SAMPE Journal - July/August 2024 - 13

The first step in the algorithm was to read in
the sample image. Examples of the images used are
shown in Figure 12. A plastic holder was created
to show contrast in the image and block shadows.
This was from a much larger single screw extruder
without a melt pump. In the figure the two examples
on the left are from a standard three-zone screw
extruder without a melt pump and the sample on
the right is from the test extruder. Since the test
extruder sample was much smaller than the threezone
screw extruder samples, the plastic holder
was not used and instead the extruded sample was
held with tweezers. It was also noted that the image
quality of the test extruder was lower than the other
samples due to the sample's size. The nozzle that
created this sample was a 2.0 mm diameter, so the
zoom required to see the porosity in the sample led
to a diminished image quality.
Once the image was read in, it was converted
to a greyscale image. Figure 13 shows the same
two samples with greyscale. At this stage it was
observed that some darker areas in the left sample
appeared to be showing porosity but when visually
inspecting that sample these were just darker
sections of the material and there was no porosity
present. This means that this method will not
be viable for combinations of base polymer and
reinforcement fibers in which the contrast between
the two elements is large.
Then, the greyscale images were thresholded
and converted to a binary image. Figure 14 shows
the three samples in binary format. This is the
image that would then be used to calculate the
porosity and a few main observations were made
at this stage including that the entirety of the
middle sample center porosity was not captured
by the black pixels meaning that the porosity
would be underestimated for that sample. Also, as
previously stated, the left samples had dark areas
of the material that were small but would lead to an
overestimate of the porosity in that sample.
Figure 15 shows the calculated porosity
percentage for each sample. This was calculated
by removing the background around the image
and dividing the porosity pixels by the total sample
pixels. Visual inspection showed that the porosity
of the second standard and test extruder samples
were likely underestimated. For the standard
extruder sample, this was likely due to some of the
central porosity catching light when the image was
taken, causing only its perimeter to show up in the
binary image. For the test extruder sample, it was
difficult to visually see if there was any porosity due
to the smaller size of the bead.
In general, a visual inspection of the bead
quality versus the standard extruder showed that
the surface quality of the bead was not significantly
different between the standard extruder samples
and the test extruder but the porosity percentage
appeared to be improved on the test extruder versus
www. sampe.org
Figure 13. Greyscale sample images (standard extruder - (left and middle),
test extruder (right).
Figure 14. Binary sample images (standard extruder - (left and middle), test
extruder (right).
Figure 15. Porosity sample images (standard extruder - (left and middle),
test extruder (right).
the standard extruder. Although its estimated
percentages were slightly different than what was
observed visually, the image processing confirmed
the same findings. The standard extruder had
less than 10% porosity in all samples that were
tested while the test extruder samples all showed
less than 5% porosity. The test extruder appeared
to show less average porosity in the samples but,
without analysis of a wide range of materials and
speeds, and with porosity depending on multiple
factors including environmental conditions, more
data is required to show that the test extruder
can significantly improve porosity. Further
investigation and improved image capturing would
be needed for this code to output the exact porosity
percentage of each sample, but it was sufficient to
put an initial scale on the porosity values without
manually measuring the size of the porous regions.
JULY AUGUST 2024
|
SAMPE JOURNAL |
13
http://www.sampe.org
SAMPE Journal - July/August 2024

Table of Contents for the Digital Edition of SAMPE Journal - July/August 2024
