Researchers at Washington State University used artificial intelligence to find a more efficient and less expensive way to 3D-print a high-performance metal alloy, freeing researchers from the need to painstakingly test more than 100 million possible combinations.
An Innovative Method
The website phys.org reports on a study whose results could one day make it possible to print an alloy widely used in the aerospace industry, but with many other potential applications, using commercially available equipment. The artificial intelligence methods used could also be applied in various fields, including the discovery of new drugs.
The research team from WSU’s School of Electrical Engineering and Computer Science and School of Mechanical and Materials Engineering published their work in the proceedings of the AAAI Conference on Artificial Intelligence and received an award for an innovative deployed application at the organization’s annual conference.
The scientists note that 90% of commercial printers cannot print this metal alloy, so the fact that suitable process parameters were found makes it possible to use these commercial printers and, in effect, democratize the printing of this alloy.
An Expensive Material for Testing
The alloy, called GRCop-42, consists of three metals — copper, chromium, and niobium. Developed by NASA, it has high thermal conductivity and remains strong at extreme temperatures, which is why it is used in the aerospace industry, for example, in the combustion chambers of liquid-fueled rocket engines. Although it is a highly desirable material with many potential applications, it is expensive and energy-intensive to print and requires significant laser power.
Researchers had unsuccessfully attempted to print the alloy at the lower power levels used by most commercial printers, but testing different configurations requires expensive materials, specialized equipment, and substantial labor. A single printing cycle can cost hundreds of dollars, while detailed post-print quality analysis can take several days.
Searching for Successful Configurations with AI
For the study, the team began with the results of 37 unsuccessful configurations that had previously been tested at the School of Mechanical and Materials Engineering. They then developed an approach that used these results to estimate the likelihood that an untested configuration would produce a successful print. Their model then selected small batches of new configurations, balancing two goals: testing promising options and exploring uncertain areas that could improve the AI model.
Working with Nathaniel Zuchscherwert, Susmita Bose, and Amit Bandyopadhyay from the School of Mechanical and Materials Engineering, the team used the AI-selected process configurations to print GRCop-42 and evaluated the resulting samples. Aaryan Deshwal of the University of Minnesota also participated in the project.
“They would send the results back to me, and I liked all of them — even if they failed — because every result improved our artificial intelligence model,” said Azza Fadhel, one of the study’s authors.
Six Successful Prints out of 40 Attempts
Printing at lower laser power can reduce energy consumption, equipment wear, and post-processing costs, while also making GRCop-42 accessible to universities, small laboratories, and companies that do not have specialized high-power equipment.
The researchers knew that only a very small fraction of more than 100 million configurations would be successful. Over three months and within a total budget of 40 experiments, the team identified six successful configurations for different laser power levels. For the first time, the team successfully printed the alloy at a power of 500 watts.
The researchers believe that the same AI-guided approach could be adapted to determine processing conditions for other metal alloys and additive-manufacturing systems. More broadly, it could support scientific research in which successful outcomes are rare and experiments are too expensive to test every possibility.
