Non-destructive testing method to quantify aging of

Accepted Manuscript Non-destructive testing method to quantify aging of materials by its apparent emissivity: case of glass-based reflectors Olivier Riou, Fabien Delaleux, Vincent Guiheneuf, Harold Espargilliere, PierreOlivier Logerais, Régis Olives, Xavier Py, Jean-Félix Durastanti PII: DOI: Reference:

S1359-4311(16)34375-7 http://dx.doi.org/10.1016/j.applthermaleng.2016.12.109 ATE 9729

To appear in:

Applied Thermal Engineering

Received Date: Revised Date: Accepted Date:

13 September 2016 23 November 2016 26 December 2016

Please cite this article as: O. Riou, F. Delaleux, V. Guiheneuf, H. Espargilliere, P-O. Logerais, R. Olives, X. Py, JF. Durastanti, Non-destructive testing method to quantify aging of materials by its apparent emissivity: case of glassbased reflectors, Applied Thermal Engineering (2016), doi: http://dx.doi.org/10.1016/j.applthermaleng.2016.12.109

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Non-destructive testing method to quantify aging of materials by its apparent emissivity: case of glass-based reflectors Olivier RIOU1*, Fabien DELALEUX1, Vincent GUIHENEUF1, Harold ESPARGILLIERE², Pierre-Olivier LOGERAIS1, Régis OLIVES², Xavier PY², Jean-Félix DURASTANTI1 1

CERTES EA 3481 – Université Paris Est Créteil

IUT Sénart-Fontainebleau, 36 rue Georges Charpak – 77567 LIEUSAINT 2

PROMES CNRS – UPR 8521 – Université de Perpignan Via Domitia

Rambla de la thermodynamique, Tecnosud – 66100 PERPIGNAN

Olivier RIOU

[email protected]

*corresponding author

Tel.: +33 1 64 13 44 79 ; Fax : +33 1 64 13 45 01

Fabien DELALEUX

[email protected] Tel.: +33 1 64 13 44 79 ; Fax : +33 1 64 13 45 01

Vincent GUIHENEUF

[email protected] Tel.: +33 1 64 13 46 86 ; Fax : +33 1 64 13 45 01

Harold ESPARGILLIERE

[email protected]

Pierre-Olivier LOGERAIS

[email protected] Tel.: +33 1 64 13 46 86 ; Fax : +33 1 64 13 45 01

Régis OLIVES

[email protected]

Xavier PY

[email protected]

Jean-Félix DURASTANTI

[email protected] Tel.: +33 1 64 13 42 33 ; Fax : +33 1 64 13 45 01

Version of

Tuesday, 03 January 2017

Page 1 out of 33

Non-destructive testing method to quantify aging of materials by its apparent emissivity: case of glass-based reflectors Abstract In this paper, the measurement capabilities of apparent emissivity are tested by means of a generic LWIR system to quantify the aging of materials. A theoretical frame permitted to elaborate abacuses and to correlate a scaling law to identify the relevant parameters, and to foretell performance in terms of absolute aging time interval resolution as well. The predictions are done for both indoor and outdoor situations. They are compared with measurements performed on glass-based mirrors of a heliostat of the solar furnace of Odeillo after having undergone a Damp Heat accelerated aging test. Apparent emissivity measurements implement an indoor home-made device which had already been characterized. The performances are in accordance with the predictions and allow to discriminate initial natural aging of glass mirror combined with accelerated aging with an interval resolution from 4 to 10 years keeping in mind that we are in quest of differentiating two mirror glasses or of quantifying the absolute aging of each mirror. Limitations of indoor measurements come mainly from the accuracy of Non-Uniformity Correction and from the internal drift compensation inherent to any IR system. Outdoor performances are foreseen to be as efficient by optimizing the contrast

. If we consider an uncertainty of emitting temperature of 0.25°C

and the one of apparent temperature of 0.1°C, a standard outdoor thermography would enable to discriminate absolute aging produced by Damp Heat test with a 10 standard year resolution.

Keywords: Glass-based reflector aging, Damp Heat test, normal LWIR apparent emissivity, apparent emissivity measurement capabilities

Page 2 out of 33

Nomenclature Temperature, °C Temperature, K Spectral bandwidth

of the IR system

Thermosignal of blackbody within the spectral bandwidth, OS Thermosignal provided by the IR system, OS sensitivity of the IR system, OS/°C Spectral response of the IR system Hemispherical spectral emissivity Width of the Gaussian anomaly, µm Amplitude of the Gaussian anomaly Amplitude of the Gaussian anomaly distributed over Spectral barycenter of the anomaly, µm Normal apparent emissivity Gap of normal LWIR apparent emissivity OS

