Quality evaluation for air plasma spray thermal barrier coatings with pulsed thermography

Quality evaluation for air plasma spray thermal barrier coatings with pulsed thermography Shi-bin ZHAO1, Cun-lin ZHANG2, Nai-ming WU3, Hua-ming WANG1 1. School of Materials Science and Engineering, Beihang University, Beijing 100191, China; 2. Department of Physics, Capital Normal University, Beijing 100048, China; 3. School of Jet Propulsion, Beihang University, Beijing 100191, China Received 5 March 2011; accepted 12 July 2011 Abstract: Pulsed thermography (PT) as a non-contact non-destructive evaluation (NDE) technique was employed to examine as-sprayed air plasma spray (APS) thermal barrier coatings (TBCs) for quality evaluation. Thickness and microstructural characteristics play a vital role in determining the quality. In this paper, PT logarithmic peak second-derivative method was adopted to measure the thickness of top coat. Time dependent thermal images were used to characterize the microstructure which was confirmed by scanning electron microscope (SEM). The results showed that there was relationship between the temperature distribution of the surface and microstructure change in TBCs. Temperature distribution in thermal images and measurement results of thickness were in fairly good agreement with the microstructure change. It can be concluded that it was possible to employ these NDE methods as quality evaluation for as-sprayed TBCs. Key words: thermal barrier coatings (TBCs); non-destructive evaluation (NDE); pulsed thermography; microstructure; thickness measurement

1 Introduction Thermal barrier coatings (TBCs) are widely used in gas turbines to provide insulation for hot sections and increase the component durability. The majority of current TBC systems are based on a duplex-coating system comprised of ZrO2 partially stabilized with Y2O3 (YSZ) top coat and MCrAlY bond coat. The YSZ top coat reduces the severity of thermal transients and lowers the substrate temperature. The MCrAlY bond coat protects the superalloy substrate from oxidation and corrosion [1−4]. Thickness and micro-structural characteristics play a vital role in determining the quality of TBCs. During the long term services, stress concentration, which is caused by thickness variation, pores and cracks stemmed from improper parameter setup or operations lead to the delamination and spallation of the YSZ top coat from bond coat [5−6]. Therefore, it is important to develop a non-destructive evaluation (NDE) method to measure the thickness and characterize the microstructure change. Several NDE techniques have been applied for

quality control of TBCs, such as photoluminescence [7−9] which can determine thermally growth oxides (TGO) residual stress and TGO phase constituents, impedance spectroscopy [10] to evaluate the thickness and composition changes of TGO during oxidation and the effect of cracks within YSZ layer and acoustic emission [11−12] which is used to identify cracking characterization. Recently, PT has been applied for evaluating the thickness and detecting subsurface damage within YSZ layer in TBCs [13−16]. SHEPARD et al [17] attempted to use PT to quantify TBC thickness. MARINETTI et al [18] and CERNUSCHI et al [19] proposed a procedure to reliably discriminate thickness changes and real defects based on the date of PT. However, abnormal microstructure with an emphasis on pores and cracks and its thermal properties had not been investigated. In this study, PT was used to evaluate the microstructural evolution of TBCs. The relationship between the microstructure of TBCs and its thermal properties that affects the temperature distribution of the surface was investigated for providing the relevant knowledge.

Foundation item: Project (61079020) supported by the National Natural Science Foundation of China Corresponding author: Cun-lin ZHANG; Tel: +86-10-82317426; Fax: +86-10-68980976; E-mail: [email protected]

Shi-bin ZHAO, et al/Progress in Natural Science: Materials International 21(2011) 301−306

302

2 Experimental 2.1 Sample preparation The substrates were cut into button-shaped coupons with 30 mm in diameter and 3 mm in height from a wrought sheet of nickel base super alloy. These coupons were grit-blasted, using 250 μm alumina grit, to obtain a sharp-peaked surface contour with a roughness average of 4−5 μm, in order to improve the adherence of the coating. The bond coat (Ni-25Cr-5Al-0.5Y) and the top coat (8%Y2O3-ZrO2) were deposited onto the surfaces of the specimen substrate by high-velocity oxygen fuel (HVOF) spray and air plasma sprayed (APS). The thickness of top coat (YSZ) is presented in Table 1. Because the YSZ layer is optically translucent [20], optical properties will also affect the PT data. To avoid the optical translucency issue, enhance infrared emissivity and optical absorption, TBCs were coated by a thin washable black paint before the test. Table 1 Thickness of YSZ Sample No.

