Journal of Acoustic Emission, Vol 14(3-4), (1996), pp. S1-S11
Advanced AE Techniques in Composite Materials Research
William H. Prosser
Abstract
Advanced, waveform based acoustic emission (AE) techniques have been
successfully used to evaluate damage mechanisms in laboratory testing of composite
coupons. An example is presented in which the initiation of transverse matrix cracking was
monitored. In these tests, broad band, high fidelity acoustic sensors were used to detect
signals which were then digitized and stored for analysis. Analysis techniques were based
on plate mode wave propagation characteristics. This approach, more recently referred to
as Modal AE, provides an enhanced capability to discriminate and eliminate noise signals
from those generated by damage mechanisms. This technique also allows much more
precise source location than conventional, threshold crossing arrival time determination
techniques. To apply Modal AE concepts to the interpretation of AE on larger composite
specimens or structures, the effects of modal wave propagation over larger distances and
through structural complexities must be well characterized and understood. To demonstrate
these effects, measurements of the far field, peak amplitude attenuation of the extensional
and flexural plate mode components of broad band simulated AE signals in large composite
panels are discussed. These measurements demonstrated that the flexural mode attenuation
is dominated by dispersion effects. Thus, it is significantly affected by the thickness of the
composite plate. Furthermore, the flexural mode attenuation can be significantly larger than
that of the extensional mode even though its peak amplitude consists of much lower
frequency components.
1. Introduction
The capabilities of AE testing in composite materials research have been
significantly improved by several recent advances. These include the development of
digital, waveform based, acquisition instrumentation with sufficient memory and
acquisition rates for AE testing. Another important development has been the
improvements in high fidelity, high sensitivity, broadband sensors. However, most
important has been the increased understanding of the nature of AE signal propagation as
guided acoustic modes in common testing specimen geometries such as thin plates and
coupons (Gorman, 1991; Gorman and Prosser, 1991; Prosser, 1991). Analysis of guided
mode AE signals has been designated Modal AE. It has led to significantly improved AE
source location accuracy (Ziola and Gorman, 1991). Modal AE has also provided the
capability to better differentiate AE signals from different source mechanisms including
extraneous noise (Ono, 1994 and Prosser et al., 1995).
Two sets of Modal AE measurements are presented herein. The first is an example
of a successful application in composite materials research to monitor the initiation of
transverse matrix cracking in cross-ply graphite/epoxy coupons. In these experiments,
noise signals created by damage in the grip region were differentiated from crack signals by
waveform analysis. The signals from matrix cracks contained a higher amplitude
extensional plate mode component with little or no flexural mode. The grip damage signals
contained significant flexural mode components.
The second set of data are measurements of the peak amplitude attenuation of the
extensional and flexural plate modes in large composite plates. The flexural mode suffered
considerably more amplitude loss even though its frequency content was much lower.
Dispersion of the flexural mode, which causes a spreading of the signal in time over
increasing propagation distance, is the dominant mechanism for this high attenuation. As
shown in the transverse matrix cracking study, simple analysis of relative plate mode
amplitudes is useful for source discrimination in coupons where the source to sensor
propagation distance is small. However, these results suggest that careful consideration of
attenuation effects will be required to extend this approach to larger specimens such as
panels or realistic structures.
As requested by the workshop organizers, additional comments on the
generalization of these results are included in the concluding remarks. In particular, it is
noted that although Modal AE analysis has been successfully used to discriminate source
mechanisms in composite materials, further research is required before the approach can be
used in arbitrary materials, laminates, and/or specimen geometries. Further developments
in modeling AE wave propagation which may provide insight into the effects of different
source mechanisms on AE waveforms, will likely speed this process. It is also suggested
that automated waveform analysis approaches such as pattern recognition and neural
networks will be most successful at source discrimination when based on knowledge of
wave propagation. Furthermore, factors such as attenuation and complicated structural
geometries must be carefully considered when extending Modal AE analysis to the testing
of large specimens and real composite structures.
