NDT&E International 44 (2011) 789796

Contents lists available at ScienceDirect

NDT&E International
journal homepage: www.elsevier.com/locate/ndteint

Ultrasonic evaluation of TiAl and 40Cr diffusion bonding quality based on time-scale characteristics extraction
Yilin Luan a,n, Tao Sun a, Jicai Feng b, Tie Gang c
a

School of Mechatronics Engineering, Harbin Institute of Technology, Harbin, China School of Materials Science & Engineering, Harbin Institute of Technology at Weihai, Weihai, China c State Key Lab of Advanced Welding and Jointing, Harbin Institute of Technology, Harbin, China
b

a r t i c l e i n f o
Article history: Received 25 September 2010 Received in revised form 12 April 2011 Accepted 13 July 2011 Available online 27 July 2011 Keywords: Ultrasonics Diffusion bonding Dissimilar materials Continuous wavelet transform Time-scale amplitude and phase

abstract
To solve the problem of ultrasonic pulse-echo method in the evaluation of kissing bond and unbond in TiAl and 40Cr diffusion bonding, a characteristics extraction algorithm was proposed. The algorithm was based on continuous wavelet transform to convert ultrasonic TiAl and 40Cr diffusion bonding interface signals into time-scale domain. The ultrasonic tests were performed by an ultrasonic C-scan imaging system using a 10 MHz focused transducer. The time-scale amplitude and phase of the interface signals were calculated and analyzed to distinguish the kissing bond and the unbond from the perfectly bonded interface. The kissing bond can be detected by the scale-dependent amplitude combined with phase variation and the unbond can be measured by the opposite phase. The amplitude and phase characteristics were extracted to reconstruct the amplitude and phase characteristics images for TiAl and 40Cr diffusion bonding specimens evaluation. The amplitude and phase characteristics images are effective in the evaluation of bonding quality. & 2011 Elsevier Ltd. All rights reserved.

1. Introduction Diffusion bonding has been considered as a potential welding method and increasingly used in the eld of aerospace and industry, which has many advantages such as high performance, signicant cost and weight savings, and low requirement for the weldability of materials [1,2]. However, imperfections such as kissing bond and unbond may occur at the interface due to improper surface preparation and upset bonding conditions [3]. These defects can degrade bonding strength, especially fracture toughness and fatigue strength [46]. Thus, it is necessary to develop non-destructive evaluation of diffusion bonding. The interfacial imperfections are parallel to the specimen surface, which is a suitable position for ultrasonic test [7]. A variety of ultrasonic methods have been applied in the evaluation of bonding quality, such as pulse-echo method, laser ultrasonic system [8], guided waves [911], and nonlinear ultrasonic measurement [1214]. Pulse-echo method is the most popular technique among these ultrasonic methods. Palmer et al. [15] described the application of ultrasonic reectivity for the characterization of copper diffusion bonds with different bonding qualities. Ultrasonic reection coefcients at 10 MHz were

Corresponding author. Tel./fax: 86045186413115. E-mail addresses: ylluan@hit.edu.cn (Y.L. Luan), taosun@hit.edu.cn (T. Sun), fengjc@hit.edu.cn (J.C. Feng), gangt@hit.edu.cn (T. Gang). 0963-8695/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.ndteint.2011.07.008

correlated with the ultimate tensile strength. Kato and Abe [16] measured diffusion bondings of steel to titanium plates to obtain the relationships among bonding strength, state of bonding interface, and two major components derived from ultrasonic testing. Considerable progress was made by Greenberg et al. [17] in developing a real-time system for the monitoring of bonding process by analyzing the amplitude ratio and attenuation of acoustic waves. In other efforts, the C-scan images at the bonding interface were used to calculate the ratio of non-bonded area of diffusion bonded joints of mild steel, combined with impact tests for threshold level determination [18]. Similar technique was applied to eld-assisted diffusion bonding joints to assess the mechanical quality by increasing the ultrasonic frequency up to 20 MHz [19]. In general, the unbonds in similar diffusion bondings are readily detectable by normal incidence wave since the ultrasonic wave will be reected at the defects whereas passing through the perfectly bonded regions. The bonding quality can be assessed by the amplitude of the reected signals. However, the kissing bonds are only a few micrometers in size, which result in weak reection. The bonding joints appear to be awless under ultrasonic inspection [20]. As for the dissimilar diffusion bondings, some ultrasonic energy is still reected from the perfectly bonded interface due to the effect of impedance mismatch between materials to be bonded [7]. It is difcult to distinguish the defect signals from the interface signals so that the bonding quality cannot be assessed by the amplitude of the reected signals.

