薄膜電晶體陣列廠生產規劃系統之構建

Abstract

TFT-LCD之製程由主要分為三部份：薄膜電晶體陣列(TFT Array)段製程、液晶面板組立(LC Cell Assembly)段製程和電路模組組立(Module Assembly)段製程。其中，TFT-Array廠其設備昂貴，且製程時間長，使TFT-Array成為三段之瓶頸段製程。 在生產規劃過程中，生產週期時間扮演一個相當重要的角色，其決定了訂單之交期時間，同時也是生產排程規劃的重要依據。在TFT-Array廠中由於玻璃基板體積龐大，常以自動化搬運系統進行運料，當系統搬運車數設計不當時，搬運時間將佔工件生產週期時間十分大的比例。 為使系統能流暢運作，及避免瓶頸工作站發生飢餓及搬運成本之浪費，本文首先估算系統內搬運車最適車數。在此搬運車數下，考量搬運時間之因素，吾人利用二種生產週期時間迴歸模式：「個別工作站等候時間迴歸式」和「整體迴歸式」，分別求算生產週期時間，並以此求算結果進行主生產排程規劃及細部排程規劃。「個別工作站等候時間迴歸式」為工件於各工作站前之等候時間迴歸式；「整體迴歸式」則為工件由投入至產出之整體生產週期時間之迴歸式。 驗證結果顯示，生產週期時間迴歸模式中，「整體迴歸式」之預估結果誤差小於「個別工作站等候時間迴歸式」估算結果之誤差。整體而言，搬運系統最適車數，結合「整體迴歸式」求得之生產週期時間，將此兩者運用於主生產排程和細部排程之預定產出量、投料時點與產出時點設定，其成果良好，可作為液晶面板組立廠之生產規劃之依據。

The manufacturing process of TFT-LCD mainly consists of three stages: TFT Array, LC Cell Assembly, and Module Assembly. The TFT-Array stage is the bottleneck among the whole TFT-LCD process flow, because this stage needs the most expensive equipments. Thus, this thesis designs the production planning system for the TFT-Array Factory. Cycle time plays an important role in production planning. It is the basis of production scheduling and due date setting. In the TFT-Array factory, the volume of glasses is quite huge, so that if the number of vehicle in automation transportion system is assigned inappropriately, the transportion time will take large portation of cycle time. In order to make the production process smooth and to avoid bottleneck starving and to waste the transportion cost, first of all, we estimated the number of automatic guided vehicle (AGV) needed. Under the number of AGV that we estimated, the cycle time incluing transportion time is estimated by two kinds of cycle time regression models, regression model for workstation queue time (RMWQT) and regression model for product cycle time (RMPCT). RMWQT refers to lots’ waiting time in front of each workstation, and RMPCT refers to cycle time of lots passing through the whole process. Then, given the throughput target and product mix, the master production schedule (MPS) module and the detail schedule (DS) module are conducted based on the cycle time derived from the regression model. Simulation experiment reveals that the estimated error of RMPCT is lower than that of RMWQT. With consideration of smooth transportation, this thesis uses RMPCT model to set the throughput, release time, and output time in MPS and DS. Simulation result shows that the performance of MPS and DS based on RMPCT can be achieved and could be used by production planning of cell stage.