备品备件
计算机科学
工程类
制造工程
制造业务
生产(经济)
可靠性工程
过程(计算)
系统工程
工程制图
运营管理
生产计划
作者
Marco Caserta,Luca D'Angelo
标识
DOI:10.1080/00207543.2026.2655026
摘要
This paper is motivated by a real-world challenge faced by Amazon in managing spare parts inventory across its extensive logistics network. Traditional single-echelon inventory systems often lead to high capital immobilisation and inconsistent service levels, creating inefficiencies in large-scale operations. To address this, we develop and implement a two-echelon inventory optimisation approach, structuring clusters of sites around central hubs to pool demand risk, reduce lead times, and enhance service reliability. The resulting mixed integer formulation is solved using a column generation algorithm that overcomes the computational limitations of standard MILP solvers. Through an extensive case study involving 24 Amazon sites across five clusters, we demonstrate significant operational and financial benefits, including a 14–27% reduction in immobilised working capital and a 0.4–3.2% improvement in service levels. These improvements correspond to projected capital savings on the order of tens of millions of dollars when extrapolated to Amazon's global network, along with reduced downtime costs. Beyond its methodological contributions, this study delivers actionable managerial insights by providing a scalable decision-support tool for firms seeking to optimise distributed inventory systems. The column generation algorithm enables real-world implementation for large-scale high-volume supply chains. By bridging optimisation theory with industry implementation, this research offers a concrete roadmap for improving supply chain resilience and cost efficiency in multi-echelon inventory management.
科研通智能强力驱动
Strongly Powered by AbleSci AI