Output Signal

SWIR

Short Wavelength Infrared

MWIR

Medium Wavelength Infrared

LWIR

Long-Wavelength Infrared

DH

Damp Heat

Subscripts Apparent temperature Reflected temperature

Page 3 out of 33

1

Introduction All the research on renewable energy systems is mainly bound to demonstrate the energy efficiency of such alternative systems in order to interest potential investors. The performance race turns out to be more expensive, the systems are physically limited and the natural aging remains difficult to quantify. In a conventional approach to quantify the durability of solar thermal power plants, one is mainly interested in the aging of the absorber. However, it must be borne in mind that reflectors play an important role in the solar concentration. Regarding reflector types, glass-based reflectors still prove to be the most durable ones in solar applications. Climatic stress factors are responsible for their degradations resulting in a decrease of the global efficiency of the plant. Contemporaneous studies on silvered-glass reflectors for solar applications report solarweighted hemispherical reflectance degradation up to 6% for thin glass reflector and less than 3% for the thick glass after more than 36 months in a climatic chamber (60°C/ 60–75 % of relative humidity) and over 6 years of outdoor conditions at various sites worldwide [1]. Sandstorms are among climatic parameters that cause a significant diminution of the mirror optical performance by generating surface erosion: Karim et al. [2] report irreversible losses in relative specular reflectivity of 0.2% and 0.4% respectively in two different sites (ocean and desert sites in Morocco) during a period of 240 days (0.66 years) of natural exposure. The morphological study of impacts highlights typical diameters 30 – 50 µm with a depth of 6 – 10 µm. It is noticeable that the impacted surface ratio in the order of 1.1 – 1.6% makes difficult to detect the incipient degradation by non-intrusive means. The effect of the combination of wind speed and aerosol concentration is also tested in controlled environment: Lopez-Martin et al. [3] found significant losses of specular reflectance up to 26% for the least favorable conditions. They pointed out that the effect of wind speed on optical degradation is more important than the effect of aerosol concentration. The connection of accelerated sandstorm tests under natural conditions has not been established yet and to date it is impossible to predict the durability of the mirror as it is dependent on a combination of climatic parameters (wind speed and dominant direction) and of aerosol properties. The challenge in the in-situ quantification of the aging of reflectors is then the definition of the associated indicators and their measuring means. Numerous aging mechanisms induce modification of the spectral emissivity of the materials due to significant changes of surface morphology (owing to corrosion, recrystallization, cracking...) or in volume (moisture ingress and chemical alteration). In a previous work, we demonstrated the connection of the apparent emissivity of common materials with their spectral emissivity obtained by IR reflectometry [4]. The apparent emissivity allows to quantify the aging only if the degradations produce spectral emissivity anomalies within the spectral band of the IR system. The apparent emissivity measurement implements non-contact temperature measurements and benefits from imaging techniques developed for infrared thermography. This non-destructive and non-intrusive approach should then be effective to quantify the material aging by mapping their apparent emissivities or reflectivities [5].

Page 4 out of 33

In this paper, in the theoretical part the sensitivity of the apparent emissivity to spectral anomalies and its measurement capabilities will be itemized. The theoretical and experimental performances will be compared thanks to a home-made device which was already characterized in a previous work [6]. The method is applied on some samples of glass mirrors of heliostats of the solar furnace of Odeillo after having undergone an IEC standard damp heat accelerated aging test [7]. The possibility of quantifying the aging of mirrors is concluded at the scale of a solar power plant.

2

Apparent emissivity Apparent emissivity appears in the radiometric equation used by infrared cameras. This parameter measures the emission thermosignal of the target compared to a blackbody at the same temperature. It is designed by

.

The expression of the radiometric equation is given by: L  ( t app. )    ( t )  L0 ( t )  (1    ( t ))  L0 ( t env. )

where

and

(1)

are the blackbody thermosignals respectively associated to the emission temperature

of the sample and to the environment temperature is associated to the apparent temperature

.

is the raw thermosignal given by the camera and

. These thermosignals are reported to the spectral band of detection of

the camera and to its spectral response. Once the validity of the model of the transmitter is admitted, the expression of apparent emissivity becomes trivial:

  ( t ) 

L  ( t app. )  L0 ( t env. ) L0 ( t )  L0 ( t env. )

(2)

Its determination can be made possible with the measurement of three temperatures (apparent, environment and emission ones) and the calibration curve of the system. The calibration of the infrared system is more precisely explained by Riou et al. [8]. This indicator is specific to the detection technique implemented in the camera, the spectral band of detection and the spectral emissivity of the target. A model of apparent emissivity was submitted by Chrzanowski in 1996 [9]: 2

  ( , T ) R

  (T) 

0

(, T) r () d

1

2

R

(3) 0

(, T) r () d

1

where

is the spectral emissivity of the target,

is the spectral response of the infrared system and

is the blackbody radiance. All the parameters are integrated over the spectral band of the camera

which

can be found with the aid of the method developed by Riou et al. [8]. The connection of normal LWIR apparent Page 5 out of 33

emissivity to the spectral emissivity ranging within [0.1 – 0.95] is fully verified for emitter temperatures varying in the range [20°C – 120°C].

3

Theoretical approach Normal LWIR apparent emissivity is connected to the spectral emissivity of materials [4, 6 and 8]. In order to illustrate the ability of the apparent emissivity to reproduce any anomaly in spectral emissivity, a gray emitter (spectral emissivity of 0.5) is processed in which two anomalies overlap respectively to the left and to the right of the center of the detection window (figure 1). These anomalies are assumed to result from structural modifications of surface which occur during the system lifetime. They lead to a gap of apparent emissivity 3.1.

Sensitivity of apparent emissivity to spectral emissivity

A Gaussian shape is employed to reproduce such a spectral anomaly. The latter is infinitely differentiable and it enables to simulate the passage from an anomaly of narrow-band type to an extended-band type one by varying its standard deviation . The width of the anomaly is set to quantified using

. The product

where

while its amplitude is

defines subsequently the typical surface of a rectangular anomaly which

differs from the Gaussian one with a ratio of

.

