A data driven artificial neural network model for predicting radiative properties of metallic packed beds

In this work, we demonstrate that machine learning methods can be reliably used to predict radiative properties of dispersed media, i.e. packed beds, as a function of packed bed geometry and material properties. The computationally expensive Monte Carlo ray tracing (MCRT) method, which is widely used in this context, is replaced by Neural Networks (NN). We demonstrate that the data-driven surrogate prediction works accurately and generally. The results of both MCRT and NN models agree well with each other and with previously measured literature results. We also measure the uncertainty of the NN results using statistical methods. It is recommended that the developed model be used for efficient inverse problems and optimizations in relevant future work.