Ni superalloys are an extremely technologically important class of material. However, the development of new superalloys by tradtional methods is a lengthy and experimentally intensive process. I investigate the use of machine learning and atomistic tools to be used in tandem to design new superalloys. My current focus is on the optimisation of thermomechanical properties, and most importantly the creep strength, of single crystal superalloys. As precipitation strengthened alloys, a crucial first step to modelling their properties is predicting their phase composition. To this end, I have developed a new Gaussian Process Regression model of phase microchemistry. My current work is on extending this machine learning approach to make use of physical representations of the alloys' input features.
In Plain English
As a jet plane takes off, the hottest parts of the jet engine can reach temperatures of 1500°C (2730°F). A special type of metal, called a superalloy, can withstand these very extreme temperature conditions. Superalloys are used in some of the most important parts of a jet engine. Inventing new types of superalloy that are stronger and lighter than those currently in use can help make jet planes more fuel-efficient. But it can take a long time to invent a superalloy, often several years. The goal of my research is to use computers to speed up this process.
- Machine learning predictions of superalloy microstructure Comp. Mater. Sci. 201 110916 (2022)