DECI NEWSLETTERS AND REPORTS
THE DECI REPORT - JANUARY 1997
ACOUSTIC EMISSION ANALYSIS OF A GRAPHITE-EPOXY HONEYCOMB PANEL
INTRODUCTION
Acoustic emission technology has been used very successfully
over the past three decades for the testing of man lift vehicle
booms, pressure vessels, and piping made from fiberglass composite
materials. There is presently a great deal of interest in using
AE technology for the testing of more exotic composite materials
such as graphite-epoxy compositions. These materials have a high
strength to weight ratio, a high elastic modulus, and are suitable
for cryogenic applications.
This report shows the results of recent acoustic emission experimental
work on a graphite-epoxy honeycomb panel of the type used for
aircraft and space applications. The overall thickness of the
panel was 0.875 in. (21.9mm) with 0.070 in. (1.8mm) graphite-epoxy
panels covering the top and bottom surfaces of the honeycomb material.
The panel was 20 inches (500mm) X 32 inches (800mm) in size. The
purpose of this experimental program was the following:
1. Compare the SE9125-MI "false aperture" transducer
performance with the SE1000-HI high fidelity transducer for composite
applications.
2. Measure the velocity and attenuation of the graphite-epoxy
material.
3. Determine the effectiveness of using a waveguide with a sharp
point for composite applications.
4. Determine whether or not one can effectively measure sources
of AE on one side of the panel when the transducers are mounted
on the opposite side.
PROCEDURE
The general procedure was to couple a trigger transducer at
the edge and midpoint of one side of the panel with petroleum
jelly as shown in figure 1. The output of this transducer was
used to trigger the sweep of a digital oscilloscope. Three data
transducers, the SE1000-HI , the SE1000-HI with 4.5 inch (113mm)
waveguide with pointed end, and the SE9125-MI were used for the
results given in this report. These transducers contain internal
electronics with approximately 10dB of gain and are designed to
drive hundreds of feet of 50 ohm cable without loss of signal.
The signals from each transducer were amplified by 40dB by a preamplifier
containing a 20Khz hipass filter before being displayed by the
digital oscilloscope. These transducers were located at different
distances from the pencil lead breaks in order to measure velocity
and attenuation of the different types of waves produced in the
panel. Out-of-plane (OOP) pencil lead breaks were made on the
top and bottom surfaces with the pencil held parallel with the
plane of the panel. In-plane (IP) pencil lead breaks were made
on the center edge of the panel again with the pencil held parallel
to the plane of the panel. Holding the pencil lead parallel to
the plane of the panel is necessary in order to prevent mixed
modes from being input into the panel. The data transducers were
coupled to the panel with petroleum jelly. There was no couplant
used for the pointed waveguide.
EXPERIMENTAL RESULTS
Figure 2 shows a comparison between the SE9125-MI and the SE1000-HI
transducers when placed at different distances from the IP pencil
lead break at the edge of the panel. Note that both the extensional
and shear wave from this IP source is detected by both transducers,
with a minimal amount of the low frequency flexural wave component
present. This is due to the attempt to break the pencil lead as
closely as possible at the center edge of the top panel. The arrival
of the shear wave (second large signal) is shown clearly by the
data at the 8 inch position for both transducers and has a very
low value of approximately 53,300in/sec. (1,330m/sec.) The extensional
velocity is approximately 200,000in/sec. (5,000m/sec.) which is
quite high for a composite material.
Note the change in the vertical scale for the three distances.
If one utilizes the data for the extensional wave in figure 2(e)
and 2(f), an order of magnitude change in amplitude is observed
for 8 inches (200mm) of travel indicating an approximate attenuation
in this particular panel of 2.5dB/inch (.1db/mm) for an IP signal.
Figure 3 is a repeat of the experimental data in figure 2 with
the exception that the data observed is a result of out-of-plane
(OOP) 0.3mm pencil lead breaks made on the top edge of the panel.
Data in figure 2 was made with 0.5mm lead breaks which give approximately
twice the signal strength of 0.3mm leads. At the time the data
was taken in figure 2, I was temporally out of 0.3mm pencil leads.
