CDS&E: Optimal control of material microstructure evolution via massively parallel computing

非技术总结该奖项支持跨学科的计算研究和教育，旨在促进对材料微观结构的理解和控制。材料的一些重要性能，如机械强度或腐蚀敏感性，在很大程度上取决于微观结构：材料成分的大小、形状和排列，例如，在微米范围内的多晶体中的颗粒。尽管微观结构很重要，但合成和加工材料以获得所需的微结构，从而获得所需的材料性能的技术仍然有限。了解材料在制造过程中微观结构的时间演变仍然是一个基本的挑战，微观结构演变对外部控制的敏感性几乎还没有被探索过。该项目研究了使微结构演化服从外部控制的基本物理特性，并探索了通过将材料建模、实时反馈和控制融入材料加工中来创建设计微结构的新途径。这项工作将填补关于微观组织可控性基本极限的深刻认识空白，从而为系统地控制组织演变奠定基础。“可控性”本身将是从支配微观结构演化的机制中涌现出来的基本物理属性。该项目提供的计算工具和基本见解将指导微结构控制器的未来开发，用于特定的材料加工应用。这个项目的计算方面要求极高，需要亿级计算机的全部能力：每秒执行10亿次浮点运算的计算机。因此，该项目的一个主要部分是有效地并行几乎每个计算阶段。技术总结该奖项支持旨在促进对材料微观结构的理解和控制的跨学科计算研究和教育。虽然在新材料的设计、发现和优化方面投入了大量的精力，但对材料加工的最优控制相对了解较少。特别是，决定对外部控制的敏感性程度的不断演变的微观结构的基本物理特征仍然没有被探索。该项目将通过以下方式为材料微观结构控制奠定基础：a)设计对微观结构敏感的反馈控制器，b)开发有效实施反馈控制的计算工具，以及c)系统地探索反馈控制器相对于目标保真度、能效和成本最优等指标的性能。这项工作是控制理论、材料建模和高性能计算的交叉点。最终，研究人员的目标是通过利用大规模并行的高性能计算体系结构来探索材料加工中的可控性基本问题，从而为设计材料的合成创造最佳控制器，例如：如何优化控制材料加工参数以制造设计微结构？使用给定的加工方法可以制造的微结构的类型的限制是什么？一种加工方法能够合成给定类型的微结构的最低要求是什么？这项工作将使用控制理论的工具来探索材料微结构可控性的基本极限：为车辆、飞机和机器人的自主操作创造最佳算法的工程学分支。为了能够快速调查这一新领域，该项目将使用计算材料模型作为物理材料的替代品。其目标是开发可在不同模型之间转移的与模型无关的控制工具。然而，最初的工作将集中在一个特定的模型问题上，即：由Allen-Cahn方程支配的微观结构演化的相场模型。由于将要开发的控制方法将足够通用，它们可以与其他材料模型一起使用。该奖项由数学和物理科学局材料研究部的凝聚态物质和材料理论计划，工程局的土木、机械和制造创新司，以及计算机和信息科学与工程局的计算和通信基础部的软件和硬件基础计划共同支持。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

NONTECHNICAL SUMMARYThis award supports interdisciplinary computational research and education aimed to advance understanding and control of the microstructure of materials. Some important properties of materials, such as mechanical strength or corrosion susceptibility, are largely a function of microstructure: the sizes, shapes, and arrangement of material constituents, for example grains in polycrystals, over distances of micrometers. Despite the importance of microstructure, techniques for synthesizing and processing materials to obtain desired microstructures, and therefore desired materials properties, remain limited. Understanding how microstructure evolves in time as a material is made remains a fundamental challenge and the susceptibility of microstructure evolution to external control is nearly unexplored. This project investigates the fundamental physical characteristics that make microstructure evolution amenable to external control and explores novel pathways for creating designer microstructures through incorporation of materials modeling, real-time feedback, and control into material processing. This work will close a deep knowledge gap concerning the fundamental limits of microstructure controllability, thereby laying the foundation for the systematic control of microstructure evolution. "Controllability" itself will be a fundamental physical property emerging from the mechanisms that govern microstructure evolution. The computational tools and fundamental insights provided by this project will guide future development of microstructure controllers to be used in specific materials processing applications. The computational aspects of this project are extremely demanding and will require the full power of exascale computers: ones that perform 1 billion billion floating point operations per second. A major part of this project is therefore to efficiently parallelize nearly every stage of computation.TECHNICAL SUMMARYThis award supports interdisciplinary computational research and education aimed to advance understanding and control of the microstructure of materials. While much effort has been invested into the design, discovery, and optimization of new materials, optimal control of materials processing is relatively less well understood. In particular, the fundamental physical characteristics of an evolving microstructure that determine the degree of susceptibility to external control remain unexplored. This project will lay the foundations for material microstructure control by: a) designing microstructure-sensitive feedback controllers, b) developing computational tools for efficient implementation of feedback control, and c) systematically exploring feedback controller performance with respect to metrics such as objective fidelity, energy-efficiency, and cost-optimality. This work lies at the intersection of control theory, materials modeling, and high-performance computing. Ultimately, the researchers aim to create optimal controllers for the synthesis of designer materials by exploiting massively parallel, high performance computing architectures to explore fundamental questions of controllability in materials processing, such as: How does one optimally control materials processing parameters to make designer microstructures? What are the limits to the types of microstructures that may be made using a given processing method? What are the minimal requirements for a processing method to be able to synthesize a given type of microstructure?This work will explore the fundamental limits of materials microstructure controllability using the tools of control theory: the branch of engineering that creates optimal algorithms for the autonomous operation of vehicles, aircraft, and robots. To enable rapid investigation of this new field, the project will use computational materials models as surrogates for physical materials. The goal is to develop model-agnostic control tools that are transferrable between different models. However, effort will initially focus on one specific model problem, namely: phase field models of microstructure evolution as governed by the Allen-Cahn equation. Since the control methods to be developed will be sufficiently general that they can be used with other materials models.This award is jointly supported by the Condensed Matter and Materials Theory Program in the Division of Materials Research in the Directorate for Mathematical and Physical Sciences, the Civil, Mechanical and Manufacturing Innovation Division in the Engineering Directorate, and the Software and Hardware Foundations Program in the Division of Computing and Communications Foundations in the Directorate for Computer and Information Science and Engineering.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.