A methodology to predict the effective thermal conductivity of a granular assembly using deep learning
Palavras-chave:
machine learning, artificial neural networks, particles, thermal conductivity, granular materials, discrete element method (DEM)Resumo
In this work, an Artificial Neural Network (ANN) is employed to predict the effective (i.e., bulk) thermal
conductivity of a granular assembly. The ANN is trained with the help of computed thermal conductivities of
various different assemblies, obtained through several simulations with our in-house DEM (Discrete Element
Method) code. Convection and radiation are not considered as to isolate the conduction problem and allow for a
better estimate of the assembly’s effective response. The methodology enables the effective thermal conductivity
of a granular assembly over a wide range of parameter values, including particles’ size and their material’s
conductivities.
