Defect Recognition in Thermosonic Imaging

Contents lists available at ScienceDirect

Chinese Journal of Aeronautics

journal homepage: www.elsevier.com/locate/cja

Defect Recognition in Thermosonic Imaging

CHEN Dapeng

, WU Naiming

* , ZHANG Zheng

a_{School of Material Science and Engineering, Beihang University, Beijing 100191, China} b_{School of Jet Propulsion, Beihang University, Beijing 100191, China}

Received 20 May 2011; revised 10 July 2011; accepted 22 October 2011

Abstract

This work is aimed at developing an effective method for defect recognition in thermosonic imaging. The heat mechanism of thermosonic imaging is introduced, and the problem for defect recognition is discussed. For this purpose, defect existing in the inner wall of a metal pipeline specimen and defects embedded in a carbon fiber reinforced plastic (CFRP) laminate are tested. The experimental data are processed by pulse phase thermography (PPT) method to show the phase images at different frequen-cies, and the characteristic of phase angle vs frequency curve of thermal anomalies and sound area is analyzed. A binary image, which is based on the characteristic value of defects, is obtained by a new recognition algorithm to show the defects. Results demonstrate good defect recognition performance for thermosonic imaging, and the reliability of this technique can be improved by the method.

Keywords: thermosonic imaging; defect recognition; Fourier transforms; characteristic value; carbon fiber reinforced plastic

1. Introduction 1

Thermosonic imaging or vibro-thermography has been proved as an effective non-destructive evaluation (NDE) method in recent decade. This technique em-ploys a pulse of low frequency ultrasound to cause the defect interfaces to clap or rub, consequently the heat-ing could be observed by an infrared camera. Subsur-face defects become visible with time delays that are determined by diffusion of heat from the defects to surface [1].

This technique was firstly developed by Fayro, et

al.[1] in Wayne State University, America. A lot of

re-searches have been done on thermosonic imaging in the 2000’s, and micron resolution of metal cracks de-tection has been reached [1-3]. In Britain, where Univer-sity of Bath and Imperial College are also strongly

*Corresponding author. Tel.: +86-10-82317426. E-mail address: [email protected]

Foundation item: Joint Funds of the National Natural Science Foundation of China (61079020)

1000-9361 ¤2012ElsevierLtd. doi:10.1016/S1000-9361(11)60431-7

committed to this technique, excitation of long pulse and low power is used to produce satisfactory impact damage detection for carbon fiber reinforced plastic (CFRP) composites, while eliminating damage at the exciter attachment point [4-5]. In Germany, University of Stuttgart has predominance at the study of frequency

modulated thermosonic imaging [6-9]. In China, as a

new NDE technique, thermosonic imaging has been gradually developed in recent years. State Key Labo-ratory of Modern Acoustics in Nanjing University, ap-plied this technique to detecting cracks in aluminum alloy, and did numerical simulation study about the

temperature field of cracks [10]. Other universities,

in-volving Beihang University and Harbin Institute of Technology, also pay more attention to this tech-nique [11-13].

Up to now, reports on this technique are more con-cerned about the qualitative defect detection but less on quantitative and recognition research. In this study, problems of defect recognition for both inner wall de-fects of a metal pipeline and CFRP composite with embedded defects are described; according to fre-quency response variation on the interfaces of different depths, pulse phase thermography (PPT) method is Open access under CC BY-NC-ND license.

applied to processing the data to obtain the phase im-ages at frequency domain. Then the characteristic of phase angle vs frequency curve is analyzed. A binary image, which is obtained by analyzing the characteris-tic of frequency spectrum, is processed by mathemati-cal morphology method to eliminate mis-recognized signals and show the defects.

2. Theory and Experimental Setup

The principle and experimental setup of thermosonic imaging are shown in Fig. 1, where the surface of the detected specimen is injected with a low frequency ultrasound pulse to cause the interfaces of defect clap or rub, the surface temperature rising is delayed by thermal diffusion from subsurface defects, which can be seen as a heat source, and d is the thermal diffusion length.

Fig. 1 Principle and experimental setup of thermosonic imaging.

