紡織業是我國創造外匯的重要來源。但是,現今,紡織業者的生存受到內部(如勞工成本提高)及外部(如大陸及東南亞等地紡織業的興起)等不利因素之衝擊,而為了提昇國內紡織業的生存及對外的競爭?,則必須從降低成本及提高效率來努力以求達成目標。生產排程扮演著降低成本的重要角色,在過去有關紡織業的生產排程,主要是仰賴有經驗的排程者在紙上進行排程規劃,因採用人工在紙上進行排程作業,所以花費的時間較長,且也較難完整保留排程後的結果。此外,也無法免除許多人為上的疏失,如遺漏訂單,重覆排單等情形。因此造成人工規劃之排程結果往往不盡理想。本研究針對紡織業之針織、染整及成衣產業,在不同的環境及需求,利用基因演算法進行求解。本研究以個案公司織布廠之針織布生產排程、染整廠之後染排程及成衣廠之成衣生產排程,在不同的排程目標下,來驗證基因演算法的效能。由於基因演算法的演化效能,會受到基因參數的影響,本研究將進一步以個案公司為例,在不同排程目標及訂單數及機台數變化,透過一連串的調整及測試後,提供管理者較佳的交配率及突變率的設定值。結果顯示,應用基因演算法僅需1~2秒即可獲得排程結果,但若採傳統人工來進行時卻需2~31分鐘。所以應用基因演算法可以大幅縮減人工排程上不必要的時間浪費,也免除了人工排程漏單及重複排單之現象。除此之外,本研究應用基因演算法與傳統派工法如先進先出及最短處理時間優先法做比較,其結果也證明本研究所使用的基因演算法在求解排程效能及結果上較為良好。
Production scheduling plays an important role in reducing production cost and increasing efficiency in textile industry. In the past, production schedules in textile industry are arranged mostly by senior managers or production controllers, and thus are significantly dependent on their experiences. Unexpected results, however, may produce as a result of the complexity of production situations and some intrinsic constraints of the company and human. It is, therefore, very important to provide useful tools that can help managers to schedule production more easily, conveniently, and flexibly. In this research, we employ genetic algorithm (GA) to solve the production scheduling problem in textile industry, including the production of garment and the knitting and dyeing of fabrics. The application of GA aims to obtain good feasible solutions within a short time. As have been demonstrated by many previous studies, how to set the values of crossover and mutation rates is a key issue when using GA. In this investigation, we try to find the optimal values of crossover and mutation rates for different order amount with different quantity of machines. The results can provide managers with suitable setting values of crossover and mutation rates for versatile production environments. Results from this study show that it just takes 1 or 2 second to get good feasible solutions by using genetic algorithm, while 2~31 minutes will be taken to get the solutions by an experienced scheduler. In addition, experimental results indicate that using GA can obtain better schedules than using the methods of first-in first out (FIFO) and shortest processing time (SPT).