The apparent emissivity is computed using relation (3) within the LWIR band [8µm – 13µm] for emitter temperatures ranging within [-20°C – 180 °C]. The spectral responsivity

implemented for the computation is

the one of an amorphous silicon IR focal plane array which equips our camera. It was shown that

can be

assumed as non-selective within the LWIR band [8]. It is interesting to quantify the typical gap

as a function of the pair

. To have an

independent estimate of the emission temperature, we must first minimize the temperature dependence of the apparent emissivity by centering the anomaly amidst the detection window. Because of the asymmetry of Planck radiance, the Gaussian shape makes it impossible to reduce in totality the temperature dependence, especially for extended anomalies. Consequently, the central wavelength

of the Gaussian anomaly is modified until the

min/max deviation of the apparent emissivity becomes less than 0.03% on the entire temperature range. Once this parameter is set, the computing is carried out by utilizing relation (3) in which the width the Gaussian anomaly changes independently. For each pair

and the amplitude

of

, the gap is computed by subtracting the mean

value of apparent emissivity to the nominal one. In figure 2, a contour plot for selected values of apparent emissivity gaps is displayed. For gaps greater than or equal to 0.25%, the characteristics are well reproduced by an empirical law of the form:

  µm     %   11.505    % 

Page 6 out of 33

(4)

meaning that the behavior is roughly hyperbolic. 3.2.

Apparent emissivity measurement capabilities

A conventional emissivity measurement method (single-heating method) consists in the measurement of the thermographic camera signal (thermosignal) when the target is kept at the uniformly known temperature t. It can be employed for both indoor [10, 11, 12 and 13] and outdoor situations [14]. This method provides generally reproducible measurements and its accuracy depends on both the emitting temperature and the apparent temperature uncertainties [6, 14 and 15]. The latter should be understood in terms of both type-A and type-B measurement uncertainties and their amplitudes are closely linked to the quality of the thermal drift compensation and of NonUniformity Corrections (NUC) inherent to any infrared systems [16]. From relation (2), we can specify the measurement uncertainty of apparent emissivity caused by any uncertainty of apparent temperature

:

  (%)  100  s ( t app. ) 

t app. L0 ( t app. )  L0 ( t env. )

(5)

The capability of apparent emissivity measurement depends obviously on the metrological characteristics of the IR system used to implement the single-heating method via its sensitivity

in OS/°C and its provided

thermosignals and through the uncertainty of apparent temperature which combines the uncertainties of the NUC and the thermal drift compensation of the system. The standard method of calibration implemented by IR system designers is based on Planck’s law by assuming a quasi-monochromatic system. The major interest of this procedure relies on its exactitude and its ease of use. Indeed, for any blackbody temperature T, Planck’s law enables us to establish a general expression for the thermosignal [11, 17]:

L  (T) 

A  B exp    D T

(6)

where A, B and D are respectively the response, the spectral and the shape factors. These factors are generally adjusted by the least square method for each measurement temperature range. Relation (6) assumes implicitly a linear correspondence between thermosignal and Planck’s law, in such a way that the spectral factor B and the shape one D converge towards identical values from one IR system to the other. This assumption was tested on several IR systems after having extracted their characteristics by the procedures detailed by Riou et al. [8]. The results are shown in table 1. The response factor A displays noticeable deviations compared to the mean value. The latter depends on numerous system parameters. The main ones are the absorption bandwidth and the integration time which produce the detector output signal and a set of readout parameters such as the dynamic of analog to digital converter which Page 7 out of 33

formats the raw signal into digital levels (thermosignals) so that the thermosignal is proportional to the radiance. Considering the available different technologies, it is not surprising to observe differences from one system to another. This is not the case for the spectral factor B for which the min/max deviation represents roughly +/- 5% of within the arbitrary bandwidth [8µm –

its mean value. The reason being that B is forced to be near

13 µm]. The shape factor D is equal to 1 in the case of a truly monochromatic system. As stated in table 1, D exhibits discernible deviations for the tested set of commercial cameras but which should not have a significant impact on the absolute value of thermosignals in the limit

. Consequently, the general form of

thermosignals can be reproduced by taking into account a generic LWIR system for which

and

. By combining relation (5) and (6), the measurement uncertainty of apparent emissivity can be restated in terms of system calibration:

  (%)  100 

L0 (Tapp. ) L0 (Tapp. )  L0 (Tenv. )



B  exp ( B Tapp. ) exp ( B Tapp. )  D



Tapp. 2 Tapp .

(7)

where the response factor A (the most dispersive parameter of calibration) is absent. Consequently, the implementation of Eq. 7 can be performed by considering a LWIR system which implements the generic calibration parameters (

).

The computation is done for a standard measurement situation in which

varies in the range [-30°C –

30°C] whereas apparent temperatures change in the range [-29°C – 200°C] and a combined apparent temperature uncertainty is set to 0.1°C. The result is displayed in figure 3 for selected values of apparent emissivity gap. A crossover study of the sensitivity of

with the variables (

) has allowed establishing an

empirical law:

  % 

t app. C ( t app.  t env. )



44 770 (301.5  t env. )

(8)

which quantifies the resolution of apparent emissivity measurements in the interval [0.1% – 1%]. To highlight the capabilities of apparent emissivity measurement, two interesting cases are detailed hereafter. The first one considers indoor environments for which the reflected temperature is close to the ambient temperature (20°C). A resolution of apparent emissivity of 0.2% can be achieved by the use of any IR system which gives combined uncertainty of apparent temperature of 0.1°C from the time the apparent temperature of the target gets above 80°C. The most restrictive parameter is then the apparent temperature uncertainty provided by the system. This case study concerns a near-black emitter viewed at a temperature neighboring the apparent one and lower-emitting materials which are lifted to higher temperatures. An asymptotic behavior is also highlighted for an

Page 8 out of 33

apparent temperature close to the surrounding one (in other words the quasi equality of thermosignals) for which any apparent temperature uncertainty contributes to the measurement dispersion. In the second case, we computed a typical outdoor application. As pointed out beforetime, a resolution on apparent emissivity of 0.2% can be achieved both by setting the pair (

) to (20°C, 80°C) and with the

equivalence to (-20°C, 50°C). For outdoor applications, the relevant parameter is the contrast

: even if

sufficient resolution in apparent temperature is not achieved, a resolution in apparent emissivity is obtained by imposing the smallest possible surrounding temperature, consistent with the environment of non-destructive testing conditions. Since the emitting temperature plays a symmetrical role with regard to the apparent temperature, the uncertainty of apparent emissivity measurement caused by any uncertainty of emitting temperature can be evaluated by relation (8) according to the testing conditions by changing emissivity is thus the quadratic sum of each measurement uncertainty of

4

by . The overall sensitivity of apparent and .

