数据驱动的镍基高温合金多目标优化设计与开发

Data-driven Multi-objective Optimization Design and Development of Ni-base Superalloy

为了研发综合性能优异的新型镍基铸造高温合金,利用机器学习和多目标优化综合策略预测优化合金成分,设计了兼具高γ′相体积分数、高γ′相固溶温度、低TCP相含量、高液相线温度和宽热处理窗口等多目标特征的候选合金。结果表明,通过在固溶和时效热处理中调控固溶温度和时间,获得了γ′相体积分数趋近65%、γ′相固溶温度趋近1 210℃、TCP相含量趋近0.01%,以及液相线温度高于1 300℃和热处理区间大于40℃的高温合金,同时满足预期设计的多个目标和约束,模型预测精度高。与具有优异高温力学性能的典型牌号K438相比,具有更高的γ′相体积分数、γ′相固溶温度,和良好的高温应用潜力。

In order to develop a new nickel-based cast superalloy with excellent comprehensive performance, the candidate alloys with multi-objective characteristics such as high γ′ phase volume fraction, high γ′ phase solid solution temperature,low TCP phase content, high liquidus temperature and wide heat treatment window were designed by using machine learning and multi-objective optimization strategy. The results show that the volume fraction of γ′ phase tends to 65%, the solution temperature of γ′ phase tends to 1 210 ℃, the content of TCP phase tends to 0.01%, and the liquidus temperature is higher than 1 300 ℃ and the heat treatment interval is greater than 40 ℃ by regulating the solution temperature and time in the solution and aging heat treatment, and the multiple objectives and constraints of the expected design can be met at the same time. The prediction accuracy of the model is high. Compared with the typical brand K438 with excellent mechanical properties at high temperature, it has higher volume fraction of γ′ phase, solution temperature of γ′ phase and good application potential at high temperature.