2. Modal AE
A number of early AE studies, including those by Pollock (1986), Stephens and
Pollock (1971), Egle and Tatro (1967), and Egle and Brown (1975), made passing
mention of the propagation of AE waves as guided acoustic modes in practical testing
geometries such as coupons, plates, shells, pipes, and rods. However, these works
offered little as to the importance of these modes on the interpretation and analysis of AE
with respect to source location accuracy and identification of source mechanisms. In fact,
Pollock (1990) raised these same questions in a review paper on critical problems for
research in AE. At about this same time, Gorman (1991) and Gorman and Prosser (1991)
published work on the effects of guided wave AE propagation in plates. It was pointed out
that in thin plates and coupons, the two observed modes of propagation in AE signals are
the extensional and flexural plate modes. The predominant particle displacement for the
extensional mode is in the plane of the plate. The largest component of the flexural mode
particle displacement is out of the plane of the plate. A source motion with predominantly
in-plane components and symmetric about the midplane generates AE signals with large
extensional mode components. Examples of such a source motion include fatigue cracking
in metals and matrix cracking in the center plies of a composite laminate. Out of plane
source motion such as delamination or impact damage produces AE signals with large
flexural mode components. This discovery led to a waveform analysis method to identify
sources and discriminate noise signals and is the basis for the Modal AE technique.
The extensional mode propagates with a faster velocity and suffers little dispersion
over the frequency range observed in most AE experiments (20 kHz to 1 MHz). It
typically contains higher frequency components than the flexural mode. The flexural mode,
however, propagates with a slower velocity and is highly dispersive with the higher
frequencies traveling at higher velocities. A typical waveform detected in a composite plate
with a broad band sensor identifying these two modes is shown in Fig. 1. The source of
this signal was a simulated AE event caused by a pencil lead fracture (Hsu-Neilsen source)
on the surface of the composite plate.
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
100 150 200 250 300 350
Flexural Mode
Extensional
Mode
Time (µsec.)
Fig. 1 Simulated AE signal in composite plate identifying extensional and flexural plate
modes.
3. Detection of Transverse Matrix Crack Initiation in Cross-Ply Laminates
The initiation and progression of transverse matrix cracking in composite materials
has been, and remains, a subject of considerable interest and importance. A vast amount of
literature on the experimental detection of matrix cracks is available, of which, a small
sampling is reviewed by Prosser et al. (1995). The improved source location accuracy and
enhanced noise discrimination capabilities of the Modal AE technique are demonstrated in
this study of the transverse matrix crack initiation in cross-ply laminates of different
stacking sequences. This work improved upon a recent similar study by Gorman and Ziola
(1991) in which and only a single cross-ply laminate was tested.
Tensile coupon specimens (2.54 cm. wide by 27.94 cm. long) of AS4/3502
graphite/epoxy composite material were loaded in tension under stroke control (0.127
mm/minute). As grip noise was eliminated by waveform analysis, specimen end tabs were
not used in these tests. Specimens from six different cross-ply laminates were tested. The
stacking sequences were [0
n
, 90
n
, 0
n
] where n ranged from one to six. Thus, the samples
varied in thickness from 3 to 18 plies.
Broadband, high fidelity sensors (Digital Wave Corporation B1000) were used to
detect the waveforms. Rather than a single sensor at either end of the specimen as in many
previous works, four sensors were used. At either end of the nominally 152 mm.
specimen gage length, a pair of sensors were positioned. The outer edge of each 6.35 mm
diameter sensor was aligned with the edge of the specimen. A diagram of a specimen
showing the sensor positions and the grip regions is shown in Fig. 2. The motivation for
this sensor array arrangement was the determination of the initiation site of the crack. Not
only could the linear location along the length of the specimen be determined, but lateral
location information was also obtained. The maximum digitization sampling frequency (25
MHz) of the digital AE acquisition and analysis system (Digital Wave F4000) was used to
provide the most accurate location results. Location was performed, post-test, using
manual, cursor based phase point matching on the extensional mode for arrival time
determination. The extensional mode velocities used for the location analysis were
measured prior to testing using simulated AE sources.
