저해상도 메쉬 기반 사출성형 Fill Time 예측을 위한 전이학습 인공신경망 모델에 대한 연구

Abstract

In injection molding simulations, accurate prediction of fill time is crucial for optimizing process conditions and ensuring part quality. However, high-resolution computational analysis is often time-consuming and resource-intensive. This study proposes a transfer learning-based artificial neural network (ANN) model that accelerates fill time prediction by utilizing low-resolution mesh data. Initially trained on a cup-shaped geometry, the ANN model is fine-tuned and adapted to a new geometry—a ricebowl shape—demonstrating its applicability across different part designs. The model effectively learns to approximate high-resolution simulation results from low-resolution inputs, significantly reducing computational cost while maintaining predictive accuracy. The proposed method shows promise for generalizing fill time predictions in injection molding CAE workflows, particularly when rapid design iteration or limited computational resources are required.