近代光學研究,逐漸因為大眾消費市場對於產品的需求,不論是外觀上的要求輕薄短小或是功能操作上的體驗提升,都成了促使光學相關產業開使對於微光學領域的重視。 微透鏡陣列是人類仿生的具體表現,是微光學領域中和人類日常生活較貼近的一環。其常被應用的領用抱括有平面顯示器的背光擴散及增效、光纖耦合,以及波前檢測等。許多製程被開發使用以提升其精密度,但能將之以較有成本大量複製的方法並不多見。 本研究即採用微射出成型、微射出壓縮成型、微熱壓成型等三種製程實驗所得的數據進行探究。以較少的實驗結果,利用倒遞類神經網路建構起一預測模式,再以基因演算法進行全域最佳解搜尋。藉由上述的流程,求解最佳的製程參數組合,以最佳化製程協助光學微透鏡陣列轉寫性特性的提升。
Because of the demand of consuming market, such as size reducing and function upgrading, modern optical research focuses on micro optical research like microlens array especially. Microlens array is from an idea that human learned from nature world, and it is one of the fields of micro optical territory, which is closer to our daily life. It could be used to enhance the efficiency of flat panel display backlight, apply on fiber coupling, and apply on wave-front detection. Many processes have been developed to enhance precision, but rare of them could be used to apply on mass production. In this study, the data obtained from the experiments of micro injection molding, micro injection compression molding, and hot embossing is be used. We apply backpropagation to build up a prediction model, and apply GA (genetic algorithm) for global optimum search with less data. By the process above, obtain the optimal process parameters combination and assist the enhancement of microlens array replication.