FMRG: Cyber: Manufacturing USA: Material-on-demand manufacturing through convergence of manufacturing, AI and materials science

人工智能最近的进步正在推动一场工业革命，导致智能、自主系统的出现。这项未来网络制造研究拨款重新设想了新一代制造机器的自主性，能够以负担得起的价格制造具有前所未有性能的先进合金产品。该项目汇集了学术界和业界在人工智能(包括机器学习、自适应控制和数据科学)、材料科学和智能制造领域的不同先驱团队，以解决基础研究和技能开发问题。该团队包括德克萨斯农工大学/德克萨斯农工工程实验站、布朗大学、德克萨斯农工大学金斯维尔、普莱里维尤农工大学、休斯顿社区学院，以及多个行业、地区政府和学术合作伙伴。这些基础使一种新的方法和演示平台能够利用3D打印、材料基因组学和传感器技术的最新进展来控制生产过程并混合多种材料以获得所需的性能。这些产品为美国经济提供了关键的竞争优势，并为战略高超声速系统和能源转换部门的国家关键材料挑战提供了有效的解决方案。该项目还将为学生和行业专业人士提供有价值的教育和技能发展的机会。该项目解决了实现未来制造机器的科学挑战，赋予了深度自主制造定制材料按需制造的能力。自主制造机器平台的设想是自适应地生成工艺计划(融合来自不同数据和知识来源的信息)-控制材料的微观结构和组成，而不仅仅是几何和形态-以生产具有显著增强的功能性能的大规模定制材料部件。这一努力将产生以下四个对自主原则的基础性贡献：(1)形状受限的机器学习。这种新颖的物理信息机器学习形式的关键思想是引入对潜在函数关系的形状/符号的约束，以对不完整的物理和经验知识进行建模。(2)利用令人惊讶的观察。从历史上看，一个实验或一个过程的意外结果会带来新的发现和见解。处理令人惊讶的观测将自动系统与自动系统区分开来。(3)保护使用数字双胞胎的外推。将开发出将物理系统与多个数字双胞胎融合的原理，每个双胞胎都以特定的保真度捕捉特定的物理。(4)知识扩张。将研究新的方法，通过创新的图形神经网络在公共制造文献/数据库中获取关于工艺链和动态工艺-材料关系的经验和深入知识。这些方法将被验证，以发现制造强度保持在1400°C以上的高熵合金的创新途径，显示出更好的加工性和减少昂贵和稀缺材料的使用。该项目将提供实践培训和教育，利用他们的专业知识以及与美国国家制造业、工业和教育网络的合作。这项未来的制造研究得到了计算机和信息科学与工程局的计算机和网络系统司(CEISE/CNS)、工程局的土木、机械和制造业创新司(ENG/CMMI)、工程局的工程教育和中心(ENG/EEC)、数学和自然科学局的数学科学部(MPS/DMS)和技术的支持，创新和伙伴关系委员会的翻译影响司(TIP/TI)。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。

Recent advances in AI are driving an industrial revolution, leading to the emergence of intelligent, autonomous systems. This Future CyberManufacturing research grant reimagines autonomy for a new generation of manufacturing machines, capable the manufacture of advanced alloy products with unprecedented performance affordably. The project brings together a diverse team of pioneers from academia and the industry in AI (including machine learning, adaptive control, and data science), materials science, and smart manufacturing, towards addressing the foundational research and skill development. The team includes Texas A&M University/Texas A&M Engineering Experiment Station, Brown University, Texas A&M University Kingsville, Prairie View A&M University, Houston Community College, and multiple industry, regional government, and academic partners. These foundations allow a new approach and demonstration platforms to harness recent advances in 3D printing, materials genomics, and sensor technologies to control the production processes and to mix multiple materials to obtain the desired properties. These products provide a critical competitive edge for the US economy and effective solutions for the national critical material challenges in the strategic hypersonic systems and energy conversion sectors. It will also provide students and industry professionals with opportunities for valuable education and skill development.The project tackles scientific challenges of realizing futuristic manufacturing machines endowed with a deep level of autonomy to make tailored materials-on-demand manufacturing. The autonomous manufacturing machine platforms are envisioned to generate process plans adaptively (fusing information from diverse data and knowledge sources) – to control material microstructure and composition beyond just geometry and morphology – to yield bulk-scale tailored material components with dramatically enhanced functional performance. The following four foundational contributions to autonomy principles would emerge from this effort: (1) Shape-constrained machine learning. The key idea in this novel form of physics-informed machine learning is to introduce constraints on the shape/sign of the underlying functional relationship to model incomplete physical and experiential knowledge. (2) Harness surprise observations. A surprise outcome from an experiment or a process has historically led to new discoveries and insights. Dealing with surprising observations differentiates an autonomous system from an automated one. (3) Safeguarding extrapolation using digital twins. The principles of fusing physical systems with multiple digital twins would be developed, each capturing certain physics with a specified fidelity. (4) Knowledge expansion. New approaches would be studied to capture experiential and deep knowledge in the public manufacturing literature/databases on process chains and the dynamic process-material relationships via innovative graph neural networks. These approaches will be validated to discover innovate new pathways to manufacture high-entropy alloys that retain strengths above 1400°C, demonstrating improved machinability and reduced use of expensive and scarce materials. The project would provide hands-on training and education, leveraging their expertise and collaborations with national Manufacturing USA, industry, and education networks.This Future Manufacturing research is supported by the Computer and Information Science and Engineering Directorate's Division of Computer and Network Systems (CISE/CNS), the Engineering Directorate's Division of Civil, Mechanical and Manufacturing Innovation (ENG/CMMI), the Engineering Directorate's Division Engineering Education and Centers (ENG/EEC), the Mathematical and Physical Sciences Directorate's Division of Mathematical Sciences (MPS/DMS), and the Technology, Innovation and Partnerships Directorate's Translational Impacts Division (TIP/TI).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.