·Home ·Table of Contents ·Materials Characterization and testing

Transient recording and analysis of acoustic emission events resulting from damage evolution in composite laminates Mikael Johnson and Peter Gudmundson Department of Solid Mechanics Royal Institute of Technology (KTH) SE-100 44 Stockholm Sweden Contact

Abstract:

Tensile testing have been performed on [0^o,90^o₂]_s ,[90^o₂,0^o]_s , [+45^o,-45^o]_s ,[0^o₄] glass fibre/epoxy laminates. All specimens were instrumented by six broad band transducers which made it possible to acquire time records of acoustic emission events until final failure of the specimens. Simultaneously, the damage evolution was visually observed and edge replicates were taken at regular load intervals. From the independent observations of damage evolution, the characteristics of the time records resulting from various acoustic emission events could be experimentally identified. Substantial differences in the transient recordings which resulted from transverse matrix cracking, local delaminations and fibre fractures were observed.

Numerical simulations of transient responses resulting from transverse matrix cracking are also presented. The results are promising and it is believed that further developments of the theoretical model would enable close predictions of time records captured by an AE-system.

1. Introduction

The high specific strength and stiffness in combination with new cost effective, environmental friendly manufacturing processes have increased the use of composites as a construction material in various applications. A particular feature of composites is that when they are subjected to mechanical loading microcracks may develop long before final failure of the structure [1]. The effects of these microcracks, which are mainly due to matrix cracking, delamination and fiber breakage, are that the stiffness of the structure will decrease [2], moisture or other chemicals will have the possibility to diffuse into the material and thereby cause the structure to leak or even trigger an earlier failure. It is therefore of great importance to have access to models which can predict the initiation and growth of these damage processes.

Acoustic emission (AE) using socalled broadband transducers is a well suited tool for experimental studies of these processes. This technique enables the AE transients to be recorded as time histories in a relatively broad frequency band [3]. It will also make it possible to make more complete and relevant comparisons between experimentally measured and numerically simulated signals

In the present paper, tensile test specimens of unidirectional, angle-ply and cross-ply glass/epoxy composite laminates, are investigated and the objective is to characterize AE signals generated from various forms of microcracks developed under tensile loading. The purpose is to study tran sient recordings from broad-band transducers generated from matrix cracking, local delaminations and fiber breakage to see if characteristic features can be recognized and used to discriminate between these different damage types. A few main characteristic signal types will be discussed here. More extensive details and other typical groups of signals are presented in [5]. Finally, some preliminary numerical results for a surface matrix crack are presented.

2. Experimental setup and procedures

2.1. Specimen preparation
Unidirectional epoxy/glassfiber prepregs were used to produce specimens with four different lay-up configurations: [0^o,90^o₂]_s ,[90^o₂,0^o]_s , [+45^o,-45^o]_s and [0^o₄]. The specimens were cut into lengths between 235-280 mm depending on lay-up configuration. In order to prevent damage initiation from edge defects introduced by the cutting process, and to enable the use of replica technique, the edges were wet sanded to a width of approximately 25 mm. Soft aluminium tabs were glued at the ends of each sample to decrease noise in the signal due to the gripping of the specimen.

2.2. Acoustic emission system
The acoustic emission (AE) signals which resulted from microcracking in the composite specimens were monitored by a digital data acquisition system (Fracture Wave Detector F4012 (FWD) system, manufactured by Digital Wave Corporation, Denver, USA). The signals were detected using six DWC B1025 piezoelectric transducers, which were attached to the specimen surface using vacuum grease as a couplant and fastened by tape. For all the tests, the low pass filter and the high pass filter of the recording system, were set to 20 kHz and 4 MHz respectively. The transducers are of broadband type and have an approximately flat response in the frequency range of 50 kHz - 1.5 MHz, which enables the time history of a transient signal to be measured.

2.3. Testing procedure


Fig 1: Specimen showing the AE - Transducer arrangement

Quasistatic tensile tests were conducted using a MTS servo-hydraulic device. The cross-head speed was set to 0.01 mm/s in each test. To capture the acoustic emissions due to microcracking, the transducers were placed as shown in Figure 1.

Transducer 1 and 2 were mounted closest to the tabs to make it possible to exclude signals coming from damage or slipping in the gripping region. In this way it was assured that only signals from within the specimen were being studied. A gauge block, designed to position two transducers on opposite sides of the specimen, was used to mount the transducer pairs 3, 5 and 4, 6 respectively. During the tests, the damage evolution was visually observed and this in combination with immediate display of the detected emission on the computer screen enabled a correlation between signal and crack mechanism. Parametric signals from load, displacement and strain gauge transducers were also recorded by the FWD unit.

