Hierarchical clustering for joint replenishment: an application to spare parts

Spare parts management is crucial in today's production environments to ensure high machine availability. The joint consumption of spare parts due to maintenance activities is often considered to optimise their inventory control performance. The design of inventory replenishment policies for clusters rather than for individual parts to deal with jointly incurred costs is known in the literature as the joint replenishment problem. We propose a two-step framework that hierarchically clusters a set of parts according to several similarity measures and optimises the inventory control policy for each cluster in each hierarchy level. Various hierarchical clustering approaches are considered, and, for each level of the generated hierarchies, two different joint replenishment policies are used, namely the continuous review (Formula presented.) policy and the periodic review (Formula presented.) policy. We conduct an empirical investigation using spare parts data from an automotive company to test the performance of the proposed framework. Through a multi-scenario analysis, we demonstrate that: (i) the clustering of spare parts with joint replenishment leads to a substantial cost reduction in comparison with the standard ABC classification, k-means clustering and single-item replenishments, regardless of the reordering policy and (ii) different clustering approaches lead to very similar performance from a total cost perspective.