15.2 cm.
27.94 cm.
2.54 cm.
Grip Region
AE Sensors
Fig. 2 Diagram of specimen showing grip region and position of AE sensors
After detection, the signals were amplified 20 dB by wide band preamps (Digital
Wave PA2040G). It was determined during the tests that the signal amplitudes were a
function of the 90 degree layer thickness, so additional system gain was varied to maintain
the signal within the dynamic range of the 8 bit vertical resolution of the digitizer. Thicker
specimens generated signals of larger amplitude. The additional system gain ranged from
as little as 6 dB for the thickest specimen to 18 dB for the nine ply specimen (n = 3). For
the three and six ply laminates (n = 1 and 2), the signal amplitudes were significantly
smaller as will be discussed below. For these, the preamp gain was increased to 40 dB and
the system gain was set as high as 18 dB in attempts to capture the much smaller amplitude
signals.
After detection of one or more transverse matrix crack AE signals, the specimen
was removed from the test machine. One edge of the specimen, which had been polished
prior to testing, was examined under an optical microscope. The specimen was mounted
on an x-y translation stage to allow measurement of crack locations for comparison with the
AE data. Backscatter ultrasonic scans were taken to further confirm the crack locations and
to provide information about the lateral extent of the cracks. This method also confirmed
that no cracks existed which were not detected at the one polished edge. In some cases,
penetrant enhanced radiography was also used as was destructive sectioning and
microscopy.
Extraneous noise signals were eliminated by post test analysis of the waveforms.
Typical waveforms from both a crack source and a noise source are shown in Fig. 3.
Because of the multiple reflections of the signals across the narrow width of the coupons,
the signals are more complicated than those presented in Fig. 1 which were detected in a
large plate. However, the high frequency extensional mode is clear in the crack signal. A
small extensional mode component is observed in the noise signal followed by a much
larger, low frequency, dispersive flexural mode signal. The source of the noise signals is
believed to be grip damage or specimen slippage in the grips as all of the noise signals
located outside the specimen gage length in the grip regions.
Time (µsec.)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
20 30 40 50 60 70 80 90 100
Amplitude (volts)
a)
20 30 40 50 60 70 80 90 100
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Time (µsec.)
b)
Fig. 3 Typical signals caused by a) transverse matrix crack and b) grip slippage or damage
For the laminates with n = 3 or larger, there was an exact one-to-one correlation
between AE crack signals and cracks confirmed with microscopy. Backscatter ultrasonics
indicated that all of these cracks extended across the full width of the specimen and that
none were present which were not observed by microscopy of the polished edge.
Destructive sectioning and microscopy of a few of these cracks also confirmed this result.
The fact that only a single AE signal was detected for each crack indicates that the cracks
immediately propagated across the width of the specimen.
Location analysis of the four sensor array data showed that all cracks initiated along
one of the specimen edges. A typical four channel set of waveforms from a matrix crack
signal is shown in Fig. 4 along with a diagram indicating the sensor positions and the crack
location. The time delay between the sensor pairs associated with the crack initiation site
being located along the edge is clearly seen. Furthermore, differences in signal amplitudes
between the sensors within a pair are the result of the increased attenuation from
propagation across the specimen width. The differences in signal amplitudes and
frequency content for signals detected at opposite ends of the specimen and thus different
distances of propagation distances should also be noted. These differences, which are
caused by attenuation and dispersion, can have significant effects on location accuracy in
threshold based arrival time AE measurement systems. Conventional amplitude
distribution analysis is also affected by this attenuation. Excellent crack location accuracy
along the length of the specimens was also obtained from the AE data as compared to
microscopy measurements. The most accurate linear location was obtained by using the
two sensors on the same edge as the crack initiation site. The average of the absolute value
of the difference in crack locations from AE and microscopy was 3.2 mm for a nominal
sensor gage length of 152 mm.