混合式多目標微粒群演算法在存貨控制模式之應用

Application of Hybrid Multi-Objective Particle Swarm Optimization Algorithm on Inventory Control Models

許晉雄(Chin-Hsiung Hsu) ； 鄒慶士(Ching-Shih Tsou)

《運籌管理評論》 4卷1期 (2010/12) Pp. 23-44

https://doi.org/10.29836/LMR.201012.0003

摘要

存貨管理對於企業來說是極為重要的工作，其目的是如何運用最少的成本維持高度的服務水準，降低缺貨的可能性以滿足顧客對產品的需求。如何在這些衝突目標間做出權衡取捨，便是多目標存貨控制所面臨的一大挑戰。本研究將Agrell（1995）提出的缺貨後補下三目標（s，Q）存貨控制模式延伸至銷售損失的情況下，運用加入區域搜尋與群集機制的混合式多目標微粒群最佳化來求解不同模式的存貨控制問題，並將結果與強健柏拉圖進化式演算法比較，發現混合式多目標微粒群最佳化的非凌越解在三項績效衡量指標上明顯的勝過強健柏拉圖進化式演算法。最後也將不同存貨模式在缺貨後補與銷售損失的狀況下求解的結果進行比較，而在銷售損失的狀況下，廠商會特別擔心因缺貨而造成的銷售損失，因此會更注重庫存的管理，讓實際銷售狀況與庫存能更緊密的配合，訂購適量的存貨，在維持高度服務水準的同時降低存貨成本。

關鍵字

存貨管理 ； 多目標最佳化 ； 混合式多目標微粒群演算法 ； 缺貨後補 ； 銷售損失

並列摘要

Inventory management is an important work to the enterprise. Traditional inventory models only involve single objective which relates to several cost concepts and/or service requirements. Even in its multi-objective formulation, most models have been solved by aggregation methods. Such solutions obtained are unsatisfactory because decision makers try to act through a surrogate variable with incomplete information. So inventory management could be regard as a multi-objective optimization problem (MOP). This work extends Agrell's inventory control problem from backorder to lost sales, and applies hybrid Multi-Objective Particle Swarm Optimization (HMOPSO), which incorporates a local search and clustering method, to an inventory planning problem. The way of multi-objective analysis can determine lot size and safety factor simultaneously under the objectives of minimizing the expected total relevant cost and some measurements about stockout. HMOPSO is compared with Strength Pareto Evolutionary Algorithm (SPEA). The comparative results show that the HMOPSO surpasses the SPEA on the three performance indexes. Finally, backorder is compared to lost sales. On the decision variables, the average safety factor in lost sales is grater than those in backorder, but lot size is smaller than backorder.

並列關鍵字

Inventory management ； Multi-objective optimization ； Hybrid multi-objective ； particle swarm optimization algorithm ； Backorder ； Lost sales