SAMPE Journal - January/February 2023 - 17

Figure 5. Surface plots generated during ML filtering of two different regions in the target area
containing a) resin contamination (yellow) and b) aged release coating (red).
no aging effects after twenty rounds of autoclave
processing. These results indicate that reapplying
one RA layer after every three processing cycles
mitigates release coating aging effects.
The
remainder of this section focuses on the tool that
received no RA reapplications and thus exhibited
significant aging after twenty processing cycles.
The results reported in this section are meant to
demonstrate the novel inspection technology's
potential to identify and evaluate aged release
coating and contamination in a production
environment.
Global Mapping
Figure 4 shows an image of the production
tool surface after undergoing twenty autoclave
processing cycles without any RA reapplications.
The image demonstrates that elevated
temperatures and pressures during processing
caused resin to bleed from the prepreg, flow to the
tool's outside edges, and cure to the surface as a
contamination border. The thickness and surface
area of resin contamination increased with each
processing cycle and consequently decreased the
area of abhesive release coating on the tool surface.
If production tools are not cleaned regularly,
contamination can continue to grow into the layup
surface and have detrimental effects on tool
www. sampe.org
abhesion, PIDs, and end-part quality. Throughout
the twenty autoclave processing cycles, the
contamination border remained outside the lay-up
dimensions of the composite parts. Therefore, the
significant decrease in tool abhesion after twenty
cycles was likely due to aging within the lay- up
boundaries rather than from the contamination
border. The remainder of the inspection was thus
conducted within the tool's lay-up boundaries.
The red box in Figure 4 outlines the target area
identified by the supervised ML model to contain
color abnormalities. After twenty processing
cycles, the tool's lay-up surface was mainly covered
by an opaque, abhesive release coating. However,
the tool section that was initially treated with only
one layer of RA contained light and dark patches
on its surface. Light patches corresponded to resin
contamination, while dark patches were aged
release coating (verified using handheld FTIR). A
lack of contamination and aging in other sections
indicates that applying more RA layers prior to
processing creates a longer-lasting barrier between
the tool and part. In other words, the number of
RA layers may directly affect the rate of aging on
a tool surface. Nevertheless, all significant aged
and contamination patches were contained in the
target area, demonstrating the ML model's ability to
effectively identify color defects on a tool surface.
JANUARY FEBRUARY 2023
|
SAMPE JOURNAL |
17

http://www.sampe.org

SAMPE Journal - January/February 2023

Table of Contents for the Digital Edition of SAMPE Journal - January/February 2023