Collaborative Research: AI-Driven Multi-Scale Design of Materials under Processing Constraints

该项目的目标是通过开发一种多尺度方法来提高材料设计的知识，该方法将基于物理的热机械加工和材料模型与人工智能（AI）和机器学习（ML）相结合。基本的假设是，金属部件可以被设计成通过优化基本的微观结构特征和加工参数来实现目标宏观尺度性质和性能。该项目将建立设计方法，使：（i）微观结构和加工参数对宏观尺度性能的影响的调查;和（ii）提供所需宏观尺度性能的多种最佳材料设计的识别。通过设计微观结构和工艺来优化宏观尺度性能的能力将提高当前和未来工程系统的性能。此外，考虑到制造约束，这种多尺度设计框架将不仅确定数学解决方案，而且确定可制造的设计。本文的研究方法和结果将与钛铝合金的实验数据进行对比。该项目将对经济产生社会影响，提高金属部件的性能，并最大限度地减少与制造相关的时间和成本。所获得的知识将通过技术活动和开放获取软件工具传播给学术界和工业界。该项目的其他成果包括本科生和研究生课程开发，学生的研究和教育经验，以及其他涉及学生和教育工作者的外展活动，特别关注来自代表性不足群体的个人。该项目的总体目标是通过开发多尺度优化策略来推进金属材料设计的知识，该策略将由物理驱动-的热机械加工和微观结构模型，以及基于AI/ML的预测建模和知识发现方法。该研究将解决旨在优化热机械加工参数（即，应变速率、温度、持续时间）以获得所需的微结构特征（即，晶体学织构、晶粒形态）和宏观尺度性质，通过研究加工-（微观）结构-性质链内的耦合、多尺度和高维相互作用。为了实现这一目标，该项目将开发基于物理的模型，使微观结构的方向和形态（晶粒尺寸和形状）的明确量化，以及ML引导的反馈感知识别策略的关键处理/（微）结构参数，这将随后通过有针对性的采样进行探索。该研究将通过将制造约束集成到设计框架中并探索提供所需宏观尺度特性的多个最佳材料解决方案来提高对逆向材料设计的理解。基于物理和AI/ML驱动的模型，以及通过多尺度设计框架获得的最佳结果，将使用钛铝合金的实验处理，微观结构和性能数据进行验证。该项目的教育和推广目标侧重于培训学生和未来的劳动力，以创造新的知识，计算和ML驱动的材料设计，这将与课程开发和广泛的传播和推广计划的支持。该奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

The objective of this project is to improve the knowledge of materials design by developing a multi-scale methodology that combines physics-based models of thermo-mechanical processing and materials with artificial intelligence (AI) and machine learning (ML). The underlying hypothesis is that the metallic components can be designed to achieve targeted macro-scale properties and performance by optimizing the underlying microstructural features and processing parameters. The project will build design methodology that enables: (i) investigation of the effects of microstructures and processing parameters on macro-scale properties; and (ii) identification of multiple optimum material designs that provide desired macro-scale performance. The ability to optimize macro-scale properties by designing microstructures and processes will improve the performance of current and future engineering systems. Additionally, with the consideration of manufacturing constraints, this multi-scale design framework will not merely identify mathematical solutions, but the designs that will be manufacturable. The researched methods and results will be tested against the experimental data of a Titanium-Aluminum alloy. The societal impacts of the project will be on the economy, with performance improvement in metallic components and minimization of the time and costs associated with manufacturing. The gained knowledge will be disseminated to academia and industry with technical activities and open-access software tools. Additional deliverables of the project include curriculum development at both undergraduate and graduate levels, research and education experiences for students, and other outreach activities involving students and educators with a special focus on individuals from underrepresented groups.The overarching goal of this project is to advance knowledge in the design of metallic materials by developing a multi-scale optimization strategy that will be driven by the physics-based models of thermo-mechanical processing and microstructures, and AI/ML-based predictive modeling and knowledge discovery approaches. The research will address the inverse design problem that aims to optimize the thermo-mechanical processing parameters (i.e., strain rate, temperature, duration) to achieve desired microstructural features (i.e., crystallographic texture, grain morphology) and macro-scale properties by investigating the coupled, multi-scale, and high-dimensional interactions within the processing-(micro)structure-property chain. To achieve this goal, the project will develop physics-based models that enable explicit quantification of microstructural orientations and morphology (grain sizes and shapes), and an ML-guided feedback-aware identification strategy for key processing/(micro)-structure parameters, which will be subsequently explored by targeted sampling. The research will improve the understanding of inverse materials design by also integrating manufacturing constraints into the design framework and exploring multiple optimum material solutions that provide desired macro-scale properties. The physics-based and AI/ML-driven models, as well as the optimum results obtained by the multi-scale design framework, will be validated using the experimental processing, microstructure, and property data of a Titanium-Aluminum alloy. The education and outreach objectives of the project focus on training students and the future workforce to create new knowledge on computational and ML-driven design of materials, which will be supported with curriculum development and an extensive dissemination and outreach plan.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.