·Home ·Table of Contents ·Materials Characterization and testing

Monitoring of Fatigue Damage in Adhesively Bonded Composite-Metal Joints by Acoustic Methods Oh-Yang Kwon, Tae-Hyun Kim, Kyung-Joo Lee Department of Mechanical Engineering, Inha University 253 Yonghyun-dong, Inchon 402-751, Korea Contact

ABSTRACT

A correlation between fatigue cycles and acousto-ultrasonic (AU) parameters has been obtained from signals acquired during fatigue loading of the single-lap joints of a carbon-fiber reinforced plastic (CFRP) laminates and an Al6061 plate. The correlation showed an analogy to those representing the stiffness reduction (E/E_o) of polymer matrix composites by the accumulation of fatigue damage. This has been attributed to the transfer characteristics of acoustic wave energy through bonded joints with delamination-type defects and its influence on the change of spectral content of AU signals. Another correlation between fatigue cycles and the spectral magnitude of acoustic emission (AE) signals has also been found during fatigue loading. Both AU and AE can be applied almost in real-time to monitor the damage evolution during fatigue loading.

Keywords: acousto-ultrasonics, acoustic emission, joints, CFRP-Al6061, adhesive bonding

INTRODUCTION

Joints with similar or dissimilar materials appear to be inevitable for many applications of polymer matrix composites (PMC) in real structures. Adhesive bonding has been widely adopted for joining of composites since the joints can extend the fatigue life, reduce the total weight of structures, and transfer stresses homogeneously. These characteristics play an important role in the highly efficient structure such as aircraft structures of aluminum-composites or aluminum-aluminum bonding. Bonded parts generally have inferior mechanical properties to base materials because of the existence of interfacial or transitional region that shows quite different gradient in chemical and microstructural aspects. Therefore, bonded parts can be more critical than base structures to determine the overall integrity of mechanical structures.
Various nondestructive evaluation methods such as coin tapping, ultrasonics, AU, AE, holography, thermography, radiography have been used for the detection of defects at the bonded region [1,2]. Especially, AU method can provide the detailed information about the bonded region by analyzing the acquired signals that have passed through defects, damage or failure of structures.
AU differs from traditional ultrasonic methods primarily in the nature of received signals. Instead of attempting to have well-defined wave propagation paths, as in the flaw detection, the AU approach requires that the signal be the result of multiple interactions with the material microstructure and the diffuse population of subcritical flaws. An objective is to produce stochastic wave propagation. Although phase information is usually lost and the signal is difficult to analyze, the signal is a rich source of information that carries the integrated imprint of material properties, microstructure, and diffuse flaw populations [3]. The result from plates is a highly modulated complex output signal that usually consists of numerous, superimposed wavelets.

DIGITAL PROCESSING OF ACOUSTIC SIGNALS

Acousto-Ultrasonic Parameters
The basic approach of AU signal analysis has been termed as the stress wave factor (SWF) analysis. Although there are many ways to define and calculate SWF, the principles underlying them are basically the same, in that SWF is a measure of the efficiency of stress wave energy transfer, as provided by numerous studies in the literature [4,5]. A more general approach based on the energy integral definition of SWF was proposed by applying statistically analyzing the power spectral density. This approach has generated an additional definition of SWF, widely known as AU parameters (AUP), which have shown very good correlation with the quality of bonded joints as well as damage of composite laminates [6-8].

AUP has been defined [7] from the power spectral moments, M_n, which can be given as

(1)

where f_N is Nyquist frequency, W(f) is power spectral density function and n is integer [9]. From real digital domain, M_n can be calculated as follows:

(2)

where D f is frequency resolution defined as D f = (1/ND t), with D t as the sampling interval.
The zero-th moment (M₀) means the total signal energy calculated as the area under spectral density curve, the first moment (M₁) means the total spectral density for each frequency, the second moment (M₂) shows the dispersion of profile slope distribution of spectral density function, and so on [9]. The odd numbered moments are related to the symmetry of and the even numbered ones are to the profile curvature of spectral density function. To apply the above quantities to AU signal analysis, M₀ has been set as AUP1, AUP2 as (M₁/M₀), and so on [8,10]. AUP2 indicates the central frequency of the received signal. As an extrapolation of such definition, the peak amplitude of AU signals may be named as AUP0.

EXPERIMENTS

Specimens and Fatigue Loading
Specimens were prepared by joining the 16-ply unidirectional CFRP laminates and 3 mm-thick Al6061 plates with a commercially available adhesive. Specimens with a single-lap (SL) joint whose bonded region was kept 25.4´ 25.4 mm were employed for all tests. Some more details can be found elsewhere [13]. High-cycle fatigue tests were carried out at 4 Hz by using a servo-hydraulic testing machine. Fatigue loading was applied in tension-tension mode to avoid any possibility of buckling effect at the joint due to compression. The maximum load was 0.4 kN, about 10 % of the nominal failure load of the joint, with R-value of 0.1. A few control specimens without having fatigued were also included as the reference of experimental measurements.

