CAREER: Optimal Engineering Decision-making Under Uncertainties for Enhanced Structural Life-cycle

该学院早期职业发展计划(CALEAR)奖的目标是促进在不确定性和风险下的最佳工程决策方面的创新，重点放在生命周期分析和结构应用上，以应对与老化的基础设施环境相关的持续和新出现的科学和社会挑战。因此，本项目建议建立一个综合的结构生命周期分析、设计、维护、翻新和恢复框架，以期降低基础设施生命周期成本，优化社会对基础设施的投资，建立更安全的结构，并协助提高国家安全和经济竞争力。除了结构工程应用是该项目的主要关注点外，这些发现还可以用于涉及智能、自动化、自主的决策支持框架的大量应用。该项目还将培养下一代多样化的工程师和科学家，他们除了需要其他技能外，还需要提高计算能力。计划中的活动影响广泛的受众，包括学校的学生和教师、本科生和研究生、研究人员、教职员工和实践工程师。在该项目中，建议从随机最优控制和强化学习的角度来处理工程决策，包括并完全集成基于预测物理的随机模型和不确定的生命周期观测。由于决策在工程问题中的重要作用，这一观点导致了结构系统的基于条件的评估规则和基于性能分析的增强的新意义。传统的结构设计、分析、改造、恢复和维护方法也被重新考虑，从传统的静态优化问题转变为对不断变化的结构的完整的终身控制过程。中心随机控制部分基于完全和部分可观测的马尔可夫决策过程、异步动态规划和深度强化学习技术。要在存在风险和不确定性的情况下对老化的结构和基础设施系统采取这种决策方法，还需要几个理论上的追求和进展。在这个项目中，将调查主要挑战的答案，例如维度和历史的诅咒，以及针对多目标和决策者的解决方案，包括非线性过滤、结构可靠性更新和广义脆弱性函数等。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。

The objective of this Faculty Early Career Development Program (CAREER) award is to advance innovation in optimal engineering decision-making under uncertainty and risks, with a focus on life-cycle analysis and structural applications, in order to address ongoing and emerging scientific and societal challenges relevant to the aging infrastructure environment. An integrated structural life-cycle analysis, design, maintenance, retrofit and recovery framework is thus suggested in this project with the aim to reduce infrastructure life-cycle costs, optimize societal investments to infrastructure, lead to safer structures and assist in improving national security and economic competitiveness. Apart from structural engineering applications that is the main focus of the project, the findings can be also used for a plethora of applications involving intelligent, automated, autonomous decision support frameworks. This project will also educate the next diverse generation of engineers and scientists, who, in addition to other skills, need increased computational competence. The planned activities impact a wide audience including school students and teachers, undergraduate and graduate students, researchers, faculty and practicing engineers.In this project, engineering decision-making is suggested to be approached from a stochastic optimal control and reinforcement learning perspective, embracing and fully integrating predictive physics-based stochastic models and uncertain life-cycle observations. Owing to the vital role of decision-making in engineering problems, this viewpoint leads to condition-based estimation rules for structural systems and an enhanced new meaning of performance-based analysis. The traditional approach to structural design, analysis, retrofit, recovery and maintenance is also reconsidered, from a conventional static optimization problem to an integrated lifelong controlling process of ever changing structures. The central stochastic control component is based on fully and Partially Observable Markov Decision Processes, asynchronous dynamic programming and deep reinforcement learning techniques. Several theoretical pursuits and advances are needed to enable such a decision-making approach to aging structures and infrastructure systems in the presence of risks and uncertainties. In this project, answers to main challenges will be investigated, such as the curses of dimensionality and history, together with solutions for multiple objectives and decision makers, incorporation of nonlinear filtering, structural reliability updating, and generalized fragility functions, among others.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.