790

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

In this paper, we focus on the evaluation of TiAl and 40Cr diffusion bonding quality. A time-scale characteristics extraction algorithm is proposed to measure TiAl and 40Cr diffusion bonding interfacial imperfections. The algorithm is based on continuous wavelet transform to analyze the amplitude and phase variation of ultrasonic interface signals in the time-scale domain. The authors shall demonstrate that the defects can be assessed by the time-scale amplitude and phase characteristics.

2. Theoretical background Classical boundary condition for ultrasonic wave interaction with welded or perfectly bonded interface assumes that stress and displacement across the interface is continuous. When an ultrasonic wave is normally incident to such an interface, the reection coefcient R12 is given by [21] R12 Z2 Z1 Z2 Z1 1

where Z1 and Z2 are the acoustic impedances of the materials on either side of the interface. Note that the reection coefcient from the perfectly bonded interface is just a function of the impedances. If the bonding is imperfect and the size of imperfections is considerably smaller than the wavelength of ultrasound, the interface can be modeled by a set of distributed springs. The ultrasonic wave interaction with such an interface can be described using spring boundary condition. The reection coefcient of normal incidence ultrasonic wave from imperfect interface is given by [22] R12 Z2 Z1 io=kn Z1 Z2 Z2 Z1 io=kn Z1 Z2 2

the result is only shown between p/2 and p/2. As kn-N, corresponding to the case of perfectly bonded, the amplitude of the reection coefcient 9R9-(Z2 Z1)/(Z2 Z1) at all frequencies. The phase of the reection coefcient tends to zero from the positive direction, which means the reected wave and the incident wave are in-phase. As kn-0, corresponding to the case of unbond, the amplitude of the reection coefcient 9R9-1 and also no frequency dependence is observed. There is an exception to the rule. 9R9 tends to (Z2 Z1)/(Z2 Z1) when f is close to zero. The frequency used in ultrasonic testing is usually greater than 2.5 MHz, even greater than 5 MHz. So this tendency has little effect on the practical ultrasonic testing. The phase of the reection coefcient tends to zero from the negative direction, which is equivalent to F-p. The reected wave is opposite in phase to the incident wave. As kn is much less than innity, corresponding to the case of kissing bond, part of the ultrasonic energy is reected from the interface and the amplitude of the reection coefcient increases with the frequency. The phase of the reection coefcient is the same at low frequencies and opposite at high frequencies. The phase transition occurs when F- 7 p/2. 3. Experimental 3.1. Specimens preparation TiAl intermetallic compound and 40Cr steel were used in the study. The specimens were a rectangular shape of 45 mm 30 mm, and the thicknesses of TiAl and 40Cr were 4.2 and 14.8 mm, respectively. TiAl specimens were given chemical cleaning by 5% hydrouoric acid, then rinsed in water and nally dried in hot airow. 40Cr specimens were cleaned using acetone. Six specimens were then bonded at various temperatures under a constant pressure of 1.33 103 Pa in a vacuum furnace. The welding temperatures were 900, 950 and 1000 1C to obtain unbond, kissing bond, and perfectly bonded joints, respectively. The welding pressure was 15 MPa for 15 min. A TiAl plate without diffusion bonding was prepared as reference specimen. 3.2. Ultrasonic measurement Ultrasonic tests were performed using ULTRAPAC C-scan immersion system produced by Physical Acoustic Corporation. The system consists of an immersion system including a scanning frame assemble, motorized axis adjusters, an immersion tank, and a computer with ULTRAWIN software to control test and provide result display. A broadband focused transducer with central

where o is the angular frequency of the ultrasonic wave and kn is the normal interfacial stiffness, which is dened as distributed spring contacts per unit area. The normal interfacial stiffness varies from innity when perfectly bonded is achieved, to zero for an unbond surface. The normal interfacial stiffness must be much less than innity when kissing bond occurs at the interface. The reection coefcient of the imperfect interface is related to three factors: the acoustic impedances of the materials on either side of the interface, the ultrasonic frequency, and the normal interfacial stiffness. The amplitude and phase of the reection coefcient of TiAl and 40Cr diffusion bonding interface as shown in Fig. 1 are calculated to illustrate the relationship among the reection coefcient and the three factors. The acoustic impedances of TiAl and 40Cr are 2.73 107 and 4.68 107 Pa s m 1, respectively. As the phase is a periodic function with period p,

