近年來國際間的生產導向以顧客需求為主要目標,所以企業生產模式轉為多樣少量的模式,導致生產者必須經常性的更換製程方法與機器模組,因此探討整備時間(set-up time)納入生產排程問題之研究也日益增加。故本論文主要解決的目標是,單機生產排程之延遲時間最小化與整備時間最小化雙目標之問題研究。希望藉由霍普菲爾-坦克神經網路來解決此雙目標之排程性的問題,在小工件模擬排程實驗時使用整數線型規劃做為最佳解之驗證,經由模擬後發現小工件時神經網路與整數線型規劃所獲得最佳解之誤差並不大,最大誤差約為2.6%,但解題時間卻比LINGO少很多倍。大工件則採神經網路與Apparent Tardiness Cost and Setup(ATCS)啟發式法則作解題品質的比較,發現神經網路所獲得的解的品質比ATCS啟發解來的更好。
The major tendency orientation of the production system has been changed to meet the customer demand in the recent years. It results in frequent changeovers in the production process. Consequently, many enterprises have switched systems from large- quantity production into the small-size large-variety mode. The research of scheduling concerning the setup time is getting more and more. According to this situation, a Hopfield-Tank Neural Network (HTNN) was applied to a single machine scheduling problem with minimizing the tardiness and the setup time is addressed in this thesis. When the problem size is small, the result of the HTNN is not far away from the optimal solutions acquired by using LINGO. The maximal error approximates to 2.6% and the computation of HTNN takes much less time. When the problem size is large, the HTNN was compared with the Apparent Tardiness Cost and Setup (ATCS) rule. The experimental result indicated that HTNN is superior to the ATCS in the quality of solution.