Fig. 7: Architecture of StressNet.
From: StressNet - Deep learning to predict stress with fracture propagation in brittle materials
The model consists of two branches. In the left branch, the Bi-LSTM encodes the temporal dependency among historical maximum stress data into a series of vectors. In the right branch, TI-CNN followed by Bi-LSTM encodes the spatial and temporal information of the damage channel into another series of vectors. By fusing features from these two branches, StressNet can predict the maximum stress at the next time step.
