CAREER: Cyberinfrastructure for Printable Multifunctional Microstructural Materials

近年来，由于先进制造的出现和先进机器学习工具和高性能计算推动的集成计算材料设计的快速发展，材料发现发生了革命性的变化。增材制造承诺对材料的微观结构和性能进行精确控制；然而，实验室到市场的发展受到计算限制的阻碍，必须解决这个问题，以保持美国在全球材料市场的竞争力。NSF CAREER项目通过开发和部署一种新的网络基础设施来设计具有理想多功能特性的可打印材料，解决了效率低下、费用高昂、过度依赖数据和制造不确定性的四个关键计算挑战。这种方法改变了当前材料开发的范式，允许新一代的微观结构几何，精确和有效的材料表征数值方法，以及强大的物理感知生成模型。除了增材制造的实际进步之外，该项目还通过预测新材料的微观结构状态，为材料科学做出了重大贡献，这些新材料的应用范围从机器人和航空航天到高频通信、传感器、电源、热管理、能量收集和医疗植入物。该项目在多学科环境中培养各级学生和专业人士，为他们在机器学习、高性能计算、材料科学、计算力学和增材制造的交叉领域为问题提供解决方案做好准备。研究结果将作为开源软件公开提供给更广泛的社区，并提供有关设计和使用的全面文档，以帮助所有领域的用户。该项目将显著增强集成计算材料工程（ICME）领域的四个关键领域。第一个研究重点是开发一个通用的、跨平台的、并行的硅体素微结构生成器，提供各种形态的数据集，从而产生不同的性能和可制造性。第二推力建立了两种数值方法的材料表征都旨在提高计算效率相比，传统的数值方法。针对压电特性，提出了利用快速傅立叶变换数值方法求解耦合机电均匀化的新能量公式。在力学性能方面，引入了一种耦合的周动力物理神经解算器。第三个项目设计了TransVNet，这是一种独特的架构，结合了变分自编码器和卷积神经层，并通过视觉转换器进行增强，用于双向结构-属性映射学习。第四个推力通过制造和测试材料的3D表示来验证材料设计网络基础设施。通过在ANSYS生态系统中实施材料微观结构探索者（PyMME）网络基础设施，开发基于项目的学习（PBL）模块，为创新（M:2:I）本科课程，研究生材料信息学课程，K-12虚拟材料探索者实验室以及吸引学生参与创新项目和产品开发，该研究被整合到爱荷华州立大学的课程中。该项目由OAC和促进竞争研究的既定计划（EPSCoR）共同资助。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

In recent years, material discovery has undergone revolutionary change due to the advent of advanced manufacturing and rapid progress in Integrated Computational Material Design fueled by advanced machine learning tools and high-performance computing. Additive manufacturing promises precise control over materials’ microstructures and properties; however, lab-to-market development has been impeded by computational limitations, and this must be addressed to maintain US competitiveness in the global materials market. This NSF CAREER project solves the four critical computational challenges of inefficiency, expense, overreliance on data, and manufacturing uncertainties by developing and deploying a novel cyberinfrastructure for designing printable materials with desirable multifunctional properties. This approach transforms the current paradigm of material development, allowing for the novel generation of microstructural geometries, precise and efficient numerical methods for material characterization, and a robust physics-aware generative model. Beyond practical advancements in additive manufacturing, this project contributes significantly to materials science by predicting the microstructure status of new materials for applications ranging from robotics and aerospace to high-frequency communications, sensors, power sources, thermal management, energy harvesting, and medical implants. The project trains students at all levels and professionals in a multidisciplinary environment that prepares them to contribute solutions to problems at the intersection of machine learning, high-performance computing, materials science, computational mechanics, and additive manufacturing. The research results will be publicly available as open-source software to the broader community, with comprehensive documentation on the design and usage to help users from all domains.This project will significantly enhance the Integrated Computational Materials Engineering (ICME) field in four key areas. The first research thrust develop a universal, cross-platform, parallelized in silico voxelized microstructure generator, offering a dataset of various morphologies that lead to distinct properties and manufacturability. The second thrust establishes two numerical methods for material characterization both aimed at increased computational efficiencies compared with conventional numerical methods. For piezoelectric property, a new energy formulation for solving coupled electromechanical homogenization through a Fast Fourier Transform numerical method is presented. For mechanical property, a coupled peridynamics physics-informed neural solver is introduced. The third thrust designs TransVNet, a unique architecture combining a variational autoencoder with convolutional neural layers, enhanced by a vision transformer, for bi-directional structure-property mapping learning. The fourth thrust validates the material design cyberinfrastructure by fabricating and testing the 3D representation of the material. The research is integrated into the Iowa State University curriculum by implementing Material Microstructure Explorer (PyMME) Cyberinfrastructure in the ANSYS Ecosystem, developing a Project-Based Learning (PBL) module for the Make To Innovate (M:2:I) undergraduate program, a graduate material informatics course, a Virtual Material Explorer Lab for K-12 and engaging students in innovative projects and product development. This project is jointly funded by OAC and the Established Program to Stimulate Competitive Research (EPSCoR).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.