T11

YSZ/μm

150

T1, T2 T4, T9,T10 190

200

T3, T5, T7, T8

T6

250

300

2.2 Pulsed thermography experiment The pulsed thermography system used in this investigation, EcothermTM, utilizes xenon flash lamps as a heat source to launch a pulse (1ms) of heat into the specimen surface. A 320×256 pixels infrared camera, operating within 8−9 μm spectral region (thermal infrared), captures the real-time surface temperature distribution and heat decay of the specimens at a rate of 60 frames/s. Acquisition time is 8 s. The thermal contrast from the raw IR image corresponds to the surface temperature and is converted to thermal response amplitude, a unit-less and relative value within the specimen, related to the temperature of the specimen. The apparatus is shown schematically in Fig. 1. After pulsed thermography measurement, microstructures of the specimens were analyzed by a J840 scanning electron microscope (SEM) equipped with energy dispersive spectroscopy (EDS). 2.3 Logarithmic peak second-derivative (LPSD) method of PT Thermographic signal reconstruction (TSR) proposed by SHEPARD et al [21−24] is a processing technique that uses polynomial interpolation to allow increasing the spatial and the temporal resolution of a thermogram sequence, while reducing the amount of data to be analyzed. Logarithmic peak second-derivative method is deduced form the TSR method [25]. Sheperd

Fig. 1 Schematic diagram of pulsed thermography, PC supplied with data acquisition card and EchoTherm 8.0 data processing and analysis software

et al determined that the second derivative of the surface temperature in the logarithmic scale contains a peak that can be used to determine the defect depth [26] and the thickness of the material [17]. TSR is based on the assumption that temperature profiles for non-defective areas follow the decay curve given by the one-dimensional solution of the Fourier diffusion equation for an ideal pulse uniformly applied to the surface of a semi-infinite body [27]. When pulsed thermal energy is applied, a thin layer of material on the TBCs sample surface is instantaneously heated to a high temperature. Heat conduction then takes place from the heated surface to the interior of the sample, leading to a continuous decrease of the surface temperature. The decay of the surface temperature T with time t was determined by PARKER et al [28] and is expressed as

T (t ) =

∞ ⎛ n 2 π 2 ⎞⎤ Q ⎡ ⎢1 + 2∑ exp⎜⎜ − 2 αt ⎟⎟⎥ ρCL ⎣⎢ L n =1 ⎝ ⎠⎥⎦

(1)

where Q is the heat deposited on the surface, ρ is density, C is specific heat, α is thermal diffusivity, and L is sample thickness. In deriving Eq. (1), it was assumed that the thermal flash is instantaneous, the heat is absorbed within a surface layer of negligible thickness, and there is no heat loss from the sample surfaces. Eq. (1) is the exact solution for one-dimensional (1D) heat conduction within a plate of uniform thickness under ideal flash thermography conditions as described above. The curve of the natural logarithm of Eq. (1) initially follows the falling straight-line behavior of the ideal (no interface) case until a particular time, where it breaks away and becomes a horizontal line, which indicates termination of the one-dimensional diffusion process [23]. The first derivative curve of the natural logarithm of Eq. (1), calculated with respect to ln t, is a straight line with derivative value of −0.5, followed by another straight line with value 0. The second derivative curve of the natural logarithm of Eq. (1) is

Shi-bin ZHAO, et al/Progress in Natural Science: Materials International 21(2011) 301−306

d 2 (ln T ) d(ln t ) 2

2

t dT t 2 ⎛ dT ⎞ t 2 d 2T = − 2⎜ ⎟ + T dt T ⎝ dt ⎠ T dt 2

where Q 2ω dT =− ρCL t dt

∞

∑ n 2 e −n ω 2

(2)

(3a)

n =1

and d 2T dt

2

=

Q 2ω 2 ρCL t 2

∞

∑ n 4 e −n ω 2

(3b)

n =1

Note that the second derivative in the logarithmic scale in Eq. (2) is dimensionless variables. And the result was plotted in Fig. 2, which shows that the second derivative is zero at early and later time periods, and reaches a maximum in between the early period, where normal diffusion reigns, and the later period, where one-dimensional diffusion has ceased [23].

303

adiabatic case, its shape and amplitude are also invariant with respect to the material composition of the sample. Only the time at which the peak occurs will vary, based on either the thickness or thermal diffusivity of the sample [23]. To verify the logarithmic derivative behavior of a slab, samples were inspected using PT. The experiment equipment and parameter settings were the same as description in section 2.2. A stainless steel plate with dimensions of 5 cm × 5 cm and thicknesses of 2.0, 2.5, 3.0 and 3.5 cm, respectively was prepared by painting the front surface with washable black poster paint to increase emissivity and optical absorption. An 8-s data sequence was acquired for the stainless steel plate, the second derivative curve was shown in Fig. 2. It is seen that the time at which the peak occurs will vary, based on the thickness of the specimen.

