Elements: Data-Science Methods for Resource Allocation During Characterization of Dynamic Systems

该项目开发软件基础设施及其与物理基础设施的集成，这是将最先进的数据科学，机器学习和人工智能（AI）工具带到国家用户设施的新材料科学实验所必需的。这项工作将创建一个开放的控制包，使动态实验能够通过实时建模得到信息。这些发展将使国家同步加速器用户设施更有效地利用稀缺的束流时间，为探测材料在负载下的机械响应的原位实验提供更高的科学吞吐量。这项工作评估了一种假设，即罕见的材料失效事件可以通过使用机器学习解决方案描述局部材料状态演变的少量特征来预测。该项目的重点是在机械载荷下对材料进行同步加速器X射线散射测量，特别是将新的和现有的工具集集成到一个控制包中，该控制包能够在康奈尔高能同步加速器源（CHESS）（一个国家用户设施）进行动态资源分配，以实现超高效的数据收集。该项目整合了这些工具集，通过实时处理来自CHESS等用户设施的数据来检测前兆特征，并建议资源分配，以促进对动态材料系统中随机事件早期阶段的研究。其目标是开发机器学习（ML）技术和软件基础设施，以便在概率意义上最好地分配材料测试设施的有限探测器资源，更好地捕获材料中罕见事件的早期阶段以及这些事件的关键因素。这项工作也有兴趣在材料系统模拟中将相同的资源分配策略应用于计算资源分配。高级网络基础设施办公室的这一奖项得到了NSF工程理事会内的土木、机械和制造创新部（CMMI）的共同支持，材料研究部（DMR）该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响进行评估来支持审查标准。

The project develops the software infrastructure and its integration with physical infrastructure that is required to bring state-of-the-art data science, machine learning, and artificial Intelligence (AI) tools to novel materials science experiments at a national user facility. The work will create an openly available control package that will enable dynamic experiments informed by modeling in real-time. The developments will make more efficient use of scarce beam-time at national synchrotron user facilities, enabling higher scientific throughput for in-situ experiments probing the mechanical response of materials under load.The effort evaluates the hypothesis that rare material failure events can be predicted from a small number of features that describe evolving local material states using machine learning solutions. The project is focusing on synchrotron x-ray scattering measurements of materials under mechanical load, and specifically integrating new and existing toolsets into a control package capable of dynamic resource allocation for hyper-efficient data collection at the Cornell High Energy Synchrotron Source (CHESS), a national user facility. The project integrates these toolsets to detect precursor signatures through real-time processing of data from user facilities such as CHESS, and suggests resource allocations to facilitate study of early stages of stochastic events in dynamic materials systems. The goal is to develop machine learning (ML) techniques and software infrastructure to inform the best, in a probabilistic sense, allocation of limited detector resources at material testing facilities, to better capture early stages of rare events in materials and the key factors for these events. The effort is also interested in applying the same resource allocation strategies to computational resource allocation in simulations of materials systems. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Division of Civil, Mechanical and Manufacturing Innovation (CMMI) within the NSF Directorate for Engineering, and the Division of Materials Research (DMR) within the Directorate for Mathematical and Physical Sciences.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.