RNNを用いたハイドロゲル４Ｄプリンティングのモデリング

―短いモデルの変形を基づいた、長いモデルの変形予測―

抄録

This study proposes a method to reduce the simulation time of the deformation of 4D printed hydrogel models using deep learning. In the proposed method, first, 4000 hydrogel models with the same shape but different expansion ratio distributions were created using Abaqus, and deformation simulations were conducted using the Abaqus thermal expansion approach. Next, from the simulation results, coordinates of 400 reference points on the edges of the deformed models were collected, and a dataset was created. Finally, the deformation of long hydrogels under the same load conditions was predicted based on the dataset from the deformation of short hydrogels using Recurrent Neural Network (RNN).