·Home ·Table of Contents ·Materials Characterization and testing

Ultrasonic Evaluation of Bond Using Segment Adaptive Filtering Jian X.M. and Guo N. School of Mechanical & Production Engineering, Nayang Technological University, Singapore Li M.X. and Zhang H.L. Institute of Acoustics, Chinese Academy of Science, Beijing, China Contact

Abstract

Ultrasonic testing is widely used in bond testing in multilayer structures such as solid fuel rocket motor, which consists of steel, rubber laminate and solid fuel. However, only a small fraction of ultrasonic waves can transmit through the steel/rubber interface due to their large difference in acoustical impedance, and very little ultrasound is reflected back by the weak bond interface within the rubber laminate, or from the interface between the rubber and solid fuel. Consequently, the interface bond degradation can only produce a very slight variation in the received ultrasonic echo sequences, which is too weak to be detected effectively. In this paper, ultrasonic pulse reflection from the interfaces is considered to evaluate the interface bond condition. A multilayer model with spring boundary condition is used to describe the weak bond, and ultrasonic reflection response is obtained for the structure in the immersion mode. After analysis of the ultrasonic echo sequences of the steel-rubber structure, the segment adaptive filtering method is used to separate the simulated echo sequences and to obtain the interface signals. With the separated interface signals, the weak bond of all the interfaces, including the one within the rubber laminate can be detected. The procedure has been applied to signals measured from the steel- rubber samples, and the amplitude and envelope of different interface signal agree well with those from the corresponding simulated signal. The disbond and weak bond can be detected and evaluated by the amplitude of corresponding signal of interfaces.
Keywords: Ultrasonic evaluation, Weak bond, Signal of interface, Adaptive filtering.

1. INTRODUCTION

A typical testing problem of multilayer structures that exists in the nondestructive testing of solid fuel rocket motor structure is to be considered here. The multilayer structure consists of a steel layer, a hardened rubber laminate and solid fuel. The laminate is formed when a number of laminae are bonded successively to the steel layer, forming a number of interfaces within the laminate. The acoustical impedance of steel is known much greater than that of the rubber, hence only a small part of ultrasonic waves can transmit through the steel/rubber interface even if the they are bonded perfectly. The received ultrasonic echo sequences of the steel-rubber laminate structure are mainly dominated by the signals reflected back by the steel/rubber interface. The laminate is also bonded to solid fuel that has approximately same acoustical impedance. Defects such as weak bond or disbond may occur at the interfaces within the laminate, or at the interface between the laminate and solid fuel.

Because the steel layer and rubber laminate are very thin, the interface signals from steel/rubber interface often superpose signals that come from the interface within the rubber laminate. As a result, the amplitude variation induced by the interface bond degradation within the rubber laminate is too weak to be detected effectively from the received total echo sequences. Furthermore, the interface bond between the steel and rubber cannot be evaluated only based on the amplitude of the measured echo sequences since the amplitude may depend on the wavelength and the location of the disbonded rubber lamina. For example, it is observed that when the thickness (from the steel/rubber interface to the disbonded interface within the rubber laminate) is 1/4 incident wavelength, the received echo sequences is weaker than that when all interfaces within the rubber laminate are perfect. However, when the thickness is half of the incident wavelength, the received echo sequences may be larger.

In order to detect interface bond degradation of all the interfaces, the signals of the rubber laminate interface should be separated. Many kinds of signal processing methods are applied in the noise removal and echo detection in noise environment of multilayer structures (Ophir et al., 1989; Zhu and Weight, 1994; Abbate, 1997). The adaptive filtering method has been applied to separate the signals of different interfaces (Jian, 1999). First, ultrasonic reflection pulse response of multilayers is deduced, and is used to simulate the reflection echo sequences of the steel-rubble laminate structure. The incident pulse used in simulation is obtained by deconvolving the measured echo sequences of a steel plate (Carpenter and Stepanishen, 1984; Sin and Chen, 1992).

In the paper, a reflection model of a steel-rubber multilayer structure based on spring boundary conditions is used to calculate the total echo sequences. The simulated echo sequences for different interface cases are then processed by the adaptive filtering method to obtain the signals of interface within the rubber laminate. According to the amplitude and envelope of the interface signal, different interface bond quality such as perfect, weak and disbond can be evaluated. Measurements are carried out on samples of steel- rubber laminate, and compared with the simulated results using the spring model. Finally, discussion is given on the shape of the envelope and amplitude of the interface signals and how they can be used to evaluate interface bonding within the multilayer structures.

2. THEORY

Reflection model
A multilayer structure with interfaces is shown in Fig 1, and the property of each layer is known, for example, r₂, c₂, h₂, and a₂are the density, compressive wave speed, layer thickness and attenuation coefficient of medium 2 respectively. A reflection model of a steel-rubber multilayer structure based on spring boundary conditions can be used to calculate the total echo sequences reflection coefficient on acoustical wave normal incidence (Jian et al., 2000),

(1)

where Y₂(w) and Y₃(w) are the ultrasonic wave beam revision function after travelling distance in medium 2 (through 2h₂) and medium 3 (through 2h₃) respectively. The reflection coefficient from each interface is a function of the acoustical impedance of the adjacent media and the interface spring stiffness that characterize the interface quality. For example, on the interface between media 1 and 2, the pressure reflection and transmission coefficients are (Tattersall, 1973; Margatan et al., 1992),

(2)

(3)

Where K_n and K_t are the normal and shear interface spring stiffness, describing the bonding quality, while Z₁ and Z₂ are the acoustical impedance of corresponding media, and w is the angular frequency of the incident ultrasonic wave.

