Due to the fierce market competition pressure brought by globalization, the paradigm of developing corporate strategy has evolved from cooperate-centric to supply chain cooperation based thought. Therefore, an avoidable challenge that enterprises face is to simultaneously reduce operation cost and raise service level in order to survive continuously. Because of the urgency of arms readiness for war, weapons should be maintained in an operable status at any time. However, most weapons, even to the spare parts, are extremely expensive, whereas purchasing budget for arms is rather tight. Therefore, how to find the appropriate quantity of spare parts for repairable items in a least cost way to reach higher availability rate is an important decision to military units, especially to those components critical to the arms system. The main objective of this research is to optimally deploy spare parts for a three-echelon logistics system with repairable items such that the total backorder delay days can be minimized but obeying a given budget constraint. The recoverable items mentioned in the thesis consist of two categories: assembly item and component. A probabilistic nonlinear integer mathematical model to the problem has been constructed initially. We then develop some heuristic algorithms based on evolution strategy to effectively and efficiently solve the intractable model. In the end, we justify the performance of our heuristics through experimental studies, and moreover the impacts on availability caused by changing the number of echelons and product structure are also addressed through examples.