塑膠的收縮性質對射出成型的產品尺寸穩定性有重要的影響，熔膠的體積變化主要源自於冷卻過程中的溫度與壓力效應。本研究透過模內感測系統，擷取半結晶塑料聚丙烯(PP)於成型過程中的溫度與壓力數據，再利用壓力-比容-溫度(PVT)關係方程轉換成比容值，由比容的變異程度與均勻性組成收縮性質的指標。接續藉由調整射出成型的製程參數優化成品品質，本研究採用實驗設計法，建構一個基於統計原理的參數最佳化流程，首先以二水準田口方法篩選顯著因子，再利用三水準田口方法與二階反應曲面法中的中央合成設計與Box-Behnken設計建構迴歸模型，預測最佳化製程參數。研究結果顯示料溫與保壓階段的效應對PP塑件的收縮性質影響最為顯著，而田口方法不能考慮每個因子之間的交互作用，導致其優化能力不及反應曲面法，此外，雖然中央合成設計與Box-Behnken設計於本實驗案例預測出相同的最佳化製程參數，但中央合成設計的預測誤差較低，顯示出較好的預測能力。相較於參數設定之初始條件，二階反應曲面法降低試片15.82%的體積收縮率；三水準田口法降低試片9.69%的體積收縮率，同時，改善了比容的均勻性，實現優化成品品質的目標。;The shrinkage of plastic has a significant impact on the dimensional stability of products manufactured through injection molding. The volumetric variation of the molten plastic primarily is caused by temperature and pressure effects during the cooling process. In this study, an in-mold sensing system was utilized to capture temperature and pressure data of semi-crystalline polypropylene (PP) during the injection molding process. The data were then converted into specific volume values using the pressure-volume-temperature (PVT) relationship equation. The changeability and uniformity of the specific volume were employed as indicators for characterizing the shrinkage properties. Subsequently, the quality of the products was optimized by adjusting the process parameters for injection molding. The design of experiments was employed to construct a procedure for optimizing parameter based on statistical principles. Initially, the significant factors were screened using the two-level Taguchi method. Regression models were then developed using the three-level Taguchi method and the second-order response surface methodology such as central composite design and Box-Behnken design, respectively, to predict the optimal process parameters. This research findings demonstrated that the effects of melt temperature and packing pressure stages had the most significant influence on the shrinkage of PP specimens. However, the interaction between individual factors is often not fully considered in Taguchi, the Taguchi method exhibited lower optimization capability compared to the response surface methodology. Although both the central composite design and Box-Behnken design predicted the same optimal process parameters in this experimental case, the central composite design exhibited lower prediction errors which indicated better predictive capability. Compared to the initial conditions of the parameter settings, the second-order response surface methodology reduced the volume shrinkage rate of the specimens by 15.82%, while the three-level Taguchi method reduced it by 9.69%. Additionally, the uniformity of specific iii volume was improved; thereby the goal of optimizing the quality of the products was achieved.