3. Experimental results and observations

Four different lay-up configurations were chosen in order to generate different damage mechanisms. Surface and internal matrix cracks transverse to the load direction were studied using the [0^o,90^o₂]_s and [90^o₂,0^o]_sspecimens respectively. The [+45^o,-45^o]_s specimens were primarily used to study the development of delaminations, but matrix cracking was also expected. The [0^o₄] specimens were used in an attempt to generate fiber breakage AE signals.

As described above, two pairs of opposite transducers were used in each test. In this way it was possible to discriminate between extensional and flexural waves, or symmetric and asymmetric waves, respectively. When an extensional (symmetric) wave is detected, it will ideally result in identical signals for the two opposite transducers. If instead, a flexural (asymmetric) wave is considered opposite signs for the measured signals on the two transducers are expected. Hence, if the recorded signals from the two opposite transducers are added or subtracted, the extensional or flexural wave will appear respectively.

3.1 Cross-ply laminate [0°,90°₂]_s
This type of specimen was loaded until final failure and the onset of acoustic emission was close to a strain value of 0.75%. The tested specimen was visually inspected by naked eye and matrix cracks were observed in the layer. Not all cracks transversed the entire width of the specimen. Many cracks extended from one edge and then arrested within the specimen. Some cracks initiated within the specimen and did not reach the edges. Microscope inspection of the replica tape confirmed that matrix cracks had developed in the layer and that the cracks extended through the whole thickness of the layer.

The most frequent type of signal during tests on this stacking sequence can be seen in Figure 2 a. It had a small positive first peak followed by a larger negative peak and a subsequent positive peak. The signal returned to zero amplitude followed by another positive peak. After that there was once again a large negative peak followed by a smaller positive peak. For several signals it was observed that in between the two distinct negative peaks, the signal showed repeated oscillations with small amplitudes. The time lag between the two large negative peaks varied from event to event. These large negative peaks separated by small peaks are characteristic for the captured emissions in this [0^o,90^o₂]_s specimen. When this signal was measured by two opposite transducers in a pair, the two signals seemed to be in phase. Hence, it can be concluded that the signals are dominated by extensional waves implying a source in the mid layer. This is confirmed by sub-tracting the signals for channels 4 and 6, which will almost make the signal disappear, see Figure 2 b. This signal is therefore believed to be due to a matrix crack in the mid layer.

Fig 2: a) Signal generated due to a matrix crack in the mid layer. b) Resulting signal from subtraction of signals measured on two opposite transducers.

3.2 Cross-ply laminate [90°₂,0°]_s
Tests were performed until final failure also for this type of specimen. The strain value was approximately 0.6% for the onset of acoustic emissions and the slope of the acoustic emission count curve was constant up to a strain value of about 1.2%, where the slope increased. The specimen was visually inspected and matrix cracks were observed in the two layers, i.e. in the two surface layers. Also for this specimen the crack lengths varied. From replica tape examinations, it was observed that the matrix cracks extended through the entire thickness of the layers. At higher loads delamination zones extended along the specimen edges. This was not observed for the specimen, which instead had a delaminated zone near the final breakage of the specimen. This indicates that matrix cracks were formed in the strain interval of 0.6-1.2%. There after the rate of matrix crack formation decreased and small delaminations as well as splitting started to form at the edges of the specimen.

An example of a captured emission, in the strain interval 0.6-1.2%, measured by transducer 4 is shown in Figure 3 a. The first arriving peak was always positive on both the transducers in a pair and the first part of the signal contains large irregular amplitudes with high frequencies, followed by smaller amplitudes and lower frequencies. In Figure 3 b, the signals measured by the two opposite transducers have been subtracted. This shows that the first part of the signal is strongly reduced and that it is followed by a part containing lower frequencies but larger amplitudes. The analysis above indicates that the first part of the signal is an extensional wave, with higher speed of propagation compared to the following wave, which presumably is a flexural mode [6]. The signal is believed to be generated by matrix cracking in the layers.

Fig 3: a) Signal generated due to a matrix crack in one of the surface layers. b) Resulting signal from subtraction of signals measured on two opposite transducers.

3.3 Angle-ply laminate [+45^o,-45^o]_s
This specimen was also tested to final failure. For this specimen type there was no significant knee in the event curve, instead the registration of AE started after a strain value of 2.0% from which the AE count rate slowly increased up to a constant value. During the test, the specimen was visually inspected, and at lower strains matrix cracks were observed in both and layers. This was also confirmed from inspection of the replica tape. For higher strains, small delaminations developed from the edges following existing matrix cracks in the surface layers. It was easy to observe the formation of these delaminations by naked eye during the test and they grew just a few millimetres from the edge before they stopped. Towards the end of the test, one of the delaminations grew, following an existing matrix crack, and when it reached the other edge the specimen failed.

