Supply Chain Decision Making Framework Considering Uncertainty

COVID-19加剧了本已脆弱的许多日常用品、食品供应、特种化学品、燃料、电子元件和药品的供应链网络。大流行考验了全球供应链的灵活性和复原力，主要国际公司经历了人员短缺和其他意想不到的业务中断。这些挑战激发了对未来重构供应链的愿景，其特点是灵活性和对突发变化的快速适应性。为了实现这一目标，在制定控制单个制造单元的策略时，必须考虑决策的时间范围、企业的复杂性和企业的目标，在制造计划中安排任务以满足产品需求，并在企业管理层进行规划以确保原材料在整个供应链中可用。该研究项目将开发以企业结构现代化为目标的方法和框架，以考虑不确定性，并更紧密地整合净化气体、制药制造和塑料回收业务的不同层次的供应链。这将允许更有效地利用资源，避免不必要的浪费，并补偿预期和意外事件，以避免供应链的崩溃。这项研究将在特拉华大学进行，并将为研究生和本科生在这个跨学科领域的教育提供资金。长期以来，PI一直在促进妇女和少数民族在其工作中的代表性不足。这项研究的结果将转化为软件工具，对行业做出更好的供应链决策有用。以前集成过程模型、调度方法、计划问题和供应链优化的尝试集中在小规模的基准问题上，并且基于使用直觉将较低级别问题的全尺寸表示合并到较高级别，或者使用较低级别的数学简化来促进集成。在本研究项目中，提出了一种整合决策过程的新方法，该方法利用企业中通常可用的大量信息，并以计算可处理的方式生成考虑不确定性的决策策略。PI在计划、调度和控制问题的集成以及可行性分析和不确定性量化方面具有广泛的专业知识。提出通过定义各决策阶段优化的数学模型，然后识别依赖于较低级问题的约束，可以实现集成和优化。在其他因素中，低级问题的可行性被确定为必须纳入高级决策的基本信息。这些约束通常形式复杂，如果不增加优化问题的维度并生成棘手的公式，就无法解释这些约束。因此，将演示数据驱动模型可以用于获得相互依赖约束的更简单形式。这些模型将利用企业提供的数据，利用大数据分析和物联网的潜力来创建。不确定性的影响是供应链中最优决策绩效的关键。在这项研究计划中，将制定策略，以最大限度地减少运营和中断不确定性带来的威胁。最终优化问题的计算复杂度将通过使用分解方法和滚动水平策略来控制。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。

COVID-19 has exacerbated an already fragile supply chain network for many of today’s everyday goods, food supplies, specialty chemicals, fuels, electronics components, and pharmaceuticals. The pandemic tested the flexibility and resilience of global supply chains as major international corporations experienced personnel shortages and other unexpected disruptions to their operations. These challenges motivate a vision for a restructured supply chain of the future, characterized by flexibility and fast adaptability to abrupt changes. To accomplish this, the decision-making time horizon, enterprise complexity, and objectives of the enterprise must be considered when developing a strategy for controlling individual manufacturing units, scheduling tasks in a manufacturing plan to meet product demand, and planning at the enterprise management level to ensure that raw materials are available throughout the supply chain. This research program will develop methodologies and frameworks that target the modernization of the enterprise structure to account for uncertainty and more closely integrate different levels of the supply chain for purified gas and pharmaceutical manufacturing and plastics recycling operations. This will allow for more efficient use of resources, avoid unnecessary waste, and compensate for expected and unexpected events to avoid the breakdown of the supply chain. The research will take place at the University of Delaware and will provide the funds to educate graduate and undergraduate students in this interdisciplinary domain. The PI has a long history of promoting women and underrepresenting minorities in their work. Results from this research will be translated into software tools useful to industry for making better supply chain decisions. Previous attempts to integrate process models, scheduling methods, planning problems and supply chain optimization focused on small scale benchmark problems and were based either on using intuition to incorporate full-scale representations of the lower-level problems into higher levels or used mathematical simplifications of the lower levels to facilitate integration. In this research program, a new approach to integrating decision-making processes is proposed that leverages the large amount of information that typically is available in enterprises and generates decision-making strategies that account for uncertainty in a computationally tractable manner. The PI has extensive expertise in the integration of planning, scheduling, and control problems, as well as in the areas of feasibility analysis and uncertainty quantification. It is proposed that integration and optimization can be achieved by first defining mathematical models for the optimization of each decision-making stage, and then identifying constraints that are dependent on lower-level problems. Among other factors, the feasibility of the lower-level problems is identified as essential information that must be incorporated in the higher-level decision. These constraints are usually complex in form and cannot be accounted for without increasing the dimensionality of the optimization problem and generating intractable formulations. Therefore, it will be demonstrated that data-driven models can be used to obtain simpler forms for the interdependent constraints. These models will be created with the data available from the enterprise, leveraging the potential of big-data analytics and the internet of things. The effects of uncertainty are key to the performance of optimal decisions in the supply chain. In this research program, strategies to minimize the threats posed by operational and disruption uncertainties will be developed. Computational complexity of the final optimization problem will be controlled through the use of decomposition approaches and a rolling horizon strategy.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.