|
|
快捷服务 |
|
|
|
|
|
|
相关链接 |
|
|
|
|
|
|
会议信息 |
|
|
|
|
|
|
友情链接 |
|
|
|
|
|
|
近期文章 |
|
|
The γ/γ′ microstructure in CoNiAlCr-based superalloys using triple-objective optimization |
发布时间:2023-09-19 |
The y/ y′ microstructure in CoNiAlCr-based superalloys using triple-objective optimization Pei Liu, Haiyou Huang, Cheng Wen, Turab Lookman & Yanjing Su npj Computational Materials 9: 140 (2023)
doi.org/10.1038/s41524-023-01090-9
Abstract: Optimizing several properties simultaneously based on small data-driven machine learning in complex black-box scenarios can present difficulties and challenges. Here we employ a triple-objective optimization algorithm deduced from probability density functions of multivariate Gaussian distributions to optimize the ′ volume fraction, size and morphology in CoNiAlCr-based superalloys. The effectiveness of the algorithm is demonstrated by synthesizing alloys with desired / ′ microstructure and optimizing ′ microstructural parameters. In addition, the method leads to incorporating refractory elements to improve / ′ microstructure in superalloys. After four iterations of experiments guided by the algorithm, we synthesize sixteen alloys of relatively high creep strength from ~120,000 candidates of which three possess high ′ volume fraction (> 54%), small ′ size (< 480 nm) and high cuboidal ′ fraction (> 77%). 摘要: 我们提出双势垒效应有望增强铁电隧道结(FTJs)的隧道电致电阻比率(TER)。为了证明这一机制的可行性,我们设计了Pt/BaTiO3/LaAlO3/Pt/BaTiO3/LaAlO3/Pt双势垒铁电隧道结(DB-FTJ)模型结构,它可以被认为是两个相同的Pt/BaTiO3/LaAlO3/Pt单势垒铁电隧道结(SB-FTJs)串联在一起。基于密度泛函理论计算,我们在双势垒铁电隧道结中获得了2.210 108%的巨大隧穿电致电阻比率,这一值比Pt/BaTiO3/LaAlO3/Pt单势垒铁电隧道结的隧穿电致电阻比率至少大3个数量级,并且该双势垒铁电隧道结的高电导状态下具有超低的电阻面积乘积(0.093 K m2)。此外,通过分别控制两个单独的铁电势垒的极化方向,我们可以在双势垒铁电隧道结中实现四种电阻状态,使其有望成为多态存储器件。 Editorial Summary Universal multi-objective active learning algorithm In the realm of constrained datasets, the efficacy of optimization algorithms wields a significant influence over the convergence and optimization prowess of machine learning models in the pursuit of optimal material solutions. This, in turn, directly impacts the efficiency of discovering novel materials and refining their designs. Professor Yanjing Su's team, hailing from University of Science and Technology Beijing and Beijing Advanced Innovation Center for Materials Genome Engineering, has propounded a universally applicable algorithm for active learning in the domain of material triple-objective optimization. Termed as the "equiprobability distribution with maximum confidence interval" (ED-MCI), this algorithm extends its reach towards collaborative optimization designs encompassing a multitude of interrelated property objectives. The ED-MCI algorithm boasts remarkable optimization efficiency, enabling the fabrication of comprehensive, exceptional materials through a relatively modest number of iterative experimental cycles. In this paper, taking superalloy design as an example, a machine learning model is trained with multiple small data sets. Via iterative feedback between model-driven recommendations and experimental validation with ED-MCI, a harmonized optimization of the volume fraction, dimensions, and morphology of the ′ phase in CoNiAlCr-base superalloys is achieved. The ED-MCI algorithm, unburdened by the limitations of material categories during mathematical derivation, introduces a novel universal approach to multi-objective collaborative optimization in the domain of materials. 材料组织-性能多参量协同优化设计利器:通用型材料多目标主动学习方法 在数据量有限的客观条件下,优化算法很大程度上决定了机器学习模型在材料最优解搜寻的收敛性和寻优能力,直接影响新材料发现和优化设计的效率。北京科技大学、北京材料基因工程高精尖创新中心的宿彦京教授团队提出了一种通用的材料三目标性能主动学习算法(equiprobability distribution with maximum confidence interval (ED-MCI)),并可拓展到更多个相互关联目标性能的协同优化设计。ED-MCI算法具有高的优化效率,可以在小样本数据情况下,通过比较少的实验迭代次数,制备出综合性能优异的新材料。本文以具有高承温能力和长蠕变寿命的高温合金为例,用多个小数据样本集训练机器学习模型,通过模型推荐和实验验证的反馈迭代,利用ED-MCI对高温合金 ′相体积分数、尺寸和形状进行协同优化,制备出了具有高 ′相体积分数、尺寸小和立方度高的CoNiAlCr基高温合金。ED-MCI算法在数学推导阶段不受材料种类的束缚,具有可拓展性,为材料多目标协同优化设计提供了通用的新算法。 |
|
【打印本页】【关闭本页】 |
|
|
|