TFT-LCD生產鏈物料與產能規劃之研究

Material and Capacity Planning for TFT-LCD Production Chain

陳子立(Tzu-Li Chen)

指導教授 ： 林則孟

國立清華大學/工學院/工業工程與工程管理學系/博士(2009年)

摘要

本研究第一目的為提出一套整合階層式生產規劃架構。隨著高科技的進步，新興的薄膜電晶體液晶顯示器(thin film transistor liquid crystal display；TFT-LCD)取代CRT顯示器，也帶動了TFT-LCD製造業的崛起。TFT-LCD面板產業是由三大生產製程階段所組成，分別為列陣(Array)製程、組立(Cell)製程和模組(Module)製程，每階段製程由多個工廠所組成，有如一小型垂直整合的供應鏈，或稱為生產鏈結構。生產鏈上具有許多特殊的生產製程特性與限制條件，而現行乃透過各生產規劃人員憑其經驗以試算表方式規劃出其所負責的規劃區塊，再透過規劃人員間彼此互相協調來完成可執行的生產計畫，不僅費時、複雜且反應慢，而造成了生產計畫與排程上的特殊問題。本研究提出的整合階層式生產規劃架構具有三大特色來整合所有規劃問題與解決上述問題：(1)階層式規劃(hierarchical planning)、(2)結合推拉式規劃流程(combined push-pull planning process) 、(3)模組化規劃方法(modular planning approach)。最後將回顧所有TFT-LCD生產規劃問題的相關文獻來確認出關鍵物料規劃(critical material planning)與產能規劃(capacity planning)的重要性。
本研究第二目的將探討在整合階層式生產規劃架構中最重要的關鍵物料規劃。模組製程是一個以物料導向為主的組裝生產系統，此段製程將前製程生產之面板與部分關鍵物料(例如：驅動IC、印刷電路板與背光模組)，根據顧客不同之需求與偏好組裝成不同類型的成品。此外，模組製程存在有許多獨特特性，例如替代物料清單(alternative bill of material; ABOM)、顧客對於ABOM喜好程度(customer preference for ABOM)與物料採購比例(purchase quantity ratio)等，因此傳統的只針對單一物料清單做來計算物料需求量的物料需求規劃方法(material requirement planning; MRP)已無法適用於此產業的物料規劃問題。因此，本研究利用網路圖表達方式針對單期與多期關鍵物料規劃問題提出了非線性規劃模式來考量上述特性與限制來決定每個供應商最佳的採購量。最後，透過實際TFT-LCD案例來驗證本研究所提出關鍵物料規劃的模式，同時透過敏感度分析來探討顧客對於ABOM之喜好程度與採購比例此兩大因素對於規劃結果的影響與管理意義。
本研究最後目的將探討在整合階層式生產規劃架構中另外一個重要的產能規劃問題。由於下列三大趨勢，而使的產能規劃問題變成了TFT-LCD產業日益重要：(1)多階層、多世代與多廠區共存的生產鏈結構、(2)複雜的產品階層結構而造成了TFT-LCD面板可生產產品種類繁多且廣泛、(3)成長快速與劇烈變動的產品需求。在這些有限產能供給、特殊生產結構與快速成長需求的特性下，TFT-LCD產業必須面臨了因供給與需求不平衡而造成的產能規劃議題，包含了(1)產能分配決策：決定每個生產廠區的最大化利潤產品族組合以及每個廠區對於每個產品的最適生產數量；(2) 產能擴充決策：決定哪種類型的產能(各廠區總產能或各廠區對各產品族產能)在哪些廠區需要擴充或是要設立新的生產廠區來增加產能、該類型的產能有多少量需要被擴充或增加以及要透過購買多少機台或是附屬資源來擴充產能。本研究特別關注的是要如何透過附屬資源來擴充各廠區對各產品族產能。本研究建構了針對多廠區產能規劃的混整數規劃模式，由於該混整數規劃模式具有高度的計算複雜度，因此同時也提出了有效率的陰影價格為基啟發演算法來求解之，透過數值實驗結果可以發現陰影價格為基啟發演算法能夠在合理的時間內快速求得近似最佳解，且明顯優於傳統分枝界限法(branch-and-bound)。此外，由於需求預測都是由公司的業務人員獲市場分析員依據經驗法則來產生，因此造成產品需求預測具有不準確性，本研究進一步提出了以情境為基礎二階段隨機規劃模式來考慮在具有需求不確定性現象下產生穩健的產能規劃結果，並同樣的提出有效率的期望陰影價格分解演算法來求解大型問題，最後透過實際TFT-LCD案例來驗證隨機產能規劃在不確性環境下的穩健性。