3 Results and discussion 3.1 Results of thickness measurement and analysis For the same TBCs specimen，the value of 1/πα is a constant but not clearly defined due to the variation of the manufacturing conditions and the material performances. 1/(πα) is treated as a coefficient k, and then Eq. (4) is simplified as t peak = kL2

Fig.2 Second derivative curves for stainless steel plates with various thickness

When Eq. (2) is solved by a numerical method, the maximum second derivative d2(ln T)/d(ln t)2 is 0.473 66 and it occurs when the non-dimensional time ω is equal to π. Therefore, the peak second-derivative time tpeak is L2 (4) πα The time regime prior to the second derivative peak describes uninterrupted diffusion, and the period after the peak describes a situation where any cooling that occurs will be via convection, radiation, or lateral diffusion. Of course, in a real sample, the back wall interface is not likely to be perfectly isolated, so that some heat will pass through the interface. However, much can be learned from this idealized scenario. The logarithmic second derivative has several remarkable properties. It is invariant with respect to the input energy amplitude (differentiation of the natural logarithm of Eq. (1) with respect to ln t removes dependence on Q). In the tpeak =

(5)

Ten standard samples with the known YSZ thickness were made to calibrate k value. The peak time obtained from the second derivatives curve and the known YSZ thickness of all standard samples are substituted into Eq. (5) to get calibration value k= 7.793×10−6. The second derivatives curves of all specimens are plotted in Fig. 3. The behavior of the second logarithmic derivative is first described for an adiabatic slab [23], in which case, the derivative peak is positive, indicating termination of the diffusion process. For a ceramic coating on a metallic substrate, the thermal conductivity of the back wall material is higher than the coating, so that heat will pass through the interface between the coating and the substrate. This case is contrary to the case of the second logarithmic derivative which is described in Ref.[25]. The peak of the second derivative curve should be the negative, which correlates well with thickness [17]. The accuracy of the derivative calculation is likely to be adversely affected by noise and the sampling interval, which will be determined by the camera frame rate. Also, amplitudes will vary in the nonadiabatic case according to the ratio of the thermal effusivity of the backing material to that of the slab material. The peak secondderivative time which occurs in the presence of the interface between YSZ and bond coat can be determined

Shi-bin ZHAO, et al/Progress in Natural Science: Materials International 21(2011) 301−306

304

deposition techniques and parameters exists at the YSZ/bond coat interface. According to the thickness of YSZ changing from 150 μm to 300 μm, the turbidity experiments show that extreme fluctuating is 15−30 μm. For this reason, it seems to be normal that the relative error is less than 10%. On the other hand, it may suggest that there exists an abnormal microstructure in the body of sample. As shown in Table 2, the relative errors of T4, T6, T7 and T10 are more than 10%. Consequently, it needs to make further investigation in order to determine the abnormal microstructure by thermal images and SEM. Fig. 3 Second derivative curves of temperature as function of time in lg-lg scale for TBC specimens

from curves are listed in Table 2 and be used to calculate thickness using Eq. (5) and k=7.793×10−6. The calculated and the actual value of YSZ thickness obtained by SEM are listed in Table 2. The inherent fluctuation of thickness influenced by

3.2 Thermal images and microstructure analysis Analysis of PT data is performed on either a local or global scale. In local analysis methods, which are simpler and more widely used, individual images taken from the cooling time sequence are examined to find anomalous abnormal spots that are indicative of subsurface defects. Figure 4 shows the thermal images collected at 0.25 s after thermal flash heating. It can be

Table 2 Peak time of second derivative curve, calculated and actual values of YSZ thickness and relative error Sample No.

