Documente Academic
Documente Profesional
Documente Cultură
NDT&E International
journal homepage: www.elsevier.com/locate/ndteint
a r t i c l e i n f o a b s t r a c t
Article history: Voids and porosity are critical imperfections in fibre reinforced composite materials. The aim of this
Received 27 July 2011 study is to assess the comparative accuracy and reliability of conventional and novel void character-
Received in revised form isation techniques for analysing voids within carbon fibre reinforced composite (CFRC) laminates.
29 November 2011
While microscopy and Archimedes density measurements can only give results of limited accuracy and
Accepted 30 November 2011
reliability, micro-computed tomography (micro-CT) analysis was found to have no significant inherent
Available online 8 December 2011
errors and is able to characterise the three-dimensional size, shape and distribution of voids in CFRC
Keywords: laminates. The accuracy of micro-CT was also verified by analysing artificial voids of a known size and
Voids shape disbursed throughout a CFRC laminate sample.
Porosity
& 2011 Elsevier Ltd. All rights reserved.
Micro-CT
Fibre reinforced composites
0963-8695/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ndteint.2011.11.011
J.E. Little et al. / NDT&E International 46 (2012) 122–127 123
distribution. It has been proven that two laminates that show the W coupon
rcoupon ¼ ð2Þ
same void content can have a varying mechanical response [8], V coupon
which highlights the important of visual analysis. Whilst micro-
The void volume fraction (VV) is able to be calculated using
scopy can detect this important visual information, it is however
Eq. (3), with the mass fraction (%m) and density (r) of the primary
limited by the information being 2D and non-representative of
constituents required. It is critical that the material properties are
the whole sample. Unless multiple images are obtained on very
accurately known, as minor variations can significantly affect the
thin slices, it also has potential for missing any pore intercon-
accuracy of the calculated void volume fraction.
nectvity between adjacent slices.
!
X-ray computed tomography is a novel non-destructive and %mmatrix %mf ibre
visual technique that can be used for three dimensional (3D) V v ¼ 100rcoupon þ ð3Þ
rmatrix rf ibre
analysis of internal structures [1,9]. Computed tomography was
conceived in the early 1960s and has been predominately used for In order to gain this visual information a sample cross-section
medical diagnosis [1]. Over the past 20 years, however, micro- is required, making this technique destructive. The testing pre-
computed tomography (micro-CT) has been developed and allows paration required for microscopy analysis of composite materials
for wider application, including the qualitative and quantitative is a time consuming and sometimes difficult process due to the
non-destructive evaluation of composite materials [1]. non-homogenous composition of composites. Microscopy pre-
Micro-CT imaging works by recording the change in attenua- paration requires the sample to be set in resin and then a flat
tion along the path of an X-ray beam fired through a sample [10]. cross-section prepared to a mirror like finish, with no visible
A cross-sectional image is then mathematically reconstructed preparation damage, such as scratches and void distortion. The
from attenuation measurements made at different angles by the high quality surface finish is achieved by sequentially grinding
detectors [10]. Each 2D slice is of a known thickness, which allows the surface with smaller sized particle paper, from 35 to 2.5 mm.
the images to be virtually stacked to create a 3D model. The image This process is difficult, as the brittle fibres regularly break free
stack can then be converted to binary (black and white), allowing and gouge large scratches in the specimen surface. It is very
for the voids to be separated from the fibre and matrix which important that the finished surface has little or no scratches,
make up the laminate. as these will be counted as a void when using image analysis
This research paper compares the aforementioned conven- software.
tional techniques to micro-CT analysis for analysing voids in CFRC The numerical analysis of microscopy images is carried out
materials. Matrix digestion and matrix burn-off were not tested, using image analysis software, which works by analysing the
as they have significant errors and/or difficulties when dealing individual pixels that make up the image cross-section. The image
with carbon fibre reinforced plastic laminates. analysis process requires the microscopy cross-section image to
be converted to binary, with the voids remaining black and the
matter (fibre and matrix) converted to white. The image is then
2. Experimental procedures ‘cleaned’, to remove unwanted noise and outliers [12]. This
conversion and cleaning process is susceptible to error only if
To compare the various composite characterisation techniques, care and common sense are not applied. The image is then able to
two 24-ply oven cured carbon fibre reinforced composite (CFRC) be analysed using built in analysis tools within the software,
laminates were manufactured. One panel was made from unidirec- which is able to calculate the total number of voids, their overall
tional (UD) preimpregnated (prepreg) carbon fibre, while the other area fraction, plus analyse individual void size, shape and two
was a twill weave carbon fibre prepreg. The laminates were oven dimensional location.
