Study on Application of Genetic Algorithms to Optimization of Steel Mill Factory By-product Transport and Logistics

[[abstract]]本論文著重於製造業廠內物流，Tompkins et al 研究指出搬運成本佔製造業總生產成本約3成左右，尤其以重工業，因能源成本高漲及供需失衡導致削價競爭，造成利潤大幅下滑，企業經營面臨嚴重考驗。隨著德國工業4 0的提出，產業逐漸進入第四次革命，加上資訊、雲端運算、大數據及人工智慧技術的飛躍成長，軟體技術扮演著關鍵角色。作者多年從事產業界廠內物流資源規劃與協調整合工作，感於實務重要決策規劃往往依賴專家經驗及既有知識，缺乏系統思考與科技應用，造成運輸物流成本的浪費，故提出應用基因演算法於鋼廠內副產物運輸物流之最佳化研究，希望藉由系統分析、資訊導入及人工智慧的優化，解決傳統決策的瓶頸，達到廠內運輸物流決策最佳化之目的。 問題定義上經由廠區路線資訊化、車輛途程系統化、並考慮搬運需求頻率，建構鋼廠內副產物運輸物流模型。基因演算法應用上，提出改良型可變長度染色體編碼技術，有效解決各分區運輸不同趟次問題，並用雙重基因演算方式，先優化整區總運輸時間與分區運輸時間標準差，再以此為基礎進行各區運輸時間最短化，同時處理最短運輸時間與最小標準差優化等問題。並依建構模型求解，進行各種分區結果比較及驗證。 提案方法實際應用C公司現階段廠內物流供需，進行分區優化，實驗結果顯示本方法所得分區結果，總運輸時間略優於現有專家排程作業，但公平性遠優於現有結果。而且決策生成速度為數十分鐘相較於專家的數日，有突破性的提升。後續研究將就即時運輸需求整體規劃，發展智慧型即時動態廠內運輸派車系統技術。[[abstract]]This study is focused on in-plant logistics in manufacturing Owing to rising energy costs and supply-demand imbalance lead to price competition resulting in a sharp decline in profits the business management is facing a severe test especially heavy industry In practice the important decision-making in C company has rely on expert experience and existing knowledge the lack of systematic thinking and technology applications resulting in a waste of logistics costs Therefore the study on application of genetic algorithms to optimization of steel mill factory by-product transport and logistics was presented It is hoped to solve the bottleneck of traditional decision-making so as to achieve the goal of optimizing transportation logistics decision-making by artificial intelligence In the definition of the problem the model of the steel mill factory by-products transport and logistics is constructed through the in-plant route information vehicle routing systematization and consideration of the transport demand frequency The modified variable length chromosome coding technique and bi-level genetic algorithm are used to effectively solve the problem of different zoning transportation on genetic algorithm application Experimental results show that the total transport time is slightly better than the existing result