現今晶圓代工廠之晶圓產出大多由生產主管、現場資深工程師或直接操作人員，根據生產排程的需求或臨時的機況，依憑個人經驗安排生產相關進度與作業，通常無法以較科學的方法來作系統性的分析、比較與評估，因此現場資深工程師或直接操作人員之能力及經驗就決定了晶圓產出順利與否，因此利用提高機台的利用率及成本降低，就成為現今晶圓產出作業上的重要議題。本研究模式以台灣一般晶圓代工廠常見之生產流程配合模式，並考量(1)不同廠區(2)負責生產之曝光機台種類及數量的限制(3)機台的利用率百分比%(4)每時段最少需運作生產之機台數量(5)分不同時段運作生產之機台數量，來進行晶圓產出最佳化為研究範圍。 本研究建立一套以整數規劃為架構之數學最佳化模式，協助晶圓代工廠之管理決策者能對曝光機台做統一規劃之編排，期使衍生之編排成本為最小化進而增加收益。當外在環境及限制條件改變時，可隨時依最新條件來調整參數，重新進行後續曝光機台及晶圓產出數量之編排。同時另外針對各種參數的敏感度分析的變化趨勢，主動調整各類相關參數，以因應實際情況變化下曝光機台數量及晶圓產出數量之編排方案。 本研究將基本資料及各限制條件訂定出相對應之各限制式參數，將資料以EXCEL2007軟體方式輸入LINGO12.0套裝軟體求解，即可得晶圓產出所求得之最小成本目標值的最佳化結果。以晶圓代工廠執行客戶需求之晶圓產出作業為範例測試，將測試結果和人工經驗決策兩種方式比較，顯示本研究模式成效明顯較優。故本研究模式適用於不同廠區多種曝光機台搭配不同時段之編排規劃，可做為決策者之重要參考依據。 ;Wafer output in today’s wafer fabrication plants mostly depends on production related schedules arranged by fab manager、senior engineers and operations performed by direct operators onsite based on production scheduling needs or machine conditions, and usually cannot be systematically analyzed, compared and assessed with scientific methods. Therefore, abilities and experiences of senior engineers or direct operators onsite determine whether wafer output will be smooth. In view of this, utilization of improved machine readiness and cost reduction has become an important issue in today’s wafer output operations. The model in this study is the common model of production process alignment in Taiwan’s general wafer fabrication plants. It takes into account (1) different plant sites, (2) limitations on type and number of exposure devices responsible for production, (3) percentage of machine readiness (%), (4) minimum number of machines running for production in each time frame, and (5) numbers of machines running for production in different time frames for wafer output optimization as the scope of research. In this study, a mathematical optimization model is created with integer programming as the framework to help decision makers in wafer fabrication plants with arrangement for unified planning of exposure devices in the hope of minimizing associated arrangement costs and thus increasing revenue. When external environments and constraints change, parameters can be adjusted anytime according to the latest conditions to rearrange subsequent numbers of exposure devices and quantities of wafer output. At the same time, various related parameters can be proactively adjusted according to trends of change in sensitivity analysis of various parameters to cope with arrangement programs for numbers of exposure devices and quantities of wafer output. In this study, corresponding constraint parameters were defined based on basic data and individual constraints, and data were input into the Lingo 12.0 software package via Excel 2007 for problem solving to obtain resultant target values of optimized minimum costs. Testing was performed on customer required wafer output operations run by wafer fabrication plants as examples. A comparison was made between the test results and human experience-based decision making and showed that the model in this study produced clearly better results. Therefore, this model is applicable to arrangement planning for multiple types of exposure devices at different plant sites in different time frames and can serve as an important reference for business decision makers.