Instrumentation and Data Acquisition
Specimens were inspected at the bonded joints by ultrasonic C-scan method to confirm they were made correctly. A table-top C-scan system were used for this purpose. A schematic diagram of experimental set-up is shown in Fig. 1. For the AU measurement, as shown in Fig. 1(a), the source signal of a 650 kHz pulse was made with an arbitrary waveform generator and filtered by a tunable band-pass filter. The signal was then amplified by 50 dB using a broadband RF amplifier to survive in CFRP, which generally exhibits considerable attenuation. To put the input signal into specimens, a 0.5 MHz ultrasonic transducer was used. Signals propagated through the bonded region were to be detected by a broadband type AE sensor. Detected signals were recorded intermittently at some particular number of fatigue cycles. They were also amplified by 40 dB at pre-amplifier with bandpass filter (100 kHz -1.2 MHz), then fed into a fully digital AE DSP unit, where they were recorded and processed to calculate AUPs.

Fig 1: Schematic diagram of experimental set-up of (a) AU- and (b) AE-measurements


For the AE measurements, two AE sensors were attached instead of the ultrasonic sensor to confirm that the detected signals were generated in the bonded region. This can be done by the so-called D T-filtering, preventing AE signals from background noises outside the region of interest based on the linear source location. Load-displacement information for each signal was also recorded as shown in Fig. 1(b).

RESULTS AND DISCUSSION

Correlation between AUP's and the fatigue damage of the bonded joints can be shown as Fig. 3. AU signals were measured at some unequal interval. The measurement was made at every 5000 cycles at the initial stage, whereas it was made at every 500 cycles at the final stage. The smoothed line was obtained by the fifth-order polynomial curve fitting. Both AUP0 (Amplitude) and AUP2 values showed virtually no change up to 300,000 cycles except the slight decrease at the very initial stage. At about 340,000 cycles, however, a significant fluctuation resulted in their values followed by abrupt decrease at the final stage up to failure. Such a trend is an analogy to the stiffness reduction (E/E_o) of polymer matrix composites by the accumulation of fatigue damage. The slight decrease at the very initial stage accounted for matrix cracking, whereas the abrupt decrease at the final stage was found dominated by growth and coalescence of the delaminations and fiber failures in the delaminated region [14].

Fig 2: Effect of fatigue damage of the bonded joints on AUP0 (Amplitude) and AUP2.

Although certain types of damage accumulation process should continued during the fatigue cycles from 50,000 to 300,000, no detectable change was observed either by acoustic monitoring or by C-scan imaging. Since the specimen was failed at about 385,000 cycles, one can first detect the fatigue damage by acoustic monitoring at about 80 % of the fatigue life. This appears to be still useful to avoid any catastrophic failure of structures of interest under cyclic loading.
The results shown in Fig. 3 are the waveforms and power spectra obtained from the AU measurement for specimens of an unfatigued and a heavily fatigued, respectively. The signal in Fig. 3(a) was recorded at the very initial stage within a few cycles of fatigue, whereas the signal in Fig. 3(b) was at the final stage at about 340,000 cycles of fatigue. First of all, one may notice that the peak amplitude decreased by about 33 %. Another fact may be noticed is the decrease in the power spectral density at 650 kHz by 27 %. Instead, the power spectral density at 400 kHz somewhat increased comparing to the initial stage. This change was firstly attributed to the development of delamination-type damage in the bonded joints with fatigue loading [15].

Fig 3: Waveforms and power spectra of AU signals detected during fatigue loading.

Results from the AE measurement, however, indicated that there had been some additional effect during the fatigue loading, especially at about 300,000-350,000 cycles. Fig. 4 shows AE signals recorded at two different stages of fatigue loading comparable to those of the AU measurement as shown in Fig. 3. Although AE signals were typically smaller than AU signals, the spectral density at about 400 kHz is significant. It has been speculated that this spectral content could contribute to the increase of power at the range of 400-500 kHz in Fig. 3(b).

Fig 4: Waveforms and power spectra of AE signals detected during fatigue loading.

Although the AU and the AE measurements were made by using the same condition but different specimens, AE monitoring data appeared to provide at least a clue for the fluctuation of AUP values preceding to the abrupt decrease at the final stage of fatigue. Shown in Fig. 5, (a) and (b) are actually the same data as in Fig. 2, (c) is the power spectral magnitude at the range of 400-500 kHz, whereas (d) is the cumulative AE events from the specimen failed at about 450,000 cycles. Besides the final fracture period where several thousand events were recorded, three distinct periods of AE activity were also detected as shown in (d). The first period at about 20,000 cycles appeared to be related to the slight decrease in initial stage, whereas the second period at about 125,000 cycles could not be related to. Finally the third one at about 340,000 cycles coincided with the fluctuation and the abrupt decrease of AUP0 and AUP2 values in (a) and (b) and the significant increase of AE magnitude as shown in (c).

Fig 5: Relationship between the decrease in (a) AUP0 and (b) AUP2 and the increase in AE activity in terms of (c) the spectral magnitude at the range of 400-500kHz and (d) cumulative AE events.

