Detection of Subsurface Defects using Active Infrared Thermography More Info at Open Access Database www.ndt.net/?id=15141 Suman Tewary 1,2,a, Aparna Akula 1,2, Ripul Ghosh 1,2, Satish Kumar 2, H K Sardana 1,2 and Ravibabu Mulaveesala 3 1 Academy of Scientific and Innovative Research, Rafi Marg, New Delhi - 110001, India 2 CSIR-Central Scientific Instruments Organisation, Chandigarh - 160030, India 3 Indian Institute of Technology Ropar, Nangal Road, Rupnagar, Panjab - 140001, India a suman.tewary@csio.res.in Keywords: Active Infrared Thermography, Polynomial fitting, Fast Fourier Transform and Non-destructive Testing. Abstract. Active infrared thermography is a widely used technique in the field of nondestructive testing. The subsurface defects produce non-uniform heat dissipation and this phenomenon is captured by an infrared camera. The phase image gives a prominent description of the defects which are not visible in the captured image sequence. The effect of non-uniform heating and the variation in surface emissivity results in noisy data acquisition. An infrared camera of mid-wave infrared range was used for the study. A frequency modulated sinusoidal signal was used for the heating of the mild steel specimen by two halogen lamps. The source control is attained using a signal generator with frequency of modulation varied from 0.01 Hz 0.1 Hz for the duration of 100 seconds. The images were acquired at a frame rate of 20 Hz. The specimen contains six numbers of circular flat bottom holes, at different depths from the sample top surface. The temporal profile of every pixel location in the infrared image sequence carries a similar pattern of heat dissipation from the surface and requires temperature compensation. The temperature compensated data is used for the Fourier analysis and pulse compression for better visualization of the defects. The correlation image and the phase of correlation image shows promising results. Experimental setup and comparative results are discussed in detail in this paper. The results from the image enhancement method are qualitatively efficient for the detection of subsurface defects. Introduction Active infrared thermography based nondestructive evaluation of structures is a widely used technique in the field of structural engineering. Non-contact and nondestructive testing for the structural evaluation is the current demand for the structural health monitoring. The applications are in mechanical industries, aerospace industries and in civil structures [1]. The thermal convection phenomenon of the material is the driving force for the inspection of structural anomalies. Any object above absolute zero temperature radiates infrared radiations and the infrared camera captures these electromagnetic radiations and produces a thermogram of the test specimen. In active thermography, the infrared images of the test specimen are acquired after excitation using an external heat source and the convection phenomenon produces the heat flow. The non-uniformity in heat dissipation produces a thermal contrast between the defective area and the rest of the specimen. By processing the acquired thermogram sequence, subsurface abnormalities are detected. Researchers across the world are coming up with a variety of approaches using the knowledge of signal processing and image processing to detect these defects. Varieties of approaches are there in active infrared thermography. Three most widely used techniques are: Pulse Thermography (PT), Lock-in Thermography (LT) and Pulse Phase Thermography (PPT) [2,3]. In pulse thermography, the test specimen is heated up with a shortduration and high power pulse. The thermal response of the surface during the cooling process is recorded, which contains the information of defects at different depths. This method is very fast, but is limited by the non-uniform surface emissivity and the requirement of high peak power for deeper defect analysis. Lock-in thermography uses a steady excitation of mono-frequency sinusoidal
thermal wave, which introduces highly attenuated, dispersive thermal waves of the same frequency inside the test specimen. The magnitude and phase information derived from the image sequence in the stationary regime of the heat cycle. The phase image is used for the analysis instead of the magnitude image as the phase image is less sensitive of the surface emissivity, non-uniform heating and extracts deeper defects [1]. In pulse phase thermography, the arrangement is same as pulse thermography, where the phase information is derived using Fourier transform on the temporal profile of each pixel in the thermogram sequence. The detection of subsurface defects depends on the quality of raw thermal image sequence. The thermograms are affected by the variation in surface emissivity and there is a need of reconstruction and filtering algorithms to enhance the subsurface defects detection [4,5]. The pattern of thermal intensity is approximately same for all pixels in the cropped area of the specimen. Temporal profile is compensated using a first order fitting. The phase information of the residual of this fitting will have prominent detection of the subsurface defects. In active infrared thermography for nondestructive testing, high power heat pulse is required for deeper penetration and it may be difficult to provide high peak power. Also defects at different depths may not be detected using a single experimentation. Low frequency and low average power for longer duration can be an alternative solution for the high peak power requirement. To overcome the above limitations frequency modulated thermal wave imaging (FMTWI) is used. In a single experimentation, a low peak power source with a suitable band of frequencies is used to thermally excite the given specimen for the detection of subsurface anomalies in a single experimentation [2,3,6]. Theory In FMTWI, thermal waves are generated using a frequency modulated heat source and these waves diffuses into the specimen. Similar time varying temperature profile is generated for the surface. The thermal diffusion for a material can be