AMHS派貨邏輯利用類神經網路結合遺傳演算法實現效能動態最佳化研究; AMHS dispatch logic efficiency dynamic optimization by neural networks and genetic algorithms

NCUIR > Engineering College > Executive Master of Mechanical Engineering > Electronic Thesis & Dissertation > Item 987654321/43500

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/43500

Title:	AMHS派貨邏輯利用類神經網路結合遺傳演算法實現效能動態最佳化研究;AMHS dispatch logic efficiency dynamic optimization by neural networks and genetic algorithms
Authors:	溫盛夫;Sheng-Fu Wen
Contributors:	機械工程研究所碩士在職專班
Keywords:	AMHS;最佳化;排程邏輯;神經網路;遺傳演算;AMHS;optimization;dispatch logic;neural networks;genetic algorithms
Date:	2010-06-28
Issue Date:	2010-12-08 13:42:50 (UTC+8)
Publisher:	國立中央大學
Abstract:	TFT-LCD面板產業近年來為台灣發展速度最快的產業之一，其產業特色為資本密集且進入與退出均具有障礙，面板廠商須不斷投資擴充產能以維持自身競爭能力，持續降低成本、提升生產效率為面板廠商成功之關鍵因素。目前工廠均導入CIM建構的全自動化生產環境，AMHS為CIM的核心之ㄧ，執行生產過程中玻璃基板的運送與儲存作業。 AMHS控制軟體派送命令先後順序的邏輯，對於整個系統的效能有相當的影響力，在本研究中以個案公司G5的In-Line式Stocker Controller為例，其特色為同一時間Controller會收到許多的搬送命令，而Stocker一次只執行一筆命令，因此，藉由調整命令派送的先後順序，就可以影響區域生產時間，進而提升產能而不用花費到硬體改造的費用。在案例公司的目前設計中，提出利用命令特徵因子加上各特徵因子權重，將其彼此相乘後的權值作為命令派送順序的依據，以兼顧效能提升與製程Q-time需求。但是，由於製程機台的生產條件會一直不斷的改變，而固定的權重未必符合所有的製程設備之生產條件，導致AMHS無法發揮最大的搬送效能。因此，本研究利用類神經網路模擬出系統搬送效能的預測模型，再搭配遺傳演算法，當生產條件改變的時候，即時的計算出適合的命令特徵因子權重，如此一來便可以使AMHS一直發揮較大的搬送效能，進而提升產能。 TFT-LCD industry in Taiwan in recent years is one of the fastest growing industries，the industry is characterized by capital-intensive and have barriers to entry and exit are，panel manufacturers have been investing to expand capacity in order to maintain their competitive edge, continue to reduce costs，improve production efficiency for the panel makers, the key success factors. Construction of the factory are now fully automated CIM into the production environment, AMHS is the core of Your CIM implementation process of production of glass substrate transport and storage operations. AMHS control software logic of sending the command sequence，the performance of the whole system has considerable influence，in this study to the case company's G5's In-Line Type Stocker Controller Example，controller of its features for the same time will receive a lot of to shift commands，and the implementation of a command stocker time only，so by adjusting the sequence order delivery，can affect the production time and increase productivity without having to spend to transform the hardware costs. Current research and design, typically made use of by various characteristics of the command characteristic factor weighting factor, to multiply each other right after the delivery of the order value as the basis for the command. However, due to the production of process equipment conditions will constantly change, and fixed weights may not meet all the conditions of production process equipment, resulting in AMHS can not get maximum performance to move. Therefore, this study is to use neural network to simulate the performance of the system to shift the forecast model, with the combination of genetic algorithm, the time when production conditions change, real-time order characteristics calculated for the weight factor, this would allow AMHS has been to shift to play a greater effectiveness, improving productivity.
Appears in Collections:	[Executive Master of Mechanical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	1248	View/Open

社群 sharing

Loading...