The general thermal diffusion equation [14] is written as ( , ) [k T( , )]t c_v T t Q( , )t t U w ’˜ ’ w r r r (1)

where T is the temperature at position r and time t, Q

the heat source function which gives how the heating is applied to a medium, and k the thermal conductivity.

The product of density U and specific heat cv is the

volumetric heat capacity which measures the ability of a material to store thermal energy. The ratio of the thermal conductivity to the volumetric heat capacity is defined as the thermal diffusivity D:

v

k c

D

U (2)

The assumption is that heat is released uniformly over the area of defect, adiabatic conditions exist at the surface and the material is isotropic and homogeneous. Then, the thermal diffusion equation can be shown as

2 1 ( , ) ( , ) ( , ) T t Q t T t t k D w ’ w r r r ₍₃₎

The surface temperature T is given as a function of time t by Eq. (4)[15]: 2 2 4 0 1 ( , ) e d 2 erf erf 4 4 d t t T x t t l x x t t D D D D ˜ S ª § · § ·º « ¨_© ¸_¹ ¨_© ¸_¹» ¬ ¼

³

where l is the length of defect, x the index in the

im-ages sequence, and erf(˜) the error function.

In thermosonic imaging, the defects can be seen as a Dirac pulse G (t) heat source, and it has an infinite flat spectrum in the frequency domain. The relationship between thermal diffusion length and the frequency of thermal wave is expressed [16] as

/

d D Sf (5)

where f is the frequency of thermal waves. For each

pixel (i, j) in the thermal image, the temporal evolution of temperature T(x) is extracted from the images se-quence. Then, discrete Fourier transform is used to analyze the experimental data by Eq. (6) [17].

1 0 1 ( ) ( ) exp[ j2 / ] Re( ) Im( ) N x F f T x fx N N f f S

¦

imaginary components of F(f), N is the number of

samples. Finally, the amplitude and phase are com-puted for each of the transformed terms using Eq. (7).

2 2 ( ) Im ( ) Re ( ) Im( ) ( ) arctan Re( ) F f f f f f f ) ­ ° ® ° ¯ (7)

The range of frequency spans from 0 to 1/'x, ('x is

the time interval between images, and 1/'x the

sam-pling rate), while the frequency increment is given by

1

f N x

' ' (8)

3. Problem Description

Figure 2 shows a steel pipeline specimen with an inner wall jelly defect, and Fig. 3, the thermal image processed by background subtracting. As shown in Fig. 3, after ultrasonic excitation, thermal anomaly appears at jelly defect location. Meanwhile, the excitation posi-tion, label on the specimen surface and edge of the pipeline all have a temperature rising, even there is thermal reflection in the inner wall. Therefore, the real defect cannot be recognized by a thermal image.

Another specimen is a CFRP laminate, which has several embedded defects and a scratch damage on the surface. Figure 4 shows subtracted background image by thermosonic imaging. Thermal anomalies appear at the location of embedded defects, as well as the scratch damage area. So, a recognition method is needed to separate deep defect signals from surface scratch dam-age.

Fig. 2 Photo of steel pipeline specimen.

Fig. 3 Subtracted background thermal image of steel pipeline.

Fig. 4 Subtracted background thermal image of CFRP laminate.

4. Defect Recognition

4.1. Phase image sequence analysis by PPT As Eq. (5) shows, thermal diffusion length and fre-quency have the relationship: d= (D/Sf )1/2. This means that the deeper of defect, the lower the corresponding frequencies it has, and only shallow defects are visible at high frequencies. Therefore, analyzing the phase images sequence obtained by Fourier transforms could separate the deep and surface signals. Figure 5 shows the phase images sequence of the steel pipeline. All the thermal anomalies, such as excitation point, defect and

label are visible in phase images at low frequency range of 0.12-0.82 Hz. Until 2.46 Hz, the defect begins to disappear, but other thermal anomalies are also visi-ble.

Fig. 5 Phase images sequence of steel pipeline specimen.

Use the same way to process the thermal images quence of the CFRP laminate. The phase images se-quence in Fig. 6 also shows that the embedded defects are only visible at low frequencies.

Fig. 6 Phase images sequence of CFRP specimen.

Although results show that PPT method can separate the signals of different thermal anomalies, too many interference signals exist in the phase images and the defects signals do not have much improvement

com-pared with the original thermal images shown in Figs. 3-4. More effective algorithm for defect recognition is needed.

