CAREER: Integrating Information Theory with Data-Driven Mechanics: Toward Predictive Modeling of Material Behavior far from Equilibrium

该学院早期职业发展（Career）奖将支持对远离平衡状态的材料行为建模的基础研究，如金属系统中的晶体塑性、固-固相变或无序介质中的粒子重排。在所有这些例子中，可观察到的宏观材料行为受到潜在的原子或粒子微观结构的影响。缺乏对微观结构和宏观力学响应之间联系的理解，阻碍了对基础设施、交通和制造业等结构和工业应用至关重要的高保真预测能力。进一步说明，在制造业中，它会导致经济损失，并阻碍颗粒加工、金属成型和增材制造的创新。因此，旨在缩小这一差距的进展对于促进国民健康、繁荣和福利以及确保国防是重要和必要的。此外，这笔拨款将支持一项教育和推广计划，以增加STEM领域的性别和种族平等，并促进公众的批判性思维和科学素养。该项目的长期目标是实现一种计算范式，提供具有原子或粒子保真度的远离平衡力学行为的连续宏观模型。为了实现这一目标，目标是双重的：1)将信息理论实践与机器学习相结合，以识别精确的宏观状态变量，这些状态变量将代表从原子或离散粒子行为开始的粗粒度多尺度过程；2)使用最先进的基于物理的变分公式来确定这些变量的连续演化方程。该方法将使用实验流变学数据进行验证，以剪切胶体颗粒的致密填料，这是一个科学丰富且技术相关的远离平衡力学行为的例子。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。

This Faculty Early Career Development (CAREER) award will support fundamental research on the modeling of material behavior far from equilibrium, such as crystal plasticity in metallic systems, solid-solid phase transformations, or particle rearrangements in disordered media. In all these examples, the observable macroscale material behavior is affected by the underlying atomic or particle microstructure. The lack of understanding of the linkage between microstructure and macroscale mechanical response hinders high-fidelity predictive capacity critical to structural and industrial applications, e.g., in infrastructure, transportation, and manufacturing. To illustrate further, in manufacturing, it leads to economic losses and barriers to innovation in processing of particles, forming of metals, and additive manufacturing. Progress aimed at closing this gap is therefore important and needed to advance the national health, prosperity, and welfare, as well as to secure the national defense. In addition, this grant will support an educational and outreach program to increase gender and racial equity in STEM fields, as well as to promote critical thinking and scientific literacy within the public.The long-term goal of this project is to realize a computational paradigm that delivers continuum macroscopic models of far-from-equilibrium mechanical behavior with atomic or particle fidelity. In pursuit of this goal, the objective is twofold: 1) integrate information theory practices with machine learning to identify precise macroscopic state variables that would represent coarse-grained multiscale processes starting from the atomic or discrete particle behavior, and 2) use state-of-the-art physics-based variational formulation to determine the continuum evolution equations for these variables. The approach will be validated using experimental rheological data for shearing of dense packings of colloidal particles, a scientifically rich and technologically relevant example of far-from-equilibrium mechanical behavior.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.