本文提出數據的取樣方法，並將其考慮為三水準的品質因子進行田口法分析，以探討取樣方法對成形螺孔之牙型填充慮及成形最大扭矩參數優化分析結果的顯著程度。為了更進一步降低隨機因素的效應，成形最大扭矩之數據採用常態分佈的概念對重覆實驗結果進行篩選，提出三種取樣方法分別為選擇一組實驗、九組實驗及標準差最小的四組實驗。牙型填充率的量測結果和取樣方法有關，且是評估螺孔品質的重要指標之一，故本文提出三種取樣方法分別為四顆螺牙、左側螺牙、右側螺牙，作為計算牙型填充率的依據，藉由比較牙型填充率計值的差異，也可以了解成形螺孔的偏心程度。 本文採用L_27(3^4)直交表進行田口分析，根據反應值的期望特性，利用不同的公式定義訊號雜訊比(S/N)，作為評估品質的指標。利用品質因子對訊噪比的效果分析結果可得到最佳解的優化條件，並繪製品質因子對訊噪比的效應圖，其中所示的線段斜率可以觀察各品質因子對反應值的影響。最後，為了探討取樣方法是否也會影響品質因子對品質特性（即實驗反應值）的效應程度，本節將視取樣方法為一獨立參數，採用L_27(3^3)直交表，重新配置成形螺孔實驗的數據進行田口分析並做討論。從訊噪比效應圖中可發現，因取樣方法的因素使得介於不同水準間各線段斜率都存在可見的差異，此意謂著取樣方法實質影響了品質因子對訊噪比的效應程度。 ;This passage discusses the sampling methods for data collection and applies Taguchi analysis with three levels of the quality factor to investigate the significance of the sampling methods on the tooth filling rate and maximum forming torque parameters of formed screw holes. To reduce the effects of random factors, the data for the maximum forming torque are filtered using the concept of a normal distribution in repeated experiments. Three sampling methods are proposed: selecting one set of experiments, nine sets of experiments, and four sets of experiments with the minimum standard deviation. The tooth filling rate measurement is related to the sampling methods and is an important indicator for evaluating the quality of the screw holes. Three sampling methods, namely, four teeth, left teeth, and right teeth, are used as the basis for calculating the tooth filling rate. By comparing the differences in tooth filling rate calculations, the eccentricity of the formed screw holes can also be understood. The study employs the L_27(3^4) orthogonal array for Taguchi analysis. Based on the desired characteristics of the response values, different formulas are used to define the signal-to-noise ratio (S/N) as an indicator of quality evaluation. Analyzing the effects of the quality factors on the signal-to-noise ratio provides the optimal conditions for optimization and the effect of each quality factor on the response values can be observed by plotting the effects of the quality factors on the signal-to-noise ratio. Finally, to investigate whether the sampling methods also affect the effects of the quality factors on the quality characteristics (i.e., experimental response values), the sampling methods are considered as an independent parameter. The L_27(3^3) orthogonal array is used to reconfigure the data from the formed screw hole experiments for Taguchi analysis and discussion. From the signal-to-noise ratio effects plot, it can be observed that the sampling methods lead to visible differences in the slope of each segment between different levels. This indicates that the sampling methods indeed affect the extent of the effects of the quality factors on the signal-to-noise ratio.