4.2. Phase angle vs frequency curve analysis The infrared radiation value vs time curve of the thermal anomalies and sound area in the steel pipeline is shown in Fig. 7. From the curve, we can hardly see any differences between the defect and other thermal anomalies. Then, data is transformed from the time domain to the frequency domain by Fourier transform. The phase angle vs frequency curve is shown in Fig. 8. As the thermal response of deep defects is always in low frequency range, 0-3.5 Hz is used. This is experi-mentally selected because the phase angles of defect in this frequency range are always above zero, while phase angles of sound area and other thermal anoma-lies are oscillated at zero. Obviously, defect signals are separated from the interference signals in the frequency spectrum.

Fig. 7 Infrared radiation value vs time curve of steel pipeline specimen.

Fig. 8 Phase angle vs frequency curve of steel pipeline.

Figures 9-10 show the infrared radiation value vs time curve and phase angle vs frequency curve about the thermal anomalies and sound area in the CFRP plate. Phase angles of defects are also above zero at the experimentally selected frequency range of 0-3.5 Hz. Thus, the deep defect signals of the CFRP laminate are also identified.

Fig. 9 Infrared radiation value vs time curve of CFRP specimen.

Fig. 10 Phase angle vs frequency curve of CFRP specimen.

4.3. Decision rules and recognition results

As discussed in Section 4.1, only the phase angles of deep defects are always above zero in the frequency range of 0-3.5 Hz. Values of other thermal anomalies and sound area are oscillated at zero. Therefore, Fou-rier transform is applied to every point Ti,j(r,t) in the

thermal image to get the value of phase angle Ii,j(r, f ).

Thus the decision rules can be designed as follows:

, , , , ( ) 1 ( ) 0 ( ) 0 ( ) 0 i j i j i j i j X f f X f f ) ) ! ­° ® _d °¯ (9) 3.5 , 0 3.5 , 0 1 ( ) 28, 0.117 BW( , ) = 0 ( ) 28, 0.117 i j f i j f X f f i j X f f ­ ! ' ° ° ® ° _{d '} °¯

¦

¦

where the frequency range of 0-3.5 Hz and the charac-teristic value 28 are experimentally selected. Then bi-nary images based on the value of BW(i, j) is obtained to show the results of characteristic recognition. Figure 11 shows the results of steel pipeline and CFRP lami-nates respectively.

As can be seen from Fig. 11, the defects are recog-nized as the white area in the binary images. However, there are some random white points mis-recognized as

Fig. 11 Binary images of defect recognition results in which the value of defects is set to 1.

defect signal. That is because typically thermosonic im-ages are quite complex, and some non-defect area may have similar temperature curve as defect signal as it ab-sorbs heat from the surrounding hot areas. The difference between real defect and those mis-recognized signal is that defect should have an area with some connected points because of heat diffusion. The mathematical morphological method could be applied to removing

those mis-recognized areas [18], and the results are

shown in Fig. 12, where Fig. 12(a) is the result of steel pipeline and Fig. 12(b) that of CFRP laminate.

Thus, the mis-recognized signals are eliminated and defect recognition (white areas in Fig. 12) is realized by the new characteristic recognition algorithm.

Fig. 12 Binary images processed by mathematical mor-phology.

5. Conclusion

In this study, thermosonic imaging is used to detect the inner wall defects of a steel pipeline and embedded defects of a CFRP specimen. According to the recogni-tion problem described in Secrecogni-tion 3, PPT method is applied to processing the data to get the phase image sequence, because deep defects are only visible at low frequency, unlike the fact that other thermal anomalies are visible at a wide frequency range. Though it sepa-rates the thermal signals of deep defects from other thermal anomalies, the defect signals in the phase im-ages do not have much improvement compared with the original thermal images. Then the phase angle vs frequency curve of thermal anomalies is analyzed. It is found that phase angles of the defects are always above zero at the low frequency range of 0-3.5 Hz, while other thermal anomalies are oscillated at zero. There-fore, a new recognition algorithm is proposed to extract the characteristic value of defects in the frequency do-main, and binary images are obtained to show the de-fects. The results demonstrate good recognition per-formance for thermosonic imaging. Further studies on more specimens with more kinds of defects and new recognition algorithms in thermosonic imaging are needed.

