薄膜液晶顯示器多階多廠排程系統之構建

Abstract

薄膜液晶顯示器製造流程依序為陣列、組立與模組製程，其中每一個製程可能包含多個廠區，形成ㄧ複雜之多階多廠生產環境。由於各製程階段各有其生產特性，如玻璃基板切割不同產品造成面板產出量不同、各製程良率不同及組立製程每兩天固定整備一次等特性。此外，隨著產品多元化的發展，且顧客需求各期不同，使得規劃期間內預估之產品需求與市場需求不逕相同。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「TFT-LCD多階多廠排程系統」，以決定出各製程階段各期各廠區之投料量與庫存量。 對此問題，本系統以滿足需求與降低存貨為原則，建構粗估產能模組、投料規劃模組與產線配置模組三個模組來加以因應。首先，在「粗估產能模組」中，依據最終產品之目標需求量推算出各製程階段內各產品之淨用料需求量，並且評估各製程階段產能是否能負荷。考量各階段之生產週期時間、換線耗用產能與廠際間的產能限制下，將上述結果帶入「投料規劃模組」中，以TDD (Throughput-Dollar-Days, TDD)與IDD (Inventory-Dollar-Days, IDD)最小化為規劃目標，建構數學規劃模式求解出各階段各廠區各期最適投料量與庫存量。最後，在「產線配置模組」中，以減少換線次數與機台產能負荷平衡為原則進行配置，將投料規劃設定之各製程階段各期各廠區之投料量分配至各產線。 實驗結果顯示，依據投料規劃模組求解之各製程階段各期各廠區之投料量，可達到滿足需求與降低存貨之目標。而產線配置模組之結果能使各期各產品之產出數量與投料規劃模組制定之規劃目標相當一致，且使各產品之生產週期時間能滿足市場生產週期時間的要求，並符合等候時間的限制。整體而言，本文提出之「TFT-LCD多階多廠排程系統」能在顧客需求變動的環境下，有效率的安排各製程階段各期各廠區之排程。

The manufacturing process of the thin film transistor-liquid crystal display (TFT-LCD) includes Array, Cell assembly and Module assembly stages. Various generation plants with different manufacturing capabilities may exist in each process stage such that a multi-stage, multi-site environment is composed. In such a complex environment, diverse production characteristics are suffered in each stage. For example, the processing unit and the product yield are different in each stage or in each site, and fixed setup time is required every two days for PI-Print process in Cell assembly, etc. In addition, predictive demand is not the same as the customer needs during planning period because the customer demand is different at each period. Under such a non-steady state environment, this thesis proposes a scheduling system for a multi-stage, multi-site TFT-LCD factory to decide releasing amount and inventory level for each period at each stage and site, and to meet the market request for the production cycle time. With the objective of satisfying demand and reducing inventory, this system contains three modules: capacity evaluation module, release planning module and capacity allocation module. In each period, the capacity evaluation module, first calculates the net demand for each product type at each stage and site, and then evaluates the loading of each stage. Then, in release planning module, with the objective of minimizing Throughput-Dollar-Days (TDD) and Inventory-dollar-days (IDD), a linear programming model is proposed to solve the releasing sizes and inventory level of each product type in each period at each stage and site. Finally, with considerations of reducing setup times and balancing load of machines of the workstation, the caoacity allocation module allocates machines to each product type, in order to achieve the production target set by release planning module. Experimental result shows, the release plant set by the release planning module supports the achievement of fulfilling demand and reducing inventory. Then the capacity allocation module makes the production quantity of each product type to match with the release plan. It also makes the cycle time of each product type to satisfy the market request, and the waiting time before the the work is under control. Overall, the proposed system can rapidly arrange the schedule for each period, each stage and each site effectively under the demand variate environment.