Microstructure-informed deep convolutional neural network for predicting short-term creep modulus of cement paste

This study aims to provide an efficient alternative for predicting creep modulus of cement paste based on Deep Convolutional Neural Network (DCNN). First, a microscale lattice model short-term is adopted build database that contains 18,920 samples. Then, 3 DCNNs with different consecutive convolutional layers are built learn from the database. Finally, performance tested unseen testing The results show can achieve high accuracy in set, R2 all higher than 0.96. distribution predicted by coincides original data. Furthermore, through analyzing feature maps, it found correctly capture local importance microstructural phases. DCNN allows therefore prediction input, which saves computational resources segmentation procedure and multiple incremental FEM calculations.