che_december-2023 - 7

Machine learning approach aids
3D-printed part qualification
M
etal components made with
laser powder-bed fusion -
an important additive manufacturing
process - must be
qualified for use in critical applications,
but the qualification relies on expensive
non-destructive evaluation (NDE) techniques.
In-process monitoring offers a
less expensive alternative, but existing
sensing and data analysis techniques
are not reliable for detecting sub-surface
flaws, such as cracks and porosity.
To help address this, researchers at
Oak Ridge National Laboratory (ORNL;
Oak Ridge, Tenn.; www.ornl.gov), in
collaboration with defense and aerospace
company RTX Corp. (formerly
Raytheon; Arlington, Va.; www.rtx.
com), have developed a framework for
quantifying the probability of detecting
flaws in the component using in-process
monitoring, and thus potentially
reducing the expense associated with
NDE and improving the confidence in a
part's safety and reliability.
The framework relies on a machinelearning
algorithm that is trained using
information collected from sensors
during the 3D-printing process, combined
with data from X-ray computed
tomography (XCT) of completed parts.
The algorithm was trained to label flaws
on the XCT scans, and was aided by
human inspection. Over time, the algorithm
recognizes flaws more accurately
and avoids " false positives. "
" We can detect flaw sizes of about
0.5 mm 90% of the time, " says ORNL
researcher Luke Scime. " We're the first
to put a number value on the level of
confidence possible for in-situ flaw detection, "
he adds.
The ORNL-developed inspection
framework could help expand additive
manufacturing applications. " With
statistically verified quality control, additive
manufacturing could become viable
for mass-producing products like
car parts, " ORNL researcher Zackary
Snow says.
25 times greater than that of CO2.
The proposed project would recover
methane gas currently released into the
atmosphere from POME, refine it into
biomethane, and supply it to customers
in Indonesia via natural gas pipelines and
other existing infrastructure.
HYBRID HEATER
Last month, Fives (Paris, France; www.
fivesgroup.com) launched e-Ductflame,
a patented electric/gas hybrid
system intended for a wide range of
heating and drying applications, and
can either replace existing duct burners
or be retrofitted into new installations.
It has been designed to cut carbon
emissions and improve the efficiency
of industrial heating processes with
minimum costs.
The e-Ductflame has the flexibility to
operate with 100% electricity or 100%
fuel gas, and has a user-friendly control
interface for easy operation and maximum
flexibility for fuel switching.
PFAS REMOVAL
Last month, Lanxess AG (Cologne,
Germany; www.lanxess.com) intro(Continues
on p. 8)
CHEMCAD process simulation
software can't mix cocktails.
But it brings your creative ideas to life.
CHEMCAD provides rigorous, fi rst- principles based
simulation for continuous and transient processes.
It is also backed by experienced and responsive
support engineers. Contact us today for a free trial.
CHEMSTATIONS.COM/MIX13
ChemE's
HAVE A NEW
SUPER
HERO.
Discover CHEMCAD NXT -
The evolution of simulation software
For details visit adlinks.chemengonline.com/84654-03
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
DECEMBER 2023
7
http://www.fivesgroup.com http://www.ornl.gov http://www.rtx.com http://www.rtx.com http://www.lanxess.com http://adlinks.chemengonline.com/84654-03 http://WWW.CHEMENGONLINE.COM
che_december-2023

Table of Contents for the Digital Edition of che_december-2023
