reected components of Stot is the spectral emissivity el dened,
for a given wavelength l, as the ratio of spectral emissive power
Ll,e(T) emitted by an object to that emitted by a blackbody at the
same temperature T (Eq. (5)):

el T

Ll,e T
Ll,bb T

In terms of signal captured by the IR camera, the global effect


of el depends on the detector spectral response R(l) and, under
the gray body approximation, can be dened as follows:
R
L TURlUel TUdl
eT lR,bb
6
Ll,bb TURlUdl
In this paper, the transient method proposed for estimating

e(T) stems from the following assumptions:


 the surroundings radiation is unknown but constant during
the test;
 emissivity does not vary in the investigated temperature
ranges;
 the target temperature is larger than Tsurr.
In such hypotheses, the total signal Stot vs. Ibb(Tt) is represented
by the well-known Eq. (7) which is a straight line whose slope is e.
Stot t eUIbb T t t 1eUIbb T surr

2. Theoretical approach
IR
[14]).
given
given

cameras are sensitive to the infrared radiation (De Witt


Each body at a temperature T can potentially emit, at a
wavelength l, a spectral blackbody emissive power Ll,bb(T)
by the Plancks law (Eq. (1)).

Ll,bb T

2pc1
5

l ec2 =lT 1

where c1 5.9635  10  17 W m2, c2 1.4394  10  2 m K.


The sensed blackbody emissive power Ibb(T) received by a
sensor with a spectral response R(l) is given by Eq. (2).
Z
Ibb T
Ll,bb TURlUdl
2
The total signal Smeas captured by the camera detector is
mainly composed of three contributions according to Eq. (3):
Smeas Se Sr Sa

The measurements were performed with a FLIR SC6000


camera operating in the mid-wave band (35 mm), with an InSb
sensor and a 640  512 pixel image format. The temperature
range of the camera was set to (  10;55) 1C. The system does
not provide calibrated information about the total amount of
radiation impinging the sensor, and only non-calibrated unit
object signal (OS) is available. OS is approximately proportional
to the amount of radiation sensed by the camera detector, and can
be used only for comparative measurements within the same
measurement range for the same camera. Since the proposed
approach requires the calculation of the sensed black-body
radiation, a calibration function was preliminary developed
according to the following Eq. (8)

where Se and Sr are the emitted and reected components coming


from the target surface, while Sa is due to atmospheric absorption
(the scattered ux is here neglected). More in detail, for an
opaque target, Eq. (3) can be rewritten in terms of black-body
radiations as in Eq. (4).
Smeas tatm eT t UIbb T t 1eT t UIbb T surr  1tatm UIatm
tatm Stot 1tatm UIatm

3. Experimental procedure

Ibb T 

A
expB=TD

where A, B and D are parameters depending on the SC6000


camera spectral response. Such a function allows to compute
the signal Stot Ibb(Tapp) (following Eq. (7)), from the apparent
black-body temperature Tapp provided by the software ThermaCAMTM Researcher [15], after compensating for the atmosphere
contribution and setting the target emissivity to 1. Moreover,

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

Ibb(Tt) can be computed from the target surface temperature


measured by a Pt100 contact probe (Eq. (8)).
A linear relationship was veried between Stot and OS, thus
conrming that Eq. (8) links temperature to a quantity proportional to the radiation actually sensed by the camera. It is worth
specifying that the use of OS in place of Ibb(T) was not possible
because the function OS(T) is not provided by the software. The
experimental setup (Fig. 1) consists of a hotplate to heat the
sample, and a hood to ensure that the reected radiation is
constant during the test.
The sample was placed in thermal contact with the hotplate
and a Pt100 probe was used to record the surface temperature.
The temperature was increased of about 20 1C above the room
temperature, and the cooling process was observed by the IR
camera with an angle of 301 with respect to the normal of the
target surface. It has been assumed that in this limited temperature range emissivity did not vary.
Sample and hotplate were surrounded by the hood, which was
placed so that the main source of reected radiation entering the
camera came from the internal surfaces of the hood itself. The hood
was an open cube of 30  30  30 cm3 (Fig. 1) made of 1.5 cm thick
aluminum plates. The internal surfaces have been treated with the
sandpaper in both x and y directions, following a random pattern,
and then painted black, for increasing surface emissivity according
to Wen and Mudawar [9,10,11] and Del Campo et al. [8] highlighted that for Ni-Co alloy samples, brushed and sandblasted
samples of the same alloy having nearly the same roughness
average value showed quite different values of emissivity. In
particular brushed samples having a strong unidirectional lay
pattern of the brushed surfaces, i.e. roughness extended only in
one direction, showed a lower emissivity value than sandblasted
samples, which were rough in both x and y directions. Moreover
the black painted hood serves as a blackbody and since its
temperature is lower (and stable) than the target, this reduces
the spurious reections since the target is at about 20 1C above the
ambient temperature. The temperature increase of the hood walls
during the test was minimized by using thick aluminum plates,
reducing the sample size, and covering the hotplate surface with a
low-emissivity aluminum tape. Hood wall thickness and sample
size were rst determined through numerical simulation, with
COMSOL Multiphysics [16] that is a FEM commercial package, and
then experimentally veried. It is worth noting that a semihemispherical hood is normally preferred to have uniform

