Collaborative Research: DMREF: Simulation-Informed Models for Amorphous Metal Additive Manufacturing

非晶态金属的添加制造是一种潜在的变革性技术，用于打印具有优异强度和韧性的三维零件。由于非晶态金属在凝固时不采用晶体结构，因此它们不会形成可限制部件性能的晶体缺陷。虽然使用激光在表面沉积金属的高冷却速度有利于避免结晶，但激光的扫描可能会导致随后的结晶和性能从一个位置到另一个位置的变化。这些问题目前将这项技术限制在小规模和特殊部件上。为了克服这一限制，机器学习方法将从电子纳米衍射和模拟数据中获得有意义的材料结构测量。在此基础上，研究小组将建立模拟信息模型，这些工具将预测加工如何产生最终材料的强度和韧性。这些模型将通过与实验的直接比较进行测试，为非晶态金属添加剂制造的计算设计工具奠定科学基础。同时，参与这项DMREF研究的三所大学将组成一个学习社区，支持研究生在线交流的专业发展。这个社区将提炼和传播调查人员的经验，为课程开发在线内容，参与公共交流，并为服务不足的社区建立外联计划。开发的模块将教会明天的研究人员如何有效地吸引不同年龄段的不同受众。总而言之，这项工作支持国家在先进制造技术和劳动力发展方面的优先事项，特别是在与数学方法和数据科学相结合的领域。DMREF项目将开发基础材料科学和计算工具，以使添加制造的非晶态金属能够设计出具有所需机械性能(包括强度和韧性)的设计。非晶态金属，也被称为金属玻璃，在添加剂制造应用中具有潜在的转化材料的潜力。与通过各向异性颗粒生长而凝固的晶体材料不同，快速冷却会导致金属玻璃凝固而没有晶体结构，这通常会导致晶界和复杂的纹理。与晶体相比，非晶态金属添加剂制造具有更好的结构均匀性，并且可以克服铸造较大结构的冷却速度限制。然而，与逐层加工相关的重新加热导致材料具有复杂的热历史和空间变化的机械性能。研究团队进行的模拟信息建模是迈向同步设计方法的第一步，以实现目标材料的性能和性能。这种方法将直接激光沉积处理与高保真物理模型相结合。机器学习将被用来从纳米分辨率的电子、纳米衍射和原子模拟数据中量化适用于预测机械性能的关键顺序参数。解决这一数据融合和推理问题将以稳健的方式将不同尺度上的实验和模拟数据与结构阶数参数联系起来。研究人员将从这些模型中建立模拟信息模型，这是一种连续的数值工具，将捕捉加工如何提高最终材料的强度和韧性。验证将通过与非现场和现场机械测试进行直接比较来实现。不确定性量化将优先包括在这些模型中。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。

Additive manufacturing of amorphous metals is a potentially transformative technology for printing three-dimensional parts with superior strength and toughness. Since amorphous metals solidify without adopting a crystal structure, they do not form crystalline defects that can limit part performance. While the high cooling rates associated with using a laser to deposit metal on a surface are favorable for avoiding crystallization, the scanning of the laser can lead to subsequent crystallization and variations in properties from one location to another. These issues currently limit the technique to small scale and specialty parts. To overcome this limitation, machine learning approaches will derive meaningful measures of material structure from electron nanodiffraction and simulation data. Upon this foundation, the research team will build simulation-informed models, tools that will predict how processing gives rise to the strength and toughness of the resulting materials. These models will be tested by direct comparison to experiment, laying the scientific groundwork for computational design tools for additive manufacturing of amorphous metals. Concurrently, the three universities engaged in this Designing Materials to Revolutionize and Engineer our Future (DMREF) research will form a learning community to support graduate student professional development in online communication. This community will distill and disseminate the investigators' experiences developing online content for courses, engaging in public communication, and building outreach programs for underserved communities. The modules developed will teach tomorrow’s researchers how to effectively engage diverse audiences of various ages. Taken together, this work supports national priorities in advanced manufacturing technology and workforce development, particularly at the intersection with mathematical methods and data science.This DMREF project will develop the underlying materials science and computational tools to enable design of additively manufactured amorphous metals with desired mechanical properties, including strength and toughness. Amorphous metals, also termed metallic glasses, have potential as a transformative material for additive manufacturing applications. Unlike crystalline materials that solidify through the growth of anisotropic grains, typically resulting in grain boundaries and complex textures, rapid cooling causes metallic glasses to solidify without crystal structure. Amorphous metal additive manufacturing is promising both for superior structural homogeneity compared to crystals and for overcoming cooling-rate limitations for casting larger structures. However, reheating associated with layer-by-layer processing results in material with a complex thermal history and spatially varying mechanical properties. The simulation-informed modeling undertaken by the research team is the first step toward a simultaneous design approach for achieving target materials properties and performance. This approach will couple processing by direct laser deposition with high-fidelity physical models. Machine learning will be used to quantify key order parameters suitable for predicting mechanical properties from nanometer-resolution electron nanodiffraction and atomistic simulation data. Solving this data fusion and inference problem will relate experimental and simulation data on differing scales to structural order parameters in robust ways. From these, the researchers will build simulation-informed models, continuum numerical tools that will capture how processing gives rise to the strength and toughness of the resulting materials. Validation will be achieved by direct comparison to ex situ and in situ mechanical testing. Uncertainty quantification will be included in these models a priori.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.