IIT Guwahati team uses AI to design high-performance alloys without critical raw materials

A team of researchers has developed a machine learning–based method to design advanced metal alloys without relying on critical raw materials, offering a potential breakthrough for sustainable manufacturing and industrial security.

The work was led by researchers at the Indian Institute of Technology Guwahati, in collaboration with teams from London South Bank University, University of Manchester and University of Leeds. The researchers used artificial intelligence to identify high-performance alloys that avoid elements such as tantalum, niobium, tungsten and hafnium, which are costly, scarce and vulnerable to supply disruptions.

Alloying has been used since the Bronze Age to improve metals by mixing a base metal with small quantities of other elements. In recent decades, attention has shifted to high-entropy alloys, a subset of multi-principal element alloys, which contain several metals in near-equal proportions. These materials often show high strength and thermal stability, making them attractive for aerospace engines, gas turbines and nuclear systems. However, many existing high-performance alloys depend heavily on critical raw materials, increasing import dependence and environmental pressure from mining.

To reduce this reliance, the researchers built a machine learning–assisted framework focused on designing CRM-free alloys. Critical raw materials were first grouped into three categories based on supply risk, economic importance and global availability. A database of 3,608 alloy compositions was then created, concentrating on simpler systems made from non-critical elements.

Among several models tested, the Extra Trees Regressor produced the most accurate predictions for Vickers hardness. This model was combined with optimisation techniques inspired by natural processes to search for compositions that could deliver high hardness without critical elements.

Using this approach, the team identified a CRM-free alloy, Ti₀.₀₁₁₁NiFe₀.₄Cu₀.₄, predicted to achieve hardness exceeding that of a widely used alloy containing critical materials, which typically measures around 480 HV. The alloy was later produced at laboratory scale at IIT Kanpur, where experimental tests showed hardness values close to the predictions, confirming the reliability of the AI-based method.

“The developed CRM-free alloy is particularly suited for applications where high hardness is a primary requirement, offering the added benefit of avoiding the use of Critical Raw Materials (CRMs),” said Shrikrishna N Joshi, Professor in the Department of Mechanical Engineering at IIT Guwahati. According to the team, potential applications include wear-resistant mechanical components, tooling and surface-contact parts, and automotive and industrial machinery components.

Highlighting what sets the framework apart, Prof Joshi said, “This is the first validated computational framework for designing critical raw material-free (CRM-free) multi-principal element alloys (MPEAs) using a unary- and binary-based compositional database, without relying on microstructural or processing parameters.” He added that the model is built entirely on compositional data and machine learning, making it transferable to other material systems with limited experimental data. The framework can also be extended to predict properties such as strength, ductility, fracture toughness, corrosion resistance, thermal conductivity and wear resistance.

The findings have been published in Scientific Reports, a journal of the Nature Publishing Group. The paper was co-authored by Prof Joshi, Dr Swati Singh of IIT Guwahati, Prof Saurav Goel of London South Bank University, Mingwen Bai of the University of Leeds, and Prof Allan Matthews of the University of Manchester.

The research team plans to work with industry partners and research laboratories next, testing the alloys under real operating conditions to move closer to commercial deployment.