藉多動態庫存管理設計閉迴路運籌

Abstract

傳統的供應鏈中，通常補貨模式是由推式管理的方式來運作，搭配預測，來盡量滿足客戶需求。然而需求的預測十分不精準，在原物料、商品種類過多的情況下，不適當的庫存管理和生產活動必將造成公司的損失，並且侵蝕其獲利水準。 本研究以動態庫存管理 ( Dynamic Buffer Management, DBM )為基礎，發展成多動態多庫存管理 (Dynamic Multi-Buffer Management, DMBM )，以 ABC存貨分析對不同的貨物和原物料分別管理，並給予不同的服務水準以及調整變數，再利用帝國競爭演算法求得最佳的管理變數，並搭配拉格朗日鬆弛法輔以偽梯度法 (Lagrangian Relaxation with Surrogate Sub-Gradient Method)，讓生產到存貨都獲得最佳控制。 希望能推廣DBM，讓企業了解DBM能有助於不同產業公司應對市場需求的變動，並且避免產生大量損失，提升公司應變能力，提高獲利能力。

Forecasting and push system are implemented to help enterprises do the replenishment in traditional supply chain. However, the forecasting is not always accurate. Once the market trend turns down, company has to face the fact of much higher inventory which can significantly affect to profit margin. The more types of products and material the company has, the more difficult to manage. Based on the Dynamic Buffer Management (DBM), this paper tries to study another way study of Demand-Pull way, Dynamic Multi-Buffer Management (DMBM), to help enterprise to manage multi-products. With ABC classification to manage different type of products and material, Imperialist Competition Algorithm (ICA) is used to find factors of service level and adjustment factors for DMBM. Moreover, Lagrangian Relaxation with Surrogate Sub-Gradient Method is used to make inventory management and manufacturing system coordinated. The proposed method has been used in a company, and the result is satisfactory.