飛機備品需求預測與存貨管理

Abstract

飛機維修備品之需求預測與存貨管理在降低航空營運成本中的重要性日益增加，飛機備品庫存量過多將導致備品屯積，使得存貨成本增加；而因缺料造成飛機營運航班延誤或取消所造成之損失更是驚人。因此，建立準確之備品需求預測系統及存貨管理系統是維修部門長期以來之目標。 飛機維修備品需求大部分屬於間歇型需求與波浪型需求，本論文透過建立倒傳遞類神經網路需求預測模式與支援向量迴歸需求預測模式以預測間歇型需求與波浪型需求兩種備品之需求，並與傳統備品需求預測方法－單指數平滑法及兩種三期加權移動平均法進行比較。結果發現，在整體上，倒傳遞類神經網路需求預測模式與支援向量迴歸需求預測模式比傳統備品需求預測方法有較好的預測結果，並且，以倒傳遞類神經網路需求預測模式的預測結果較好。 最後，根據預測間歇型需求備品及波浪型需求備品之最佳預測結果計算再訂購水準及經濟訂購量以建立飛機維修備品存貨模式。

The importance of demand forecasting and inventory management for aircraft spare parts for reducing the airline operating cost is increasing. It will not only cause the increasing of inventory cost when spare parts are excessive but also great loss result from flight delay or cancellation when those are in shortage. Therefore, the department of maintenance expects to establish the accurate demand forecasting system and inventory management system. The types of aircraft spare parts demand are mostly intermittent demand and lumpy demand. This thesis applies Back-Propagation Neural Network (BPN) and Support Vector Regression (SVR) to forecast the intermittent demand and lumpy demand of aircraft spare parts. In addition, the forecasting results of BPN and SVR are compared to that of traditional methods, Single Exponential Smoothing (SES) and two types of Weighted Moving Average (WMA). From the results, BPN and SVR outperform the traditional methods, particularly, the accuracy of BPN is better. Finally, according to the best result of intermittent demand forecasting and lumpy demand forecasting, the reorder level and economic order quantity are computed to establish the inventory model of aircraft spare parts.