Researchers from the Skolkovo Institute of Science and Technology (Skoltech) and the Moscow Institute of Physics and Technology (MIPT) have presented an innovative method for accelerated search of high-performance metal alloys using machine learning. This approach significantly reduces the time required to select promising compositions for further laboratory testing.
Traditionally, modeling metal alloys is a lengthy and complex process involving many variables: element composition, proportions, crystal structure. However, the new method helps to quickly iterate through all possible combinations and identify potential candidates for further research.
Unlike traditional approaches based on complex quantum mechanical calculations, the new method uses potentials obtained through machine learning. This allows for faster calculations and efficient scanning of many options. The algorithm has already demonstrated its effectiveness on two sets of metals: refractory (vanadium, molybdenum, niobium, tantalum, and tungsten) and noble (gold, platinum, palladium, copper, and silver).
The test results were surprising: 268 new stable alloys were obtained that were not listed in popular industry databases. For example, 12 previously unknown alloys were discovered in the niobium-molybdenum-tungsten system.
These new alloys need to be thoroughly tested to understand their potential applications. It is expected that their use may lead to the creation of new materials for the aerospace, mechanical engineering, construction, electronics, and medical technology industries.
In the future, scientists plan to expand the algorithm to explore other compositions and crystal structures, which will open up new opportunities for various industries.
Read related materials:
Систему машинного обучения применили для изучения свойства металлического водорода в МФТИ
В Казани научились плавить метал при помощи звуковых волн
В Москве адаптировали нейросети для прогнозирования дефектов в стали для АЭС