Accelerated design of L12-strengthened single crystal high entropy alloys based on machine learning and multi-objective optimization

Abstract

A comprehensive strategy of machine learning and multi-objective optimization based on thermodynamic simulation data was proposed to accelerate the composition design of L1₂-strengthened single crystal high entropy alloys (SX-HEAs). This approach simultaneously optimized various alloy microscopic parameters, such as volume fraction and solvus temperature of L1₂ phase, TCP phase content, liquidus–solidus temperatures, and density, using a machine learning model. From a pool of 2 515 661 candidates, a composition of Ni_41.5Co₃₁Cr₈Ti₄Al₁₀W_0.5Mo₂Ta₃ (at%) was selected and then the related microscopic parameters were verified experimentally with a high accuracy. Furthermore, the designed alloy reached a yield strength of 873 MPa at 800 °C and 503 MPa at 1000 °C, as well as a creep life of 111.65 h at 1038 °C/158.6 MPa and 62.67 h at 1038 °C/172 MPa. This material design strategy based on machine learning and multi-objective optimization extends the L1₂-strengthened SX-HEAs design method to optimize simultaneously multiple objectives, rather than one by one.