AI Finds a Cheaper Way to 3D Print NASA's GRCop-42 Rocket Alloy

Why GRCop-42 Has Been a Printer's Nightmare

GRCop-42 is the kind of material aerospace engineers love and print operators dread. NASA developed the copper-chromium-niobium alloy to survive the brutal heat of rocket combustion chambers while still pulling heat away efficiently. That combination makes it ideal for spacecraft. It also makes it difficult and expensive to 3D print, because the process normally demands so much laser power that most commercial metal printers simply cannot handle it.

Researchers at Washington State University have now found a workaround, and the key was not a bigger laser. It was AI. The team, led by computer scientist Jana Doppa and mechanical and materials engineers including Susmita Bose and Amit Bandyopadhyay, developed a method to search more than 100 million possible printing configurations without running every one of them. They published the work in the Proceedings of the AAAI Conference on Artificial Intelligence and picked up the conference's Innovative Deployed Application Award.

From 37 Failures to Six Working Recipes

The project started with 37 configurations that had already failed. Using those results, the AI estimated how likely any untested combination was to succeed. It then recommended small batches of new settings to try, balancing high-confidence candidates against riskier ones that would teach it more. The loop ran for three months and only 40 actual experiments. At the end, the team had six working configurations, including one that printed GRCop-42 at just 500 watts of laser power for the first time.

That matters commercially. Lower laser power means less energy, less wear on the machine, and lower post-processing costs. It also opens the alloy to universities, smaller labs, and companies that do not own specialized high-power systems. As Doppa put it, the finding helps democratize the printing of this alloy.

A Template for Other Hard Problems

The WSU group thinks the same approach could work for other alloys and additive manufacturing systems. More broadly, it is a template for scientific discovery when successful outcomes are rare and experiments cost real money. The team is essentially treating the 3D printer as a testbed for active learning: ask the right questions, run only the most informative experiments, and stop guessing.