GOALI: Inventory Management in Assemble-to-Order Systems: Analysis, Policies, and Asymptotic Optimality

该学术与工业联络资助机会（GOALI）奖的目标是开发具有一般结构和参数值的组装订单制造系统库存管理的优化框架。其基本方法是利用随机程序将一个难以解决的控制问题转化为一个简单得多的可解优化问题。随机规划的最优解设定了长期平均期望总库存成本的可证明下界，将用于激励管理动态系统的一系列库存控制策略。这些策略将通过扩散尺度上的渐近最优性准则进行检验。换句话说，当组件的交货时间增加，从而使系统成本增加时，建议的政策将使库存成本与其下限之间的百分比差为零。这些政策在不同参数区域的表现也将通过数值实验、模拟和案例研究进行经验评估，包括与其他方法的比较以及在最不利条件下的压力测试。将开发涉及近似的算法，以使该方法能够在工业规模上实现。如果成功，本研究的结果将有助于制造企业实现库存成本的显著节省，从而提高广泛采用的组装到订单制造策略的效率。由于这些政策对交货时间较长的系统特别有效，它们将减轻制造商与其海外供应商之间长时间运输延迟对库存系统的负面影响，从而导致有利于内包的成本效益计算的变化，这将有助于美国制造业的振兴。随机规划和渐近分析的结合使用也为库存管理优化提供了一个有希望的新方法，这涉及到许多众所周知的难题。

The objective of this Grant Opportunity for Academic Liaison with Industry (GOALI) award is to develop an optimization framework for inventory management in Assemble-to-Order manufacturing systems with general structures and parameter values. The basic approach is to employ a stochastic program to transform an intractable control problem into a much simpler solvable optimization problem. The optimal solution of the stochastic program, which sets a provable lower bound on the long-run average expected total inventory cost, will be used to inspire a family of inventory control policies for managing dynamic systems. These policies will be tested by the criterion of asymptotic optimality on the diffusion scale. In other words, when component lead times increase, thus making the system more costly, the proposed policies will drive the percentage difference between the inventory cost and its lower bound to zero. Performance of these policies in various parameter regions will also be evaluated empirically by numerical experiments, simulations, and case studies, including comparisons with other approaches and stress-tests under the least favorable conditions. Algorithms involving approximations will be developed to allow the approach to be implementable on an industrial scale.If successful, the results of this research will help manufacturing companies to achieve significant savings in inventory cost, thus improving the efficiency of the widely-adopted Assemble-to-Order manufacturing strategy. Because the policies are particularly effective for systems with long lead times, they will mitigate negative impacts on the inventory system of long transportation delays between manufacturers and their oversea suppliers and thus lead to changes in the cost-benefit calculation that favor insourcing, which will contribute to the US manufacturing sector's revitalization. The combined use of stochastic programs and asymptotic analysis also points to a promising new approach for inventory management optimization, which involves many notoriously hard problems.