DMREF: Data-Driven Integration of Experiments and Multi-Scale Modeling for Accelerated Development of Aluminum Alloys

传统上，材料设计是一个迭代的过程，通过调整材料的制造工艺，逐渐提高材料的性能。通过将计算模型引入循环，并将输出与每一步材料结构和性能的实验测量结果进行比较，可以加速这种方法。然而，由于需要人类参与数据收集、分析和简化的几乎每一步，这个周期被放慢了。该设计材料以革新和工程我们的未来（DMREF）奖支持研究，通过利用现代数据科学工具开发数据框架和基础设施，以便在复杂材料开发问题中无缝集成实验和模型，从而促进先进材料在工程应用中的快速采用。该框架将广泛适用，允许研究人员采用它来解决特定问题，并允许将这些想法整合到广泛的材料和应用领域。所开发的概念和工具将通过免费提供的软件、开放源码模块、在线教程、开放获取的数据集以及为用户提供在新环境中应用这些工具的培训来传播。这项工作旨在为材料设计循环建立一个新的范例，其中数据流，而不是个人建模或实验任务，被视为中心。将利用现代数据科学工具来创建语义框架和物理基础设施，以便在复杂材料开发问题中无缝集成实验和模型。这个框架将定义和描述材料数据的类别，以及这些类别之间的联系，这些类别需要创建一个自动化的数据流，在这个数据流中，设计循环中的每个实验和计算任务都会自动将信息推送到所有任务共同的数据层，或者从中提取信息。这将通过最大限度地减少所需的人为干预来加速材料设计过程，同时仍然允许人为输入必要的循环。该方法的具体实例将是实现商用铝合金抗剥落性的多尺度建模框架，并通过集中数据层自动连接到先进的微观结构表征和高通量激光冲击测试。通过进一步耦合加工过程中的微观结构发展模型，通过训练机器学习算法来优化微观结构，以抵抗碎片空洞的成核和生长，从而闭合设计回路。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。

Traditionally, materials design is an iterative process in which the properties of the material are gradually improved by adjusting process by which the material is made. This approach can be accelerated by introducing computational models into the loop and comparing the output with experimental measurements of the structure and properties of the material at each step. The cycle is slowed, however, by the need for humans to be involved at almost every step for data collection, analysis, and reduction. This Designing Materials to Revolutionize and Engineer our Future (DMREF) award supports research to facilitate rapid adoption of advanced materials into engineering applications by leveraging modern data science tools to develop a data framework and infrastructure for seamless integration of experiments and models in complex materials development problems. This framework will be broadly applicable, allowing researchers to adopt it to particular problems and allowing the ideas to be integrated across a wide range of materials and application areas. The concepts and tools developed will be disseminated through freely-available software, open source modules, online tutorials, open access data sets, and training for users to apply the tools in new contexts.This work seeks to establish a new paradigm for the materials design loop in which the flow of data, rather than individual modeling or experimental tasks, is viewed as central. Modern data science tools will be leveraged to create the semantic framework and physical infrastructure necessary for seamless integration of experiments and models in complex materials development problems. This framework will define and describe classes of material data, and connections among these classes, that are required to create an automated data flow in which each experimental and computational task in the design loop automatically pushes information to, or pulls information from, a data layer common to all tasks. This will accelerate the materials design process by minimizing the required human intervention, while still allowing human input to the loop where essential. A specific instantiation of this approach will be the implementation of a multi-scale modeling framework for the resistance of commercial aluminum alloys to spall failure, with automated connections to advanced microstructural characterization and high-throughput laser shock testing through a centralized data layer. The design loop will be closed by further coupling to models of microstructure development during processing, with machine-learning algorithms trained to optimize the microstructure to resist the nucleation and growth of spall voids.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.