AE magnitude shown in Fig.5(c) is actually the sum of AU contribution and AE contribution at the range of 400-500 kHz since it was simply calculated from AU signals by filtering out all other frequency component except for the range. Nevertheless, it is quite obvious that the spectral density at the range of 400-500 kHz in Fig. 3(b) was likely to be enhanced by AE signals shown in Fig. 4(b). The enhancement was significant only for the particular stage of fatigue loading, 300,000-350,000 cycles in the tests. It appeared that the most prominent AE wave energy took place at about 80 % of the fatigue life although the apparent density of AE events was the highest at the final fracture.
In order to verify the development of fatigue damage at the interface, the joints was examined by ultrasonic C-scan after a certain number of fatigue cycles. The image shown in Fig. 6(a) was obtained at the very initial stage, whereas that in Fig. 6(b) was observed after 340,000 cycles of fatigue loading. The blue color, dark if printed in black white, means the perfect bonding, whereas the green color, light in black and white, means the delaminated interface. Edge delamination was detected in the unfatigued specimens. The delamination started from the CFRP side, then propagated toward the aluminum side, which is partly because of the geometry of bonded joints.

Fig 6: Ultrasonic C-scan images indicating the delamination-type damage developed at the bonded interface by fatigue loading.

CONCLUSION

Fatigue damage of the adhesively bonded, composite-metal joints appeared to be in the form of delamination at the interface. Acousto-ultrasonic combined with acoustic emission was found useful to detect the fatigue damage as early as 80 % of the fatigue life. The abrupt change of AUP2 at about 80-90 % of the fatigue life has been attributed to the prominent AE wave energy at the onset of delamiantion. This can also be confirmed with the monotonic decrease of AUP0 or Amplitude. The correlation between AUP's and AE activity and fatigue cycles accounts for the development of delmination-type damage at the interface. Both AU and AE can be applied almost in real-time to monitor the damage evolution during fatigue loading.

ACKOWLEDGEMENT

This work was supported by the Grant (ME-97-C-29) from the Ministry of Education, Government of Korea.

REFERENCES

1. Chapman, G. B., Jr. (1981) A nondestructive method of evaluating adhesive bond strength in fiberglass reinforced plastic assemblies, In: Joining of Composite Materials, ASTM STP 749, Kedward, K. T. (Ed.) ASTM, Philadelphia, PA, pp. 32-60.
2. NDT Abstracts:(1992) Acoustic Techniques for Adhesive Bond Evaluation, NDT&E International, 25(2), 95-105.
3. Vary, A. (1990) Acousto-Ultrasonics, In: Nondestructive testing of fiber reinforced plastics composites, Volume 2, Summerscales, J. (Ed.) Elsevier Applied Science, London, pp. 1-54.
4. Vary, A. (1988) The acousto-ultrasonic approach, In: Acousto-Ultrasonics: Theory and Application, Duke, Jr., J. C. (Ed.) Plenum Press, New York, pp. 1-22.
5. Drouillard, T. F. and Vary, A. (1994) AE Literature - Acousto-Ultrasonics, Journal of Acoustic Emission, 12(1/2), 79-100.
6. Kiernan, M. T. and Duke, Jr., J. C. (1988) Acousto-ultrasonics as a monitor of material anisotropy, Materials Evaluation, 46(7), 1105-1113.
7. Kiernan, M. T. and Duke, Jr., J. C. (1988) PC analysis of an acousto-ultrasonic signal, Materials Evaluation, 46(9), 1344-1352.
8. Kiernan, M. T. and Duke, Jr., J. C. (1990) A physical model for the acousto-ultrasonic method, NASA Contractor Report 185294, October 1990.
9. Hearn, G. E. and Metcalfe, A. V. (1995) Spectral analysis in engineering: concepts and cases, Arnold, London.
10. Dickstein, P. A. et al. (1990) Ultrasonic feature-based classification of the interfacial condition in composite adhesive joints, Research in Nondestructive Evaluation, 2(4), 207-224.
11. Mallat, S. G. (1989) IEEE Trans. Pattern Analysis and Machine Intelligence, 11, 674-693.
12. Bruce, A. and Gao, H.-Y. (1996) Applied Wavelet Analysis with S-Plus, Springer, New York.
13. Kwon, O.-Y. and Lee, S.-H. (1999) Acousto-ultrasonic evaluation of adhesively bonded CFRP-aluminum joints, NDT&E International, 32(3), 153-160.
14. Jamison, R. D., Schulte, K., Reifsnider, K. L. and Stinchcomb, W. W. (1984) In: Effects of Defects in Composite Materials, ASTM STP 836, ASTM, Philadelphia, pp. 21-55.
15. Kwon, O.-Y. and Kim, T.-H. (1999) A real-time evaluation technique of fatigue damage in adhesively bonded composite-metal joints, J. of Korean Soc. for Nondestructive Testing, 19(6), 439-447.

© AIPnD , created by NDT.net

|Home|    |Top|