Washington State University researchers trained an AI to cut through 100 million possible laser settings and make a NASA rocket alloy printable on commercial 3D printers.
The Bottleneck Was the Search Space
NASA's GRCop-42 alloy is built for extreme heat. The copper-chromium-niobium material lines rocket engine combustion chambers and withstands temperatures near 6,000 degrees Fahrenheit. It performs well. The problem is printing it.
Traditional manufacturing for GRCop-42 depends on high-powered lasers that exist on roughly 10 percent of commercial metal 3D printers. Finding the right combination of laser power, speed, and material placement requires testing more than 100 million parameter combinations. Individual attempts cost hundreds of dollars each. Most end as scrap metal.
AI Cuts the Search to 40 Experiments
A team led by Washington State University computer scientists and materials engineers trained an algorithm on data from 37 previously failed print attempts. The AI evaluated the parameter space and recommended small batches of settings to test physically, balancing high-probability options with uncertain ones that would improve the model. Every result, good or bad, fed back into the system.
After three months and only 40 physical experiments, the team identified six successful printing configurations. The most significant finding: GRCop-42 printed successfully at just 500 watts of laser power, a record low for the alloy.
What Lower Wattage Means
500 watts is a practical threshold. Most commercial metal 3D printers can operate at that level, which means universities, smaller research labs, and emerging aerospace companies can now print GRCop-42 without specialized industrial equipment. Lower wattage also extends printer lifespan, reduces energy use, and cuts post-processing costs.
The team published its findings in the Proceedings of the AAAI Conference on Artificial Intelligence and received the conference's Innovative Deployed Application Award.
A Path Beyond Rocket Parts
The researchers note that the AI-guided testing framework applies to any field where physical experiments are expensive or time-consuming. Pharmaceutical drug discovery and novel industrial materials development are two obvious candidates.
NASA already proved GRCop-42 works in flight. In March 2023, Relativity Space launched Terran 1, a 100-foot rocket built almost entirely from 3D-printed parts. Its nine engines used GRCop-42 combustion chambers produced with laser powder bed fusion. That flight showed government materials research can transition to commercial use. The WSU work suggests that transition is about to get cheaper and faster.
Comments (0)
No comments yet. Be the first!
Leave a Comment