Deep Neural Network-based View factor Modelling of Radiative Heat Transfer between Particles

Abstract

We present a Deep Neural Network (DNN)-based view factor model to calculate radiative heat transfer between particles. Different neural net model parameters like: the number of hidden layers, the number of nodes per layer, optimization algorithms and techniques to prevent overfitting of the DNN are investigated. Using a simple regressor showed that one input marker, namely the surface distance between the interacting particles, cannot accurately predict the widespread view factors for increasing surface distance. Therefore, the effect of additional input markers like the solid angle between particles or a derivate of a local particle volume fraction is examined. The outcome of the DNN-model is compared to literature models and a simple regression analysis. It is demonstrated that the pretrained DNN-model can model view factors at higher accuracy and with significantly less computational effort than other literature models.