Interface disbond in thermal barrier coatings (TBCs) is one of the key issues that cause their premature failure. In general, blind hole defects are often used as substitutes in transient thermography. The linear laser fast scanning thermography (LLFST) method was developed in this study and combined with several post-processing algorithms to accurately detect blind hole defects in TBCs. Through numerical simulation and experimental verification, a unique thermal response characteristic of blind holes in the cooling phase, namely a distinct “tailing” phenomenon, was summarized and utilized to recognize small defects. Validation tests indicated that blind holes with diameters of 1, 2, and 3 mm and artificial disbonds with diameters of 2 and 3 mm in TBCs are detected with high efficiency.
© 2017 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Thermal barrier coatings (TBCs), as effective thermal protection, allows some parts and structures of an aircraft to work in its complex high-temperature environment [1–4]. This coating structure is a typical multilayer system, which usually consists of a super-alloy substrate, bond coating, and ceramic coating. Each coating has significantly different physical, thermal, and mechanical properties. However, the fabrication procedure of such complex structures and the extremely harsh working conditions may result in the presence of disbond defects in TBCs, which is one of the reasons causing their premature failure. Therefore, nondestructive testing (NDT) of defects in TBCs has attracted more and more attention of researchers, and relevant detection technologies have been introduced.
Due to the advantage of large-area inspection, infrared thermography [5–7] is widely used in nondestructive testing of disbond defects in TBCs. For different types of defects and tested materials, various thermal excitation sources have been developed. Using ultrasound infrared thermography (UIRT), Choi  measured the size of crack defects. Microwave thermography  has a good thermal excitation effect on the ceramic and wood products, but it will be mostly reflected by metal interfaces, which restrict its application in NDT of TBCs. Biju  used eddy current thermography (ECT) combined with a genetic algorithm to inspect the defect size, but deeper defects are not detectable due to skin effect. Compared to other thermal excitation methods, optical excitation is simple and practical. IR pulsed thermography is a universal method for detecting defects [11–13]; however, it must be pointed out that, when a pulsed heat source (like a quartz lamp) is switched off, and thermal image data is collected and analyzed, the inevitable afterglow effects will have varying degrees of impact on the final experimental results . In addition to the plane heat source, the 3D modeling of line heating source was also thoroughly described  and this thermal excitation has been used for large-area testing of aircraft fuselage delamination . What’s more, the line scanning can add spatial information, and the problem that the full-field image in the flash thermography is affected by varying luminosities will not occur. But in order to improve detection sensitivity, some post-processing algorithms need to be developed to eliminate noise. In this case, a laser provides an ideal heat source with a small area and high heat flux. Although there are few applications for TBCs, the laser spot thermography has demonstrated clear results in the detection of surface cracks [16–18]. In addition, the moving heat sources generated by lasers can physically solve the afterglow effect and reduce the requirements of post-processing.
In addition, since real disbonds in TBCs cannot be easily obtained, blind hole defects are often used to substitute disbonds in transient thermography [19, 20]. In particular, Ptaszek  compared flat-bottomed holes (FBHs) with artificial disbonds and concluded that the thermal response of a blind hole defect is significantly stronger than the thermal response of disbonds of the same size, so a real disbond can be substituted with a smaller-diameter FBH. Tang  successfully detected FBH defects with an area of 6.25 mm2 and a depth of 0.2 mm by using principal component analysis algorithm in pulse light incentive method.
In this paper, with a linear laser fast scanning system and a long-wavelength IR camera, the linear laser fast scanning thermography (LLFST) NDT method is proposed to rapidly detect artificial blind hole defects in TBCs. Through numerical simulation, and experimental verification, the unique thermal response characteristic of the cooling phase, an obvious ”tailing” phenomenon stimulated by the linear laser is summarized and utilized. Thus, using our method in combination with several post-processing algorithms developed here to improve the signal-to-noise ratio of thermal images, blind hole defects can be accurately inspected by analyzing this thermal response characteristic. Compared to excitation with an area light source, the power of the linear laser is adjustable and more concentrated, so this method has a higher sensitivity and does not require spraying black paint to enhance optical absorption. In addition, since the linear laser targets different areas at different moments of time, the afterglow effect is avoided. Furthermore, a fabrication method for artificial disbond defects is developed. Validation experiments show that blind hole defects with diameters of 1, 2, and 3 mm and artificial disbonds with diameters of 2 and 3 mm in TBCs can be quickly detected.
