TFT-LCD製程重工決策之研究

Abstract

在TFT-LCD產業中，無論是Array製程所製作的薄膜電晶體基板(TFT基板)或是CF製程所製作的彩色濾光片基板(CF基板)，通常皆被設計為可切割成數個面板，切割數依顯示器應用產品需求不同而有所不同。因為新製程導入階段，製程良率會偏低，所以一般而言，TFT-LCD廠會在Array製程及CF製程階段中，於某些作業與作業中間加設檢測站，依據基板上所包含的良好面板個數，來決定此基板是否可繼續加工或是逕行報廢再加以重工，但如何訂定良好面板個數的門檻值，也就是如何訂定重工決策(rework decision)，是目前TFT-LCD廠經常面臨的一個決策問題。有鑑於過去研究的不足，以及目前工業界多採用直覺性的思考來訂定各檢測站的重工決策，並未考量到整體績效，且並非站在整個生產系統的觀點來訂定，所以本研究建構一數學模式，考量範圍包含整個TFT-LCD廠生產流程中Array製程、CF製程及Cell製程的特性，以產品的獲利(profit)最大化為目標，利用基因演算法，針對不同的產品，訂定個別產品在Array製程及CF製程階段中，各檢測站的最佳重工決策。

In the manufacturing of TFT-LCD, the array process is for producing TFT-plates, the CF (color filter) process is for CF-plates, and the cell process is for assembling a TFT plate with a CF plate to form a single one—called TFT-LCD plate. A TFT/CF plate is composed of several panels and so is an assembled TFT-LCD plate. A TFT-LCD panel is good only when its two composing panels are good. In the ramp-up stage of manufacturing TFT-LCD, the yields of array/CF processes may be so low that many panels on a plate are defect ones. Two questions might be raised: Should we hold the manufacturing of a low-yield plate and send it to rework? If so, how to determine the threshold values for rework? This research mathematically formulates the rework decision problem and develops a genetic algorithm to solve the problem. Experiment results indicate that the proposed method significantly outperforms the heuristic method used in industry.