explained by the heat conduction equation given by (1) Where α is the thermal diffusion coefficient, T is the instantaneous temperature, x is the direction of heat flow and t is the time. Solving the equation with the stipulated boundary conditions the temperature can be shown as ( ( )) (2) Where B/τ is the frequency sweep rate of the chirp and τ is the duration of excitation. The material of the specimen under test attenuates the penetrating thermal waves. The depth at which the energy of the thermal wave attenuates to 1/e times the energy at the surface is called the thermal diffusion length which can be given by The depth resolution for defects at different depth can be improved by using a frequency range in a single experimentation cycle of FMTWI. The diffusive thermal wave encounters reflections due to the presence of defects which produce the mismatch in thermal impedance of the material. The temperature response on the surface due to the reflected and surface existing waves results in a delayed thermal wave depending on the depth of the defects, with respect to the incident thermal wave [2,3]. The defects produce hot or cold spots and using a suitable post processing a thermal contrast can be generated to observe the defects accurately. The defect detection is carried out using the phase analysis where the Fourier transform of the temporal profile gives the frequency domain information. The conventional method of trend removal is to fit a first order polynomial and subtract it from the original temporal profile for every pixel of the infrared image sequence. This method is easy to compute and carries the exact information of the temperature profile which is used for further processing. Phase based analysis is (3)
carried out using the fast Fourier transform (FFT) of the temporal thermal profile of every pixel information in the sequence. (4) Where n is the number of sample in FFT and resultant real and imaginary components are Re and Im. The phase information is calculated using these components. Pulse compression has been widely used to enhance the target detection in radar [7]. The same idea has been utilized in active infrared thermography. In active infrared thermography the thermal response gets attenuated and delayed depending upon the depth of subsurface defects. The pulse compression approach uses the cross correlation of the temporal profile of a non-defective pixel with the temporal profiles of all other pixels [8]. The cross correlation is defined as (6) The pulse compression approach concentrates the energy in the main lobe by reducing the sizes of the side lobes and improves the resolution of detection. The resultant compressed profiles concentrate most of the energy into the main lobe of the pseudo pulse with a delayed response based on the depth of defects. The pulse compression approach also improves the signal to noise ratio (SNR). (5) Methodology A mild steel specimen of thickness 2 cm with cylindrical voids of diameter 1 cm and thickness 1mm is used as shown in Fig. 1. This specimen has been excited with the frequency modulated heat source producing a heat flux of 2000 W/m 2 with a linear frequency variation from 0.01 Hz to 0.1 Hz in 100 seconds duration. Figure 1: The mild steel specimen with cylindrical voids numbered as 1-6 at different depths The setup as shown in Fig. 2 contains a function generator to provide the required modulated frequency for the halogen lamps. The infrared camera from CEDIP infrared system of mid-wave infrared region acquires the thermal profile of the specimen for the required time duration of 100 seconds. The data acquisition and processing is carried out using a computer. The infrared image sequence is acquired at 320 256 resolution and contains 2000 frames with the frequency variation from 0.01 Hz to 0.1 Hz. The frames are cropped with 260 225 resolution to select the specimen region only and reduce the time complexity to process the full frame sequence. The temporal thermal profile is compensated by subtracting the first order polynomial fitting.
Figure 2: The experimental setup for active infrared thermography The temporal profile of every pixel location is used for the phase extraction using the Fourier transform. The defects are visible in the phase image, but have low resolution to identify all defects. The same temporal profile is used for the pulse compression approach where the temporal profile of each pixel is cross correlated with the temporal profile of a reference non-defective pixel. In comparison with the phase image, the pulse compression approach has high resolution to detect the deeper subsurface defects. The energy spectral density is calculated using the Fourier transform of the autocorrelation of energy signal. Here, the defects in the specimen add delay in the reflected thermal wave, so the result from the pulse compression approach can be used for the Fourier analysis to find the energy spectral density. The result of the energy spectral density is effective in discriminating the non-defective and defective regions. Results The image sequence is cropped by selecting the region of interest. Preprocessing of the data is required to compensate the thermal effects. A first order polynomial fit trend removal is used for every pixel of the image sequence [2, 3]. The trend in all pixels follows a similar pattern and with the subtraction of the fitted data, the effects of surface artifacts are minimized. Fig. 3 shows the temporal profile of a reference pixel and its corresponding trend removed profile. Figure 3: Temporal thermal profile of a reference pixel and the corresponding trend removed profile The temporal thermal profile of the image sequence cannot provide the details about the subsurface defects. Phase image provides better information about the subsurface defects as it is free from the surface abnormalities [1]. The phase image is obtained by using the Fourier transform on the temporal profile of every pixel locations. The detected defects in phase image are shown in Fig. 4. The defects close to the surface is prominent in comparison to the deeper defects.