Experimental approach Our experimental approach is to demonstrate that apparent emissivity can be used as an aging indicator of materials. Tests are carried out on three samples of soda-lime glass with a thickness of 6 mm. These samples come from the heliostats of the 1 MW solar furnace of Odeillo in the South of France and are respectively named Odeillo #1, Odeillo #2 and Odeillo #3. These mirrors were in operation for a certain number of years. The age of Odeillo #1 is of five years, Odeillo #2 of 10 years and Odeillo #3 of 20 years. Tests are carried out by characterizing the mirror glasses in spectral and apparent emissivities in their initial states and after undergoing the standard damp heat accelerated aging protocol, i.e. 85°C/85% of relative humidity for up to 1000 hours. 4.1.

Experimental facilities

The apparent emissivity is characterized by utilizing the homemade device displayed in figure 4. The latter implements the single-heating method. The emitter temperature is controlled in three stages: 1. An isothermal baseline is provided by a thermostated steel block of 150×150×50 mm3 sealed with refractory cement with a thickness of 150 mm. A PID controller permits to fix the temperature setpoints in the interval [20°C – 120°C] with a min-max non-uniformity of 0.5% of the temperature; 2. A second stage is done by using a thick 50×50×6 mm3 expanded natural graphite (ENG) pellet which covers the entire backside of the sample to assure a great mechanical and thermal contact; 3. The third stage is realized by thick spacer made of paper. The spacer has a circular hole of 8 mm diameter in its center. It covers the front face of the sample and it thermally insulates the sample of the mounting plate. The latter is made of iron having a thickness of 1.5 mm. Two thermocouples are glued on the spacer. Hot junctions are positioned on the lips of the opening and are pressed to the sample. Radiometric temperatures are evaluated by means of a Thermacam E300 IR camera which implements a 320×240 amorphous Page 9 out of 33

silicon FPA technology detection. Its spectral range is [7.9-12.7µm]. The characterization of apparent emissivity is done in the temperature range [20°C – 120°C]. The reflected temperature is recorded according to standard test methods [18] by means of an optically polished aluminium plate. To prevent any irradiance non-uniformities due to the camera, the latter is inclined by 15° from the normal direction of the aperture. The device permits to quantify the normal LWIR apparent emissivity with type B measurement uncertainty ranging within [0.7% – 4.2%]. The reproducibility is found within [0.5% – 16%] whereas the repeatability is comprised between 0.2% and 10% for materials of spectral emissivity ranging respectively within [0.95 – 0.1]. A complete description and study of the performances of the measurement setup were conducted by Riou et al. [6]. A climatic chamber Binder® regulated in temperature (- 40°C to 180°C) and humidity (0 to 99%) allows to realize Damp Heat accelerated tests. It consists in damp heat exposure during 1000 hours at 85°C and 85% relative humidity corresponding to a natural aging of 20 years in the typical environmental conditions of Miami [19]. This test is normalized and used for the certification of solar thermal and photovoltaic panels. The spectral emissivity of the three raw samples (Odeillo #1, #2 and #3) is quantified from hemispherical directional reflectance measurement in the wavelength range [8µm – 13µm]. The spectrometer used is a HDR SOC 100 reflectometer coupled with a FT-IR Nicolet 6700 spectrometer using a ETC (Electronically Temperature Controlled) as IR source, a Ge/KBr beam splitter and a DTGS detector at an angle close to the normal (around 10°). The measurements were performed with an instrumental resolution of 15 cm-1 at the ambient temperature of 20°C thanks to the PROMES-CNRS lab facilities. The measurement uncertainty ranges from 1.6 10-3 to 0.9 10-3 in the emissivity interval [0.233–0.979] within the LWIR band [20] 4.2.

Initial characterizations of mirror glasses

The spectral emissivity and the apparent emissivity of raw samples have been characterized to obtain a reference before accelerated aging tests. The three samples having different ages and operation times, another aim of the initial characterization is to know if it is possible to discriminate them with regard to their emissivities by comparing the spectral emissivities with the apparent ones. The spectral emissivity of the three raw samples (Odeillo #1, #2 and #3) is displayed in figure 5. It can be noted that the spectral emissivities for the three glasses follow very close trends with a value of 0.98 at 8 µm wavelength and 0.92 at 13µm with an important decrease at about 9.5 µm, attributable to Si-O stretching and bending by network modifiers [21]. Spectral emissivity can discriminate samples at different states of aging (Odeillo #3) but it is hard to apply the same distinction for ages close from one to the other (Odeillo #1 and Odeillo #2). The spectral emissivity gap within the range [9.2µm – 11µm] seems to be characteristic to the aging state of the glasses. Compared to Odeillo#1, its rectangular extent is roughly of about 3.1%.µm (7.7%.µm of equivalent Gaussian shape area). Outside this domain, spectral emissivity does not differ with age. A precursor effect of aging is observed for Odeillo # 2: it results in a diminution of the absorption within the band [10.3μm - 10.9µm] counterbalanced by an increase in the band [8μm – 9.5µm]. Page 10 out of 33