Peak time/s

Calculated thickness, D1/μm

Actual thickness, D2/μm

Relative error, [(D1 − D2 ) ⋅ D2−1 ] / %

T1

0.289

192.57

200.41

−3.9

T2

0.340

208.88

224.30

−6.9

T3

0.527

260.05

288.37

−9.8

T4

0.561

268.31

168.09

59.6

T5

0.374

219.07

205.14

6.8

T6

0.697

299.06

252.46

18.5

T7

0.459

242.69

285.29

−14.9

T8

0.340

208.88

203.12

2.8

T9

0.306

198.16

195.78

1.2

T10

0.272

186.82

235.78

−26.2

T11

0.323

203.59

200.37

1.6

Fig. 4 Thermal images of TBCs specimens (T4, T6, T7, T10, and T11) after 0.25 s thermal injection

Shi-bin ZHAO, et al/Progress in Natural Science: Materials International 21(2011) 301−306

noticed that temperature distribution on its surface is abnormal in some parts of T4, T6, T7 and T10 samples; however, the relative error of the temperature distribution of the other sample (for example T11) less than 10% is normal. Because of the surface treatment before testing, these areas of abnormal temperature distribution are not affected by the surface condition. According to the thermal conduction theories of isotropic materials, it is concluded that the thermal conductivity of these areas is higher than that of normal areas. Thermal contact resistance between YSZ and bond coat can vary with presence of significant pores and cracks, which can give rise to variation of heat transfer rate. The surface temperature distribution of the YSZ topcoat will be abnormal, which is caused by the local change of the

305

heat transfer rate. During the spraying, improper experimental parameters and operations can easily result in pores and cracks at the YSZ/bond coat interface. In order to further investigate the defects, microstructure analysis was carried out by scanning electron micrographs. For sample T11, whose surface temperature distribution is normal, displays no indication of pores and cracks. Except for samples T4, T6, T7, and T10, no large porosity, pore size and cracks are present in the cross-sectional micrographs (Fig.5). The thermal images of TBC samples are in good agreement with the corresponding cross-sectional micrographs. This shows that pores and cracks can cause abnormality temperature distribution, and also identifies that the PT method is reliable for detecting pores and cracks.

Fig. 5 Cross-section morphologies of TBCs (Cracks and pores close to the interface are clearly visible of samples, as pointed out by arrows): (a) T4; (b) T6; (c) T7; (d) T10; (e) T11

Shi-bin ZHAO, et al/Progress in Natural Science: Materials International 21(2011) 301−306

306

4 Conclusions Non-destructive evaluation of as-sprayed APS TBCs was carried out by pulsed thermography. The results showed that pulsed thermography can be employed to evaluate the quality. Thermal images corresponding to subsurface pores and cracks documented by microstructural analysis contain significant information about defective position and degree. By means of random sampling, Logarithmic peak second-derivative method will be applied to direct, quantitative measurement of the thickness during production. Homogeneity for a large area of as-sprayed TBCs can be qualitatively checked with the aid of thermal images. Combined methods of qualitative analysis and quantitative measurement of pulsed thermography can realize the target of quality evaluation for as-sprayed TBCs. It provided a fast and convenient non-contact way to control TBCs quality, and guide for fabrication of the TBCs. However, more research work is needed, pulsed thermography can be utilized in industrial and engineering fields.

Acknowledgements The TBCs samples and SEM images were supplied by Chinese Academy of Agricultural Mechanization Sciences.

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

References [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

GUO H B, VAßEN R, STOVER D. Thermophysical properties and thermal cycling behavior of plasma sprayed thick thermal barrier coatings [J]. Surf Coat Technol, 2005, 192: 48−56. BEELE W, MARIJNISSEN G, VAN LIESHOUT A. The evolution of thermal barrier coatings-status and upcoming solutions for today’s key issues [J]. Surf Coat Technol, 1999, 120/121: 61−67. XU Qiang, PAN Wei, WANG Jing-dong, et al. Preparation and thermophysical properties of Dy2Zr2O7 ceramic for thermal barrier coatings [J]. Mate Lett 2005, 59: 2804−2807. NI Li-yong, WU Zi-long, ZHOU Chun-gen. Effects of surface modification on isothermal oxidation behavior of HVOF-sprayed NiCrA1Y coatings [J]. Progress in Natural Science: Materials International, 2011, 21: 173−179. BAO Z B, WANG Q M, LI W Z, et al. Preparation and hot corrosion behavior of an Al-gradient NiCoCrAlYSiB coating on a Ni-base superalloy [J]. Corros Sci, 2009, 51: 860−867. SREEDHAR G, RAJA V S. Hot corrosion of YSZ/Al2O3 dispersed NiCrAlY plasma-sprayed coatings in Na2SO4-10wt.% NaCl melt [J]. Corros Sci, 2010, 8: 2592−2602. SELCUK A, ATKINSON A. The evolution of residual stress in the thermally grown oxide on Pt diffusion bond coats in TBCs [J]. Acta Mater, 2003, 51: 535−549. NYCHKA J A, CLARK D R, SRIDHARAN S. NDE assessment of TBCs: An interim report of a photo-stimulated luminescence ‘round-robin’ test [J]. Surf Coat Technol, 2003, 163/164: 87−94. ZHOU Bo, XIAO Zhi-song, HUANG An-ping, et al. Effect of Tm-Er concentration ratio on the photoluminescence of Er-Tm:Al2O3 thin