Figure 4: Detected defects in the phase image and the corresponding locations of the defects In the present context, the pulse compression approach is carried out using the correlation method. The temporal profile of a reference non-defective pixel is chosen and is cross correlated with the temporal profile of every other pixel in the image sequence. The cross-correlation result produces a compressed profile with most of the energy is concentrated in the narrow main lobe of the pseudo pulse [8]. The results are promising and the defects are visualized in the correlation images. Fig. 5 shows the results of the pulse compression approach using correlation. Figure 5: Detected defects in the correlation image using pulse compression and the corresponding locations of defects The phase information from this pulse compression information or the energy spectral density is extracted which gives a smooth visualization of the subsurface anomalies. Image processing tools such as median filter and circular average filter are used to reconstruct the phase information and visualize the subsurface defects. Figure 6: Detected defects in the phase image of the correlation coefficients and the corresponding locations of defects
By qualitative analysis of phase image and the surface profile it can be observed that the pulse compression based result is more effective than the phase image. The subsurface defects appear even better if the energy spectral density which is the phase information of the pulse compression results is observed. Fig. 6 shows the detected defects in the specimen. The result is smoother than the normal phase image and pulse compression results, but unable to detect the deeper defects.the detection result from all the above discussed methods are shown in Fig. 4, Fig. 5 and Fig. 6. The result in Fig. 6 shows a smooth discriminating the defective and non-defective regions, but unable to detect the deeper defects. In comparison, the pulse compression result is effective in terms of localizing the defective regions as shown in Fig. 5. Conclusions The limitations of PT and PPT are the requirement of high power pulse and repetition of experiment cycle. FMTWI overcomes these limitations by using a modulated frequency range and in a single experiment cycle the defects are observed. Both phase analysis and correlation based pulse compression methods have been implemented and results are shown. The phase image is unable to detect deeper defects. The energy spectral density or the phase information of pulse compression results gives the best visualization of the defects, but unable to detect deeper defects. The pulse compression approach provides an effective solution to localize the subsurface defects. Image processing tools such as median filter and circular average filter of different diameters improves the results by enhancing the thermal contrast. FMTWI is a promising method to detect the subsurface defects with low peak power and single experimentation. References [1] X. Maldague, Applications of infrared thermography in nondestructive evaluation. Trends in optical nondestructive testing (2000): 591-609. [2] V. K. Ghali and R. Mulaveesala, Comparative data processing approaches for thermal wave imaging techniques for non-destructive testing. Sensing and Imaging: An International Journal 12, no. 1-2 (2011): 15-33. [3] R. Mulaveesala et al. Non-Destructive Evaluation of Concrete Structures by Non-Stationary Thermal Wave Imaging. Progress In Electromagnetics Research Letters 32 (2012): 39-48. [4] K. Wan and M. Qishuang. Data-fitting reconstruction for defect inspection of airplane aluminum structure in infrared thermographic ndt. Industrial Electronics and Applications (2009). ICIEA 2009. 4th IEEE Conference on. IEEE, 2009. [5] S. M. Shepard et al. Enhancement and reconstruction of thermographic NDT data. In AeroSense 2002, pp. 531-535. International Society for Optics and Photonics, 2002. [6] R. Mulaveesala and S. Tuli. Theory of frequency modulated thermal wave imaging for nondestructive subsurface defect detection. Applied Physics Letters 89.19 (2006): 191913-191913. [7] E. C. Farnett and G. H. Stevens. Pulse compression radar. Radar Handbook (1990), 2, 10-1. [8] R. Mulaveesala, J. S. Vaddi and P. Singh. Pulse compression approach to infrared nondestructive characterization. Review of Scientific Instruments, 79(9) (2008), 094901-094901.