13.8
13.6
Stot [W/m2]

130

13.4
Stot = 0.901 * Ibb + 0.908

13.2

R2 = 0.9990
13
12.8
12.6

13.2

13.4

13.6

13.8

14

14.2

14.4

Ibb [W/m2]
Fig. 2. Plot of Stot vs. Ibb(Tt).

reections. In this case, since the main hypothesis is the stability


of the hood temperature, the non uniform reections due to the
abrupt angles of the box do not affect the measurement accuracy.
Samples of different materials have been tested: white paper,
black painted paper, black insulating tape, and polished aluminum. These materials are usually used as emissivity references for
thermographic inspection.
The sample has been heated up to 40 1C, and 20 thermograms
have been acquired during the cooling stage with a sampling
interval of 30 s, corresponding to about 8 1C cooling. Simultaneously, the surface temperature values Tt, measured by two
Pt100 sensors, have been averaged and stored. The air temperature and relative humidity have been measured before and after
each test, and the mean values have been set in the software to
take into account the atmosphere contribution. Setting the
emissivity parameter to 1, ThermaCAMTM Researcher provided
apparent black-body temperature prole Tapp of the sample
surface. Starting from this prole, Stot Ibb(Tapp) has been evaluated with a proprietary MATLAB [17] code implementing Eqs.
(2) and (8). The same procedure allowed one to evaluate the
sensed blackbody emissive power of the sample surface starting
from the temperature prole Tt measured by the Pt100 probes.
Fig. 2 shows the linear relationship in Eq. (7), as expected.
The slope of the curve in Fig. 2 is e, while the intercept is the
absolute contribution of reections. For each material, 10 measurements have been performed.

4. Results and discussion

Fig. 1. Overview of the experimental setup.

In Fig. 3 results for black paper sample are reported. Uncertainty


bands refer to the variance of e, evaluated from the regression
analysis of Stot vs. Ibb(T), multiplied by a coverage factor equal to 2,
corresponding to the 95% condence level. It can be observed that
the mean value falls within all the error bars. The same behavior
has been veried for the other tested materials.
In Table 1 the mean emissivity value and the standard
deviation of all tests are reported.
Two of the previous samples were tested also with the FLIR
SC3000 camera operating in the long-wave (89 mm), with a
QWIP sensor, and a 320  240 pixel image format. In Table 2
results of measurements are shown. In particular it can be seen
that, in this case, both black and white papers show almost the
same emissivity value.

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

Finally, in Table 3 emissivity data available from the infrared


literature and measurements made by FLIR Systems are shown.
From Table 3, it can be seen that signicant differences are
reached in emissivity of the same sample in different wavelength
bands, thus remarking the importance of setting the right emissivity value for the specic IR camera spectrum. Furthermore, it
can be observed that black and white papers have signicantly

different emissivity in the 25 mm band (i.e. respectively 0.86 and


0.760.78), while in the 814 mm band they show almost the
same values (i.e. about 0.90). Nevertheless, a direct comparison
with values reported in Tables 1 and 2 is not possible because
Table 3 refers to a testing temperature of 70 1C which is higher
than that considered in this paper.

