應用蟻群演算法於整合生產與配送之排程問題

Abstract

面對日益激烈的競爭環境以及消費市場特性的改變，愈來愈多企業選擇採用訂單式生產(make-to-order)與直接銷售(direct-order)的營運模式。在此模式之下，企業被迫減少存貨但仍須維持快速回應顧客需求之能力，此種現象導致生產與配送兩階段的互動更為緊密，提昇了整合這兩階段研究的實用性。此外，顧客服務水準與成本亦為此營運模式下的兩大主要考量，但彼此間存在互償(trade-off)關係，為達到整體系統之最佳化則必須在兩者之間取得平衡。因此本研究針對整合生產與配送之排程問題進行探討，以非等效平行機台模擬產品製造階段的生產環境，而成品配送階段則模擬為車輛途程問題(vehicle routing problem, VRP)，並將此問題視為一個以整體最佳化為目標的系統問題。而本研究的績效衡量指標則同時考量顧客服務水準與成本，以訂單完成時間作為顧客服務水準的具體表現，而成本部分則以配送成本作為衡量。本研究應用蟻群演算法(ant colony optimization)於求解此整合生產與配送之排程問題，並加入動態規劃演算法進行配送批次分配與費洛蒙回饋機制，以提升演算法的求解品質與兩階段間的整合程度，由測試結果顯示此設計對於演算法的求解品質確有助益。而本研究亦針對整合生產與配送進行排程安排之效益以整合考量與順序式考量進行比較，其測試結果平均相對差距約為18.04%，顯示整合考量在整合生產與配送之排程問題中顯著優於順序式考量，而此測試結果驗證了整合兩階段進行排程之價值與重要性，並可作為相關產業實務上的參考依據，讓業者能在合理的時間內作出較佳的決策以同時達到滿足顧客服務水準與降低成本的目標，進而達到最大化整體利潤的目的。

More and more enterprises have chosen to use make-to-order or direct-order business models in order to be competitive in the demanding market. In such business models, enterprises are forced to reduce their inventory but still have to respond quickly to customer’s requirements. The reduction of inventory results in closer interaction between production and distribution activities and thus increases the usefulness of an integrated model. This study considers a problem in which orders are first processed by a set of unrelated parallel machines and then distributed to the corresponding customers directly by vehicles with limited capacity without intermediate inventory. This objective is to find a joint schedule of production and distribution such that an objective function which takes into account both customer service level and total distribution cost is optimized. A heuristic algorithm based on ant colony optimization (ACO) is developed to solve such a problem. Computational experiments illustrate that the algorithm developed is capable of generating near-optimal solutions in reasonable computational times. Furthermore, we also investigate the benefits of using the integrated model relative to a sequential model where production and distribution operations are considered separately and sequentially. The test results show the value of integration and the importance of integrated model for an integrated production and distribution scheduling problem.