MIT Researchers Sharpen Alloy Predictions With ML, Beating 100,000-Hour Brute-Force Training
Updated
Updated · MIT News · Jun 19
MIT Researchers Sharpen Alloy Predictions With ML, Beating 100,000-Hour Brute-Force Training
3 articles · Updated · MIT News · Jun 19
Summary
MIT researchers built machine-learning models that more accurately predict how chemically disordered metal alloys behave, aiming to cut the time and cost of developing materials for aerospace, energy and computing.
The advance comes from new training datasets designed to capture a wider range of local atomic environments; current brute-force methods can take more than 100,000 computing hours for a single material and still transfer poorly across compositions.
Science Advances results showed the models outperformed versions trained with random or common sampling methods, and MIT said they were more accurate than much larger models developed by companies including Google and Microsoft.
Tests across multiple alloys linked the simulations to real-world use: the models reproduced phase diagrams close to experimental data, helping predict which phases form during welding, casting and heat treatment.
The team is now applying the method to mechanical strength and radiation tolerance and says it could extend beyond alloys to materials such as semiconductors.
If big tech's AI struggles with materials discovery, what is MIT's secret to modeling chemically chaotic alloys?
How does an abstract math theory enable AI to predict the real-world properties of next-generation sustainable materials?
MIT’s AI-Driven Alloy Design Achieves Record-Strength, Heat-Resistant 3D Printable Aluminum: A New Era for Materials Science
Overview
MIT researchers have introduced a groundbreaking AI-powered method for alloy design, leading to the creation of a super-strong, heat-resistant, and 3D-printable aluminum alloy. This innovation marks a major leap in how new materials are discovered and developed, setting new standards for efficiency in manufacturing and research. The team, led by Mohadeseh Taheri-Mousavi, used artificial intelligence to dramatically speed up the discovery process, resulting in an alloy that can withstand extreme conditions and is ideal for demanding applications like aerospace. This AI-driven approach opens up new possibilities for material engineering across many industries.