Collaborative Research: RETRO: Toward Safe and Smart Operations via REal-Time Risk-based Optimization

过程安全管理（PSM）旨在防止在异常过程条件下发生危险事件，通常依靠被动保护机制（如减压阀）。工业数字化和智能制造的持续趋势给PSM带来了新的挑战，PSM拥有更加复杂、动态和集成的工艺工厂。因此，迫切需要解开安全关键决策与基于系统的实时操作之间的联系，从而主动减少化学过程的安全损失。为了实现这一目标，本研究项目旨在通过整合在线过程安全监测、基于模型的异常预测和预后风险控制来实现范式转变。统一的理论、概念框架和软件原型将基于对过程和安全系统动力学的基本理解而开发。这些方法的发展将在氢燃料电池实验原型上进行演示，这将作为下一代智能PSM系统设计的具体指南，用于广泛的制造业，以避免美国每年因过程事故造成的数十亿美元的经济、社会和环境损失。该项目的研究成果将被纳入为本科生、研究生和高中生量身定制的课程材料、在线学习模块和研讨会。该项目还将利用西弗吉尼亚大学、德克萨斯农工大学现有的STEM项目和区域联盟，招募不同群体的弱势群体和第一代学生。该项目将开发一种在线过程安全管理策略，将适合目的的风险控制的离线计算与实时优化相结合，同时考虑过程安全性、可操作性和经济性的相互作用和权衡。计划研究活动的主要支柱具有：(i)统计动态风险建模，它明确考虑了安全关键过程变量的非线性物理相互作用；（ii）基于风险的多参数模型预测控制，提供风险控制可调、过程运行路径有界的双层预测安全管理设计；（iii）容错过程安全控制，理论上保证了模型逼近、实时测量和状态估计中对动态误差的鲁棒性；（iv）故障预测设计、控制和实时优化，通过单一混合整数动态规划公式同时解决这些多个决策层。该项目的一个关键创新在于对这个多时间尺度决策问题采用了一种新颖的基于多参数优化的表示，这导致了一种时间可扩展和自适应的过程安全管理策略，允许在具有快速、缓慢或混合动态的所有类型的过程系统中高效、敏捷和灵活的应用。硅方法的发展将应用于实验室规模的聚合物电解质膜氢燃料电池的网络物理原型系统，以实现在市场需求变化下安全、健康和可持续运行的最佳需求驱动电力生产。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。

Process safety management (PSM) aims to prevent the occurrence of hazardous events under abnormal process conditions, typically relying on passive protection mechanisms (e.g., pressure relief valves). The ongoing trends toward industrial digitalization and smart manufacturing have posed new challenges to PSM with substantially more complex, dynamic, and integrated process plants. Thus, there is an imperative need to unravel the link between safety-critical decision making and systems-based real-time operation which can proactively reduce chemical process safety losses. Toward this goal, this research project aims to create a paradigm shift by integrating online process safety monitoring, model-based abnormality prediction, and prognostic risk control. A unified theory, conceptual framework, and software prototype will be developed based on a fundamental understanding of process and safety system dynamics. These methodological developments will be demonstrated on a hydrogen fuel cell experimental prototype, which will serve as a concrete guide for next-generation smart PSM system designs for a broad range of manufacturing industries to circumvent the annual billion-dollar financial, societal, and environmental losses across the US due to process incidents. The project findings will be incorporated to course materials, online learning modules, and workshops tailored to undergraduate, graduate, and high school students. This project also will be used to recruit a diverse group of underrepresented and first-generation students by leveraging existing STEM programs at West Virginia University, Texas A&M University, and regional alliances.This project will develop an online process safety management strategy coupling offline computation of fit-for-purpose risk control with real-time optimization to simultaneously account for the interactions and tradeoffs of process safety, operability, and economics. The major pillars of planned research activities feature: (i) Statistical dynamic risk modeling, which explicitly considers the nonlinear physics-based interactions of safety-critical process variables; (ii) Risk-based multi-parametric model predictive control, which provides a dual-layer predictive safety management design with adjustable risk control and bounded process operation path; (iii) Error-tolerant process safety control, which offers theoretically guaranteed robustness against dynamic errors in model approximation, real-time measurement, and state estimation; and (iv) Fault-prognostic design, control, and real-time optimization, which addresses these multiple decision layers in a simultaneous manner via a single mixed-integer dynamic programming formulation. A key innovation of this project lies in a novel multi-parametric optimization-based representation to this multi-time-scale decision making problem, which results in a temporally scalable and self-adapting process safety management strategy allowing for efficient, agile, and flexible application in all types of process systems with fast, slow, or hybrid dynamics. The in silico methodological developments will be applied to a cyber-physical prototype system of lab-scale polymer electrolyte membrane hydrogen fuel cell to achieve optimal demand-driven power production with safe, healthy, and sustainable operations under market demand changes.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.