Adaptive Robust Optimization with Endogenous Uncertainty and Active Learning in Smart Manufacturing

该奖项将通过提供严格的方法来管理复杂系统的不确定性，特别是制造系统，为国家的繁荣做出贡献。不确定性在制造业中普遍存在，不确定性下的安全高效运行，以及有效的风险管理，对国家制造业竞争力至关重要。本项目开发的方法支持明确考虑内生不确定性的操作控制，即直接依赖于决策的不确定性。在制造环境中，很大的不确定性是由于实物资产的健康状态，有效管理这些资产构成了一个重大挑战。所得到的方法，虽然在制造系统上得到了验证，但预计将广泛适用于其他复杂系统，如能源系统。该项目集成了一种自适应的、基于集合的鲁棒优化方法，将内生不确定性与主动学习方法结合起来，持续收集和分析数据。鲁棒优化是一种易于处理的方法，通过解决各种类型的内生不确定性来提高运行控制的水平。该项目有望加深我们对不确定性对决策的依赖性的理解，为内生不确定性提供基于集合的鲁棒优化视角，并引入一种新的方法来建模主动学习和优化勘探与开发之间的权衡。该理论的进步将受到智能制造挑战的推动，即在网络化生产环境下的集成生产和维护计划，该PI将包括研究生和本科生参与研究，并将优化方法纳入化工顶点设计课程。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。

This award will contribute to the national prosperity by providing rigorous methods to manage uncertainty in complex systems, with a particular focus on manufacturing systems. Uncertainty is pervasive in manufacturing, and safe and efficient operation under uncertainty, as well as effective risk management, is critically important to the nation's manufacturing competitiveness. This project develops methods to support operational control that explicitly consider endogenous uncertainty, that is, uncertainty that depends directly on decisions. In the manufacturing environment, significant uncertainty is due to the health states of physical assets, and managing these assets effectively constitutes a major challenge. The resulting methods, while validated on manufacturing systems, are expected to be widely applicable to other complex systems, such as energy systems. This project integrates an adaptive, set-based robust optimization approach incorporating endogenous uncertainty with active learning methods that collect and analyze data on an ongoing basis. Robust optimization presents a tractable approach to advance the state of the art in operational control by addressing various types of endogenous uncertainty. The project is expected to deepen our understanding of the dependence of uncertainty on decisions, provide a set-based robust optimization perspective on endogenous uncertainty, and introduce a new approach to modeling active learning and optimizing the trade-off between exploration and exploitation. The theoretical advances will be motivated by challenges in smart manufacturing, namely in integrated production and maintenance planning in networked production environments, The PI will include graduate and undergraduate students in the research, and will incorporate optimization methods in capstone design courses in chemical engineering.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.