Predicting peak stresses in microstructured materials using convolutional encoder–decoder learning

Journal: Mathematics and Mechanics of Solids

Dear colleagues,

We invite you to read our paper on "Predicting peak stresses in microstructured materials using convolutional encoder-decoder learning," published in Mathematics and Mechanics of Solids. Here, we propose a data-driven-based deep learning model to predict peak-stress clusters inside microstructures in the framework of heterogeneous linear elasticity. The deep-learning model uses convolutional filters to model local spatial relations between microstructures and stress fields. The model prediction accuracy was analyzed using the cosine similarity and by comparing the peak-stress clusters' geometric characteristics. It was observed that the model could predict the location and size of the peak-stress clusters. Furthermore, during visualization of the features extracted by the first layer of the model, it was observed that different convolutional filters capture different but distinguishable local features, such as grains and grain boundaries of the microstructures.