The rest of the paper is organized as follows. Thermal response characteristics of disbonds obtained by linear laser scanning are discussed in Section 2. The post-processing algorithms for thermal images are presented in Section 3. The validation tests, results, and discussion are provided in Section 4. Finally, the conclusions are outlined in Section 5.
2. Thermal response characteristics of disbonds obtained by linear laser scanning
When a line heat flow transfers from the surface to the bottom of a TBC system, the surface temperature distribution at the TBC defects will be abnormal due to the fact that defects change the thermal parameters of specimens . To summarize the thermal response characteristics at the defects, the process of linear laser scanning of the specimen surface with a blind hole [Fig. 1(a)] and an artificial disbond [Fig. 1(b)] are simulated based on the heat conduction theory using Abaqus. The size of these two TBC models is 30 × 30 × 2.5 mm3. The thicknesses of the substrate, ceramic coating, and bond coating are 2 mm, 400 μm, and 100 μm, respectively. The thermal and physical parameters of the TBCs are shown in Tab. 1 . Figure 1(a) shows a 2 mm diameter blind hole defect on the TBC specimen with a depth of 1.9 mm. The remaining wall thickness of the blind hole is therefore 0.1 mm. A sufficiently long linear laser is then fast scanning on the coating structure with an FBH. In Fig. 1(b), the blind hole in the above model is replaced by an air layer that is used to simulate a disbond, with the diameter and thickness of the air layer of 2 mm and 0.1 mm, respectively.
The two simulation results are shown in Fig. 2. Based on the analysis of these figures, two distinct thermal response characteristics in time and space can be summarized. First, the temperature at the defect region is higher than that of the other (normal) region when the linear laser is scanning over the defect. Second, the temperature image shows a dragging effect near the defect after scanning and a “tailing” phenomenon is distinctly observed behind the linear laser for a period of time. From the results, both types of the defects visually produce the same temperature distributions and their thermal response characteristics are alike, which also proves that blind hole defects can be used to substitute disbonds in transient thermography, but the temperature at the blind hole is slightly higher than that at the artificial disbond defect (about 1.4°C temperature difference), which may result in the detection sensitivity of blind hole defects superior to that of disbond defects.
3. Post-processing algorithms for thermal images in the LLFST method
3.1 Constructing and subtracting an oriented carrier temperature field
For better identification of the defect information, the raw thermal images of the thermal response of the defects are not sufficient, and the corresponding reference images without the defects are required. Therefore, the post-processing algorithm of constructing and subtracting an oriented carrier temperature field is developed here to acquire the reference images corresponding to the raw thermal images of each frame, such that no additional experiment to obtain the reference thermal images without defects is needed to be carried out.
The detailed process of this algorithm is as follows. A raw thermal image with the thermal response of the defects is selected as an example. The temperature values of all pixels in this image along the linear laser direction are added up and averaged. Since the defect area is much smaller than the normal region, the calculated average can be considered as the temperature values for the linear laser scanning the normal area. The average values are also the temperature values of the corresponding points in the carrier temperature image. Thus, an oriented carrier temperature field of the selected image is constructed. The position of the blind hole defect can be obtained by subtracting the oriented carrier temperature field from the raw thermal image. We use this algorithm on all the raw thermal images in the time series to obtain their respective carrier temperature images and subtract them.
An Abaqus simulation is performed to validate the effectiveness of this post-processing algorithm. Figure 3(a) is the raw temperature image with the thermal response of the defects, and Fig. 3(b) is the oriented carrier temperature field reconstructed by using the developed algorithm. The position of the defect is then clearly displayed in Fig. 3(c) by subtracting the image in Fig. 3(b) from that in Fig. 3(a).
Before each raw thermal image is processed, pre-excitation image is subtracted to remove the background noise. After the respective carrier temperature fields are subtracted from all the raw thermal images in the time series, the resultant thermal response image of the blind hole defect in the whole field can be obtained by accumulating the subtracted thermal images at all moments.
