整合TFT-LCD多階多世代多廠之生產策略

Abstract

薄膜液晶顯示器製造流程依序為陣列、組立與模組製程，其中每一個製程包括不同世代之多廠區，形成ㄧ複雜之多階多世代多廠生產環境。 除此之外，整個生產鏈具有許多特有之生管特性，如各製程良率不同、陣列階段不同世代廠之玻璃基板大小不同所造成面板產出量不一、組立製程階段不同世代廠間可互相支援等特性。隨著產品多元化的發展與需求的不確定性，TFT-LCD產業的生產規劃需以企業整體的角度出發，根據需求波動進行各階層各世代廠區最適生產的產品組合之規劃，在必要時更能截長補短相互支援，才能達到TFT-LCD產業所追求最大化利潤之生產鏈目標。 因此，本文針對非穩態的生產環境，因應顧客需求變動的情境，且同時滿足市場對於生產週期時間的要求，提出「最適產品組合評估系統」，以篩選出對各廠區最能維持經營績效之產品組合，提供給管理者作一接單之參考。本系統為遴選最適之產品組合，建構總體產能粗估模組、產量配置模組與產品組合評估模組來加以因應。 首先，在「總體產能粗估模組」中，依據最終產品之目標需求量推算出模組製程階段之毛用料需求量，並且計算各製程階段之產能。接著將上述結果輸入至「產量配置模組」中，建構數學規劃模式求解出每一階段不同世代中各廠區各期最適產品組合與數量。最後，在「產品組合評估模組」中，以資料包絡分析法(Data Envelopment Analysis，DEA)進行整體評估，決定各廠區具有長期競爭優勢之產品組合，以提升整體生產效率。 模擬結果顯示，本文所提之最適產品組合評估系統確實可以提昇整體生產效率。

The manufacturing process of the thin film transistor-liquid crystal display (TFT-LCD) includes Array, Cell assembly, and Module assembly stages. Each process stage contains various generation plants with different production capability such that a multi-echelon environment is comformed. Besides, there are many particular production properties inherited in the whole process. For example, each stage provides different yield rates; in the Array stage, a glass panel from different generation can be scribed into different sizes; and plants of different generation could backup each other. In addition, the production planning in TFT-LCD industry should consider the benefits of the whole enterprise, based on demand fluctuation to achieve optimal product mix at the aggregate planning. Under such a dynamic environment, this thesis proposed an optimal product mix selection system to generate the most efficient product mixes. This system contains three modules: total capacity evaluation module, production planning module, and product mix evaluation module. First, we calculate the gross demand at the module assembly stage and the capacity of each stage. Using the obtained gross demand information, a mixed integer programming model is then proposed for generating the optimal product mix and quantities at each site in each planning period. Finally, the product mix evaluation module uses data envelopment analysis (DEA) to evaluate and select sets of product mixes with the most competitive advantage in the long term. Validated by simulation methods, the proposed mechanism is capable in improving the overall manufacturing efficiency.