Fig. 1. Calculated reection coefcient of TiAl and

40

Cr diffusion bonding interface: (a) amplitude of reection coefcient and (b) phase of reection coefcient.

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

791

45
9

17 13 9 5 1

18 14 10 6 2

19 15 11 7 3

20 16 12 8 4

following equation:   Z 1 Z 1 1 t b Wf a, b dt f t ca, b t dt q f t c a 1 9a9 1 t A 0, n 4

30

Fig. 2. Plan view of the shear test specimens.

where Wf(a,b) is the continuous wavelet transform of function f(t), f(t) is the interface signal or the reference signal, t is the time variable, (0,n] is the sampling interval of f(t), c(t) is the basic wavelet, and a and b are referred to as the scale and time parameter, respectively. The complex morlet wavelet is employed in the continuous wavelet transform for its similarity to the ultrasonic signal and linear phase. The complex morlet wavelet is dened as

frequency at 10 MHz was used. The ultrasonic wave at normal incidence from TiAl side was focused on the TiAl and 40Cr diffusion bonding interface. The C-scan images of the specimens and the interface A-scan signals were obtained at the same time with sampling frequency of 100 MHz. The ultrasonic wave was also focused on the bottom of the reference TiAl plate and the reference signal was collected from the TiAlair interface, which had 99.99% amplitude and opposite phase of the incident wave. Phase inversion was performed on the reference signal.

ct p et

=fb 2ipfc t

pfb

3.3. Shear test The central part of each diffusion bonding specimen was cut to obtain 20 shear test specimens with the dimensions of 4 mm 9 mm 19 mm. A plan view of the shear test specimens cutting method are illustrated in Fig. 2. The area with section lines was not used for shear test to eliminate the edge effect caused by focused transducer in the ultrasonic measurement. The numbers were the serial number of shear test specimens. The shear test specimens were subjected to shear tests at TiAl and 40Cr diffusion bonding interface in a universal testing machine, providing information on the shear strength of 20 areas of the diffusion bonding specimen. The shear strength tb was obtained by

where fb and fc are the bandwidth parameter and the central frequency of the basic wavelet, respectively. It is necessary to optimize the time-scale resolution and determine the scale parameter a and the step of the scale s prior to the continuous wavelet p transform. The time-scale resolution is related to 1=2pfc fb , and therefore, fc is set as 1 Hz to adjust fb. The optimal time-scale is obtained when fb equals 0.8. The scale parameter a of the continuous wavelet transform is determined by the following equation: a fc fs fa 6

tb

Fb A0

where fs is the sampling frequency of the ultrasonic measurement and fa is the central frequency of the wavelet corresponding to scale a. As can be seen, the scale is related to fa when fc and fs are known, which is determined by the useful bandwidth of the transducer. The useful bandwidth of the transducer is ranging from 6 to 15 MHz. The scale parameter is 16.76.7 according to Eq. (6). The scale parameter is rounded to integers as 176 and the correspondence bandwidth is ranging from 5.8 to 16.7 MHz. The step of the scale s is set as 0.2 considering the computation efciency. (2) The time-scale ratio of the interface signal and the reference signal R(a,b) is obtained by Ra, b Wf a, binterface Wf a, breference 7

where Fb is the loading of nal failure and A0 is the area of shear test specimen.

