在現今的全球化競爭環境下，越來越多企業從以往的單廠區生產規劃走向跨國多廠區的生產模式，形成一個供應鏈網絡的供需模式，隨之而來的壓力已經不再只是企業與企業的競爭，而是供應鏈與供應鏈間的競爭，所以，在未來面臨的課題是如何獲取整體供應鏈的最佳利潤。 過去的許多文獻提及，供應鏈規劃為一NP-Complete的問題，針對此問題，過去有學者提出利用模擬法以及啟發式演算法來解決之，這些研究往往可以節省許多運算時間，但所求結果皆非最佳解。本研究以製造廠為中心，考量不同特性如產能、物料限制、運輸等，建立適合的數學模型，並且利用LINGO10.0數理規劃軟體進行最佳結果規劃。 在規劃NP-Complete問題時，時常會花費許多的時間，本研究利用分散式帄行計算，將建構的數學模型分割成許多組合，藉由Java RMI(Java Remote Method Invocation )將切割後的組合進行帄行計算，以達降低運算時間之目的。實驗結果顯示，與運用單一處理器進行運算比較，透過分散式帄行計算可以大幅降低運算時間。 Under the global competition, more and more enterprises change their single-plant production planning to multinational-plant production system. As the supply and demand model of supply chain network, the following pressure is no longer the competition between enterprises but supply chains, so the future issue is how to obtain the best profits in the overall supply chain. A lot of past literature mentioned that supply chain planning is an NP-Complete problem. For the problem, the past researcher proposed some methods to solve it such as simulation or developing a heuristic algorithm. Those researches often save much computation time, but the result they found is not optimal. This research is manufactory-centralized, which considers different features such as manufactory capacity, material limit, transportation and so on. Using those features construct appropriate mathmetical model and planning the optimal solution by using mathmetical planning software. It often spends a lot of time on planning NP-Complete problem. This research uses distributed parallel computing. It divided the mathmetical model into a lot of combinations. Then by using Java Remote Method Invocation technique pass the divided combinations to the parallel computers. Parallel computing will achieve the goal ‘saving a lot of time.’ Experimental results demonstrates that it reduce a lot of solving time by parallel computing significantly. The results are compared with using single computer to solve the mathmetical model.