3.2 Three-moment windowing amplitude method for noise suppression
After the linear laser scans over a certain region, the heat within this region begins to dissipate, so that the final temperature in that region is the same as that of the non-scanned region when the linear heating source is far away. If the temperature field of this region is still superposed with the entire field at all following moments, an additional background noise will be accumulated. Therefore, the three-moment windowing amplitude method (before scanning, scanning, and after scanning) is developed to eliminate such noise accumulation. This method is applied only near the linear heat source, and when the linear heat source is far away, the superposition algorithm is stopped.
The implementation of this three-moment windowing amplitude method is as follows. Consider a raw thermal image with the thermal response of a defect as an example, as shown in Fig. 4. A fixed rectangular region near the linear laser is selected as a computational window in this image, with an area larger than the area of the linear laser, indicated as “scanning” thermal image. The raw thermal images of the other two moments are selected at a fixed interval t0 according to the location of the preselected computational window. In the “before scanning” thermal image, the linear laser is below the computational window, while at the “after scanning” thermal image, the linear laser is above the computational window. The temperature amplitude inside the computational window is related to the temperature of the corresponding pixels in the two thermal images. The amplitude field outside the window is set as empty. We use this method for all the raw thermal images in the time series to obtain their respective temperature amplitude distribution. The calculation formula is shown in Eq. (1).
Correspondingly, the oriented carrier temperature field for the respective temperature image is modified to the oriented carrier amplitude field along the linear laser direction. The specific formula is expressed asEq. (1), N is the number of pixels, m and n are the pixel numbers in the x and y coordinate axes, respectively, and Tfit-a(t)(x, y) is the oriented carrier temperature amplitude distribution corresponding to the raw temperature image of each frame.
After using the raw temperature distribution and subtracting the corresponding carrier amplitude fields, the resultant thermal amplitude image of the blind hole defect in the whole field can be accurately obtained by superposing the subtracted temperature amplitude images at all moments.
3.3 Variable weight edge filtering window to eliminate the edge effect
The edge effect causes a larger error at the edge of an object, which has significant influence on the experimental results. This is because the thermal boundary of the edge position is different from that of the internal area, and the temperature distribution near the edge area is more complex, with higher temperatures closer to the boundary. Given that there are no defects at the edge, a variable weight edge filtering window algorithm is used to eliminate the abnormal high temperature of the edge area in the thermal images. The resultant thermal amplitude image in the whole field is modified as follows:
4. Validation tests and discussion of the results
4.1 Specimens and the LLFST system
In this experiment, a series of TBC specimens are used, and one 30 mm diameter specimen is selected as an example to display the result. It consists of a nickel-based alloy substrate, bond coating, and ZrO2 coating, with thicknesses of 2 mm, 100 μm, and 400 μm, respectively. Several FBHs with diameters of 1, 2, and 3 mm are drilled in the alloy substrate using a flat drill to substitute disbonds. The diagram of the specimen is shown in Fig. 5. The depth of the blind hole is controlled to be below 1.9 mm to avoid destroying the coating. Thus, the remaining wall thickness of the hole is about 0.1 mm. The material parameters are the same as for the simulation, as shown in Tab 1.
An LLFST system is established in this investigation with a semiconductor laser marking machine and a long-wavelength IR camera. The experimental apparatus is shown in Fig. 6. Fast linear laser excitation can be applied to the TBC specimen surface by the laser marking machine, and, when the linear laser moves on the specimen surface, the IR camera begins to collect thermal images.
The power of the laser is adjustable, and the maximum power is 10 kW. The laser spot with a diameter of ~50 µm is controlled to move at a speed of 1000 mm/s in a straight line and can be viewed as a linear laser because the moving speed is sufficiently high. The length of this linear laser is 40 mm, and the scanning speed in the direction perpendicular to the line is ~1.25 mm/s. The scanning speed of the linear laser is much smaller than the moving speed of the laser spot, so, ignoring the movement time of the laser point, thermal excitation by the linear laser scanning can be achieved.