3.4. Metallographic analysis The cross-sections of TiAl and 40Cr bonding joints were polished for metallographic analysis. TiAl specimens were lightly etched with 2% nitric acid. Micrographs of the diffusion bonding interfaces were obtained by an optical microscope.

where the subscript interface and reference correspond to the interface signal and the reference signal, respectively. (3) The time-scale amplitude 9R(a,b)9 and the time-scale phase F(a,b) are obtained by   q Ra, b R2 a, b R2 a, b 8 R I

Fa, b +Ra, b arctan

RI a, b RR a, b

4. Time-scale characteristics extraction algorithm As we know from Section 2, the ultrasonic amplitude and phase after interacting with the interface are affected by the bonding quality of the interface. Therefore, the interfacial imperfections can be evaluated by the ultrasonic amplitude and phase characteristics. The time-scale characteristics extraction algorithm is proposed to assess the bonding quality. The algorithm procedures are as follows: (1) The continuous wavelet transforms are performed on both the interface signals and the reference signal according to the

where the subscript R and I correspond to the real and the imaginary part of R(a,b), respectively. Concerned only with the same or opposite of the time-scale phase, 1 is employed to represent the same phase, and 1 is employed to represent the opposite phase. (4) The time-scale amplitude 9R(a,bj)9 of every time parameter bj is linear tted along the scale parameter a decreasing direction to obtain the tting curve yj according to the following equation: yj Aj 9Ra, bj 9 Kj , j l, l 1=fs , . . ., m 10 where Aj and Kj are the tting slope and the tting constant of the tting curve yj, respectively, and l and m are the scope of the time parameter b, respectively. The amplitude

792

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

characteristic CR is obtained by CR
m X jl

Aj

11

(5) The phase characteristic CF is calculated by CF

v X m X s 1 Fai , bj vu s mfs lfs 1 i u j l

i u, u s, . . ., v,

j l, l 1=fs , . . ., m

12

where u and v represent the scope of the scale parameter a,

F(ai,bj) is the time-scale phase of every scale parameter ai and

every time parameter bj. The reason of amplitude and phase characteristics extraction will be explained in results and discussion section.

5. Results and discussion Fig. 3 shows the reected signals of the perfectly bonded interface, the kissing bond, and the unbond and the corresponding shear strengths of the joints are 246.1, 44.7, and 6.7 MPa, respectively. As can be seen, there are reected signals not only from the kissing bond and the unbond but also from the perfectly bonded interface due to the effect of impedances mismatch (the reectivity of the TiAl and 40Cr diffusion bonding interface is 26.3%). Moreover, there is no apparent difference between signals from the perfectly bonded interface and from the kissing bond. It is difcult to detect the kissing bond by the signals and the bonding quality cannot be accessed by the amplitude of the reected signals. Differences are illustrated clearly after the signals are processed by the algorithm described in Section 4. Fig. 4 shows

the time-scale amplitude of the signals shown in Fig. 3. The signal-to-noise ratio of time-scale amplitude is low when time is less than 0.1 ms and greater than 0.4 ms, so that the signal analysis was performed at the interval of [0.1 ms, 0.4 ms]. Four time parameters were extracted and the curves of the amplitude changing with the scale were illustrated on the right side of the time-scale amplitude. The extracted time parameters were 0.1, 0.2, 0.3, and 0.4 ms. As can be seen from Fig. 4(a), the amplitude was low and remained roughly constant from 0.1 to 0.4 ms at every time parameter for the perfectly bonded interface. This phenomenon was conrmed by the extracted time parameters for the curves of the amplitude changing with the scale were almost straight lines. As the scale is related to the frequency, the amplitude does not vary with the frequency. For the kissing bond as shown in Fig. 4(b), the amplitude increased with the scale decreasing at every time parameter. According to (6), the scale is in inverse proportion to the central frequency of the wavelet, that is to say, the amplitude increases with the frequency. The amplitude was high and did not vary with the scale for the unbond as shown in Fig. 4(c). The reason of the scale-dependent amplitude may be explained as follows. The length of the kissing bond is much smaller than the wavelength of the ultrasonic wave. The ultrasonic wave of low scale (high frequency) is much more sensitive to the interfacial imperfection than that of high scale (low frequency). The lower the scale (the higher the frequency is), the higher the amplitude after interacting with the defect. Thus, the scale-dependent amplitude is observed for the kissing bond. Though the kissing bond cannot be detected by the amplitude of the reected signals, one can reliably identify the defect using the scale-dependent amplitude. As for the kissing bond, the time-scale amplitude 9R(a,bj)9 of every time parameter bj increased with the decrease in scale , whereas that of the unbond and the perfectly bonded interface did not vary with the scale. Therefore, the tting slope Aj of the tting curve yj along the scale decreasing direction was chosen to