Fig 1: Ultrasonic propagation in multilayers

For the configuration shown in Fig. 1, it can be shown that the total received ultrasonic signal in the medium 1 can be divided into three parts as,

Rtotal = R' + R'' + R'''

(4)

where R',R'' and R''' are designated as the signals of interface I, II and III respectively. According to the definition of interface signal and Eq (1), R' and R'' are respectively given by,

(5)

R ''

(6)

where all items in Eq (6) have components of but no component of . Finally, signals of interface III should have component and can be found from equation (4).

Five cases of different bond quality of interface II and III under consideration are summarized in Table 1. Corresponding interface signals can be transformed to echo sequences in time domain. Assuming x(t) is echo sequences for case #5; y(t) is the echo sequences of the multilayer to be tested. Correspondingly, x"(t) and y"(t) are the signal of interface II in x(t) and y(t) respectively.

(7)

where x''_i(t) is echo i of x"(t) and y''(t) is echo i of y"(t). It can be seen that interface signal y''_i(t) is directly correlated to x''_i(t) since R₂₃ is a constant depending on interface stiffness. When the signal to noise ratio in the received ultrasonic echo train is not too low, y''_i(t) can be separated using x''_i(t) as a reference signal by adaptive filtering method. The signal of interface II, y''(t) can then be used to evaluate the quality of interface II, while the signal of interface III, [ y(t)- y''(t)- y'_i(t)] can be used to evaluate the quality of interface III, where y'(t) is the sign of interface I in y(t) respectively.

Segment Adaptive Filtering
Signal segment adaptive filtering is used. Reference signals, designed as x(n), are echo sequences of multilayers in which the interface II is disbonded, and echo sequences of multilayers to be tested are designed as y(n). The echo sequences are segmented into x_i(n) and y_i(n), (i=1,2,1/4). Each segment has time length dt, where dt=2h₂/c₂

Fig 2: Segment Adaptive Filtering Flow Chart

Fig 2 is the segment adaptive filtering flow chart. x_i(n) and y_i(n), (i=1,2,1/4)are present to an adaptive filter segment by segment. The signals of interface II in x_i(n) and y_i(n) are designated as x_i"(n) and y_i"(n)respectively. Generally, x_i"(n) doesn't correlate to y_i"(n). Therefore, z_i"(n) is the signal of interface III. If the interface III doesn't exist, z_i(n) should be zero in the case that noise is neglected.

The envelope of z(n) is different from that of x(n). When a weak interface bond occurs among rubber laminates or between rubber layer and solid fuel, z(n) corresponds to the signals of interface III. When all interfaces between rubber laminates and between rubber and solid fuel, z_i(n) is nothing but random noise.

3.SIMULATED ECHO SEQUENCES

For solid fuel rocket motor, the interface 1 is between water and steel, and should remain perfect. The interface 2 is between steel and rubber. The interface III is possible weak/disbond interface within the rubber laminate or that between the rubber layer and solid fuel.

Simulated echo sequences for the 5 structure cases in Table 1 are calculated. Fig 3a and 3c are the simulated echo sequences of case #3 and case #4 respectively. The signals of interface II, which are much larger than that of interface III, superpose the signals of interface III. Though interface bond quality of interface II differ, it is hard to detect imperfect bonds from these echo sequences directly.

Interface Case #	Interface Bond Quality
	Interface I	Interface II	Interface III
1	Perfect	Perfect	Perfect
2	Perfect	Perfect	Weak
3	Perfect	Perfect	Disbonded
4	Perfect	Weak	Disbonded
5	Perfect	Disbonded	Disbonded
Table 1: The steel- rubber laminate structure and different interface cases

Case #5 in Table 1 is used to obtain the reference signal. By segment adaptive filtering, the signals of interface III are separated as shown in Fig 3b for case #3 and 3d for case #4 respectively. It can be seen that the signal amplitude in Fig. 3b is much smaller than that in Fig. 3d which is predominately the signals of the interface II. More acoustical energy transmits through interface II into medium 3 in case #3 than in case #4, as a result, the signals of interface III in case #3 are stronger, and the signals of interface II are weaker and attenuate more quickly. The separated interface signals could be employed to evaluate interface effectively.

Fig 3: Simulated echo sequences and signals of interface III. (a), (b) case 3; (c), (d) case 4.(a), (c) signal of interface II, (b), (d) signal of interface II

4. MEASUREMENTS

A wideband transducer with nominal center frequency of 2.25 MHz was used to generate ultrasonic wave in the experiment, and the test was carried out in normal incident and pulse-echo immersion mode. Ultrasonic plane waves were incident to the multilayer sample, reflected by the interface, received and transformed to electrical signal by the same transducer. The received signal was amplified, digitized and transferred to a computer for storage and processing.