In most thermographic NDT applications, the wavelength of the IR camera is not an important consideration, since sample surfaces are generally IR opaque. However, TBCs are atypical, which is caused by part translucency to visible, near and middle IR radiation . So the incident pulse may penetrate the coating and the photo-thermal effect happens through the volume of TBCs. Under the circumstances, a long-wavelength camera can be able to mitigate the effects of TBC translucence [25, 26]. Therefore in the LLFST system, a long-wavelength (7.5–14 µm) IR camera (VarioCAM® hr) with a maximum resolution of 640 × 480 pixels is used to record thermal images in a time series. The maximum acquisition frame rate of the IR camera is 50 Hz, exactly 20 ms per frame, and the measured temperature range is −40 to 1200 °C. Semiconductor laser used here with a wavelength of 1064nm is not harmful to the IR camera, so that the camera does not require additional protection.
4.2 Experimental results and discussion
Due to the presence of blind holes, the thermal diffusion at the specimen is changed, and the temperature at the defect is higher than that of the normal area when the linear laser is scanning over the specimen surface. Figure 7(a) shows a collected raw thermal image with the defect during the linear laser scanning. As shown in the picture, a unique “tailing” phenomenon in the cooling phase can be clearly observed. To confirm this phenomenon, two points, A and B, are selected in the defect and normal regions. In the temperature–time plots of A and B [Fig. 7(b)], a large temperature difference is seen during the temperature dissipation, which confirms the “tailing” phenomenon at the defect.
Based on the unique thermal response characteristic, the constructing and subtracting an oriented carrier temperature field post-processing algorithm is used to obtain the positions of the defects. The detailed process of the algorithm is as follows: (1) the raw temperature image of each frame is collected by the IR camera, and the pre-excitation image is subtracted [Fig. 8(a)]; (2) the linear laser center line is fitted [Fig. 8(b)] to correct the modified shape of the linear laser due to the too high capturing rate of the IR camera; (3) the area above the fitted center line is clipped [Fig. 8(c)] to retain only the useful data points in the cool phase; (4) the oriented carrier temperature image corresponding to the raw temperature image of each frame is constructed [Fig. 8(d)]; (5) the constructed carrier image is subtracted from the raw thermal image [Fig. 8(e)]; (6) the resultant thermal image with the defects in the whole field is produced by superposing all such subtracted images at all moments [Figs. 8(f) and 8(g)]; (7) the final black and white binary image after threshold extraction is obtained [Fig. 8(h)].
As shown in Fig. 8(h), the defects with diameters of 2 mm and 3 mm can be clearly displayed in the final binary image, but the 1 mm diameter defect is ambiguous and cannot be fully detected. This is because the area of the first two defects is larger, and the excited thermal response characteristic is more obvious. On the other hand, the area of the third defect is so small that its weak thermal response is dominated by the background noise. In general, the temperature value is closely related to the diameters of blind hole defects. The bigger the diameter of a blind hole defect, the higher the temperature.
In order to eliminate such noise and detect smaller defects, the three-moment windowing amplitude method is used to further process the thermal images. The temperature field of each image is no longer available, and the calculated temperature amplitude distribution in the thermal images of each frame is used. Meantime, the calculation area is not the entire thermal image, but a computational window below the linear laser (the “tailing” phenomenon appears in the cooling phase). The detailed process is as follows. After calculating the temperature amplitude of every pixel and constructing the oriented carrier amplitude in the computational window, the amplitude image with the thermal response of the defect is obtained by subtracting the oriented carrier [Fig. 9(a)]. By superposing such amplitude images at all moments, the resultant thermal image with the defects in the whole field is obtained [Figs. 9(b) and 9(c)]. The final black and white binary image after threshold extraction is shown in Fig. 9(d).
Comparing the images shown in Fig. 8(h) and Fig. 9(d), the 1 mm diameter defect that belongs to the suspected defect region in Fig. 8h can be clearly displayed in Fig. 9(d). Obviously, the ability to detect defects is improved through the use of the three-moment windowing amplitude method.
4.3 Experimental contrast analysis between blind holes and artificial disbonds
In order to verify that the above algorithms developed to inspect blind hole defects are also applicable to artificial disbond defects, multiple linear laser scanning thermography experiments were performed on blind holes and artificial disbonds, and the experimental results were compared.
