A copper alloy engineered by NASA to survive the searing heat inside rocket engines has long been notoriously difficult to 3D print, requiring specialized equipment that puts it out of reach for most manufacturers. Researchers at Washington State University say they have found a workaround: letting artificial intelligence do the searching.
The alloy, GRCop-42, is a blend of copper, chromium and niobium developed at NASA's Glenn Research Center for use in the combustion chambers of liquid-fuel rocket engines, where it must carry heat away from surfaces reaching thousands of degrees while holding its structural strength. Those same properties make it hard to melt and fuse reliably with a laser, and according to Washington State University, roughly 90% of commercial metal 3D printers cannot process it at all.
Forty Experiments Instead of Millions
Rather than testing laser power, scan speed and other print parameters by trial and error, a team led by computer scientist Jana Doppa and graduate student Azza Fadhel built an AI model that used a set of 37 previously failed print attempts to estimate which untested combinations were most likely to succeed. That let the researchers shrink a search space of more than 100 million possible parameter combinations down to a small number of settings worth physically testing on real hardware.
Over about three months, the team ran 40 print experiments and identified six configurations that reliably produced usable GRCop-42 parts, including one that worked at just 500 watts of laser power, a record low for the alloy and within range of far more common commercial printers. The work, described in the Proceedings of the AAAI Conference on Artificial Intelligence, received the conference's Innovative Deployed Application award.
"Ninety percent of commercial printers cannot print this metal alloy, so given that we were able to find these feasible process parameters, it allows us to use those commercial printers," said Doppa, the Huie-Rogers Endowed Chair Professor of computer science at WSU. "We are essentially democratizing the printing of this alloy."
The project drew on WSU's mechanical and materials engineering labs, including researchers Susmita Bose and Amit Bandyopadhyay, alongside collaborator Aryan Deshwal of the University of Minnesota. The team says the same AI-driven search approach, rather than the alloy-specific settings themselves, is the more transferable result: it could in principle be pointed at other hard-to-print aerospace and defense alloys that face similar bottlenecks.
The findings so far come from a peer-reviewed conference paper rather than a mechanical-testing study, and the researchers note that further work is needed to characterize the strength and durability of parts printed under the new settings before they could be used in flight hardware. Still, cutting the number of required physical experiments from a search space in the hundreds of millions to just 40 marks a sharp change in how such alloys are qualified for cheaper machines, and the team says it is now applying the same method to other difficult metals.