cured under vacuum to ensure that voids were present within the The micro-CT scanning and image reconstruction was conducted
laminates for analysis. Samples measuring 15 by 7 by 5 mm were at the Auckland Bioengineering Institute. The micro-CT scanning
cut from the middle of the panels for testing. The size of the reconstruction parameters were kept constant for all samples, while
specimens was a compromise between the different techniques’ the alignment parameter was tailored for increased accuracy. Micro-
requirements, with micro-CT requiring small samples to reduce data CT scanning requires little specimen preparation, as this technique is
file size. Carbon fibre epoxy laminates were used for the samples as non-destructive with respect to the sample specimen. The scanning,
characterisation interest primarily lies with this material. image reconstruction and the data acquisition processes can be time
The determination of the void volume fraction using the consuming steps and vary with respect to the sample size and the
Archimedes Theoretical versus Actual Density technique was micro-CT scan settings, most notably the scanning resolution. A
carried out in accordance with ASTM D3171. Due to size restric- constant scanning resolution of 7 mm was used for each sample and
tions of the micro-CT technique, the samples were smaller than represents the physical dimension of the voxels (3D pixel) making
the recommended testing standard sample volume of 1 cm3. This up the image slices.
variation is necessary and is assumed to have negligible impact The micro-CT data acquisition process was completed using
on the accuracy of the Archimedes results. software provided by Skyscan, the micro-CT scanner’s manufac-
The testing procedure requires the density of the sample to be turer. This software has the ability to be able to calculate overall
measure using Archimedes’s Principle, where the specimen is void fraction and the individual void size, shape and distribution
weighed in air (Wspecimen) and then weighed while submerged in three dimensions (3D), while also being able to produce 3D
(Wsubmerged). It is critical that no air bubbles are present on the rendered models of the voids.
coupon while submerged, as this will decrease the measured
mass and thus inaccurately increase the measured void volume
[11]. The coupon volume (Vcoupon) can be calculated in conjunc- 3. Results and discussion
tion with the submerging liquid density (Eq. (1)), which in this
case is distilled water (rwater). The actual density of the sample The following sections concentrate on the characterisation
specimen is then calculated using Eq. (2). techniques themselves, while using the sample laminates as an
example of the capabilities and comparative accuracy of each
W coupon W submerged technique. The overall void volume fraction calculated for each
V coupon ¼ ð1Þ
rwater technique is compared in a later section.
124 J.E. Little et al. / NDT&E International 46 (2012) 122–127
The Archimedes Theoretical versus Actual Density testing There are various kinds of microscopes that can be used for
technique is a non-destructive and non-visual technique that imaging material cross-sections, with optical and environmental
is able to calculate the overall void volume fraction within a scanning electron microscopy (ESEM) the most common for void
specimen, providing accurate material properties and mass frac- analysis [13].
tion values are known and the testing procedure is accurately Fig. 1 shows the same area of a twill weave laminate cross-
performed. Since this technique is non-visual, the only quantita- section under both optical and ESEM microscopy. The optical image
tive data ascertained is the total void volume fraction of the (Fig. 1a) shows the fibres as an illuminated white, the matrix as grey
sample. and the voids black. While the ESEM image (Fig. 1b) shows the fibres
This Archimedes based characterisation technique requires as a light grey, the matrix as a slightly darker grey and the voids
accurate knowledge of the material’s physical properties, includ- black. The most notable and critical difference between the two
ing both constituents’ densities and mass fractions. The necessity microscopy images is the superior contrast between void and matter
of having to know physical properties and make-up of the (fibre and matrix) of the ESEM image in comparison with the optical
laminate is a considerable weakness of the Archimedes based image. When utilising image analysis software, it is important that
testing procedures, and limits it to samples of known material the contrast between void and matter is distinguishable so that an
properties. Small variations in densities or the mass fractions can accurate binary conversion can occur for image analysis purposes.