1.0 Normalized amplitude 0.5 0.0 -0.5 -1.0 0.0 Normalized amplitude 0.1 0.4

1.0 0.5 0.0 -0.5 -1.0 0.0

0.2 0.3 Time (s) 1.0 Normalized amplitude 0.5 0.0 -0.5 -1.0 0.0

0.5

0.1

0.2 0.3 Time (s)

0.4

0.5

0.1

0.2 0.3 Time (s)

0.4

0.5

Fig. 3. Ultrasonic signals reected from TiAl and

40

Cr diffusion bonding interface: (a) perfectly bonded interface, (b) kissing bond, and (c) unbond.

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

793

17 15 Scale 13 11 9 7 0 0.1 0.2 0.3 Time (s) 0.4 0.5

1.0 0.8 Amplitude 0.6 0.4 0.2

t=0.1s t=0.2s t=0.3s t=0.4s

0.5

0.0 17 16 15 14 13 12 11 10 9 8 7 6 Scale 1.0 0.8 Amplitude 0.6 0.4 0.2

t=0.1s t=0.2s t=0.3s t=0.4s

17 15 Scale 13 11 9 7 0 0.1 0.2 0.3 Time (s) 0.4 0.5

0.5

0.0 17 16 15 14 13 12 11 10 9 8 7 6 Scale 1.0 0.8 Amplitude 0.6 0.4 0.2

t=0.1s t=0.2s t=0.3s t=0.4s

17 15 Scale 13 11 9 7 0 0.1 0.2 0.3 Time (s) 0.4

40

0.5

0.5

0.0 17 16 15 14 13 12 11 10 9 8 7 6 Scale

Fig. 4. Time-scale amplitude of signals from TiAl and

Cr diffusion bonding interface: (a) perfectly bonded interface, (b) kissing bond, and (c) unbond.

represent the scale-dependent feature of 9R(a,bj)9. The tting slopes Aj were then summed as the amplitude characteristic CR. The calculated amplitude characteristics were 6.30, 0.02, and 0.01 for the kissing bond, the unbond, and the perfectly bonded interface, respectively. The kissing bond was distinguished from the unbond and the perfectly bonded interface by the extracted amplitude characteristics. A special phenomenon for the unbond was observed in the experiment as shown in Fig. 5. The time-scale amplitude of the unbond decreased with the scale at every time parameter when the ultrasonic signal was too high to overow the oscilloscope. This was because the signal was distorted by the saturated sampling. However, this phenomenon has no effect to distinguish the unbond from the kissing bond and the perfectly bonded interface. Fig. 6 shows the time-scale phase of the signals reected from the perfectly bonded interface, the kissing bond, and the unbond. As can be seen from Fig. 6, the phase was almost the same for the perfectly bonded interface at the interval of [0.1 ms, 0.4 ms]; the time-scale phase presented the same at high scale whereas opposite at low scale for the kissing bond; the time-scale phase was opposite for the unbond. The opposite phase can be explained by considering the difference in the acoustic impedances on either side of the interface. The phase is the same after

the ultrasonic wave interacting with the interface in the case of the acoustic impedance of the top material is less than that of the bottom material, whereas the phase is opposite in the case of the acoustic impedance of the top material is greater than that of the bottom material. The acoustic impedance of the interface layer is much less than that of the upper materials for the kissing bond and the unbond, so that the opposite phase occurs. Although the bonding quality is difcult to be evaluated by the amplitude of the reected signal, the kissing bond can be detected by the phase variation, and the unbond could be measured by the opposite phase. The time-scale phase becomes another useful tool to assess the kissing bond and the unbond. The extraordinary feature of the time-scale phase of different bonding quality was the same and opposite differences. The mean value of the time-scale phase was able to represent the phase differences. This is the reason for selecting the mean value as the phase characteristic. The phase characteristic CR of the kissing bond predicted to be between 1 and 1 for the time-scale phase F(ai,bj) presented the same at high scale whereas opposite at low scale. The time-scale phase F(ai,bj) for the unbond and the perfectly bonded interface were opposite and the same, so that the predicted phase characteristics CR should be 1 and 1. The calculation result showed that phase characteristics were

794

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

1.0 Normalized amplitude 0.5 0.0 -0.5 -1.0 0.0 Scale

17 15 13 11 9 7 0.1 0.2 0.3 Time (s) 0.4 0.5 0 0.1 0.2 0.3 Time (s) 0.4 0.5

0.5

Fig. 5. Saturated sampling signal from unbond interface and corresponding time-scale amplitude: (a) saturated sampling signal and (b) time-scale amplitude.

