WSU AI Finds Settings That Let Common 3D Printers Print Rocket-Grade Metal

From 100 million options to six working recipes

Researchers at Washington State University have used an AI-guided experimental design method to print a high-performance NASA alloy on lower-power commercial equipment. The alloy, GRCop-42, is a copper-chromium-niobium mix originally developed for rocket engine combustion chambers. It handles extreme heat, but it is famously picky to print. The team now says it has identified six workable parameter sets, including one at just 500 watts of laser power.

Why GRCop-42 matters

GRCop-42 stays strong where most metals soften. Rocket builders want it for regeneratively cooled combustion chambers because it conducts heat well and resists thermal fatigue. The catch is that printing it reliably usually demands high-power, specialized equipment that few universities or small shops can afford. Most common commercial metal 3D printers simply lack the muscle.

Teaching AI to learn from failure

The WSU group, led by computer scientist Jana Doppa and first author Azza Fadhel, started with 37 failed print configurations that mechanical engineering collaborators had already collected. Their model, called BEAM, used those failures to estimate which untested settings were most likely to work. It then ran small batches of experiments that balanced exploiting promising settings with exploring uncertain ones.

"They would give me back the results, and I liked all of them, even if they failed, because every result improved our AI model," Fadhel said in the university release.

Over three months and 40 experiments, the team landed on six successful configurations across 950W, 700W, 600W, and 500W power levels. The 500W result is the headline: it suggests GRCop-42 can be printed on far more accessible machines.

What changes if this scales

If the findings hold beyond the lab, GRCop-42 could move from defense primes and well-funded aerospace firms to university labs, startups, and smaller contractors. Lower laser power also means less energy use, less wear on optics, and simpler post-processing. The same approach could in principle tune other tricky alloys, not just this one.

The bigger picture

The paper, published in the Proceedings of the AAAI Conference on Artificial Intelligence, picked up an Innovative Deployed Application Award. The work was done in WSU's School of Electrical Engineering and Computer Science and School of Mechanical and Materials Engineering, with collaborators at the University of Minnesota.

The practical takeaway is clear: AI here is not generating hype. It is cutting the number of expensive test prints needed to tame a difficult material. For an industry that still loses too much time and money on trial-and-error parameter hunts, that is a genuinely useful step.