lead to noticeable variations in the overall result, which can lead An example of this contrast difference can be seen around the large
to calculating the impossibility of a negative void fraction. circular void in Fig. 1, where the optical image shows a dark and less
The material properties, fibre and matrix density, of both well defined area to the right of the void. Optical microscopy,
laminates were acquired from the carbon fibre prepreg manufac- however, has a greater contrast between fibre and matrix, which
turer. Due to cure variables the exact matrix density is unknown, is beneficial if the primary constituents are being examined. Another
with a provided range between 1.15 and 1.18 g/cm3. Due to this benefit of optical microscopy is the financial investment required, as
variation only a range of the sample’s void volume fraction could optical microscopes are generally much cheaper to own, operate or
be calculated (Eq. (3)), with the exact value remaining unknown. lease. In this study, an ESEM microscope was solely used for the
The pre-cure weight fractions were acquired from the material cross-sectional analysis, since it is the preeminent microscopy
data sheets that are produced for each batch of prepreg material technique for the imaging of voids in CFRC laminates.
by the manufacturer, with the final values taking into considera- An advantage of ESEM is the capability to easily and relatively
tion the resin bleed during cure. The complications of having to efficiently image the entire specimen cross-section. Since an ESEM
accurately know the material properties is an inherent error of detector is computerized, it is able to incrementally change location
Archimedes based characterisation techniques (including matrix during imaging, which allows the whole cross-sectional surface to
digestion and burn-off), which limit their effectiveness and use to be systematically imaged. The sequential images are able to be
known materials. meshed together to create a complete image of the specimen cross-
In some cases, however, the Archimedes Theoretical versus section (Fig. 2). The cross-sections for this study, 15 by 5 mm,
Actual Density technique is more suitable than others, as visual required twenty one images (Fig. 2) at 100 magnification.
information is not always, at least initially, required. This techni- Fig. 3 is an example of the numerical analysis that is able to be
que is suited for quality control purposes, as it is able to handle performed using image analysis software. This graph shows the
relatively large samples and is less time consuming than other
techniques. An example application of this technique is the
inspection of autoclaved cure products for manufacturing defects,
such as bag rupture and corner bridging. These particular man-
ufacturing defects generally cause significant increases in the
overall void volume fraction, so providing the material properties
and constituent volume fractions are within reasonable accuracy
these faults are able to be identified using this technique. If a fault
is found the cause of the increased void volume fraction is then
able to be investigated further, using more time consuming and
destructive visual techniques, or simply, the specimen can be
discarded. Fig. 2. An example of ESEM microscopy cross-section.
Fig. 1. (a) Optical image and (b) SEM image of the same area of a twill weave CFRC laminate cross-section.
J.E. Little et al. / NDT&E International 46 (2012) 122–127 125
cumulative number of voids and their cumulative area fraction Microscopy analysis allows important individual void charac-
plotted against void projected area diameter for the unidirec- teristics to be assessed. Fig. 3 for example shows that E45% of
tional laminate cross-section shown (Fig. 2). the total number of voids with a projected area diameter between
To limit the effects of section-bias three cross-sections were 15 and 200 mm have a projected area diameter of less than 50 mm,
analysed per sample, making a total of fifteen cross-sections per while only accounting for E6% of the total void area fraction. On
laminate. This large number makes the analysis process very time the other hand, this graph shows that voids with a projected area
consuming, with each surface requiring careful preparation to a diameter greater than or equal to 200 mm account for only 1.5% of
mirror like finish. the total number of voids greater than 15 mm, but contribute to
E24% of the overall void area fraction. This example highlights
the benefits of visual analysis, where the individual character-
istics of the voids are able to be evaluated against each other.
As can be seen in Fig. 2, the voids observed within a single
cross section of a laminate are not consistent, but have variable
shape and distribution due to the section cutting through pores at
different orientations. Due to this fact, microscopic characterisa-
tion of composite laminates has an inherent section bias, where
one cross-section is not necessarily an accurate representation of
the complete sample. In order to limit the effects of section bias, it
is important that multiple cross-sections are analysed, which is a
time consuming and costly process. It is also difficult when
obtaining multiple sections of the same sample to know exactly
where those sections are being taken from within the sample.
3.3. Micro-CT
Fig. 3. The void number and voidarea cumulative distribution curves from the The limitation of current micro-CT technology means that only
ESEM cross-section in Fig. 2. small samples are able to be scanned. Small increases in the
Fig. 4. Orthographic snapshots of a 3D unidirectional model from micro-CT analysis, with (a) top, (b) 3D model and (c) front and (d) side.
126 J.E. Little et al. / NDT&E International 46 (2012) 122–127
4. Conclusion
Acknowledgement