17 15 Scale 13 11 9 7 0 0.1 0.2 0.3 Time (s) 0.4 0.5

1 0.5 Scale 0 -0.5 -1

17 15 13 11 9 7 0 0.1 0.2 0.3 Time (s) 1 0.5 0 -0.5 -1 0.4 0.5

1 0.5 0 -0.5 -1

17 15 Scale 13 11 9 7 0 0.1 0.2 0.3 Time (s) 0.4 0.5

Fig. 6. Time-scale phase of signals from TiAl and

40

Cr diffusion bonding interface: (a) perfectly bonded interface, (b) kissing bond, and (c) unbond.

0.68, 0.99, and 0.97 for the kissing bond, the unbond, and the perfectly bonded interface, respectively. The extracted phase characteristics distinguished the kissing bond and the unbonded from the perfect bonded interface. The characteristics extraction was performed on the diffusion bonding specimens. The amplitude and phase characteristics images were reconstructed according to the position of the ultrasonic C-scan images using color indicating their values. Results of one specimen are shown in Fig. 7. Fig. 7(a)(d) are the ultrasonic C-scan image, the shear strength of 20 areas of the specimen, the amplitude characteristics image, and the phase characteristics image, respectively. The lines on Fig. 7(b) were manually added according to the dimension of the shear test specimens. The elliptical regions A, B, and C were discussed. The amplitude of the C-scan image is approximately 40% in region A, which is close to region C. However, the shear strength of region C is 173.1 MPa, whereas that of region A is averagely 19.5 MPa. There must be some imperfections on the interface and their characteristics match those of the kissing bonds. However, it is too difcult to identify the kissing bonds by the ultrasonic C-scan image. Differences are illustrated clearly in the amplitude and phase characteristics images. The amplitude characteristics approach to 5 and the phase characteristics are approximately 0.7 in region A. The amplitude

and phase characteristics of region C are approximately 0.02 and 0.98. The amplitude of the ultrasonic C-scan image in region B is too high to overow the oscilloscope and the shear strength is zero. These defects belong to the unbond. The amplitude characteristics are less than zero due to signal saturation sampling and the phase characteristics are approximately 1. The microstructures of the regions A and B are shown in Fig. 8. Small areas with the length of a few micrometers in which diffusion process is inhibited can be seen at the kissing bond interface; and a narrow long gap with the width of approximately 89 mm is located at the unbond interface. The amplitude and phase characteristics images are effective to assess the kissing bonds and the unbonds in the TiAl and 40Cr diffusion bonding specimens. Analogous signal analyses were performed on ve other specimens, and the same results were obtained.

6. Conclusions Ultrasonic interface signals of the TiAl and 40Cr diffusion bondings are transformed in the time-scale domain to analyze the time-scale amplitude and phase and extract characteristics for the bonding quality assessment. The algorithm proposed by the

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

795

[%] 100

50

0 10 5 0 -5 -10
40

1 0.5 0 -0.5 -1

Fig. 7. Characteristics extraction of TiAl and Cr diffusion bonding specimen: (a) Ultrasonic C-scan image, (b) shear strength of specimen, (c) reconstructed image of amplitude characteristics, and (d) reconstructed image of phase characteristics.

Fig. 8. Microstructures of TiAl and

40

Cr diffusion bonding specimen: (a) kissing bond and (b) unbond.

paper differs from conventional ultrasonic evaluation for it utilizes the scale-dependent amplitude instead of the amplitude of the reected signal. Another difference of the algorithm is the application of the phase information. From the above study we arrive at the conclusion that the kissing bond can be detected by the scale-dependent amplitude combined with phase variation and the unbond can be measured by the opposite phase. The kissing bonds and the unbonds exist not only in diffusion bondings but also in other solid-state welding methods, such as high-frequency induction brazing and friction welding. The algorithm shall be applied in other solid-state welding methods to analyze its universaliability in our further study. Acknowledgment The authors are grateful to all the members of non-destructive testing research team of the State Key Lab of Advanced Welding Production Technology in the Harbin Institute of Technology for their help. References
[1] Tuppen SJ, Bache MR, Voice WE. Structural integrity of diffusion bonds in Ti6Al4V processed via low cost route. Mater Sci Tech 2006;22:142330.

