Predicting the effective thermal conductivity of geo-materials using artificial neural networks

Marine and Land Geomechanics and Geotechnics, Kiel University, Kiel, Germany

^* Corresponding author: dinesh.shrestha@ifg.uni-kiel.de

Abstract

Soil thermal conductivity is an important thermal property used in heat transfer modelling and geo-energy applications. Because of its complex nature and depending on several factors such as porosity, moister content, structure, etc., it is always challenging to predict the thermal conductivity of geo-materials. In the past, many predictions models like theoretical, semi-empirical, empirical models have been proposed based on the experimental data. However, these models are more specific to certain boundary conditions. Therefore, in this study, an artificial neural network (ANN) approach was used to predict the thermal conductivity of geo-materials as a function of porosity, gradation and mineralogy. A comparison between existing prediction models and the developed ANN model for predicting thermal conductivity is also given.