There are several interesting observations one can make when comparing
the data in figure 3 for OOP sources as opposed to the IP data
in figure 2. First note that there is a lack of the extensional
wave component at this amplification in Figure 3 for the small
aperture SE1000-HI and only faint evidence of this wave in the
data from the SE9125-MI. Secondly the shear wave component is
observed in all of the signals in figure 3 as well as in figure
2. Third, the zero order antisymmetrical lamb wave (flexure wave)
is the predominate feature of the data from the SE1000-HI in figure
3. Fourth, note the change of vertical scale between the data
for the SE1000-HI and the SE9125-M in figure 3. The SE1000-HI,
a small aperture mass loaded transducer is much more sensitive
to large displacements produced by a flexure wave, while the larger
aperture partially mass loaded SE9125-M has greater sensitivity
than the SE1000-HI to extensional and shear waves.. Sixth, previous
reports have shown that splitting the signals from the output
of the SE9125-M into a low and high frequency components and taking
the ratio of the high frequency/low frequency peak voltages resulted
in voltage ratios of less than one for OOP sources and greater
than one for IP sources for experiments conducted on steel bars.
The data in figures 2 and 3 confirm that these ratios are also
valid for this composite panel.
I was confident that the extensional and shear waves recorded
in these experiments were propagating in the panel adhesively
bonded to the honeycomb structure which the trigger and data transducer
were mounted on. I was not so certain that the low frequency flexure
wave was propagating in this panel alone. I was also interested
in determining if a transducer mounted on one panel could detect
IP and OOP waves that might be generated in the panel on the other
side of the honeycomb structure. The following experiment was
performed to answer these questions. Data was taken for the condition
of the trigger and data transducer mounted on the same panel for
IP and OOP signals. The data transducer was left in place and
the trigger transducer was moved to the opposite panel, and pencil
lead breaks were made adjacent to the trigger transducer. The
SE9125-MI transducer was used for the IP experiment since it has
a much better response for this type of signal. The SE1000-HI
was used for the OOP experiment since it has much better response
to this type of signal. Figure 4 shows the results of this experiment.
Note from the SE9125-MI results that moving the source and trigger
transducer to the bottom plate, while leaving the data transducer
in place, results in approximately 10dB of attenuation and a slight
time delay. This time delay corresponds to the amount of time
it takes for the signal to travel through the honeycomb structure
to the top panel.
Note from SE1000-HI that there is no attenuation of the OOP signal
when the source and trigger transducer are moved to the bottom
panel. In fact this signal is higher amplitude than when source
and trigger sensor were on the same panel as the data transducer.
I ascribe this difference in amplitude to the lack of exact repeatability
of the source input when breaking pencil leads. Note that the
two signals are 180 degrees out of phase. The lack of attenuation
and the 180 degree phase shift are positive evidence that the
whole structure rather than a single panel is sustaining the low
frequency flexure wave due to the OOP source.
In previous reports
I have shown that a SE1000-HI mounted on a 1/4 inch diameter stainless
steel rod with a pointed end, could be used effectively in detecting
very small displacements in metal bars. This experiment was repeated
in this work on the graphite-epoxy panel. An aluminum cylinder
1.5 inch in diameter and 2 inches long with a 1/4 inch diameter
hole was used to support the waveguide in the vertical direction
(figure 1). The waveguide transducer combination was placed 8
inches from OOP 0.3mm pencil lead breaks and lead breaks were
made on the top and bottom surface of the panel. The waveguide
transducer was always positioned on the top panel, but the trigger
transducer was placed at the location of the pencil lead breaks.
No couplant was used between the tip of the waveguide and panel.
Figure 5 shows the results of this experiment. A lot of "ringing"
of the signal occurs due to reflections in the wave guide, but
the sensitivity and phase information (notice the 180 degree phase
shift for the two conditions) is maintained.
DISCUSSION OF RESULTS
There were three types of waves observed in this panel during
the course of these experiments: extensional and shear waves in
the top and bottom cover plates, and a low frequency flexure wave
in the honeycomb panel as a whole. Each of these waves travels
at a different velocity. It was found that at least 8 inches between
the source and transducer was required in order to get good time
separation of the three different wave types. The SE9125-MI data
for the 8 inch distance shown in figure 3 is a good example of
the three types of waves present. Note that the peak voltage of
the high frequency shear wave is approximately the same value
as the low frequency flexure wave for this OOP signal. This transducer
was designed to produce a voltage ratio of 1 or less for these
two frequency components for OOP signals based on experiments
on metal bars. It appears that this feature also is valid the
composite specimen used in these experiments. Note that the ratio
of high frequency/low frequency (HF/LF) in figure 2 from the SE9125-MI
for IP signals is much greater than 1. Therefore one should be
able to identify whether or not an AE signal is due to fiber breaking
or matrix cracking (IP sources) as opposed to delamination, impact
or friction (OOP) sources by simply looking at the peak voltage
ratio (HF/LF) from the SE9125-MI.