A type of emission which was not observed in other tests, but for this stacking sequence is shown in Figure 4 a. These emissions were captured during the part of the test where small delaminations at the edges were formed and characteristic for these signals were the smooth oscillating waveform. The peak arriving first was positive on all transducers, and on the triggering pair the signal starts with a small segment of high amplitudes followed by smaller amplitudes which slowly decreases. This phenomenon was accentuated for emissions toward the end of the test, which is shown in Figure 4 b. Here the signal starts with four or five periods of high amplitudes, followed by a part with almost constant amplitude before it was damped out.

Fig 4: a) Signal generated due to initiation of a delamination at the edge and b) due to a delamination growing over the specimen width.

If the fast Fourier transform (FFT) is applied to the signals, it is seen that signals of the type shown in Figure 4 a have a sharp peak around 390 kHz and for the type shown in Figure 4 b this peak is near 280 kHz. The first described signal is thought to be generated from initiation of a small delamination at the edge, while the second signal is thought to be due to a larger delamination growth prior to the final failure.

3.4 Unidirectional laminate [0^o₄]
This four layer unidirectional laminate was used in an attempt to study emissions generated by fiber breakage. During the test, the specimen was ruptured with many close splitting damages, and broken fiber bundles were also observed. The stress-strain curve had an almost constant slope up to a strain value of about 2.6%, where the stress started to decrease in a stepwise fashion with increasing strain, until final failure occurred. It is believed that during the part of the tensile test where the slope was constant, splitting damages developed in the specimen. When the strain was increased, fibres and fibre bundles started to break which resulted in the step-wise behaviour, due to the reduced stiffnes. In Figure 5 an emission measured just before a decrease in force for increasing strain can be seen, and it is thought to be due to a fiber bundle breakage. It can be seen that the emission consists of two parts, the first one with high amplitudes and high frequencies but Channel 5 with short duration. The first and the second part of the signal are thought to be due to an extensional wave and a flexural wave respectively.

Fig 5: Signal generated due to rupture of fiber bundle.

Fig :6 Numerical calculated signal due to a surface matrix crack. Numerical signal

3.5 Numerical simulation
In Figure 6 an example of a numerically calculated signal is presented. The time history of the vertical velocity due to matrix cracking in one of the surface layers of a specimen is simulated. The calculations are based on the numerical model presented in [7] which is based on a finite element discretization of the cross section of a specimen. Fourier transforms in the axial coordinate as well as in time are applied and the solution is obtained by superposition of eigen-modes followed by inversion of the transform using residue calculus and FFT.

For the signal in Figure 6, the source is modelled as a matrix crack starting at one edge of the specimen and running over the whole width (25 mm) with a crack growth velocity of 976 m/s. The distance between the source and the point where the vertical velocity is calculated, i.e. the corresponding location of a transducer is 50 mm. There are however several uncertainties which must be encountered before a direct comparison can be made to experimentally acquired time histories. The frequency characteristics of the transducers and the data acquisition system have for example a large influence on the time records. Further investigations must be made before any correct comparisons can be made between a numerically calculated and experimentally measured signals.

4. Conclusions

The presented time records show that different crack mechanisms produce characteristic AE signatures. The signals show clearly that different frequencies appear and that these propagate with different velocities. Resonance transducers used in traditional AE-systems would not be able to Channel 2 capture these effects. It is therefore believed that a robust classification system should be based on features from broad band time records.

To be able to increase the understanding about crack initiation and crack growth processes, the broad band AE signals have to be compared to numerically calculated signals. Preliminary numerical results were presented. There are however uncertainties left, primarily related to the frequency characteristics of the data acquisition system, before direct comparisons to experimental time records can be made.

5. References

1. M.G. Bader, J.E. Bailey, P.T. Curtis, A. Parvizi, "The mechanisms of initiation and development of damage in multi-axial fibre-reinforced plastics laminates", ICM 3 1979;3:227-239.
2. E. Adolfsson, P. Gudmundson, "Matrix crack induced stiffness reductions in [(0_m /90 _m /+q _p /-q _q ) _S ] _M composite laminates", Composites Engineering 1995;5:107-123.
3. W.H. Prosser, K.E. Jackson, S. Kellas, B.T. Smith, J. McKeon, A. Friedman, "Advanced waveform-based acoustic emission detection of matrix cracking in composites", Materials Evaluation 1995;53(9):1052-1058.
4. D. Guo, "Lamb waves from microfractures in composite plates", PhD thesis, University of California, Mechanical Engineering, 1996.
5. M. Johnson, P. Gudmundson, "Broad-band transient recording and characterization of acoustic emission events in composite laminates", Report 255, Department of Solid Mechanics, Royal Institute of Technology, Stockholm, Sweden, 1999.
6. M.R. Gorman, S.M. Ziola, "Plate waves produced by transverse matrix cracking", Ultrasonics 1991;29:245-251.
7. M. Åberg, "Numerical modeling of acoustic emission in laminated tensile test specimen", Report 239, Department of Solid Mechanics, Royal Institute of Technology, Stockholm, Sweden, 1999.

© AIPnD , created by NDT.net

|Home|    |Top|