References

[1] Favro L D, Thomas R L, Han X Y, et al. Sonic infrared imaging of fatigue cracks. International Journal of Fa-tigue 2001; 23(1): 471-476.

[2] Han X Y, Loggins V, Zeng Z. Mechanical model for the generation of acoustic chaos in sonic infrared imaging. Applied Physics Letters 2004; 85(8): 1332-1334. [3] Favro L D, Han X Y, Ouyang Z, et al. Sonic IR imaging

of cracks and delaminations. Analytical Sciences 2001; 17: s451-s453.

[4] Barden T J, Almond D P, Pickering S G, et al. Detection of impact damage in CFRP composites by thermosonics. Nondestructive Testing and Evaluation 2007; 22(2): 71-82.

[5] Barden T J, Almond D P, Cawley P, et al. Advances in thermosonics for detection impact damage in CFRP composites. Quantitiative Nondestructive Evaluation. AIP Conference Proceedings 2006; 820(1): 550-557. [6] Dillenz A, Busse G, Wu D. Ultrasound lock-in

thermo-graphy: feasibilities and limitations. SPIE 1999; 3827: 10-15.

[7] Riegert G, Zweschper T, Gleiter A, et al. Acoustic lock-in thermography of imperfect solids for advanced defect selective NDE. Innovations in Nonlinear Acous-tics. AIP Conference Proceedings. 2006; 838: 108-111. [8] Zweschper T, Dillenz A, Riegert G, et al. Ultrasound

excited thermography using frequency modulated elas-tic waves. Insight: Non-Destructive Testing and Condi-tion Monitoring 2003; 45(3): 78-182 .

[9] Zweschper T, Riegert G, Dillenz A, et al. Ultrasound burst phase thermography (UBP) for applications in the automotive industry. Review of Progress in Quantita-tive NondestrucQuantita-tive Evaluation. AIP Conference Pro-ceedings 2003; 657: 531-536.

[10] Liao P C, Mi X B, Zhang S Y, et al. FEM analysis of transient temperature fields of samples with defects during ultrasonic pulse excitation. Journal of Nanjing University: Natural Sciences 2005; 41(1): 98-104. [in Chinese]

[11] Guo X W, Liu Y T, Guo G P, et al. Pulsed phase ther-mography and its application in the NDT of composite materials. Journal of Beijing University of Aeronautics and Astronautics 2005; 31(10): 1049-1053. [in Chinese] [12] Guo X W, Ding M M. Simulation of thermal NDT of thickness and its unevenness of thermal barrier coatings. Acta Aeronautica et Astronautica Sinica 2010; 31(1): 198-203. [in Chinese]

[13] Liu H, Liu J Y, Wang Y. Detection of contacting

inter-face-type defects using ultrasound lock-in thermogra-phy. Optics and Precision Engineering 2010; 18(3): 654-670. [in Chinese]

[14] Zhang H J. Heat conduction. Beijing: Higher Education Press, 1992. [in Chinese]

[15] Ouyang Z, Favro L D, Thomas R L, et al. Theoretical modeling of thermosonic imaging of cracks. Quantitia-tive NondestrucQuantitia-tive Evaluation. AIP Conference Pro-ceedings 2002; 615: 577-581.

[16] Maldague X P V, Patrick O M. Nondestructive testing handbook: infrared and thermal testing. Columbus: The American Society for Nondestructive Testing, 2001. [17] Li Y H, Zhao Y J, Feng L C, et al. Measurement of

defect depth by infrared thermal wave nondestructive evaluation based on pulsed phase. Optics and Precision Engineering 2008; 16(1): 55-59. [in Chinese]

[18] Serra J, Soille P. Mathematical morphology and its applications to image processing. Proceedings of the 2nd International Symposium on Mathematical Mor-phology, 1994; 396.

Biographies:

CHEN Dapeng is a Ph.D. student of Beihang University. His main research interest is thermal wave imaging.

E-mail: [email protected]

WU Naiming is an associate professor of Beihang University. His main research interest is heat conduction and thermal wave NDT.