Researchers in the USA have used artificial intelligence to find laser settings that make a copper-chromium-niobium alloy used in rocket engine combustion chambers printable at power levels available on standard commercial machines.
The work, carried out at Washington State University (WSU) in the USA, targeted GRCop-42, an alloy developed by NASA for environments that demand both high-temperature strength and efficient heat transfer.
GRCop-42 combines high thermal conductivity with retained strength at extreme temperatures, which suits it to liquid rocket engine combustion chambers. It is difficult and costly to 3D print, however, because the process typically requires substantial laser power and energy.
Earlier attempts to print the alloy at the lower wattages available on more common machines had not succeeded. Exhaustive testing is impractical: a single print can cost hundreds of dollars, and analyzing a finished sample can take several days.
Searching for settings
The team faced a search space of more than 100 million possible printing configurations, within which workable settings were expected to be rare.
“It’s a very challenging case for AI,” said Jana Doppa, professor of computer science, who led the research. “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.”
The researchers began with data from 37 configurations that had already failed in earlier experiments at WSU’s School of Mechanical and Materials Engineering.
From those results they built a model that estimates how likely an untested combination of settings is to produce a successful print, then used it to recommend small batches of new configurations for testing.
The selection balanced two objectives, targeting configurations that appeared promising while also sampling less certain regions of the search space to improve the model.
Six workable configurations were identified at different laser power levels over three months, within a total of 40 experiments. One produced the first successful print of GRCop-42 at 500W of laser power.
Lower power advantages
Lower laser power reduces energy consumption and wear on printing equipment, and cuts the cost of processing samples after printing, said the researchers. It would also put the alloy within reach of universities, smaller laboratories and companies without access to high-power printing systems.
“90% 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 Doppa.
The team says the same approach could be applied to processing conditions for other alloys and additive manufacturing systems, and to scientific problems where successful outcomes are rare and each experiment carries high material, financial or time costs.
The work was published in the Proceedings of the AAAI Conference on Artificial Intelligence Conference on AI and received the Innovative Deployed Application Award. Aryan Deshwal of the University of Minnesota also worked on the project, alongside Azza Fadhel, first author, and Nathaniel Zuckschwerdt, Susmita Bose and Amit Bandyopadhyay of the School of Mechanical and Materials Engineering.
