本文主要目的是應用田口實驗計畫法與Trace-Pro光學模擬軟體，來研究LED燈具之設計與分析，並以提高最小輻射照度(Radiant Irradiance)為設計目標。 首先我們應用田口實驗計畫法，其採用Ｌ25的直交表設定各項參數，含6個可控制因子：導光管路徑橢圓雙軸比、導光管路徑夾角、導光管截面上之散射面中心位置、導光管散射面於截面之夾角、導光管之材質、LED之空間輻射分佈曲線，並各含5個水準。 接下來利用Trace-Pro光學模擬軟體進行模擬，來取得25組模擬結果，再做變異數的分析，來找出實驗中主要效應分子為何及各因子間交互關係對實驗結果的最佳化，用以得出一組最佳組合。 最後，最佳組合經Trace-Pro模擬可得該狀況之最小輻射照度為0.11625 watt/m^2 ，跟原本未經田口實驗計畫法優化之最小輻射照度為0.0012037watt/m^2 一比較，為其96.5772202倍，最小輻射照度獲得明顯提升。 由此可知田口實驗計畫法能有效減少模擬次數，將原本6因子5水準需做15625次模擬簡化為25次，讓設計者可以在較短的時間內得到較佳的設計參數。 The main purpose of this paper is to apply the application of Taguchi method and optical simulation software ‘Trace-Pro’ for the optoelectronics industry in LED-lighting design and analysis, and getting higher min-irradiance for the design. First, we use the orthogonal table ‘Ｌ25 ’of Taguchi method to set the factors: The elliptical ratio of lighting-pipe’s path, the included angle of lighting-pipe’s path, the center of scattering surface on lighting-pipe’s cross-section, the included angle of scattering surface on lighting-pipe’s cross-section, the materials of lighting-pipe, and the distribution curve of radiation for LED in space. In addition to previous writing, anyone factor also has five levels. Next simulates it in Trace-Pro, and obtained 25 results. In order to find best design parameters, we do variance analysis to identify not only the main effects of elements in this experiment but also the interactive relationship of various factors to optimize the experimental result. End, the simulate result for the design by using Taguchi method is 0.11625 watt/m^2 is 96.5772202 times higher in comparison with the simulate result for the original design without the optimization by Taguchi method is 0.0012037 watt/m^2 . Thus that proofs the min-irradiance was improved significantly. It can be seen Taguchi method can effectively reduce the simulate times from 15625 times to 25 times, so that designers can get better the design parameters in a short time. By the way, 15625 times come from arrangement and combination of the 6 factors and the 5 levels.
