深層学習を用いた編み構造を持つ繊維強化複合材の代表体積要素の物性値予測モデル構築

抄録

A surrogate model that predicts results of simulation using machine learning is expected to be put into practical use because it can predict faster than simulation calculations. Many machine learning algorithms such as deep learning require a lot of training data for learning. However, it is difficult to create a lot of training data because a simulation requires a lot of computational resources to calculate. In this study, we constructed a surrogate model that predicts rigidity of a representative volume element of fiber reinforced composite material with a woven structure. We confirmed the possibility of creating a prediction model with a small amount of data by having information of a woven structure and shape in a two-dimensional matrix and creating a model using a convolutional neural network.