A clustering-based repair shop design for repairable spare part supply systems. (November 2018)

• Record Type: Journal Article
• Authors: Turan, Hasan Hüseyin; Sleptchenko, Andrei; Pokharel, Shaligram; ElMekkawy, Tarek Y.
• Is Part Of: Computers & industrial engineering. Volume 125(2018)
• Page Start: 232
• Page End: 244
• Publication Date: 2018-11
• Subjects: Spare part logistics -- Repair shop -- Pooling -- K-median -- Queuing approximation -- Heuristic; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2018.08.032
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: The design problem of a single repair shop in a spare part supply system is studied. The joint problem of pooling, inventory allocation and capacity level designation is solved. A novel sequential heuristic integrating queuing theory and K-median is developed. Approach provides ∼11% and ∼34% cost reductions compared to fully flexible and dedicated designs, respectively. The advantages of clustered design diminish with decreasing cross-training/server cost ratio. Abstract: In this study, we address the design problem of a single repair shop in a repairable multi-item spare part supply system. We propose a sequential solution heuristic to solve the joint problem of resource pooling, inventory allocation, and capacity level designation of the repair shop with stochastic failure and repair time of repairables. The pooling strategies to obtain repair shop clusters/cells are handled by a K-median algorithm by taking into account the repair time and the holding cost of each repairable spare part. We find that the decomposition of the repair shop in sub-systems by clustering reduces the complexity of the problem and enables the use of queue-theoretical approximations to optimize the inventory and capacity levels. The effectiveness of the proposed approach is analyzed with several numerical experiments. The repair shop designs suggested by the approach provide around 10% and 30% cost reductions on an average when compared to fully flexible and totally dedicated designs, respectively. We also explore the impact of several input parameters and different clustering rules on the performance of the methodology and provide managerial insights.