A Novel Design for Refractory Complex Concentrated Alloys Based on Multi-Objective Bi-Level Optimization

Refractory complex concentrated alloys (RCCAs) have great potential for applications in the nuclear industry due to their exceptional strength at high temperatures, as well as their fair corrosion resistance and abrasion resistance. However, the limited ductility of RCCAs at room temperature hinders their processing and engineering applications. To cope with this challenge, a novel design framework was utilized, which integrated machine learning and bi-level optimization techniques, to design novel RCCAs for cladding materials. In order to address the challenge of a large composition space of RCCAs and the difficulty of establishing a direct model between composition and properties, bi-level optimization was employed. This was accomplished by establishing mapping relationships between composition-microparameters and microparameters-properties, respectively. Furthermore, Pareto optimization was implemented to achieve a compromise between high-temperature ultimate strength and ambient temperature ductility. Ultimately, according to experimental validation, A1 alloy exhibited a superior compression strain limit (49.8 %) and ultimate strength (1994.6 MPa) at ambient temperature. The desired RCCA was found to outperform existing materials in terms of target mechanical properties. This approach can not only uncover the indirect relationship between composition and properties but also achieve a trade-off between strength and ductility.