Plate Nº 54 · recorded October 13, 2026
Chemistry & MaterialsReported finding
AI Finds Six Ways to 3D-Print a NASA Rocket Alloy in 40 Tries
Across 40 experiments, an AI system at Washington State University identified six configurations for 3D printing GRCop-42, the NASA alloy used in rocket engines. One succeeded at 500 watts of laser power.
By Priya Raman3 min read617 words
In brief
- At Washington State University, an AI system identified six working 3D-printing configurations for the NASA alloy GRCop-42 across 40 physical experiments.
- For the first time, one successful print of GRCop-42 ran at 500 watts of laser power, a record low for the alloy.
- The AI narrowed a configuration space of more than 100 million possibilities down to a handful of successful recipes.
- The study appeared in the Proceedings of the AAAI Conference on Artificial Intelligence and received the Innovative Deployed Application Award.
- Roughly 90 percent of commercial 3D printers cannot print GRCop-42 because the alloy demands substantial laser power.
Forty physical trials produced six successful prints, including one at a record-low 500 watts of laser power. That ledger comes from a Washington State University project in Pullman, where researchers in a materials engineering laboratory waited for each new AI-predicted configuration to emerge.
The work targeted a notoriously difficult manufacturing problem surrounding GRCop-42, a copper-chromium-niobium alloy that NASA developed for rocket engine combustion chambers. Earlier attempts had tried 37 configurations for printing the alloy, a process known as additive manufacturing (building objects by depositing material layer by layer), and none succeeded.
The number of untested configurations exceeded 100 million. A single print cost hundreds of dollars, and analyzing each finished sample took days.
What does GRCop-42 do, and why is it hard to print?
GRCop-42 pairs high thermal conductivity (the ability to move heat efficiently) with strength at extreme temperatures. That combination makes it valuable for aerospace parts exposed to severe thermal stress. Printing the alloy, however, demands substantial laser power, and roughly 90 percent of commercial 3D printers cannot produce it.
"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, and we are essentially democratizing the printing of this alloy," said Jana Doppa, who led the work as Huie-Rogers Endowed Chair Professor of Computer Science and Berry Distinguished Professor in Engineering.
How did the AI choose what to print next?
Azza Fadhel, first author of the paper and a PhD student in computer science, built the AI system with Doppa and collaborators. Starting from the 37 prior failures, the model learned to estimate the probability that any new combination of parameters would succeed. It then proposed small batches of experiments, balancing high-confidence candidates against less certain settings that might teach the model something new.
"Sometimes they printed a certain configuration, and the product just melted," Fadhel said. "It wasn't really printable, and even with time and money, they wouldn't be able to try all 100 million options. What we were doing in our collaboration is to apply the AI so that we efficiently choose candidates from this very large search space."
What did the experiments find?
Over three months, the team attempted 40 prints. Six succeeded. For the first time, one of them ran at 500 watts of laser power, low enough to fall within the range of many commercial machines. Lower power could trim energy use, reduce wear on printing equipment, and lower post-processing costs. It could also make the alloy more accessible to universities and smaller laboratories that lack specialized high-power systems.
"It's a very challenging case for AI," Doppa said. "Every time you try, you basically get a binary success or failure signal, and you are trying to minimize the number of tries that you have so that you get to those successful needles very quickly."
Could the method reach beyond metal printing?
The researchers see broader applications. Any problem in which successful outcomes are rare and individual experiments carry steep material, financial, or time costs could benefit, including drug discovery. Aryan Deshwal of the University of Minnesota contributed to the project.
"There's always uncertainty when you are deploying something where real people, materials, and physical costs are involved," Doppa said. "We didn't know whether we would succeed or not, and there is always that risk. There are real stakes. I was very surprised that we were able to do this so well."
The study appeared in the Proceedings of the AAAI Conference on Artificial Intelligence. The project received the organization's Innovative Deployed Application Award.
via news.wsu.edu (Original)
More from Priya Raman
Show full bio
Senior reporter covering people, launches and day-to-day practice in general science news at SciBeat.
222 articles
Nearby plates
- A New Recipe Makes 'Impossible' Metal Nitride Nanocrystals
- MIT Makes Flexible, Transparent Silicon Photonics Chips at Wafer Scale
- Molten Salt Process Turns Plastic Bags Into Gasoline at Oven Temperatures
- X-rays and AI Reveal Spacecraft Heat Shields Burning in Real Time
- Sulfur-Powered Material Splits Water Into Hydrogen Using Sunlight