應用兩階段解碼之遺傳演算法於 TFT-LCD陣列製造之動態排程問題

隨著需求的快速變化和智慧製造的彈性決策，TFT-LCD廠面臨巨大的客戶群及多樣的產品別。因此，在維持產品品質情況下如何提升生產力成為一個重要的議題。由於黃光區為瓶頸站點，本研究以黃光區的排程為目標，同時考慮產品的來到時間。為了解決來到時間的不確定性，本研究發展了兩階段解碼之遺傳演算法(TDGA)並結合滾動式的排程，以解決在黃光區的動態排程問題。透過染色體解碼設計，TDGA也可避免重工與產能不均的問題。為了驗證其效度，本研究透過台灣某實際的TFT-LCD 廠資料作為實證研究，分析TDGA與有左移機制之遺傳演算法的表現。實驗結果顯示，TDGA可以縮短在工單之間的閒置時間以提升機台利用率到99%以上，進而獲得較高品質的解。因此在所有情境之下，TDGA的表現皆比左移機制之遺傳演算法來得好。

關鍵字

訂單式生產；遺傳演算法； TFT-LCD ；黃光區；動態排程；滾動式排程

並列摘要

Due to the rapid change of the market and decision flexibilities of intelligent manufacturing, TFT-LCD industries are facing the challenges of a huge number of customers and different kinds of products. Therefore, it is important to enhance productivity as well as remain product quality. Because photolithography stage is the bottleneck, this study focuses on photolithography scheduling which considers job arrivals. To deal with the uncertainty of arrival time, this study develops Two-phase Decoding Genetic Algorithm (TDGA) combined with rolling strategy for dynamic scheduling in photolithography stage under complex restrictions. TDGA can also avoid the reworked problem and load unbalancing through the design of chromosome. For validation, TDGA is also compared with GA which has the left-shift mechanism through empirical data from a leading TFT-LCD industry in Taiwan. The experimental result shows that TDGA can shorten the idle time between jobs. It can obtain a high quality solution with 99% machine utilization. Thus, TDGA performances better than GA in all scenarios.

並列關鍵字

Job shop scheduling ； Genetic algorithms ； Thin-film transistor-liquid crystal display (TFT-LCD) ； Photolithography ； Dynamic scheduling ； Rolling Scheduling

參考文獻

Lee, J., Kang, M., Park, G. L., & Shin, S. Y., (2007), “Design of a reliable real-time scheduling policy for dual-channel networks,” Journal of Information Science and Engineering, Vol. 23, No. 5, pp. 1407-1419.

辛宛珉(2012)，半導體廠黃光區派工紫式決策架構，清華大學工業工程與工程管理學系學位論文。

Wu, S. D., Erkoc, M., & Karabuk, S., (2005), “Managing capacity in the high-tech industry: A review of literature, ” The Engineering Economist, Vol. 50, No. 2, pp. 125-158.

Pinedo M., (1995), “Scheduling theory, algorithms, and systems,” IIE Transactions, Vol. 28, No. 8, pp. 695-698.

A.A. Fredericks, (1986), “Performance analysis modeling for manufacturing lines,” AT&T Technical Journal, Vol. 65, No. 4, pp. 25-34

國際替代計量

應用兩階段解碼之遺傳演算法於 TFT-LCD陣列製造之動態排程問題

全文下載

主題瀏覽