廠商為了市場競爭會不斷推出新產品,在此同時仍然必須對既有產品提供一定期間的售後維修服務,隨著產品愈來愈多,零件庫存的規模也會跟著愈大,造成存貨管理上的負擔,因此如何規劃一個良好的庫存量在實務上是值得關注的。 本研究針對機車銷售與機車維修時對備用零件產生的需求,考量零件失效率因素,假設備用零件之失效型態吻合浴缸形失效率曲線,運用韋伯分配之特性配適浴缸曲線各階段失效率函數,依此建立備用零件需求預測模式,並使用基因演算法幫助模式求取全域最佳解。然後利用個案公司實際資料進行實證,以驗證備用零件需求預測之準確度。 從實證研究可以發現,本研究提出之預測模式之預測效果較個案公司現行使用之方法來得優良,且在面對零件需求有較大幅度的變動情況時,本研究提出之方法較個案公司現行使用之時間序列方法更能夠準確預測備用零件未來的需求數量。
Manufacturers have been rolling out new products constantly for the market competition, and at the same time they must provide the maintenance service of the products after selling in a period of time. As many as the products, the scales of spare parts inventories are larger, and definitely become a burden of inventory management. Hence, how to scheme out a method well for stock is deserved to be concerned in practice. This research is directed to the demand of spare parts brought by selling and mending of motorcycles. Considering the factor of the hazard rate of parts, and supposing that the hazard function of spare parts conform to the bathtub curve, the study applies the property of Weibull distribution to fit each stage of the hazard function into bathtub curve. By this means we construct a demand forecasting model of spare parts and use genetic algorithm to help find out global optimum. Then we use the real data from case company to carry on the examination to verify the accuracy of demand forecasting of the parts. From the empirical study we can find that the result of the forecasting model presented in the research is better than current methods used by case company. And while there is a wider range of demand variations, the forecasting model can make a much more exact future demand forecast than time series models used by case company.