[2] Kurt B, Orhan N, Kaya M. Interface characterisation of diffusion bonded Ti6Al4V alloy and austenitic stainless steel couple. Mater Sci Tech 2009;25:55660. [3] Wang ZC, Ridley N, Lorimer GW, Knauss D, Briggs G. Evaluation of diffusion bonds formed between superplastic sheet materials. J Mater Sci 1996;33: 5199206. [4] Achenbach JD, Xu Y. Reection by defective diffusion bonds. In: Proceedings of IEEE Ultrasound Symposium, Montreal; 1989. [5] Xuan FZ, Zhang B, Tu ST. Interfacial resistance method for quality evaluation of diffusion bonded joints. Key Eng Mater 2007;353358:19447. [6] Tuppen SJ, Bache MR, Voice. WE. A fatigue assessment of dissimilar titanium alloy diffusion bonds. Int J Fatigue 2005;27:6518. [7] Nagy PB, Adler L. Ultrasonic NDE of solid-state bonds: inertia and friction welds. J Nondestr Eval 1988;7:199215. [8] Hutchins DA, Saleh C, Moles M, Farahbahkhsh B. Ultrasonic NDE using a concentric laser/EMAT system. J Nondestr Eval 1990;9:24761. [9] Rose JL, Zhu WH, Zaidi M. Ultrasonic NDT of titanium diffusion bonding with guided waves. Mater Eval 1998;56:5359. [10] Rose JL. Guided wave nuances for ultrasonic nondestructive evaluation. IEEE Trans Ultrason Ferrelectr Freq Control 2000;47:57582. [11] Lee BC, Palacz M, Krawczuk M, Ostachowicz W, Staszewski WJ. Wave propagation in a sensor/actuator diffusion bond model. J Sound Vibr 2004;276:67187. [12] Ulrich TJ, Sutin AM, Guyer RA, Johnson PA. Time reversal and non-linear elastic wave spectroscopy (TR NEWS) techniques. Int J Nonlinear Mech 2008;43:20916. [13] Kawashima K, Murase M, Yamada R, Matsushima M, Uematsu M, Fujita F. Nonlinear ultrasonic imaging of imperfectly bonded interfaces. Ultrasonics 2006;44:132933. [14] Ryuzo Y, Koichiro K, Morimasa M. Application of nonlinear ultrasonic measurement for quality assurance of diffusion bonds of gamma titanium aluminum alloy and steel. Res Nondestr Eval 2006;17:22339.

796

Y.L. Luan et al. / NDT&E International 44 (2011) 789796

[15] Palmer DD, Rehbein DK, Smith JF, Buck O. Nondestructive characterization of the mechanical strength of diffusion bonds. I. Experimental results. J Nondestr Eval 1988;7:15366. [16] Kato H, Abe S. Ultrasonic evaluation of the bonding strength of dissimilar metal bonds. NDT&E Int 1996;29:35561. [17] Greenberg Y, Itzhak D, Kohn G. Ultrasonic monitoring of a low temperature diffusion bonding process. J Test Eval 2000;28:8895. [18] Katoh M, Nishio K, Yamaguchi T. Materials evaluation of diffusion bonded steel bar and its impact characteristics. NDT&E Int 2002;35:26371.

[19] Cao ZJ, Chen HD, Xue J, Wang Y. Evaluation of mechanical quality of eieldassisted diffusion bonding by ultrasonic nondestructive method. Sensor Actuat APhys 2005;118:448. [20] Oosterkamp A, Oosterkamp LD, Nordeide A. Kissing bond phenomena in solid-state welds of aluminum alloys. Weld J 2004;83:225S31S. [21] Tattersall HG. The ultrasonic pulse-echo technique as applied to adhesion testing. J Phys D (Appl Phys) 1973;6:81932. [22] Lavrentyev AI, Beals JT. Ultrasonic measurement of the diffusion bond strength. Ultrasonic 2000;38:5136.