UNS: Improved Risk Mitigation Strategies for Industrial Process Scheduling

本课题研究了在问题参数不确定的情况下，过程调度优化（PSO）的风险降低问题。PSO一词指的是在化学加工工业中普遍存在的一系列决策问题，这些问题通常嵌入到监督工厂运营的制造执行系统中。原型设置是指一组有限的可用资源（例如，设备，人员，原材料，公用事业）需要在一个时间范围内进行协调（“计划”），以满足许多生产目标。识别PSO实例的最优解，有时甚至获得单个可行解，通常是一项具有挑战性的任务。这是由于所涉及的各种复合组合复杂性，包括来自工厂拓扑结构（流程图），复杂的生产配方或其他操作限制（市场相关，监管等）的复杂性。PSO的目标通常是利润最大化（“在有限的时间内尽可能多地生产”）或makespan最小化（“尽快生产固定数量”），尽管也可以考虑其他目标，例如平衡资源利用负载或最小化环境足迹。技术目标是开发一个可调鲁棒优化（ARO）框架，用于系统处理PSO中的不确定性。PSO涉及在一定时间范围内协调有限的可用资源，以实现若干生产目标。确定PSO实例的最佳解决方案通常是一项具有挑战性的任务，这一事实进一步复杂化，因为这种生产管理系统考虑输入数据中的不确定性是实际利益，因为不这样做可能导致解决方案不可行或高度次优。本项目采用ARO，这是一种风险缓解方法，扩展了鲁棒优化（RO）范例，旨在根据不确定性集规定的“最坏”情况优化问题。但是，与RO不同的是，RO的结果是静态的、“此时此地”的解决方案，通常过于保守，而ARO的结果是更灵活的——通常更有利可图的——解决方案策略，通过调整决策时已经发生和观察到的不确定参数的实际实现的决策。有效的算法可以降低过程工业的技术和财务风险，在美国制造业基地的竞争力、产品质量和可持续性方面发挥重要作用。提高工艺操作的效率限制了对环境的影响，并促进了职业健康和安全。采用这些创新可以通过提高工艺设备、原材料和人员的利用效率，为个别公司带来切实的利益。这对小公司尤其有用，因为它们不容易开发出适合自己环境的“内部”框架。在制造和企业资源规划领域，也有可能增强软件供应商的产品。潜在的教育效益将是为相关课程生成材料，并创建一个以教育为重点的pso主题软件applet。所有学生将接受生产管理、优化方法和算法、不确定度量化和分析以及科学计算方面的培训。

Gounaris, 1510787This project addresses the mitigation of risk in the context of Process Scheduling Optimization (PSO) in view of uncertainty in problem parameters. The term PSO refers to a family of decision-making problems that are prevalent in the chemical process industries and which are typically embedded in the Manufacturing Execution System supervising a plant's operations. The archetypal setting is the one where a set of limited available resources (e.g., equipment, personnel, raw materials, utilities) needs to be coordinated ("scheduled") along a time horizon so as to meet a number of production goals. Identifying optimal solutions, and sometimes even obtaining a single feasible solution, of a PSO instance is generally a challenging task. This is due to the various compounding combinatorial complexities involved, including complexities stemming from the plant's topology (flowsheet), complicated production recipes, or other operational restrictions (market-related, regulatory, etc.). The objective in PSO is typically the maximization of profit ("produce as much as you can within a limited amount of time") or the minimization of makespan ("produce a fixed amount as soon as possible"), though additional objectives, such as the balancing of resource utilization load or the minimization of environmental footprint, can also be considered.The technical objective is to develop an Adjustable Robust Optimization (ARO) framework for the systematic treatment of uncertainty in PSO. PSO involves the coordination of limited available resources along a time horizon so as to meet a number of production goals. Identifying optimal solutions of a PSO instance is generally a challenging task, further complicated by the fact that it is of practical interest that such production management systems take into account uncertainties in input data, since failure to do so may lead to solutions that are infeasible or highly suboptimal. This project applies ARO, a risk mitigation methodology extending the paradigm of Robust Optimization (RO) that seeks to optimize the problem in view of a "worst-case" scenario, as dictated by an uncertainty set. But unlike RO, which results in a static, "here-and-now" solution that is often overly conservative, ARO results in a more flexible--and generally more profitable--solution policy by adjusting the decisions on the actual realizations of the uncertain parameters that have already occurred and been observed by the time of the decision. Effective algorithms to mitigate technical and financial risk in the process industries can play an important role in the competitiveness, product quality and sustainability of the U.S. manufacturing base. Exploiting efficiencies in process operations limits environmental impact as well as promotes occupational health and safety. Adopting these innovations could provide tangible benefits to individual companies by materializing efficiencies in their utilization of process equipment, raw materials and personnel. This could be particularly useful for small companies, which cannot readily develop an "in-house" framework suitable to their setting. There is also the potential to enhance products of software vendors in the sector of manufacturing and enterprise resource planning. Potential educational benefits will be in generating material for a relevant course and creating an educationally-focused PSO-themed software applet. All students will receive training in production management, optimization methods and algorithms, uncertainty quantification and analysis, and scientific computation.