綠色混合型循環物流模式之路徑最佳化

Abstract

近年來，綠色物流開始被越來越多學者所研究，本研究希望將傳統常應用於汽車業之循環物流模式，結合綠色物流之概念，發展出能夠同時降低環境及經濟成本的綠色混合型循環取貨物流模式。 循環物流模式相較於傳統直運(Direct line haul)系統，此模式強調的少批量、多批次的配送或取貨，這可以讓每台貨車在旅行時都能維持較高的乘載率，因為頻繁的補貨，也使場站倉庫不用維持太高的存貨水準，連帶降低存貨成本，所以許多汽車製造業者皆使用循環物流模式來達成及時(JIT)的概念，但過去研究對於循環物流模式之模式建構，皆未能完整呈現其特性，像是多批次的取貨就代表循環物流須面對多期訂單的問題，而且一般汽車組裝工廠的周圍除了有上游衛星零件工廠，當汽車組裝完成後也需要銷售至顧客節點，所以混合型取配是需要被考慮的，綜觀以上原因，本研究決定將循環物流模式以結合”含有時窗限制之混合型車輛多次使用車輛途程問題(VRP with pickup and delivery and multiple use of vehicles and time windows)”與”存貨路徑問題(Inventory routing problem; IRP)”兩種概念的方式呈現。 本研究模式主要在決定路線規劃及取配排程，但傳統路徑規劃問題即屬於NP-HARD的問題，所以本研究採用全域型啟發式解法中的基因演算法做為核心求解工具，並設計出能將基因演算法有效應用於本研究問題之基因編碼方式、初始解產生方式、交配及突變機制，以證明本研究模式能以基因演算法求解。 最後本研究參考國際範例設計一適當規模之情境範例模擬求解，由範例求解結果中可發現本研究模式可依工廠組裝速率安排取配貨排程，並於一次求解產生多期路徑規劃結果，由於本研究模式的目標式設計是屬於多成本項目，所以各成本項目間的權衡情形則左右結果的變化，於是在最後本研究設計兩個情境，來證明當目標間的權重關係改變後，模式求得之結果也將不同。

In recent years, more and more scholars began to study green logistics. This research hopes to combine the green logistics with milk run model and develops the green mixed milk run model. Comparing to the traditional direct line haul system, milk run model emphasizes the small quantities and high frequencies pickup and delivery. This allows each vehicle can maintain a higher loading rate. Because of the frequent replenishment, the warehouse of depot does not have to keep a high inventory level and it brings the inventory cost reduction. That’s why many of the automobile manufacturers achieve the concept of just in time by using milk run model. However, past studies are unable to show the feature of milk run completely when building this model, such as frequent pickup, it represents that milk run model needs to face the problem of multi-period demand. In general, the satellite plants which supply the assembly parts will be set around the automotive assembly plant. Besides, when the car is assembled, it needs to sell to the customer, so mixed delivery and pickup needs to be considered. According to the reasons mentioned above, this study decided to present the milk run model in the way of combing two concepts, VRPPDMTW and IRP. The model of this study mainly determines the routes and the schedule of pickup/delivery. However, the VRP is the NP-Hard problem, so this study uses genetic algorithm as core solver and this study designs those GA mechanisms to prove this study establishes a model allowing the use of genetic algorithm to solve. Finally, this study refers to the international example to design the data input. From the solution, we can find the mixed milk run model can arrange the pickup and delivery schedule according to the assembly speed rate and generate the multi-period routes planning results. Due to the construction of objective function of this study belongs to multiple cost items. So the trade-off effect of each cost items will manipulate the results. To conclude, this study designs two scenarios to prove that when the weight among cost items changed, the model results will be different.