Repair priorities in repairable spare part supply systems: A simulation-optimization approach

Abstract

Industries employing expensive assets maintain extensive repair facility operations and keep spare stocks. This type of logistics system, where both forward and reverse logistics systems are required, in addition to repair facilities, is known as a repairable system. We study a repairable spare part supply system consisting of one repair facility and one stock point, where repairables are kept on the stock to serve expensive capital assets in order to prevent downtime. We set up the objective of our model to minimize the expected total inventory holding costs of spare parts and costs for the downtime of assets over an infinite h orizon. In this study, we particularly analyze the effect of static repair priorities on the expected total cost. To achieve this, we seek optimal values of the repairable spare parts stocks and the assignment of different repairable types into priority classes. We model the repair facility as a multi-server multi-class queue, where failed repairable parts are repaired based on priority classes. It is generally difficult to analyze this type of queuing systems with analytical methods even for a small size problem with the limited number of priority classes and repairable types. Therefore to alleviate this difficulty, we develop a two-stage sequential simulation-optimization a lgorithm. In the first stage, the set of all feasible priority assignments is searched by a Genetic Algorithm (GA) meta-heuristic to find an assignment that achieves the minimum c ost. In the second s tage, a discrete event simulation (DES) is run for the given priority assignment provided by the GA to analyze the multi-class multi-server queueing model. The probability distribution for the number of failed spare parts in the repair facility is obtained as an output of the DES. We use probability distributions to calculate the optimal level of repairable spare part stocks to keep in the inventory. We compare the performance of the simulation-optimization algorithm with a First-Come First-Served (FCFS) service discipline since FCFS reflects the common way of working in practice. The conducted computational experiments show that the proposed approach yields a significant a mount o f t otal c ost r eduction i n some extreme cases reaching up to 90%.