隨著全球化市場的競爭壓力下，許多企業於各地區設有配銷中心及製造廠，傳統中企業僅追求本身的淨利極大化的目標已嫌不足，已經轉變為整體供應鏈的總淨利最大化，所面臨的生產活動，亦從過去的單階單廠區生產演變成為供應鏈中多階多廠區生產模式。因此，在目前的許多探討議題中，有關供應鏈已聚焦在整體實務面的操作，藉由供應鏈全面性的考量以創造整體價值最大化來取代傳統的目標。 過去的文獻中，經常提及排程規劃為NP-Complete的問題，因此，本研究提出符合記憶體模組產業的供應網絡生產規劃模式，考慮物料供給限制、原物料採購價格、各廠區產能限制、生產及運輸前置時間等特性，於多階多廠供應網絡環境中，建構出一個非線性整數規劃 (Integer Non-Linear Programming, INLP)的數學模型，並配合LINGO 10.0 extended求解多限制式，多目標的排程問題。 針對數學模型的NP-Complete問題，本研究建構出分散式平行系統進行數值運算，而在分散式平行計算的技術方面，採用Java遠程方法調用（Java Remote Method Invocation, JavaRMI）和推演出來的演算邏輯來實現高速平行計算，達到計算時間上的改善。經由實驗結果顯示，本研究方法相較於數學模型求解的作法，可在短時間內達到最佳的規劃結果，所求得的結果可以讓規劃人員在短時間內做出決策。 With competive pressure of globalization, many companies have distribution centers and manufacturing plants at various locations. Tranditional enterprises only persue their goal of profit maximization had been not enough. They have been transformed into the whole supply chain to maximize profit. The production activities has also altered from the previous single-stage and single-plant production into a multi-stage supply chain and multi-plant production. Therefore, many of the current issues in the supply chain have been focused on the whole practical side. Through a comprehensive consideration of the supply chain to maximize the overall value to replace the traditional goal. The past literature often referred to schedule planning is a NP-Complete problem. Therefore, this study found the memory module industry supply network planning model of production, considering material supply constraints, the purchase price of raw materials, the plant capacity constraints, transportation lead time, the multi-stage, multi-factory supply network environment and other features. This paper has to propose a non-linear integer programming mathematical model and uses LINGO 10.0 extended to solve multi-constraint and multi-objective scheduling problem. For NP-Complete problem of Mathematical model, this study constructs a distributed parallel system for numerical computing. In distributed parallel computing technologies, using Java Remote Method Invocation and the deduced algorithm to achieve high-speed parallel computing and improve computing time. The experimental results show that our method compared with solving mathematical model can achieve the good planning results in a short time. The results obtained allow planners to make decisions in a short time.
