Predicting thickness impregnation in a VaRTM resin flow simulation using machine learning

Abstract This study proposes a method for predicting the impregnation time in the thickness direction of vacuum-assisted resin transfer molding (VaRTM) with flow media using 2D impregnation analysis and machine learning to reduce the computation time. We first let the machine learn the relationship between the impregnation time and the state of the flow media using a VaRTM 3D analysis. To predict the impregnation time in the thickness direction with low computation cost, a 2D analysis of the flow media portion was performed, and the impregnation time was predicted based on the 2D results and pre-trained machine learning program. For the prediction of a flat plate model with injection points that were different from the training model, the impregnation time error was 6.31%. Further, when using the machine learning model for the flat plate model, the analysis time was reduced to approximately 0.74% of the time required for the 3D analysis.