在本文中，傳統的田口直交法與倒傳遞類神經網路，被使用來評估且決定在氣體輔 助射出成型製程中參數的最佳數值。在田口直交法的應用中，L18直交表首先被使用來收集 實際觀察值。相同的資料集亦被用來建構一適當的類神經網路模式，以比較類神經網路方法 是否能較傳統的統計技術，提供出較佳之結果。為能進一步瞭解類神經網路模式的建構程序 ，針對模式中的學習步伐量及隱藏層的神經元數目有進一步的探討。由研究結果發現，類神 經網路模式比傳統的田口直交法，有更高的機會能成的辨識尋找出參敷的最佳數值。

　　　　　A statistical Taguchi approach and a backpropagation neural network model were utilized to evaluate the effect of various parameters and to determine the optimal parameter setup values in a gas-assisted injection molding process. In the application of the Taguchi approach, an L18 orthogonal array was used to collect the observations. The same data set was utilized to construct a neural network model and to determine if the utilization of a neural network would provide improved generalization capability over the statistical method. The effect of the learning rate and the number of hidden nodes on the efficiency of the neural network learning algorithm was extensively studied to identify the learning rate and the number of hidden nodes that resulted in the best network performance. In addition, to verify the generalization capability of neural model, eight different parameter setups, not all included in the full factorial design, were constructed for network testing. The results show that the neural network has a higher chance than the Taguchi experimental design approach of finding the optimal parameter setup.