Collaborative Research: A Metamodeling Machine Learning Framework for Multiscale Behavior of Nano-Architectured Crystalline-Amorphous Composites

协作研究：纳米结构晶体非晶复合材料多尺度行为的元建模机器学习框架

• 批准号: 2132383
• 负责人: Lin Li
• 金额: $ 21.47万
• 依托单位: University of Alabama Tuscaloosa
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-01-01 至 2023-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2132383&HistoricalAwards=false
• 关键词: Collaborative; Research; Metamodeling; Machine; Learning

Nanostructured composites may achieve the preferred combination of strength, ductility, toughness, and irradiation tolerance through microstructure tailoring and chemistry; thus, they are highly desired structural materials in harsh environments. In accelerating the design of nanostructured composites, the need is for models that can capture microstructure-dominated deformation mechanisms from the atomic scale to the structural scale. This award supports the development of a meta-modeling framework that enables the incorporation of atomic-level knowledge of microstructure-dominated deformation behaviors to predict structural response. This framework will be applied to design amorphous ceramic reinforced nickel alloys used in advanced nuclear reactors and high-temperature environments. The interdisciplinary nature of the research will provide graduate students a diverse training in computational and experimental mechanics, collaborative teamwork experience, as well as research experience at Los Alamos National Laboratory. Research opportunities and mentorship programs will be created at the University of Nebraska–Lincoln and the University of Alabama for undergraduate students, especially for women and underrepresented minorities. Additionally, outreach activities at university museums will attract local K-12 students towards STEM careers.The accelerated design of nanostructured composites with desired properties needs sophisticated models that can capture microstructure-dominated deformation mechanics at multiple scales. In this project, a metamodeling machine learning based framework that enables the incorporation of microstructure-dominated deformation mechanics into a predictive macroscale model will be developed though an integrated characterization, experimental, and computational approach. Nanocrystalline nickel with high crystallization temperature amorphous ceramic SiOC at the grain boundaries will be the model material. Microstructures, deformation mechanisms, and mechanical properties of the Ni-SiOC nanocomposites will be characterized using advanced microscopes and in situ micromechanical testing. The atomic details of amorphous boundaries-mediated deformation in the nanostructures will be revealed through the combination of density functional theory calculations and large-scale molecular dynamics simulations. A dual-phase micromechanics model will be developed by coarse-graining the key deformation mechanisms identified at the atomic level. Surrogate models of the activation functionals for various mechanisms in the micromechanics model will be derived by training machine learning models from atomistic simulation data. At the macroscopic scale, a physics-informed neural network model in which the hybridized machine learning models will be used as surrogates to bridge scales will complete the metamodeling framework.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

纳米结构的公司可以通过微观结构调整和化学来实现强度，延性，韧性和辐射耐受性的首选组合；因此，它们是Harmsh环境中高度期望的结构材料。在加速纳米结构公司的设计中，需要模型，这些模型可以捕获以微观结构为主的变形机制，从原子量表到结构量表。该奖项支持了一个元模型框架的发展，该框架可以使原子级知识的基础结构对微观结构为主的变形行为进行预测。该框架将应用于设计用于高级核反应堆和高温环境中的无定形陶瓷镍合金。该研究的跨学科性质将为研究生提供计算和实验机制，协作团队合作经验以及Los Alamos国家实验室的研究经验的潜水员培训。内布拉斯加州 - 林肯大学和阿拉巴马大学的研究机会和心态计划将针对本科生，尤其是女性和代表性不足的少数民族而创建。此外，大学博物馆的外展活动将吸引当地的K-12学生进入STEM职业。具有所需物业的纳米结构公司的加速设计需要复杂的模型，这些模型可以在多个尺度上捕获以微观结构为主导的变形力学。在这个项目中，将开发基于机器学习的基于机器学习的框架，该框架将开发出微观结构主导的变形力学为预测性宏观模型，尽管综合表征，实验性和计算方法将开发为预测性的宏观尺度模型。具有高结晶温度在晶界处的具有高结晶温度的纳米晶镍将是模型材料。 Ni-SIOC纳米复合材料的微观结构，变形机制和机械性能将使用高级显微镜和原位微机械测试来表征。纳米结构中无定形边界介导的变形的原子细节将通过密度功能理论计算和大规模分子动力学模拟的结合来揭示。双相微力学模型将通过在原子水平上鉴定的关键变形机制来开发。微力学模型中各种机制的激活功能的替代模型将通过原子模拟数据训练机器学习模型得出。在宏观量表上，物理信息的神经网络模型，其中杂交机器学习模型将用作替代物来桥梁尺度，将完成元模型框架。该奖项反映了NSF的法定任务，并通过使用基金会的知识分子优点和更广泛的影响评估标准来评估值得支持。