以決策為基礎之智能製造系統架構

A framework for decision-based intelligent manufacturing system

周哲維(Che-Wei Chou)

指導教授 ： 簡禎富

國立清華大學/工學院/工業工程與工程管理學系/博士(2016年)

摘要

隨著消費性產品市場需求模式的改變，消費者追求物美價廉且客製化的產品。傳統的製造策略主要是追求經濟規模，面對市場需求變動快與產品生命週期短的改變，傳統生產系統的集中和分級控制的結構設計，已無法滿足複雜且變動迅速的製造環境，並有效支援因產品多樣化與複雜化而衍生的生產製造問題。
因應產品需求變化與複雜產品線衍生的生產規劃、排程規劃、製造資源配置等問題，本研究發展以決策為基礎之智能製造系統架構，建構具備穩定與效率的製造系統模組，以及具備決策能力的代理人系統。其中，第一部分是階層式的製造系統，追求穩定與效率，以達生產最佳化。從客戶實際訂單與潛在需求開始，透過訂單管理模組彙整，決定總需求量，作為後續生產規劃模組、多目標排程規劃模組的依據；進而展開至製造現場的資源配置與派工，以及機台的智能整合感控功能，即時取得產品生產資訊與機台狀況，掌握生產進度。第二部分則是具備可適性、靈活性、協同自主性與決策能力的代理人製造系統。透過代理人的自主協同溝通，在突發事件發生時，回饋最新的即時生產資訊；並提出解決方案，賦予智能製造系統決策能力，處理干擾變異事件，即時將系統回歸正常運作。
本研究利用三個實證研究驗證效度：第一個為TFT-LCD模組組裝多目標排程問題，考量實際面板產業的後段模組組裝線生產限制，利用多目標基因演算法找出滿足多目標的最佳妥協排程；第二個是TFT-LCD模組組裝的塗膠製程變異造成的生產擾動案例。在異常發生時，啟動決策代理人的異常處理機制。透過代理人之間的資訊交換，即時了解問題狀況，並提出解決方案給予決策者參考，選出最適方案作為異常處理的生產決策。第三個為半導體8吋晶圓代工廠的人工晶圓搬運系統之人員配置問題，透過本研究發展的人員配置模式與搬運途程規劃，提出最佳人力配置，提升人員生產力。

關鍵字

製造決策 ； 人工晶圓搬運系統 ； 智能製造系統 ； 多目標排程規劃

並列摘要

Consumers pursue affordable and customized products and services with the changes of consumer products market. Manufacturing industry is facing rapid changes in market demand, short product life cycle, fast technology migration and high capital investment, resulting in difficult to accurately predict demand, lengthen production lead time, and low production yield and other operational challenges.Traditional manufacturing strategy is focusing on scale economies. Facing the changes of various marketing demands and short product life cycles, traditional production system has been unable to meet the complex and rapid changes in the manufacturing environment, and can not effectively deal with derived manufacturing problems.
In order to cope with production planning, scheduling, and manufacturing allocation problems that caused by various product demands and complex product lines. This study aims to develop a framework for decision-based intelligent manufacturing system that including production system modules and decision agent system to pursue quickly response, high productivity, flexible and robust production.
First, the proposed hierarchical manufacturing system is pursuing stability and efficiency to achieve production optimization. The order management module will accumulate all customers' forecast and orders to integrate total demands that as a follow-up demand input for production planning module and multiobjective scheduling module. Further, the manufacturing resource allocation module and cyber-physical system function module will provide promptly manufacturing and machine information to control production status.
Second, the proposed decision-based manufacturing agent has the capability of adaptive, flexibility, coordination, decision-making to provide promptly production while abnormal events arise. The decision-based manufacturing agent could coordinate and communicate with others function modules and agents to integrate information and provide possible solutions for those disturb events. Thus, the decision maker could consult the possible solutions and do the final decision to access those disturb events, and let the production system return to normal operation.
Three empirical cases are demonstrated to validate the proposed decision-based intelligent manufacturing system in this study. First, a multiobjective scheduling problem in TFT-LCD module assembly process is proposed that consider manufacturing conditions in shop floor. The multiobjective scheduling module used multi-objective hybrid genetic algorithm (MO-HGA) to solve multiple and conflicting production objectives, minimizing makespan, minimizing the weighted number of tardy jobs, and minimizing the total machine setup time, that are directly related to productivity and customer satisfaction. Second, a manufacturing disturb case in TFT-LCD module assembly process. When an abnormal event happened in tuffy glue coating process, decision agent was trigged that communicated with other agents to exchange information and understand the latest status. Then, decision agent could provide the possible solutions for decision makers for final judge. Third, a manpower allocation and route planning problem for manual material handling system (MMHS) in a 200mm wafer fab is proposed. The manufacturing resource allocation module used a novel route planning approach to utilize the routes that reduce the technician traveling distance and transportation time. And it also used manpower loading evaluation model is developed for determining the appropriate number of technicians that all improve the MMHS efficiency and productivity.

並列關鍵字

Manufacturing decision ； Manual material handling system ； Intelligent manufacturing system ； Multiobjective scheduling planning

參考文獻

簡禎富、林國義、許鉅秉、吳政鴻 (2016)，台灣生產與作業管理之相關期刊文獻回顧與前瞻：從工業3.0到工業3.5，管理學報，33卷1期，頁87-103。

周哲維、簡禎富 (2016)，需求不確定性之產能擴充決策分析－TFT-LCD產業之個案研究，管理與系統，23卷1期，頁1-29。

Holdren, J., E. Lander, S. Jackson, and E. Schmidt. (2011). Report to the President on Ensuring American Leadership in Advanced Manufacturing. Washington, DC: Executive Office of the President President’s Council of Advisors on Science and Technology.

Agrawal, G. K. and S. S. Heragu (2006), “A survey of automated material handling systems in 300-mm SemiconductorFabs,” IEEE Transactions on Semiconductor Manufacturing, vol. 19, no. 1, pp. 112-120.

Akanle, O. M. and D. Z. Zhang (2008), “Agent-based model for optimising supply-chain configurations,” International Journal of Production Economics, vol. 115, no. 2, pp. 444-460.