Machine learning accelerates discovery of high-temperature alloys
In a study recently published in Engineering, scientists from the University of Science and Technology Beijing, Guangdong Ocean University, and AiMaterials Research LLC have demonstrated a novel method to accelerate the discovery of refractory high-entropy alloys (RHEAs) compositions optimized for extreme conditions.
The research, titled “Machine-Learning-Assisted Compositional Design of Refractory High-Entropy Alloys with Optimal Strength and Ductility,” outlines how machine-learning (ML), genetic search, cluster analysis, and experimental design were employed to sift through billions of possible compositions and identify those with superior mechanical properties.
The research team, led by Turab Lookman and Yanjing Su, synthesized and tested 24 different alloy compositions through a rigorous iterative process involving six feedback loops. Their efforts resulted in four compositions demonstrating remarkable high-temperature yield strength and room-temperature ductility. Among these, the ZrNbMoHfTa alloy system, specifically the composition Zr0.13Nb0.27Mo0.26Hf0.13Ta0.21, stood out with a yield strength approaching 940 MPa at 1200 °C and a room-temperature fracture strain of 17.2%.
A leap forward in high-temperature materials
The exceptional performance of the ZrNbMoHfTa alloy marks a significant advancement in materials science. Its yield strength at 1200 °C exceeds that of previous RHEAs and traditional nickel-based superalloys, which are typically limited to lower temperatures. This enhancement opens up new possibilities for high-temperature structural applications, including in gas turbines, aerospace propulsion systems, and nuclear reactors.
The integration of machine learning with traditional alloy design methods has allowed researchers to rapidly identify and optimize compositions that were previously unimaginable. This breakthrough not only addresses the limitations of existing materials but also sets a new standard for high-temperature alloys.
A new paradigm for material design
The researchers’ approach represents a paradigm shift in material design by effectively managing the vast compositional space of RHEAs and addressing multiple performance objectives simultaneously. By leveraging ML algorithms, the team was able to predict alloy properties with unprecedented accuracy and efficiency, overcoming common challenges such as limited data and complex optimization tasks.
The study also highlights the importance of incorporating multi-objective optimization (MOO) techniques to balance various material properties, including strength, ductility, and oxidation resistance. The proposed framework’s adaptability to other alloy systems demonstrates its potential to revolutionize the design of materials across different applications and industries.
While the current study has achieved remarkable results, the researchers emphasize that there is still room for improvement and further exploration. Future work will focus on integrating additional elements to enhance properties like oxidation resistance and refining ML models to manage uncertainties and improve prediction accuracy. The study also underscores the need for efficient selection strategies, such as cluster analysis, to optimize experimental and computational costs.
“The success of this research opens new avenues for material innovation,” noted Nan Zhang, editor of Engineering. “As research scientists continue to refine their approach and explore new compositions, we anticipate even greater advancements in high-temperature alloys that could transform a wide range of engineering applications.”
The paper was authored by Cheng Wen, Yan Zhang, Changxin Wang, Haiyou Huang, Yuan Wu, Turab Lookman, Yanjing Su.
More information:
Cheng Wen et al, Machine-Learning-Assisted Compositional Design of Refractory High-Entropy Alloys with Optimal Strength and Ductility, Engineering (2024). DOI: 10.1016/j.eng.2023.11.026
Citation:
Machine learning accelerates discovery of high-temperature alloys (2024, September 27)
retrieved 30 September 2024
from
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.
link