AI Institute: Planning: Novel Neural Architectures for 4D Materials Science

非技术描述：极端条件下（高温、高应力、腐蚀性环境等）复杂材料的高保真预测建模对于加速材料设计和优化以应对我们世界面临的紧迫挑战至关重要。该项目旨在利用基础和用途启发的人工智能（AI）研究，再加上尖端实验，彻底改变和改造传统的材料科学与工程（MSE）。这种新颖的方法根植于MSE的基本原理，即微观结构控制性能，专注于新型神经结构的发展，这种神经结构可以在多个长度和时间尺度上自然地捕捉关键微观结构特征之间的物理因果关系，用于预测建模和优化材料设计。本项目开发的用于构建新型物理学习模型的方法和实验框架将应用于复杂材料系统中的各种引人注目的问题，包括陶瓷、金属和金属合金、复合材料和多孔材料。预计该项目将影响航空航天、微电子、石油工业、消费品等多个领域。技术描述：该研究所的主题涉及通过基础和使用启发的人工智能研究开发革命性方法，再加上4D x射线微断层扫描和相关显微镜，开发和理解截然不同的材料系统中的结构-性质关系，用于预测建模和优化材料设计。该研究所的目标将是加速新的学习理论、实验方法和验证协议的研究，这些研究将促进复杂材料系统的进化和分层结构-属性映射的科学建模。在这个规划项目中，研究人员用数学方法阐述了在关键应用领域（金属和金属合金、多功能复合材料、多孔地质材料、核燃料等）复杂材料结构-属性映射建模中普遍存在的挑战，论证了机器学习和人工智能在解决这些挑战方面的必要性和初步可行性，并通过x射线微断层扫描和相关显微镜与4D实验相关联。将开发一个工业合作者联盟，将该计划知识的基础知识转化为实际解决方案，并在劳动力中培养一批新的熟练从业者。这个项目将激发人们重新思考机器学习在材料科学中的效用：从知识不可知的特征学习到适应特定领域知识的推理机制。它将为潜在的自动化材料表征、研究和发现提供关键基础设施。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。

Non-technical Description: High-fidelity predictive modeling of complex materials under extreme conditions (high temperature, high stress, corrosive environment etc.) is crucial for accelerating material design and optimization to address the pressing challenges in our world. This project will aim to leverage both fundamental and use-inspired artificial intelligence (AI) research, coupled with cutting-edge experiments, to revolutionize and transform traditional materials science and engineering (MSE). The novel approach, rooted in the fundamental principle in MSE, that microstructure controls properties, focuses on the development of novel neural architectures that naturally capture the physical causal relations across key microstructural features at multiple length and time scales for predictive modeling and optimal material design. The methodologies and experimental frameworks for constructing novel physics-based learning models developed in this project will be applied to a variety of compelling problems in complex material systems including ceramics, metals and metallic alloys, composites, and porous materials. It is expected that this project will impact many areas including aerospace, microelectronics, petroleum industry, and consumer products. Technical Description: The theme of this institute involves the development of revolutionary approaches enabled by fundamental and use-inspired AI research, coupled with 4D X-ray microtomography and correlative microscopy, to develop and understand structure-property relationships in vastly different materials systems for both predictively modeling and optimal material design. The goal of the institute will be to accelerate converging research on new learning theories, experimentation methodologies, and validation protocols that will facilitate scientific modeling of the evolutionary and hierarchical structure-property mappings of complex materials systems. In this planning project, researchers mathematically formulate the ubiquitous challenges in modeling complex material structure-property mappings across critical application domains (metals and metallic alloys, multi-functional composites, porous geo-materials, nuclear fuels, etc.), demonstrate the necessity and preliminary feasibility of machine learning and AI in addressing these challenges, and correlate with 4D experiments through x-ray microtomography and correlative microscopy. A consortium of industrial collaborators will be developed to transfer the fundamental knowledge from this program knowledge into practical solutions and to educate a new class of skilled practitioners in the workforce. This project will inspire one to re-think the utility of machine learning in materials science: From knowledge-agnostic feature learning to reasoning mechanisms adaptive to domain-specific knowledge. It will provide the key infrastructure for potential automated materials characterization, research, and discovery.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.