[20]

[21]

[22]

[23]

[24]

[25]

[26] [27] [28]

films fabricated by pulsed laser deposition [J]. Progress in Natural Science: Materials International, 2008, 18: 661−664. HUANG Hui, LIU Chao, NI Li-yong, et al. Evaluation of microstructural evolution of thermal barrier coatings exposed to Na2SO4 using impedance spectroscopy [J]. Corros Sci, 2011, 53: 1369−1374. MA X Q, TAKEMOTO M. Quantitative acoustic emission analysis of plasma sprayed thermal barrier coatings subjected to thermal shock tests [J]. Mater Sci Eng A, 2001, 308: 101−110. KUCUK A, BERNDT C C, SENTURK U, et al. Influence of plasma spray parameters on mechanical properties of yttria stabilized zirconia coatings: II. Acoustic emission response [J]. Mater Sci Eng A, 2000, 284: 41−50. BISON P G, MARINETTI S, GRINZATO E, et al. Inspecting thermal barrier coatings by IR thermography [C]// Proceedings of SPIE. Bellingham, 2003, 5073: 318−327. NEWAZ G, CHEN Xiao-qun. Progressive damage assessment in thermal barrier coatings using thermal wave imaging technique [J]. Surf Coat Technol, 2005, 190: 7−14. CHAUDHURY G M, NEWAZ S Q, NUSIER T A. Interfacial damage in EB-PVD thermal barrier coatings due to thermal cycling [J]. Mater Sci Eng A, 1997, 231: 34−41. CHAUDHUP Z A, AHMED T, NEWAZ G M, et al. Evaluation of bond integrity of TiO2-Y2O3 sensor using thermal wave imaging technique [J]. Mater Lett, 1998, 34: 76−80. Shepard s m, HOU Yu-lin, LHOTA J R, et al. Thermographic measurement of thermal barrier coating thickness [C]// Proceedings of SPIE. Bellingham, 2005: 407−410. MARINETTI S, ROBBA D, CERNUSCHI F, et al. Thermographic inspection of TBC coated gas turbine blades: Discrimination between coating over-thicknesses and adhesion defects [J]. Infrared Phys Technol, 2007, 49: 281−285. CERNUSCHI F, MARINETTI S. Discrimination between over-thickness and delamination of thermal barrier coatings by apparent thermal effusivity thermographic technique [J]. J Therm Spray Technol, 2010, 19: 958−963. MANARA J, ARDUINI-SCHUSTER M, RÄTZER-SCHEIBE H J, et al. Infrared-optical properties and heat transfer coefficients of semitransparent thermal barrier coatings [J]. Surf Coat Technol, 2009, 203: 1059−1068. SHEPARD S M, AHMED T, RUBADEUX B A, et al. Synthetic processing of pulsed thermographic data for inspection of turbine components [J]. Insight: Non-destructive Testing and Conditioning Monitoring, British Inst of NDT, 2001, 43(9): 587−589. SHEPARD S M. Flash thermography of aerospace composites [C]// Proceedings of the 4th Pan American Conf for NDT. Buenos Aires, Argentina-Open archives on http://www.ndt.net/article/panndt2007/ papers/132.pdf. 2007−10−22. SHEPARD S M, HOU J, LHOTA J R, et al. Automated processing of thermographic derivatives for quality assurance [J]. Opt Eng, 2007, 46(5): 0510081−6. SHEPARD S M, LHOTA J R, AHMED T. Measurement limits in flash thermography [C]// Proceedings of SPIE Thermosense 31th Conference, 2009: 72990T1−T7. SHEPARD S M, LHOTA J R, RUBADEUX B A, et al. Reconstruction and enhancement of active thermographic image sequences [J]. Opt Eng, 2003, 42(5): 1337−1342. SUN J G. Analysis of pulsed thermography methods for defect depth prediction [J]. Journal of Heat Transfer, 2006, 128: 329−338. CARSLAW H S, JAEGER J C. Conduction of heat in solids [M]. 2nd edition. Oxford: Clarendon Press, 1959: 231−232. PARKER W J, JENKINS R J, BUTLER C P, et al. Flash method of determining thermal diffusivity, heat capacity, and thermal conductivity [J]. J Appl Phys, 1961, 321: 1679−1684.