- 學位論文
研究可拓神經網路及可拓理論於最佳化之應用:以生醫奈米結構薄膜製程為例
Study on Processing Optimization for Extension Neural Network and Extension Theory: Take Processing of Thin Film of Bio-medical Nano-structure for Example
摘要
製程參數組合的選擇，對於產品最終品質特性與製程中各項成本都佔有極重要之影響力，因而於工業、製造業…等各產業中，其最佳化製程參數之探討，一直以來都是一重視課題。 生醫材料產品，由於其將使用於生物體，其產品之特性，多是期望能達到理想品質與特性，托奈米科技的演進，因而提供了不少材料更多優異的特性效應，因此不少奈米新製程也隨之開發。 關於最佳化領域之探討，近年間，有不少新方法逐一浮現，如本論文提到的可拓理論，本論文將探討其方法於最佳化研究之成效，並將可拓理論加以修正改進比較，並針對可拓理論其優缺點提出相關論點。 本論文結果顯示，可拓理論擁有快速且簡易的計算過程，在效率上有特別的優勢，但由於可拓理論經典域之決定，往往取決於專家之意見，有失客觀性，因此提出修正可拓理論來補強此點，可藉此提高經典域判定準確率，並提供較為精確的製程參數水準數值。在論文最後，提出可拓理論對於多峰數據資料判別上的敏感性並不佳，針對此一問題，未來可加以探討。
並列摘要
The quality and cost of final product are influenced by a combination of processing parameters, so it is important to choose the right combination of process parameters. Because the Bio-medical material will be used in organisms, the quality tests have to be rigorous. Nanotechnology can increase material’s quality, so more and more processes are created with Nanotechnology. In recent year, many new methods have been optimized in different fields. This study will refer to a new method, extension theory and discuss how this method can efficiently solve problems and improve shortcomings. This study concludes that extension theory can lead to faster and easier calculations. But the decision of classic region always needs professional suggest, it is subjective. This study uses the Extension Neural Network to increase objectivity and provide more accurate computation. Finally, this study discusses how the extension theory is not properly sensitive to the structure of data.