| Summary: | 碩士 === 國立勤益科技大學 === 機械工程系 === 102 === In this study injection molding process window was identified for optical lenses with specific form accuracy; an inverse model was constructed using artificial neural network (ANN) method with genetic algorithm (GA). Taguchi parameter design was used to perform screening experiments, and them the significant factors affecting form accuracy of lens were identified as mold temperature, cooling time, packing pressure and packing time. Full factorial experiments were implemented using the four significant factors, and the result was used for training and checking samples for constructing artificial neural network model. The genetic algorithm was then integrated into the artificial neural network model to construct an inverse model for injection molding process. Inverse model predictions were implemented for lenses with form accuracies of 0.5, 0.7 and 1μm respectively (within 2% error), and the result indicated that the set of 26, 17, 6 conformed the injection molding processing conditions are obtained. Result of confirmation experiments indicated that average error was 8.27%, which implies that the inverse model proposed in this study is accurate and reliable. Moreover, the optimized process parameters identified through comprehensive searching with genetic algorithm improved the Taguchi optimal result still further; the confirmation experiments revealed that the form accuracy of lens was improved to 0.415μm from the original 0.479μm, a 13.36% improvement was achieved.
|
|---|