Figure 6 shows the apparent emissivity of the three raw samples of soda-lime glass depending on the emission temperature in the range [35°C – 105°C]. Above the temperature of 70°C, the age discrimination between the three samples seems to be conceivable by a careful selection of the test temperature: the apparent emissivity values augment monotonously with the initial aging state of the samples of glass mirrors. The raise of apparent emissivity with temperature is due to the shifting of Wien’s wavelength towards the wavelength of maximum absorptivity of SiO4 band which is more pronounced for Odeillo #1 and #2 owing to their lower absorptivity than the one of Odeillo #3. Below 70°C, the discrimination between Odeillo #1 and #2 poses difficulties as long as the precursor effect seen for Odeillo #2 counterbalances its lower absorptivity at 9.5 µm. On the contrary, Odeillo #3 is clearly discriminated: the apparent emissivity gap is of 0.8%. The latter is in good agreement with the scaling law (4) which predicts 4.3.

.

Effects of accelerated aging test

Figure 7 presents spectral emissivity of Odeillo #1 in its raw state and after 1000 hours in climatic chamber for the Damp Heat test. Effects of aging are clearly observed in the same band [9.2µm – 11µm] like for Odeillo #3 in the raw state. Figure 8 details the gap

of spectral emissivity for the three Odeillo glasses within [3µm – 20

µm]. Aging effects of glasses are exclusively observed within the LWIR band. This confers to any long wave IR systems the possibility of quantifying them, excluding de facto the other normative spectral bandwidth SWIR and MWIR. Effects of aging confirm the process of natural aging: a. a global raise of the absorption within the [9.5µm – 10.5µm] band which is more pronounced for Odeillo #1 and appears to be in the same order of magnitude for Odeillo #2 and #3; b. a lessening of absorptivity within the band [8µm – 9 µm] and c. a surge of absorptivity in the band [11µm – 13µm] which is progressively counterbalanced by the enhancement of reflectivity within the band [8µm –9 µm]. Counterbalancing is achieved in the case of Odeillo #3 for which

and is in progress for

the two other samples. To connect the spectral emissivity to the previous section, each anomaly of amplitude equivalent rectangular gap

within the spectral bandwidth of the camera (

100 1  rect. %   min ( initial )  The product

is quantified by its ) using the relation:

2

  () d

(9)

1

defines the equivalent rectangular extent of the anomaly which serves to calculate its

equivalent Gaussian extent by multiplying the latter by emissivity gap using relation 4. The central wavelength

. This result enables us to predict the apparent is specified by the spectral barycenter of the anomaly.

The latter is calculated by:

Page 11 out of 33

2

 rect.     c 

    () d

(10)

1

The position of

with respect to the middle of the detecting window (

the temperature dependence of the apparent emissivity [15]: a. as long as temperature; b. it is steady with temperature

) determines , apparent emissivity grows with

) and c. it diminishes with temperature when

. As the

central wavelength of the anomaly can change during the aging process, the gap does not necessarily follow the temperature dependence of the apparent emissivity. The characteristics of each glass sample after accelerated aging test are displayed in table 2. In figure 9, the normal LWIR apparent emissivity of Odeillo #1 is presented before and after the Damp Heat test. Error bars indicate the min/max repeatability of six successive measurements for each setpoint (± 0.002). Figure 10 depicts the gap % for each glass sample. The latter continuously runs down with the initial aging state in accordance with their spectral emissivity. Contrary to Odeillo #1, Odeillo #2 and Odeillo #3 samples exhibit quasiconstant values for emission temperature ranging [40°C – 110°C] which means that the temperature dependence of apparent emissivity is preserved within the accelerated aging test. It can be concluded that the aging processes have different natures according to the initial aging state of the glass, as found for spectral emissivity. Nevertheless, the gap % faithfully reports the final aging state and its values are in accordance with the predicted ones if taken into consideration an uncertainty of measurements of ± twice the repeatability (table 2).

5

Discussion The LWIR apparent emissivity is a relevant indicator to quantify aging of glass. Its measurement implements non-contact temperature measurements and benefits from imaging techniques developed for infrared thermography. Mapping the apparent emissivity is feasible by developing a specific post-processing code for raw data. Finally, it is needless to develop a specific IR system insofar as commercial camera meets the requirements of reproducibility (dispersion of the measurements achieved for different measurement conditions) and combined measurement uncertainty. Typical performances of indoor apparent temperature measurements of soda-lime glasses are detailed in table 3 (reproducibility, repeatability and combined uncertainty). Values are done for Odeillo#1 at the emitting temperature of 100°C. Normal LWIR apparent emissivity is provided by implementing the Thermacam E300 IR camera. a. Measurements are undertaken for an ambient air free flow with ambient temperature reproducibility of 1.4°C. It indicates that measurements are done in different environmental conditions. Ambient temperature affects mainly the reflected temperature and, to a lesser extent, the emitting and the apparent temperatures due to the efficiency of the three stages of temperature regulation employed in the experimental device. Nevertheless, the Page 12 out of 33

experimental reproducibility is not impacted as these temperatures are routinely taken at each measurement point. It highlights the robustness of the single-heating measurement method. b. Combined uncertainty of measurement of apparent and emitting temperatures dominate for high emitting temperatures. The reason being that the thermosignal of environment