This could have important implications for monitoring of space
vehicles constructed from composite materials. For example assume
that a composite tank is used in a space vehicle to carry liquid
hydrogen, and one wishes to use AE to determine if meteorite impacts
occur and if so, has any damage occurred. A meteorite impact would
act as an OOP source and a SE9125-MI mounted to the pressure vessel
would read a HF/LF ratio of 1 or less. If the meteorite penetrates
the shielding and impacts on the composite pressure vessel and
causes fibers to be broken, a HF/LF ratio of greater than 1, indicating
an IP source will be produced which indicates that damage has
occurred. Once an IP source has been identified assessment of
the damage can estimated by the energy, amplitude, or counts of
the AE signal produced by the IP source. The risetime of the AE
signals produced by IP sources (figure 2) is quite sharp and conventional
time of arrival from several transducers could be used to accurately
locate the source. The IP signals in figure 2 were produced by
breaking pencil leads at the center edge of the panel and due
to symmetry very little flexure wave component was produced. In
real life the signals will not be produced at the exact center
of the panel, and therefore some flexure wave component will be
present and will arrive at a later time. The SE9125-MI is designed
such that for an IP signal the higher frequency faster traveling
wave will always be of higher amplitude than the slower flexure
wave and will prevent inaccuracies in measuring time of arrivals
encountered with standard AE transducers. One would have problems
locating IP sources in the conventional way from signals produced
by the high fidelity SE1000-HI. Note in figure 1 that the extensional
wave produced by this transducer is much lower in amplitude than
the following wave forms and thus could produce large time difference
errors if one triggered on the extensional wave from one transducer
and the flexure wave from another. Time difference errors in locating
OOP sources with the SE9125-MI are significantly reduced due to
the fact that the slower flexure wave is approximately the same
amplitude as the faster wave (figure 3). This is not the case
for the SE1000-HI (figure 3). The low frequency flexure wave produced
by OOP sources can be detected very well with the pointed waveguide
configuration (figure 5). This zero aperture configuration is
not well suited for detecting IP sources. In our space application
example, if one were only interested in measuring and locating
impacts from meteorites, this system could be used for this purpose.
The advantage in using this type of configuration is that the
transducer and associated electronics would be protected from
the large thermal fluctuations that occur if bonded directly to
the pressure vessel.
Both the SE9125-MI and the SE1000-HI have frequency response up
to 1 Mhz. The instrumentation bandwidth used in these experiments
were 20Khz to 1Mhz. FFT frequency analysis was performed on signals
from both transducers. The frequency components above 200Khz were
negligible compared to those below 200Khz. From a practical standpoint
it appears that it makes no sense to attempt to use high frequency
transducers and instrumentation for monitoring structures of this
type of material because of the high attenuation in the high frequency
range. Use of higher frequencies might be beneficial for small
coupon testing in the laboratory, but it will be difficult to
translate any information gained to practical field applications.
CONCLUSIONS
Acoustic Emission testing of the graphite-epoxy panel in this
experimental yielded the following information. 1. The panel is
anisotropic in that a 10% difference in velocity was observed
between 0 and 90 degrees. For berivity only data in one direction
was presented in this report. 2. The extensional and shear velocity
in the panel was 5000m/sec and 1,330 m/sec respectively. 3. Attenuation
for extensional waves was 0.1dB/mm and for flexure waves 0.05dB/mm.
4. OOP pencil lead breaks resulted in a low frequency flexure
wave propagating in the whole panel including the honeycomb portion.
This shows that it makes no difference which side of the panel
a transducer is located on for detecting this type of signal.
Approximately 10dB of attenuation was observed for extensional
waves when the IP source is on the bottom panel and the transducer
is on the top panel. 5. The SE9125-MI transducer was equally sensitive
to both IP and OOP generated signals. 6. The AE signals had very
little frequency content above 200Khz for transducer spacing used
in these experiments. 7. A pointed waveguide mounted on a SE1000-HI
could be used for detecting flexure waves due to OOP sources on
either the bottom or top panel with very little attenuation when
compared to the transducer mounted directly on the panel. 8. A
shear wave was always detected by the SE9125-MI for both IP and
OOP source inputs.