隨著工業及經濟的發展，危險物品於製造的過程中是不可或缺的一環，並且數量持續增加中。由於可及性因素，國內危險物品運送方式仍以公路運輸為主，公路上車流量大，公路旁人口密度較高，倘若不幸發生意外後果甚大，因此如何降低公路危險物品運輸的風險為政府及運送業者重要的課題之一。 前人對於危險物品運送路徑選擇上已有較完整的理論，而本研究將以政府單位進行災後緊急應變之觀點，考量系統總成本最小、場站涵蓋總風險最小、場站涵蓋風險差異最小、場站服務範圍差異最小等目標對公路危險物品救援站建構多目標最佳區位指派。問題的求解則是以C語言自型撰寫免疫演算法配合模糊理論進行求解。 本研究以免疫演算法來求解本研究的區位指派問題，其獨特的記憶機制能保留較佳的抗體反覆演算，改善了傳統遺傳演算法容易因初始解不佳影響求解能力及因交配、突變、而損失優秀親代之問題，而加入混沌方程式尋找初始解的混沌免疫演算法由於搜尋能力更為全域性，因此求解能力較一般免疫演算法較佳。 以自行設計的小型範例配合窮舉法來確認模式與演算法的正確性，並以免疫演算法求解蔡麗敏(2000)的區位問題，測試結果發現本研究演算法求解品質較佳。最後進行大型問題求解，結果顯示系統總成本與場站涵蓋總風險量成反向關係；若提高風險參數中的風險值時，則因整體風險提高而使整體滿意度下降；在不改變路網特性增大路網時，其結果滿意度相近，驗證本模式具有穩定求解能力而不受路網大小影響。由權衡結果可知合乎一般的邏輯性，更可證明本模式求解大型問題之正確性，具有應用價值，可以作為實務單位營運決策之參考。 With the progress of the economics in Taiwan, the dangerous goods production is inevitable function of the manufacturing industries. The transportation of dangerous goods is a critical issue in traffic safety since any contingent incident could have catastrophic impact to local residents and environment.
A series of studies are devoted in the transportation of dangerous goods. Most of the studies are focus on the route selection. There are few studies have addressed the location related issues in this area. The purpose of this study is to develop a methodology for the allocation of the rescue stations in the emergency response network. A nonlinear integer mathematical model is developed for this methodology with multiple objectives, which include minimum system total cost, minimum total risk of service area coverage, minimum risk difference of service area among stations, and minimum difference in service area among stations. To solve this problem, solution procedure is developed based on artificial immune algorithm using programming language C.
The artificial immune algorithm has its unique mechanism of memory to retain the better repeated antibody calculation; would improve the traditional generic algorithm that uncertain initial solution could wrongfully influence its final solving capability; and would improve losing of excellent parental generation due to recombination and mutation. Further, adding of chaos formula would also improve problem solving capability due to its holistic searching capability than that of traditional immune algorithm.
First, we used a small scale problem to test the result of the propose solution procedure with the result from enumeration. The differences of these two solutions are within acceptable range. We then applied the procedure on the location-allocation problem (Tsai, 2000). The results of the proposed procedure in this study have turned out to have a better solution. Subsequently, we tested the procedure in a large scale problem. The results of this large scale problem indicated that the overall system cost has a conflict relationship with overall risk coverage of emergency response network. Overall satisfaction rate of multiple objectives decreased when the value of risk parameters increased, this is a direct result of the increased in overall risk. However, overall satisfaction rate remains the same when network is expanded and parameters are kept the same. The results of these tests indicate that the proposed solution procedure can provide valid and logical outputs for the problem addressed, therefore, can also be used as decision support process for the associated government agencies.