is generally much lower than the

one of the apparent temperatures and of the emitting temperatures which are in the same order of magnitude as in the case of near-black surfaces. The accuracy of measurement can then be improved by a careful management of the thermal drift compensation. For instance, the thermal drift compensation uncertainty for the Thermacam E300 camera is typically of 0.3°C at 100°C. c. The repeatability mainly depends on the NETD (0.12°C @ 35°C). The latter can then be upgraded by an informed choice of the IR system used for inspection. d. by comparing the apparent emissivity gap of 1.8% to the combined uncertainty, indoor measurements can discriminate absolute aging produced by Damp Heat test with a resolution of 6 equivalent years of exploitation (10 years with reproducibility included). Indoor measurements can also discriminate alterations equivalent to 3 standard years on the same reflector. Is it possible to obtain the same performances in outdoor conditions? For a non-intrusive approach, regulating the emitting temperature must not be sought. The latter is imposed by ambient conditions. Two key parameters to reach similar performances are then the contrast

along with the uncertainty of emitting

temperature measurement. The use of a standard camera ( uncertainty of 0.5°C can attain a resolution of

) associated with an emitting temperature without the need to heat the target (

) by

setting the reflected temperature in the range [-30°C – -20°C]. These conditions are achievable in the case of a solar power plant by performing measurements at summer time, ambient air free flow and during a clear night. A way to improve performances is based on the choice of the thermal imager which procures accurate thermal drift compensation. For example, the triplet (

,

and

) permits to quantify

the apparent emissivity gap with a resolution of 1.7%. Once this parameter is set, the next barrier lays on the uniformity of the emitting temperature and in its measurement accuracy: an uncertainty of emitting temperature of 0.25°C induces a resolution of 0.9% and so on. An important point is the understanding of the applied aging test. It only implies chemical alterations owing to forced hydration of the glass. The effects on the reflectivity of the mirrors are weak but significant effects are noticed over large periods of time generally above 20 years of exploitation. Apparent emissivity should be capable of quantifying the effects of erosion over shorter times. Erosion produces a surface flaking, cracks and porosity of the glass at the vicinity of aerosol impacts. The damages viewed on the glass after a sandstorm test [22] are similar to those noted on alumina ceramics with increased porosity [23]. Porosity is known to change emissive properties of the material, especially in the LWIR band [23]. For example, the spectral emissivity of high porosity alumina ceramic differs massively from those of alumina crystal. The product Page 13 out of 33

is evaluated at 80 %.µm

within the LWIR band and should produce an apparent emissivity gap of an order of 18%. If a detection threshold of 1% is considered, it would enable to quantify percussive effects due to erosion equivalent to 5% of the surface of the glass reflector (3 years of natural erosion in oceanic/desert sites in Morocco according to Karim et al. [2]). It is also important to mention alterations of mirror surfaces induced by UV radiations which can be very significant in some locations with high radiation level like in the desert of Puna region (Andes mountains)

6

Conclusion This paper proves the relevance of the LWIR apparent emissivity in the aging of materials and more precisely in the case of glass used for solar applications. First, different values of apparent emissivity were noticed for the three glass-based reflectors which were in operation in the heliostats of the solar furnace of Odeillo for shorter or longer durations. Secondly, after application of the Damp Heat accelerated aging test in the climatic chamber, the apparent emissivity of the three samples augments substantially. However, the accelerated aging test implies at least two different aging processes following the initial aging state of the samples. The latter can be distinguished both in terms of spectral and apparent emissivity measurements. The comparison between the performances of apparent and spectral emissivities as aging indicator for some samples of glass was made. As expected, the spectral emissivity of glass shows important variations during aging but its characterization requires an operation stop and a sampling of the tested material. This paper demonstrates variations of apparent emissivity of the glass samples during accelerated aging test in the range 0.6% to 1.8%. By comparing the apparent emissivity gap with the combined uncertainty, indoor measurements can discriminate absolute aging produced by a Damp Heat test with a resolution of 6 equivalent years of exploitation (10 years with reproducibility included). Indoor measurements can also discriminate alterations equivalent to 3 standard years on the same reflector. The method can be employed both in indoor or outdoor situations. The latter is predicted to be sensitive by maximizing the contrast

such as in the case of a clear sky. Its potentialities are studied and a set of

abacuses are ready to be used and directives are done to enhance the detectivity on intervals of aging times. Damp Heat test implies only chemical alterations due to forced hydration of the glass. The effects on the reflectivity of the mirrors are weak but significant effects are observed on large periods of time generally above 20 years of exploitation. It is more pertinent to quantify the effects of erosion. If a detection threshold of 1% is considered, it would be possible to quantify percussive effects owing to erosion equivalent to 5% of the surface of the glass reflector or 3 years of natural erosion in oceanic/desert sites. Another interesting perspective is to correlate the apparent emissivity with the functional optical properties in the visible range. This correlation is essential for justifying the use of apparent emissivity as an aging indicator for solar applications.

Page 14 out of 33

Acknowledgement The authors are thankful to Mrs. Wilhelmina Logerais, a mother tongue speaker, for her assistance in writing this paper. The authors remain in gratitude for the financial support received from the French government through the ANR within the frame of the DryRSP project. This work was backed by the French "Investments for the Future" program managed by the National Agency for Research under the contract ANR-10-LABX-22-01 (Labex SOLSTICE). The authors also wish to express their thankfulness to Christophe Escape for his technical aid.