In addition, a fabrication method for artificial disbond defects was developed. The detailed process is as follows: (1) a threaded hole is drilled in the TBC substrate; (2) a screw matching the threaded hole is used, with the excess removed by mechanical cutting; (3) after polishing and ultrasonic shot peening, the bond coating and ceramic coating are sprayed on the specimen surface; the thicknesses of the bond coating and ceramic coating are 100 μm and 400 μm, respectively; (4) the screw is loosened to form the air coating, and a TBC specimen with artificial disbond defects is manufactured (Fig. 10). When the screw is fully removed from the threaded hole, a TBC specimen with a blind hole defect is also prepared.
Using this manufacturing method, the sample with artificial disbonds and the sample with blind holes are made, as shown in Fig. 11. The dimensions of their substrates are 50 × 40 × 2 mm3, and the diameters of the defects are 2 mm and 3 mm. Under the same experimental conditions, the two samples are scanned with the scanning system developed in this paper. The detection results of the blind hole and disbond defect measurements are shown in Figs. 11(d) and 11(e).
According to the experimental results, the methods and algorithms developed here can be used for both kinds of defects. The experimental results show that the temperature of the disbond defects is less than that of the blind hole defects, meaning that blind hole defects correspond to a higher detection sensitivity than disbonds defects.
Aiming at NDT of artificial disbonds defects in TBCs, the linear laser fast scanning thermography (LLFST) method is established in this study in combination with some simple but practical post-processing algorithms. The main advantages are as follows.
First, through numerical simulation and experimental verification, a unique thermal response characteristic of blind holes in the cooling phase, namely a distinct ”tailing” phenomenon induced by linear laser fast scanning, is summarized and utilized.
Second, by combining the developed LLFST method with several post-processing algorithms, such as the constructing and removing an oriented carrier temperature field method, three-moment windowing amplitude method, and variable weight edge filtering window, blind hole defects with diameters of 1 mm or more and artificial disbonds defects with diameters of 2 mm or more in TBCs can be accurately detected.
Third, compared with the existing flash thermography, the power of the linear laser is adjustable and more concentrated, so that its sensitivity is higher, and there is no need to spray black paint on the ceramic surface to enhance its optical absorption. In addition, the IR camera does not require additional protection.
Finally, previous studies have proved that blind hole defects can be used to replace disbonds in transient thermography, and our experiments also show that the methods developed here for blind hole defects can be used to detect real disbonds in TBCs, albeit with slightly reduced sensitivity.
National Natural Science Foundation of China (11232008, 11372037, and 11572041).
The authors are grateful all those people involved in the work.
References and links
1. W. Beele, G. Marijnissen, and A. van Lieshout, “The evolution of thermal barrier coatings — status and upcoming solutions for today’s key issues,” Surf. Coat. Tech. 120–121, 61–67 (1999).
2. H. B. Guo, R. Vaßen, and D. Stöver, “Thermophysical properties and thermal cycling behavior of plasma sprayed thick thermal barrier coatings,” Surf. Coat. Tech. 192, 48–56 (2005).
3. R. Vaßen, M. O. Jarligo, T. Steinke, D. E. Mack, and D. Stöver, “Overview on advanced thermal barrier coatings,” Surf. Coat. Tech. 205, 938–942 (2010).
4. N. I. Li-Yong, W. U. Zi-Long, and C. G. Zhou, “Effects of surface modification on isothermal oxidation behavior of HVOF-sprayed NiCrA1Y coatings,” Progress in Natural Science: Materials International 21, 173–179 (2011).
5. B. Franke, Y. Sohn, X. Chen, J. Price, and Z. Mutasim, “Monitoring damage evolution in thermal barrier coatings with thermal wave imaging,” Surf. Coat. Tech. 200, 1292–1297 (2005).
6. B. Ghiassi, S. M. Silva, D. V. Oliveira, P. B. Lourenço, and L. Bragança, “FRP-to-masonry bond durability assessment with infrared thermography method,” Journal of Nondestructive Evaluation 33, 427–437 (2014).
7. S. Marinetti, D. Robba, F. Cernuschi, P. Bison, and E. Grinzato, “Thermographic inspection of TBC coated gas turbine blades: Discrimination between coating over-thicknesses and adhesion defects,” Infrared Phys. Technol. 49, 281–285 (2007).
