Machine Learning Drives AM Research and Enables Resilience
ADAPT’s work on behalf of the US Department of Defense Office of Economic Adjustment (DoD OEA) began in 2017 and has been extended into 2020. This collaborative effort with the University of Utah, Colorado School of Mines, and Carnegie Mellon University is focused on diversifying the manufacturing supply chain of the future.
A key part of the project, both in Colorado and Utah, is educating manufacturers – especially defense contractors – about metal AM. A strong foundation of metal AM knowledge and capability offers a path to economic resilience for manufacturers. AM provides flexibility to enable manufacturers to diversify and gives the DoD a way to build a more resilient defense supply chain based on AM.
“This collaboration sets the foundation of how we diversify the manufacturing supply chain of the future,” emphasized ADAPT Executive Director Aaron Stebner. “We’re moving away from the assembly line to a distributed network of AM machines. That promotes a more even keel for the defense manufacturing supply chain and provides opportunities for others to get involved.”
Through workshops and training sessions, ADAPT and the collaborating institutions on the DoD OEA program have been able to help local manufacturers. “We invite them to learn about the technical and business case for using metal AM to enhance and diversify their businesses,” said Bart Raeymaekers, project lead with University of Utah. “That helps make their companies and the workforce as a whole more resilient and able to deal with the ups and downs in defense spending.”
In Utah, work with Hill Air Force Base has focused on new approaches to keep the aging A-10 Warthog and the brand-new F-35 Lightning II fleets in the air. “The supplier that made a part may no longer exist, or we may have better ways to build a part today than when the plane was originally manufactured. Using AM is proving effective in sustainment for old aircraft and in field repairs for very new aircraft,” said Raeymaekers.
Together, the team is printing parts with titanium and Inconel using different process parameters to characterize material and mechanical properties, which are stored in a cloud-based database. Machine learning algorithms from Citrine Informatics then mine the database produced.
The result is data-driven models that can direct future research. Continue reading for more program details »
Researcher of the Month
Sen Liu is an ADAPT PhD student focused on finding ways to accelerate the processes that enable additive manufacturing using machine learning. He worked to collect and integrate information from literature that was used to develop Ti-6Al-4V build parameters necessary to achieve both a target porosity and a target hardness. Sen showed that his data-driven model, trained only on data available from literature, could be used to predict process parameters that were able to hit specific target properties with excellent accuracy the first time, even on machine types that were not included in the original data. Extracting information from literature is only the first step. Sen is also working to develop computer vision algorithms for extracting features from microscopy, thermal analysis, and mechanical test data that will enable the high-throughput materials characterization necessary for verification and validation of AM. Sen anticipates completing his PhD in Fall 2021.
Christopher Lefky, Thomas Gallmeyer, Senthamilaruvi Moorthy, Aaron Stebner, and Owen Hildreth: "Microstructure and Corrosion Properties of Sensitized Laser Powder Bed Fusion Printed Inconel 718 to Dissolve Support Structures in a Self-Terminating Manner," Additive Manufacturing (in press).