Page 15 out of 33

References [1]

A. García-Segura, A. Fernández-García, M.J. Ariza, F. Sutter; L. Valenzuela, Durability studies of solar reflectors: A review. Renewable and Sustainable Energy Reviews, volume 62, september 2016, pages 453–467

[2]

M. Karim, S. Naamane, C. Delord, A. Bennouna, Study of the Surface Damage of Glass Reflectors Used in Concentrated Solar Power Plants. Energy Procedia, volume 69, may 2015, pages 106–115

[3]

López-Martín R, Caron S, Sutter F, Meyen S, Fernández-García A. Accelerated aging of solar reflectors under sandstorm conditions. In: Proceedings of SolarPACES 2011, concentrating solar power and chemical energy systems, Granada; 2011.

[4]

Riou O., Logerais P. O., Durastanti J. F, Quantitative study of the temperature dependence of normal LWIR apparent emissivity. Infrared Physics & Technology, 60 (2013), 244-250

[5]

Silvana Flores Larsen, Marcos Hongn, Determining the infrared reflectance of specular surfaces by using thermographic analysis, Renewable Energy, Volume 64 (2014), Pages 306– 313

[6]

Riou O., Guiheneuf V., Delaleux F., Logerais P.O., Durastanti J.F., Accurate methods for single-band apparent emissivity measurement of opaque materials, Measurement, 89(2016), 239–251

[7]

Standard IEC 61215, Crystalline silicon terrestrial photovoltaic (PV) modules – Design qualification and type approval.

[8]

Riou O., Logerais P.O., Delaleux F., Durastanti J.F., A self-method for resolving the problem of apparent LWIR emissivity for quantitative thermography up to 130 °C, Infrared Physics & Technology, 67(2014), 504–513

[9]

Chrzanowski K., Problem of determination of effective emissivity of some materials in MIR range. Infrared physics & technology, 36(1995), 679-684

[10] Riou O., Logerais P.O., Froger V., Durastanti J.F.& Bouteville A. Thermal study of an aluminium nitride ceramic heater for spray CVD on glass substrates by quantitative thermography. QIRT Journal vol. 10(2), 2013 [11] Marinetti S., Cesaratto P. G., Emissivity estimation for accurate quantitative thermography. NDT & E International, 51 (2012), 127-134. [12] Walach T. (2008), Emissivity measurements on electronic microcircuits, Measurement, 41(2008), 503-515

Page 16 out of 33

[13] Vainer B.G., Focal plane array based infrared thermography in fine physical experiment. J. Phys. D: Appl. Phys. 41 (2008) [14] Schurer K., A method for measuring infrared emissivities of near-black surfaces at ambient temperatures, Infrared Physics, 16(1976), 157-163 [15] Richard R. Corwin and Adrianus Rodenburgh , Temperature error in radiation thermometry caused by emissivity and reflectance measurement error, Applied Optics Vol. 33, Issue 10 (1994), pp. 1950-1957 [16] Riou O., Berrebi S., Bremond P., Non uniformity correction and thermal drift compensation of thermal infrared camera. Proc. of SPIE Vol. 5405 (2004) [17] Chrysochoos A. (2012). Infrared thermography applied to the analysis of material behavior: a brief overview. Quantitative InfraRed Thermography Journal, 9(2), 193-208. [18] ASTM E1862-97e1 (2002). Standard Test Methods for Measuring and Compensating for Reflected Temperature Using Infrared Imaging Radiometers [19] Wohlgemuth, John H., Cunningham, Daniel W., Monus, Paul, et al. Long term reliability of photovoltaic modules. In : 2006 IEEE 4th World Conference on Photovoltaic Energy Conference. IEEE, 2006. p. 2050-2053 [20] Van Nijnatten, P. A., Hutchins, M. G., Kilbey, N. B., Roos, A., Gelin, K., Geotti-Bianchini, F. & Turner, P., Uncertainties in the determination of thermal emissivity by measurement of reflectance using Fourier transform spectrometers, Thin solid films, 502 (2006), 164-169 [21] K.M. Davis, M. Tomozawa, An infrared spectroscopic study of water-related species in silica glasses, Journal of Non-Crystalline Solids Volume 201, Issue 3, 2 June 1996, Pages 177-198 [22] Kolli, M., Hamidouche, M., Bouaouadja, N., & Fantozzi, G. (2009). HF etching effect on sandblasted soda-lime glass properties. Journal of the European Ceramic Society, 29(13), 2697-2704. [23] Rozenbaum O., De Sousa Meneses D., Echegut P. (2009). Texture and Porosity Effects on the Thermal Radiative Behavior of Alumina Ceramics, Int J Thermophys 30 (2009) 580-590.

Page 17 out of 33

List of figures Figure 1

Connection of normal LWIR apparent emissivity to anomaly of spectral emissivity

Figure 2

Sensitivity of normal LWIR apparent emissivity done by an a-Si IRFPA to a Gaussian shape anomaly of spectral emissivity centered close to the middle of the detecting window ∆λ

Figure 3

Measurement capabilities of normal LWIR apparent emissivity implementing a generic IR system by considering a combined uncertainty

Figure 4

Apparent emissivity measurement setup (2: ENG pellet, 3: glass sample, 4: thermocouples, 5: spacer, 6: mounting plate)

Figure 5

Hemispherical spectral emissivity of raw Odeillo glass mirrors

Figure 6

Thermacam E300 normal LWIR apparent emissivity of raw Odeillo glass mirrors. Measurements are done for

. Error bars indicate the repeatability of six successive measurements per

setpoint. Figure 7

Hemispherical spectral emissivity of Odeillo #1 in the raw state and after 1000h of Damp Heat test