8. M. Choi, J. Park, H. Park, K. Kwon, and W. Kim, “Study of micro crack detection using a variable ultrasound Infrared thermography.” 6th Pan American Conference for NDT 12-14 August 2015, Cartagena, Colombia (2015).
9. Y. K. Zhu, G. Y. Tian, R. S. Lu, and H. Zhang, “A review of optical NDT technologies,” Sensors (Basel) 11(8), 7773–7798 (2011). [PubMed]
10. N. Biju, N. Ganesan, C. V. Krishnamurthy, and K. Balasubramaniam, “Defect Sizing Simulation Studies for the Tone-Burst Eddy Current Thermography Using Genetic Algorithm Based Inversion,” Journal of Nondestructive Evaluation 31, 342–348 (2012).
11. S. B. Zhao, C. L. Zhang, W. U. Nai-Ming, and H. M. Wang, “Quality evaluation for air plasma spray thermal barrier coatings with pulsed thermography,” Progress in Natural Science: Materials International 21, 301–306 (2011).
12. P. G. Bison, E. G. Grinzato, and V. P. Vavilov, “Inspecting thermal barrier coatings by IR thermography,” Proceedings of SPIE - The International Society for Optical Engineering 5073, 318–327 (2003).
13. J. G. Sun, “Analysis of Pulsed Thermography Method for Defect Depth Prediction,” J. Heat Transfer 128, 329–338 (2006).
14. V. P. Vavilov and R. Taylor, “Theoretical and practical aspects of the thermal nondestructive testing of bonded structures,” Academic Press, Research Techniques in Nondestructive Testing 5, 239–279 (1982).
15. D. F. Woolard and K. E. Cramer, “The thermal photocopier: a new concept for thermal NDT,” Proc. SPIE 5405, 366–373 (2004).
16. A. Rashed, D. P. Almond, D. A. S. Rees, S. Burrows, and S. Dixon, “Crack detection by laser spot imaging thermography,” AIP Conf. Proc. 894, 500–506 (2007).
17. L. Teng, D. P. Almond, and D. A. S. Rees, “Crack imaging by scanning pulsed laser spot thermography,” NDT Int. 44, 216–225 (2011).
18. S. E. Burrows, A. Rashed, D. P. Almond, and S. Dixon, “Combined laser spot imaging thermography and ultrasonic measurements for crack detection,” Nondestruct. Test. Eval. 22, 217–227 (2007).
19. Y. P. Pan, R. A. Miller, T. P. Chu, and P. Filip, “Detection of defects in commercial C/C composites using infrared thermography,” in SEM Annual Conference and Exposition on Experimental and Applied Mechanics (2009).
20. Z. Xu, L. Cheng, H. Mei, X. Chen, and X. Deng, “Thermographic measurement methods for depths of the blind holes drilled in C/SiC composites,” Acta Materiae Compositae Sinica 34, 937–956 (2011).
21. G. Ptaszek, P. Cawley, D. Almond, and S. Pickering, “Artificial disbonds for calibration of transient thermography inspection of thermal barrier coating systems,” NDT Int. 45, 71–78 (2012).
22. Q. Tang, J. Liu, Y. Wang, S. Hao, and L. Chen, “Experimental study of inspection on SiC coated high-temperature alloy plates with defects using pulsed thermographic technique,” Infrared Phys. Technol. 57, 21–27 (2013).
23. China Aviation Materials Handbook (China Standard, 2002).
24. A. Salazar, A. Mendioroz, E. Apiñaniz, C. Pradere, F. Noël, and J. C. Batsale, “Extending the flash method to measure the thermal diffusivity of semitransparent solids,” Meas. Sci. Technol. 25, 035604 (2014).
25. S. M. Shepard and J. R. Lhota, “Thermographic measurement of thermal barrier coating thickness,” Proceedings of SPIE - The International Society for Optical Engineering 5782, 407–410 (2005).
26. G. Ptaszek, P. Cawley, D. Almond, and S. Pickering, “Transient thermography testing of unpainted thermal barrier coating (TBC) systems,” NDT Int. 59, 48–56 (2013).