Researchers at the University of Toronto Engineering have used artificial intelligence (AI) to discover six new metal alloys that remain strong under extremely high temperatures.
The new materials were designed for metal 3D printing and could eventually be used to make stronger, lighter and more durable parts for jet engines, power plants and other equipment that must operate in harsh conditions.
The research was published in npj Advanced Manufacturing.
Many modern industries need materials that can survive intense heat and pressure without losing their strength. For example, jet engines, gas turbines and nuclear power plants all contain parts that are exposed to temperatures that ordinary steel simply cannot handle.
At the same time, manufacturers increasingly want materials that can be produced using metal 3D printing, which builds objects one thin layer at a time.
This method allows engineers to create complex shapes that are difficult or impossible to make using traditional manufacturing.
Professor Yu Zou, who led the research, says metal 3D printing also makes it possible to create parts with different properties in different areas. For example, the outside of a component could be made from a very hard, durable material, while the inside could be lighter to reduce overall weight.
Finding the right metal mixture has always been a major challenge. High-performance alloys often contain nickel or cobalt along with many other elements. With thousands of possible combinations available, testing each one by hand would take many years.
To speed up the search, the research team developed what they call a self-driving laboratory. The system combines computer simulations, machine learning and robotic manufacturing into one automated process. Instead of scientists testing every possible alloy themselves, the AI predicts which combinations are most promising.
Robots then produce small samples and test their performance. The results are immediately fed back into the AI, helping it decide which materials to investigate next. This continuous learning process is known as active learning.
One challenge in using AI for materials discovery is that machine learning usually needs large amounts of existing data. When exploring completely new materials, very little information is available.
The team’s active learning approach solves this problem by carefully choosing only a small number of experiments that provide the most useful new information, allowing the system to improve rapidly without requiring huge databases.
To demonstrate the approach, the researchers focused on alloys made from three main elements: nickel, cobalt and chromium. Although this is a relatively simple combination compared with many commercial alloys, the AI-guided system identified six promising new materials in only a few weeks.
One of the new alloys, made from 12% nickel, 62% cobalt and 26% chromium, performed exceptionally well at temperatures of about 600°C. During laboratory tests, it stayed harder than Inconel 625, one of the industry’s most widely used high-temperature alloys, outperforming it by 4.5%.
Another alloy, containing 36% nickel, 14% cobalt and 50% chromium, was designed for even hotter environments, reaching around 1,000°C. At these temperatures, metals often react with oxygen and slowly burn away through a process called oxidation. The new alloy showed outstanding resistance to this damage, outperforming Inconel 625 by an impressive 85%.
The researchers say these six alloys are only the beginning.
Their current work focused on just three elements, but future studies will explore much more complex mixtures containing 10 or even 12 different elements. As more ingredients are added, scientists may discover entirely new combinations with even greater strength, durability and heat resistance.
The success of this AI-powered discovery platform shows that intelligent automation could dramatically speed up the search for the next generation of advanced materials, helping industries develop safer, longer-lasting and more efficient products much faster than ever before.
Source: KSR.