Figure 8

Gaps of hemispherical spectral emissivity of aged Odeillo glass mirrors: a. Odeillo #1; b. Odeillo #2 and c. Odeillo #3. In solid line: gap [%]; horizontal dashed line: equivalent rectangular gap [%] within the spectral bandwidth  vertical dashed line: barycenter

Figure 9

–

of the Thermacam E300 camera;

of the anomaly

Thermacam E300 normal LWIR Apparent emissivity of Odeillo #1 in the raw state and after 1000h of Damp Heat test

Figure 10

Gap % of Thermacam E300 normal LWIR Apparent emissivity of aged Odeillo glass mirrors: a. Odeillo #1; b. Odeillo #2 and c. Odeillo #3

Page 18 out of 33

List of tables Table 1

Main characteristics of several commercial IR cameras

Table 2

Comparison between prediction and measurement

Table 3

Type A and B uncertainty of indoor LWIR apparent emissivity measurements of glass reflector Odeillo#1. LWIR apparent emissivity is provided by using the Thermacam E300 IR camera. The predicted gap of apparent emissivity

is done by relation (8) for each temperature uncertainty

Page 19 out of 33

Figure 1

Connection of normal LWIR apparent emissivity to anomaly of spectral emissivity

Page 20 out of 33

Figure 2

Sensitivity of normal LWIR apparent emissivity done by an a-Si IRFPA to a Gaussian shape anomaly of spectral emissivity centered close to the middle of the detecting window

Page 21 out of 33

Figure 3

Measurement capabilities of normal LWIR apparent emissivity implementing a generic IR system by considering a combined uncertainty

Page 22 out of 33

Figure 4

Apparent emissivity measurement setup (2: ENG pellet, 3: glass sample, 4: thermocouples, 5: spacer, 6: mounting plate)

Page 23 out of 33

Figure 5

Hemispherical spectral emissivity of raw Odeillo glass mirrors

Page 24 out of 33

Figure 6

Thermacam E300 normal LWIR apparent emissivity of raw Odeillo glass mirrors. Measurements are done for

. Error bars indicate the repeatability of six successive measurements per

setpoint

Page 25 out of 33

Figure 7

Hemispherical spectral emissivity of Odeillo #1 in the raw state and after 1000h of Damp Heat test

Page 26 out of 33

Figure 8

Gaps of hemispherical spectral emissivity of aged Odeillo glass mirrors: a. Odeillo #1; b. Odeillo #2 and c. Odeillo #3. In solid line: gap [%]; horizontal dashed line: equivalent rectangular gap [%] within the spectral bandwidth  vertical dashed line: barycenter

– of the anomaly.

Page 27 out of 33

of the Thermacam E300 camera;

Figure 9

Thermacam E300 normal LWIR Apparent emissivity of Odeillo #1 in the raw state and after 1000h of Damp Heat test

Page 28 out of 33

Figure 10

Gap % of Thermacam E300 normal LWIR apparent emissivity of aged Odeillo glass mirrors: a. Odeillo #1; b. Odeillo #2 and c. Odeillo #3

Page 29 out of 33

Operating temperature range

IR system

Operating spectral bandwidth

NETD

A

B

C

OS

K

K

-

Thermacam E300

[-20°C - 120°C]

0.1 °C

[7.9µm – 12.7µm]

170 000

1436

1397

1.0010

Thermacam E300

[80°C - 500°C]

-

[7.7µm – 12.1 µm]

213 200

1545

1453

1.0000

Thermacam E4

[-20°C - 250°C]

0.12 °C

[8.3µm – 13.2 µm]

148 800

1396

1338

0.9998

Thermacam P65

[0°C - 500°C]

0.1 °C

[7.6µm – 11.9 µm]

186 000

1494

1476

1.4990

NEC TH 5100

[-40°C - 2000°C]

0.1 °C

[6.9µm – 13.7 µm]

259 700

1472

1397

0.995

Average value:

195 540

1 469

1 412

1.1

Min/max deviation:

110 900

149

137

0.5

Table 1.

Main characteristics of several commercial IR cameras

Page 30 out of 33

Table 2.

Glass mirror

Odeillo #1

Odeillo #2

Odeillo #3

Natural aging

 5 years

 10 years

 20 years

%

1.48

0.78

0.25

µm

4.8

4.8

4.8

µm

9.17

9.70

12.33

%.µm

17.87

9.37

2.95

(scaling law)

%

1.6±0.1

0.8±0.1

0.3±0.1

(measure)

%

1.8±0.5

1.4±0.5

0.6±0.5

Comparison between prediction and measurement

Page 31 out of 33

Temperature

measurand

repeatability

combined uncertainty

Ambient

21°C

0.2°C

0.2°C

Reflected

25°C

0.2°C

0.2°C

Emitting

100°C 90°C -

0.09°C 0.2% 0.1°C 0.2%

0.26°C 0.5% 0.27°C 0.5%

1.8%

0.3%

0.7%

Apparent

Table 3

Type A and B uncertainty of indoor LWIR apparent emissivity measurements of glass reflector Odeillo#1. LWIR apparent emissivity is provided by using the Thermacam E300 IR camera. The predicted gap of apparent emissivity

is done by relation (8) for each temperature uncertainty.

Page 32 out of 33

Highlights Aging of glass-based reflectors used in concentrated solar plants. Effects of damp heat protocol on infrared spectral emissivity of glass-based reflectors. Effects of damp heat protocol on normal LWIR apparent emissivity of glass-based reflectors. In-situ mapping of the aging of glass-based reflectors by utilizing apparent emissivity.

Page 33 out of 33