5. Thermographic temperature measurements

single test
mean value

0.96

0.92

0.88

0.84

0.8

131

5
6
test number

10

Fig. 3. Results of tests on the black paper sample.

Table 1
Results of emissivity measurements with FLIR SC6000 IR camera.
Material

Black paper
Black PVC tape
White paper
Aluminum (polished)

0.901
0.902
0.719
0.022

0.004
0.006
0.002
0.003

Table 2
Results of emissivity measurements with FLIR SC3000 IR camera.
Material

Black paper
White paper

0.910
0.908

0.006
0.002

Simulations in MATLAB environment were performed to assess


the sensitivity of thermographic temperature measurements to e
and Tsurr using a SC6000 camera. Two case-studies have been
considered in the eld of building inspection: outdoor and indoor
measurements (Fig. 4). In the hypothesis of absence of direct solar
radiation these two situations differ mainly for the temperature
difference between the target and the surroundings. In particular
the simulated conditions were Tt 15 1C, e 0.85 and Tsurr  10 1C
in the outdoor scenario, while in the indoor case Tt 19 1C, e 0.85
and Tsurr 20 1C.
The signals Stot were computed using Eqs. (7) and (8), then Tt
was recomputed assuming an uncertainty on e of 70.05 and on
Tsurr of 72 1C, which appear to be reasonable values in practice,
and the absolute error DTt has been evaluated. Results of the
simulations are summarized in Table 4 and Fig. 5.
It can be noticed that in the outdoor case, where Tsurr is quite
different from Tt, e has a strong inuence on the nal result; the
uncertainty of 70.05 on e caused a DTt of about 71 1C, while the
uncertainty of 72 1C on Tsurr caused a DTt of about 70.2 1C. For
the indoor case e has a little inuence (DTt varies in a range
of 70.06), while an uncertainty of 72 1C on Tsurr leads to a DTt of
about 70.4 1C.
A further analysis was made for measurements of indoor
temperature difference. The scenario was the same as Fig. 4b,
but in this case two targets at Tt1 19 1C and Tt2 19.5 1C were
considered, and the sensitivity study was made on the temperature difference (Tt2 Tt1) 0.5 1C assuming the uncertainties
previously mentioned on e and Tsurr. Results in Fig. 6 shows how
in this case the knowledge of the true value of e and an uncertainty
of 72 1C on Tsurr estimate leads to an error of about 70.005 1C,
while the correct estimate of Tsurr and an uncertainty of 70.05 on
e lead to an error of about 70.03 1C on the nal result.
Afterwards, experimental measurements of absolute temperature and temperature difference have been performed. In particular, tests have been done for two couples of the same kind of
samples (white paper and black paper), for which emissivity
values were assumed to be those obtained with the proposed

Table 3
Emissivity data for the considered samples from the infrared literature and measurements made by FLIR Systems (ThermaCAMTM
Researcher).
Material

Spectrum

Reference
temperature (1C)

Black paper
Black, dull paper
Paper coated with black lacquer
Black, dull paper
Black, dull paper
White paper
White paper, 3 different glosses
White paper, 3 different glosses
White bond paper
Polished aluminum
Aluminum sheet, 4 samples differently scratched
Aluminum sheet, 4 samples differently scratched
Heavily weathered aluminum

Total
Total
Total
814 mm
25 mm
Total
814 mm
25 mm
Total
Total
814 mm
25 mm
25 mm

70
70
20
70
70
20
50100
70
70
17

0.90
0.94
0.93
0.89
0.86
0.700.90
0.880.90
0.760.78
0.93
0.040.06
0.030.06
0.050.08
0.830.94

132

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

Table 4
Results of the sensitivity analysis for absolute temperature measurements.
Outdoor

Absolute
error
DTt
(1C)

Indoor

e known;
uncertainty on
Tsurr: 7 2 1C

Tsurr known;
uncertainty on
e7 7 0.05

e known;
uncertainty on
Tsurr: 7 2 1C

Tsurr known;
uncertainty
on e7 70.05

0.15
 0.16

1.07
 0.98

0.35
 0.38

0.06
 0.06

1.5

Tsurr=12C
Tsurr=11C
Tsurr=10C

Tt [C]

Tsurr=9C
Tsurr=8C

0.5

0.5

1
0.8

0.82

0.84

0.86

0.88

0.9

1.5

Tsurr=18C
Tsurr=19C
Tsurr=20C

Tt [C]

Tsurr=21C
Tsurr=22C

0.5

0.5

1
0.8

0.82

0.84

0.86

0.88

0.9

Fig. 5. Effects of e and Tsurr uncertainties on absolute temperature measurement.


