類神經網路應用於透明導電膜連續濺鍍法之配方最佳化搜尋

Optimal Recipe Search Applying Neural Network for the Successive Deposition of Transparent Conductive Thin Film

林永威(Yung-Wei Lin)

指導教授 ： 張耀仁

中原大學/工學院/機械工程研究所/碩士(2010年)

https://doi.org/10.6840/cycu201000960

摘要

近幾年來，國內的觸控面板產業發展的相當迅速，雖然目前市場上觸控面板已發展出電阻式(Resistive)、表面電容式(SCT)、投射電容式(PCT)、紅外線式(Infrared Touch)、聲波式(Acoustic Wave)、電磁感應式(EMR)與數字式(Digitizer)等。但是，就國內市場上來看，電阻式及電容式為目前市場上的兩大主流，其一，電阻式的製程技術已相當成熟，且生產成本較低，故可以大量生產 ; 反觀電容式的製程技術仍有許多問題需要克服 ，故無法像電阻式一樣，可以大量生產。本論文所探討的氧化銦錫磁控真空濺鍍低阻抗模組最佳化，是建立在電容式的觸控面板上，其相關技術也可以運用到各式TFT面板的磁控真空濺鍍上。
真空濺鍍在觸控面板製程中占了極為重要的地位，如何利用最短的時間及最低的成本耗損來完成ITO濺鍍鍍膜，相當值得研究地課題。目前國內許多科技產業大廠，為了能更快速尋找到製程最佳化的條件，都已建立一套完整的最佳配方搜尋方式，目前國內大多觸控面板廠卻沒有建立搜尋規則，導致浪費生產成本，本論文主要目的，期望能運用最少的實驗成本取樣，利用類神經網路之智慧型學習方式，可獲取最佳化配方。運用徑向基底網路(Radial Basis Function Network，簡稱RBFN)以及田口基因演算法(Taguchi Genetic Algorithm，簡稱TGA)來搜尋最佳化配方，不僅可以迅速製程穩定、減少人為變因，更可以降低生產成本。

關鍵字

觸控面板 ； 類神經網路 ； 主成份分析 ； 田口基因演算法

並列摘要

Developing the TFT-LCD technology is quite rapid in these years. Though there are some latest productions of TFT-LCD have been developed like Resistive, SCT, PCT, Infrared Touch, Acoustic Wave, EMR, Digitizer and so on that have been on the market so far, no one really can say which is really the one hundred percent successful of the main stream at the market actually . If we just focus on the domestic market and try to find out what is the most of all about the main stream in Taiwan, I think there are two items that can be the main streams of domestic market. At first, the Resistive process integration’s technology has been approaching mature nowadays. The prime cost of the Resistive process integration is less than any others products, so it’s can be made a plenty of productions of the Resistive. However, like SCT or PCT, it has a few essential problems will be needed to overcome. No matter SCT or PCT that can’t be made a great quantity like Resistive . This topic of the thesis will probe into the Indium Tin Oxide magnetron sputtering low resistivity to resist the best modeling capability, and it can be used on the Capacitive Touch Panel. Therefore, the others techniques are being to related to all kinds of the TFT-Panel’s magnetron sputtering.
The magnetron sputtering is extremely important status of the touch panel’s process. How could we get a nice using way with the lowest cost to finish the ITO in a short time 。That’s quite available topic to study . Nowadays, there are many famous science and technology of domestic industries in order to search for the best condition of processing, and they have set up the complete and best searching method for using. Even though most of domestic touch-penal companies have no any complete searching rule like those famous company like so that most of companies waste their productive cost . The main purpose of this thesis, is expected to use the less experiment cost by using the Artificial Neural Network (ANN) of Industry Standard Commercial Identifier of studying way to get the optimization. To use the Radial Basis Function Network(RBFN) and Taguchi Genetic Algorithm(TGB) to search the optimization are not only to steady processing and to reduce human
influence but also to decrease the cost of production。

並列關鍵字

Taguchi Genetic Algorithm ； Neural Network ； Touch Panel ； Principal Component Analysis

參考文獻

9. 張益瑞, “銅化學機械研磨之製程參數最佳化”, 中原大學機械所碩士論文, 2005.

10. 達虹科技股份有限公司內部技術資料,2007-2010.

19. 呂秀英,”變數分析在農業科技之應用”,政院農委會農業試驗所農藝組,2006

18. 李柏甫,”TFT-LCD濺鍍製程之智慧型診斷系統發展”,國立成功大學,2004

Algorithm for Global Numerical Optimization,”IEEE Transactions