A Bayesian multiscale CNN framework to predict local stress fields in structures with microscale features.

The purpose of this work is to train an Encoder-Decoder Convolutional Neural Networks (CNN) to automatically add local fine-scale stress corrections to coarse stress predictions around unresolved microscale features. We investigate to what extent such a framework provides reliable stress predictions inside and outside particular training sets. Incidentally, we aim to develop efficient offline data generation methods to maximise the domain of validity of the CNN predictions. To achieve these ambitious goals, we will deploy a Bayesian approach providing not point estimates, but credible intervals of the fine-scale stress field to evaluate the uncertainty of the predictions. This will automatically encompass the lack of knowledge due to unseen macro and micro features. The uncertainty will be used in a Selective Learning framework to reduce the data requirements of the network. In this work we will investigate stress prediction in 2D porous structures with randomly distributed circular holes.