Outdoor (a) and indoor (b) measurements.
Fig. 4. Case studies for error analysis: outdoor (a) and indoor (b) inspections.

method and reported in Table 1. The samples were placed on


aluminum plates with different thickness and heated in such a
way to generate a temperature difference DT (Fig. 7). Tests have
been done for different values of mean temperature and temperature difference between the samples. The cooling stage was
observed by the SC6000 camera. Moreover the surface temperature of the targets was monitored by two Pt100 sensors. Thirty
thermograms with a sampling interval of 10 s were acquired, and
a region of interest on each sample near the Pt100 sensor was
selected.

Analyses were made with ThermaCAMTM Researcher software;


Tsurr was approximated with the air temperature and the
measured values of e were set according to Table 1. It is here
remarked that there were no external walls inside the laboratory,
no air conditioning devices were operating during the test, and
the adjacent rooms were at approximately the same room
temperature of the laboratory itself. Hence setting Tsurr equal to
the air temperature was a reasonable assumption. Comparisons of
absolute temperature and temperature difference DT, measured
by IR thermography (TIR and DTIR) and by Pt100 probes (TPt100 and
DTPt100), are shown in Tables 58 for all the samples.
From Tables 5 and 6, it can be seen that for white
paper samples the thermographic measurement of the absolute

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

Tsurr=19C

0.03
(Tt2Tt1) [C]

Table 7
Results of temperature measurements on black paper samples.

Tsurr=18C

0.04

Sample 1

Tsurr=20C
Tsurr=21C

0.02

133

Tsurr=22C

0.01
0

Sample 2

TPt100 (1C)

TIRTPt100 (1C)
e 0.90; Tsurr Tair

TPt100 (1C)

TIRTPt100 (1C)
e 0.90; Tsurr Tair

29.41
29.41
31.74
38.92

0.03
 0.00
0.06
0.08

29.72
32.78
32.06
39.88

0.03
 0.04
0.03
0.10

0.01
Table 8
Results of DT measurements on black paper samples.

0.02

DTPt100 (1C)

0.03
0.8

0.82

0.84

0.86

0.88

e 0.90; Tsurr Tair

0.9

Fig. 6. Effects of e and Tsurr uncertainties on indoor differential temperature


measurements.

DTIRDTPt100 (1C)

0.31
3.37
0.31
0.96

 0.03
 0.03
 0.02
0.02

temperature gives deviations up to 0.4 1C with respect to the


Pt100 readings, while the DT measurement leads to a deviation up
to 0.03 1C between IR thermography and Pt100 readings. For black
paper samples (Tables 7 and 8), the higher e leads to lower
deviations (less than 0.1 1C) between IR thermography and Pt100
measurements, while deviations on DT measurements are the
same obtained for white paper.

6. Conclusions

Fig. 7. Experimental layout for validation tests.

Table 5
Results of temperature measurements on white paper samples.
Sample 1

Sample 2

TPt100 (1C)

TIRTPt100 (1C)
e 0.72; Tsurr Tair

TPt100 (1C)

TIRTPt100 (1C)
e 0.72; Tsurr Tair

27.62
30.61
33.07
36.92

0.32
0.35
0.35
0.38

28.01
33.67
37.55
44.92

0.29
0.38
0.38
0.38

Table 6
Results of DT measurements on white paper
samples.

