NSF Young Investigator: Analysis and Simulation of Multi-Stage Inventory Systems

本研究主要针对有限生产能力下的多级库存系统进行分析与仿真。重点和共同的主题是使用新技术来限定和近似这些系统的性能。这些方法包括利用分析罕见事件概率的技术，在一个适当设计的系统中，这些事件对应于库存缺货、最后一次销售、长响应时间等。要研究的问题的不同子类包括有能力的多阶段系统、多产品之间的容量分配、部件共性存在下的库存水平、多产品的循环调度以及罕见事件的有效模拟。在有能力的多阶段系统中，研究重点是研究许多行业的装配和供应链的特征，目的是开发设计和控制工具来协调各个阶段，以满足外部需求的服务水平并限制库存成本。产能分配问题的研究将寻求将解决方案扩展到涉及多产品订单库存环境下的产能分配问题。库存水平研究的目的是为多种产品所需的部件制定库存水平问题的解决方案。本研究还将考虑循环计划的研究，其中产品在固定的计划中生产，每次访问产品将其库存恢复到预定的目标水平，这取决于中间的需求。在复杂的库存系统中，诸如缺货、销售损失等事件的罕见性使得使用模拟作为此类系统的评估工具非常困难。本研究也将著重于发展基于重要抽样的方差缩减技术，以提高模拟分析的效率。本研究的预期结果包括：为有容量库存系统的控制提供渐近最优基本库存水平的特征、产品间渐近最优产能分配的识别，以及更好地分析、理解和设计多产品的周期计划。此外，还将开发通过模拟更好地分析库存系统的工具。这将使通过仿真技术分析复杂的库存系统变得更容易、更有吸引力、计算效率更高。

9457189 Glasserman This research is focused on the analysis and simulation of multi-stage inventory systems with limited production capacity. The emphasis and common theme is on the use of new techniques for bounding and approximating the performance of such systems. The methodologies involve draw upon techniques for analyzing probabilities of rare events, which, in a properly designed system, correspond to inventory stockouts, last sales, long response times, etc. The different subclasses of problems to be studied include capacitated multi-stage systems, capacity allocation among multiple products, stocking levels in the presence of component commonality, cyclic schedules for multiple products, and efficient simulation of rare events. In capacitated multi-stage systems, the research emphasis is on the study of features of the assembly and supply chains of many industries with the objective of developing design and control tools to coordinate the various stages to meet service levels for external demands and to limit inventory costs. Research in capacity allocation problems will seek to extend the solution to capacity allocation problems that involve multiple products in make-to-order stock environment. Research in stocking levels is aimed at developing solution to the problem of setting stock levels for components that are required by multiple products. The study of cyclic schedules in which products are made in a fixed schedule and each visit to a product restores its inventory to a preset target level that depends on intervening demand will also be considered in this research. The rarity of events such as stockouts, lost sales, etc. in complex inventory systems make the use of simulation as an evaluation tool of such systems very difficult. This research will also focus on the development of variance reduction techniques, based on importance sampling, to increase the efficiency of simulation analysis. Expected outcomes of this research inclu de the characterization of asymptotically optimal base-stock levels for control of capacitated inventory systems, identification of asymptotically optimal capacity allocation among products, and better analysis, understanding, and design of cyclic schedules for multiple products. Furthermore, tools to better analyze inventory systems through simulation will be developed. This will make it easier, more attractive, and computationally more efficient to analyze complex inventory systems through simulation technique.