GOALI: Digital Technologies for Manufacturing Production Systems

该奖学金的目标是为车间规划和调度问题开发数字技术。PI将与工业合作伙伴Kimberly-Clark合作，调查工厂范围内的计划和调度模型，包括每月批量问题，每周装运协调问题和每日生产调度问题。目前，大多数车间调度是手动或使用电子表格进行的。制造业高层和车间之间的这种脱节意味着人们不了解他们的决策对生产的真正影响。缺乏洞察力通常会导致整个生产系统的效率低下。PI计划通过将企业范围内的规划问题根据规划的时间框架分解为多个层次的模型，以协调的方式解决这些模型，并重新组合解决方案以获得一个集成的计划来解决这个问题。建立从最高层企业范围管理到低层车间调度的协调的主要原因是需要可行的调度决策来支持和改进更广泛的操作和经济目标。通过建立顶层到车间的通信，工业将能够显著提高其生产效率，同时实现对制造环境中的变化和干扰的更快响应。这一研究将推动车间规划和调度优化问题的解决。数字技术将在车间和顶层之间建立联系，以配合需求和入境供应、生产和出境货物的计划。该方法有望推动解决商业制造应用中的车间计划和调度优化问题的边界。为了确保这种更广泛的影响是可实现的，PI将与金佰利密切合作，以验证这些技术。为了解决规划和调度问题，提出了两种新的优化方法：嵌套两界法（NTB）和阈下峰值法（PUT）。NTB方法在利用给定领域（上界）的专家知识的同时，实现了将下界纳入该框架的好处。PUT方法利用经典极值理论，该理论表明，低于高阈值的超出值可以用广义帕累托分布来近似。该阈值包含有关实际下限超过某个值的可能性的关键信息。因此，PUT方法有望在寻找最优解方面非常有效，正如我们的初步工作所证明的那样。PI计划研究NTB和PUT方法，以有效地解决每月批量，每周装运协调和每日调度问题。NTB和PUT方法将极大地促进对大规模规划和调度优化的理解。该项目的研究成果可应用于许多制造生产系统中常见的问题，所产生的方法将适用于工厂范围的优化问题。优化技术和统计理论的结合有望在研究界产生更广泛的影响，因为其他这样的组合可能会从这项工作中受益。本研究涵盖运筹学、工程学、计算机科学和统计学的计算和理论方面。因此，预计该研究将在几个社区产生重大的学术兴趣。

The objective of this Grant Opportunity for Academic Liaison with Industry (GOALI) award is to develop digital technologies for shop floor planning and scheduling issues. Collaborating with the industrial partner, Kimberly-Clark, the PI will investigate plant-wide planning and scheduling models which include monthly lot-sizing issues, weekly shipment coordination issues, and daily production scheduling problems. Currently, most shop-floor scheduling is conducted either manually or using spreadsheets. Such a disconnect between top floor and shop floor in manufacturing means that people do not understand the true impact of their decisions on production. That lack of insight typically leads to significant inefficiencies across the production system. The PI plans to address this issue by decomposing the enterprise-wide planning problem into multiple layered models according to the time frame of planning, solve the models in a coordinated manner, and recompose the solutions to obtain an integrated plan. The main reason for building coordination from top level enterprise-wide management to lower level shop-floor scheduling is the need for feasible scheduling decisions to support and improve the broader operational and economic objectives. By establishing top floor to shop floor communication, industries will be able to significantly improve their production efficiency while achieving a faster response to changes and disturbances in a manufacturing environment. This research will advance the science of solving shop floor planning and scheduling optimization. The digital technologies will establish a connection between shop floor and top floor to align with demand and plan of inbound supply, production and outbound goods. This methodology is expected to push the boundaries of solving shop floor planning and scheduling optimization problems in commercial manufacturing applications. To ensure that this broader impact is realizable, the PI will collaborate closely with Kimberly-Clark to validate these technologies.To solve the planning and scheduling problems, two new optimization methods are planned: the Nested-Two-Bounds (NTB) and the Peak-Under-Threshold (PUT) method. The NTB method realizes the benefits of incorporating a lower bound into this framework while utilizing the expert knowledge of a given domain (upper bounds). The PUT method utilizes classical extreme value theory which suggests that exceedance values below a high threshold value can be approximated by a generalized Pareto distribution. This threshold value contains key information on the likelihood of the actual lower bound exceeding a certain value. Thus, the PUT method is expected to be very efficient in finding optimal solutions as demonstrated in our preliminary work. The PI plans to investigate both NTB and PUT methods for efficiently solving monthly lot-sizing, weekly shipment coordination and daily scheduling problems. The NTB and PUT approaches will significantly advance the understanding of large-scale planning and scheduling optimization in empirical situations. The research findings in the project have applications to problems common in many manufacturing production systems and the resulting methodology will be applicable to plant-wide optimization problems. The combination of optimization technique and statistical theory is expected to have a broader impact in the research community as other such combinations may benefit from this work. This research spans computational and theoretical aspects in operations research, engineering, computer science and statistics. It is thus expected that the research will generate significant academic interest in several communities.