DTPt100 (1C)

DTIRDTPt100 (1C)

e 0.72; Tsurr Tair


0.39
3.06
4.47
8.00

 0.03
0.03
0.02
 0.01

A transient method for emissivity estimate is proposed, which is


based on the spectral response of the IR sensor and does not require
emissivity references and reected temperature knowledge.
The surroundings radiation was kept constant by positioning
the sample inside a black painted hood with a high thermal
inertia. Tests were performed with a MWIR camera on materials
commonly used as emissivity references in thermographic building inspections, like black and white paper, polished aluminum
foil, and black adhesive tape. Furthermore, some of the mentioned
materials were tested with a QWIP camera and signicant
differences in the results have been observed, as expected. Moreover, simulations were performed to assess the sensitivity to e
and Tsurr in outdoor and indoor conditions for absolute thermographic temperature measurements. Results showed that for
outdoor measurements, the inuence of e was stronger than in
the indoor case, where the knowledge of the reected temperature is more important. A further simulation was carried out for
indoor differential measurements. In such a case emissivity is a
crucial factor to obtain a high accuracy. Finally, experimental tests
were performed, comparing the IR thermography results with
Pt100 readings. Emissivity values were set to those obtained with
the proposed method while the reected temperature was set
approximately equal to the room temperature. For white paper
samples, absolute temperature measurements exhibited higher
deviations (0.4 1C) with respect to those obtained with black
paper (0.1 1C). Different behavior was observed for differential
temperature measurements, where in both cases (white and black
paper) the deviations from Pt100 readings were the same
(0.03 1C). This last result shows how indoor thermographic
inspections can take advantage from the accurate knowledge of
the surface emissivity for a given IR camera, which can be
preliminarily assessed in the lab with the proposed methods,

134

S. Marinetti, P.G. Cesaratto / NDT&E International 51 (2012) 127134

and a contact sensor used as reference. In such a way, the twodimensional thermal eld on a wall can be measured with a high
accuracy. Research is going on to exploit this method for in-situ
emissivity estimation.
References
[1] Especel D, Matte S. Total emissivity measurements without use of an
absolute reference. Infrared Phys Technol 1996;37:77784.
[2] Schurer K. A method for measuring infrared emissivities of near-black
surfaces at ambient temperatures. Infrared Phys 1976;16:15763.
[3] Walach T. Emissivity measurements on electronic microcircuits. Measurement 2008;41:50315.
[4] Mazikowski A, Chrzanowski K. Non-contact multiband method for emissivity
measurement. Infrared Phys Technol 2003;44:919.
[5] Chrzanowski K. Problem of determination of effective emissivity of some
materials in MIR range. Infrared Phys Technol 1995;36(3):67984.
[6] ASTM E 1862-97. Standard test methods for measuring and compensating for
reected temperature using infrared imaging radiometers; 2002.
[7] Datcu S, Ibos L, Candau Y, Matte S. Improvement of building wall surface
temperature measurements by infrared thermography. Infrared Phys Technol
2005;46:45167.

[8] Del Campo L, Perez-Saez RB, Gonzalez-Fernandez L, Esquisabel X, Fernandez I,


Gonzalez-Martn P, et al. Emissivity measurements on aeronautical alloys. J
Alloys Compd 2010;489:4827.
[9] Wen C, Mudawar I. Emissivity characteristics of roughened aluminium alloy
surfaces and assessment of multispectral radiation thermometry (MRT)
emissivity models. Int J Heat Mass Transfer 2004;47:3591605.
[10] Wen C, Mudawar I. Emissivity characteristics of polished aluminium alloy
surfaces and assessment of multispectral radiation thermometry (MRT)
emissivity models. Int J Heat Mass Transfer 2005;48:131629.
[11] Wen C, Mudawar I. Modeling the effects of surface roughness on the
emissivity of aluminium alloys. Int J Heat Mass Transfer 2006;49:427989.
[12] Inagaki T, Okamoto Y. Surface temperature measurement near ambient
conditions using infrared radiometers with different detection wavelength
bands by applying a grey-body approximation: estimation of radiative
properties for non-metal surfaces. NDT E Int 1996;29(6):3639.
[13] Eto M, Ishii T, Inagaki T, Okamoto Y. Infrared radiation properties of the
carboncarbon composite and their application to nondestructive detection
of its defects. Carbon 2002;40:28594.
[14] De Witt DP. Theory and practice of radiation thermometry. Wiley-Interscience;
1988 (ISBN 10: 0471610186/0-471-61018-6, ISBN 13: 9780471610182).
[15] ThermaCAMTM Researcher users manual, professional edition. Version 2.8.
FLIR Systems; 2006.
[16] COMSOL MultiPhysics users guide. Version 4.2; 2011.
[17] MATLAB R2011b online documentation. The MathWorks. /www.mathworks.
com/helpS; 2011.

You might also like