關鍵字

薄膜電晶體液晶顯示器 ； 階層式規劃架構 ； 關鍵物料規劃 ； 產能規劃 ； 陰影價格 ； 混整數規劃 ； 隨機規劃

並列摘要

The first purpose of this dissertation is to present a hierarchical planning framework for a thin film transistor–liquid crystal display (TFT-LCD) production chain in an assembly-to-order (ATO) environment. The TFT-LCD manufacturing process, or the TFT-LCD production chain, comprises three sequential steps, namely: the Array, Cell and Module processes. Many special characteristics and constraints, such as product grade constraints, site-eligibility constraints and key material availability constraints, are inherited in this production chain. The globally distributed nature of production planning and scheduling activities of TFT-LCD production leads to an urgent need for an integrated planning and scheduling framework to balance supply and demand problems. There are three main features of this integrated framework: (1) hierarchical planning (2) combined push-pull planning process (3) modular planning approach. A review of OR methods applied in TFT-LCD planning and scheduling for various planning modules are given and the importance of critical material planning and capacity planning is identified.
The second purpose of this dissertation is to study the critical material planning which is one of the most important planning modules in the proposed hierarchical planning framework. The module process is the material-oriented and discrete production environment that combines many critical materials such as integrated circuit (IC), printed circuit board (PCB), and backlight (BL) into the final TFT-LCD products according to different customer requirements and preferences. There exist many specific characteristics such as alternative bill of material (ABOM), customer preference for ABOM, and purchase quantity ratio occurring in the TFT-LCD industry. The traditional MRP mechanism that calculates the purchase quantity by exploding a fixed BOM structure is difficult to apply to CMP. In response to this challenge, this study formulates a single-period and multi-period CMP problem as nonlinear programming models represented by the network graph to optimize the purchase quantities of all suppliers under the existence of specific characteristics. Finally, a numerical example from a real TFT-LCD manufacturer is illustrated to clarify the multi-period CMP model and the managerial implications of the customer preference factors and the purchase quantity ratio are also correspondingly discussed.
The final purpose of this dissertation is to study the capacity planning problem in the hierarchical planning framework. Capacity planning is critical to TFT-LCD industry due to its complex product hierarchy and increasing product types; the coexistence of multiple generations of manufacturing technologies in a multi-site production environment; and the rapidly growing market demands. One key purpose of capacity planning is to simultaneously determine the profitable “product mix” and “production quantities” of each product group across various generation sites in a particular period and the optimal “capacity expansion quantity” of specific product groups at a certain site through the acquisition of new auxiliary tools. This study proposes a mixed integer linear programming model for the multi-site capacity planning. A shadow-price based heuristic is developed to find a near-optimal solution. Our computational study indicates that the proposed method performs better and is more robust than the conventional branch-and- bound (B&B) algorithm. In addition, since the demand forecasts faced by TFT-LCD manufacturers are usually inaccurate, we develop a scenario-based two-stage stochastic programming model to seek a robust capacity allocation and expansion policy against demand uncertainties. An expected shadow-price based decomposition is constructed to obtain a near-optimal solution in efficient manner. According to our numerical study using real industry cases, the stochastic capacity planning model significantly improve system robustness and outperforms traditional deterministic model under demand uncertainties.

並列關鍵字

TFT-LCD manufacturing ； Hierarchical planning framework ； Critical material planning ； Capacity planning ； Shadow price ； Mixed integer linear programming ； Stochastic programming

參考文獻

[105] Park, B. C., Park, E. S., Choi, B. K., Kim, B. H. and Lee, J. H., “SIMULATION BASED PLANNING AND SCHEDULING SYSTEM FOR TFT-LCD FAB”, Proceedings of the 2008 Winter Simulation Conference, 2008.

[78] Lin, J. T., Chen, T. L. and Lin, Y. T., “A Hierarchical Planning and Scheduling Framework for TFT-LCD Production Chain”, 2006 IEEE International Conference on Service Operations and Logistics, and Informatics, pp.711-716, 2006a.

[85] Lin, J. T., Chen, T. L. and Lin, Y. T., “Critical material planning for TFT-LCD production industry”, International Journal of Production Economics, to appear, 2009a.

[82] Lin, J. T., Chen, T. L. and Chen, W. J., “Capacity and product mix planning problem for TFT Array multi-plant”, Journal of the Chinese Institute of Industrial Engineers, 24(6), pp.489-504, 2007a.

[76] Liang, Y. Y. and Chou, Y. C., “Option-based Capacity Planning for Semiconductor Manufacturing”, 2003 IEEE International Symposium on Semiconductor Manufacturing, pp.77-80, 2003.