Three samples, named as sample #5, #3, and #1, are used in experiment. Sample #5, #3 and #1 corresponds to interface case #5, #3 and #1 in Table 1. In all cases, the dimension of steel layer, the rubber laminate and the depth of interface III are same as those used in the theoretical simulation.

Echo signals from sample #5 was first measured as x(n). It was segmented as x_i(n) and present to the adaptive filter as a reference signal. The total echo sequences from the disbond sample #3 and good sample #1 were measured as y(n) respectively, then segmented as y_i(n) and presented to adaptive filter to obtain the separated signals of interface III.

Fig 4: Real signals of interface II and interface III for case 1and 3. (a),(b) case 3; (c),(d)case 1;(a), (c) signals of interface II; (b), signal of interface III; (d) noise.

The measured echoes of interface II and the separated signals of interface III of samples #3 and sample #1 are shown in Fig 4 for comparison. From Fig 4a and 4c, the measured signals of interface II are hardly different between the case for disbond sample #3 and good sample #1 as both exhibit almost same amplitude and similar shape of envelope. However, their separated signals of interface III are very different as shown in Fig 4b and 4d. For sample #3 that is disbonded at interface III, there is distinguished envelope in the processed signal as shown in Fig 4b. While for sample #1, all rubber laminates are bonded perfectly, so there would ideally be no signal from these interfaces as expected. This is evident from the processed signal of the interface III as shown in Fig 4d, where only some noise remains.

5. CONCLUSION

Ultrasonic evaluation of bonding interfaces in a multilayer structure has been investigated, and a segment adaptive filter method has been used to obtain the signals of the interface in a multilayered structure such as solid fuel motor. It is shown that both simulation and measurement results agree to each other, and it is evident that the method is effective in evaluating the quality of bonding interface that is deeply embedded. Signals from the disbonded interface are shown predominately with a distinctive envelope with large amplitude, and signals from a weak bonding have a reduced envelope amplitude, while only noise occurs in the case of perfect bonding. Furthermore the depth of the interface can be determined by the arrival time of the corresponding interface signal. Finally the quality of bonding interface can be quantified by the varying shape and amplitude of the corresponding interface signal in comparison with that from the disbonded and perfect interfaces.

The simulation and measurement studies are performed on the sample structures of steel- rubber laminate. However, the method can be applied to evaluate bond and interface of other multilayer structures in which there is multiple interfaces and acoustical impedance may differ greatly between the successive layers or approximately same. For example, aluminum- based composite laminates in airspace industry, and silicon/mold compound, and silicon/die attach/copper structures in electronic packaging applications.

REFERENCES

1. Abbate, A., 1997, "Signal detection and noise suppression using a wavelet transform signal processor, application to ultrasonic flaw detection", IEEE Trans. Ultrason. Ferroelect. and Freq. Contr., vol.44, pp14-26.
2. Carpenter, R.N. and Stepanishen, P.R., 1984, "An improvement in the resolution of ultrasonic pulse echo system by deconvolution", J. Acoust. Soc. Am., vol.75, pp1084-1091.
3. Jian X.M., Guo, N., Li, M.X., and Zhang, H.L., Characterization of bonding quality in multilayer structure using adaptive filtering , submitted for publication, 2000.
4. Kazys, R. and Svilainis, L., 1997, "Ultrasonic detection and characterization of delaminations in thin composite plates using signal processing techniques", Ultrasonics, vol.35, pp367-383.
5. Margatan, F.J., Thompson, R.B., Rose, J.H., and Gray, T.A., 1992, "The interaction of ultrasound with imperfect interfaces: experimental studies of model structures", J. Non-Destruct. Eval., vol.11, pp109-125.
6. Ophir, J., Yazdi, Y., Lin, T.S., and Shattuck, D.P., 1989, "Optimization of speed-of-sound estimation from noisy ultrasonic signals", IEEE Trans. Ultrason. Ferroelect. and Freq. Contr., vol.36, pp16-.
7. Rokhlin, S.I. and Wang, Y.J., 1991, "Analysis of boundary conditions for elastic wave interaction with an interface between two solids", J. Acoust. Soc. Am., vol.89, pp503-515.
8. Sin, S.K. and Chen, C.H., 1992, "A comparison of deconvolution techniques for the ultrasonic nondestructive evaluation of materials", IEEE Trans. Image Processing, (1), pp3-10.
9. Tattersall, H.G., 1973, "The ultrasonic pulse-echo technique as applied to adhesion testing", J. Phys. D: Appl. Phys., vol.6, pp819-832.
10. Zhu, Y. and Weight, P., 1994, "Ultrasonic NDE of highly scattering materials using adaptive fieltering and detection", IEEE Trans. Ultrason. Ferroelect. and Freq. Contr., vol.41, p26-33.

© AIPnD , created by NDT.net

|Home|    |Top|