Machine learning-based prediction of the thermal conductivity of filling material incorporating steelmaking slag in a ground heat exchanger system

This study used machine learning models to predict the thermal conductivity of heat-transfer materials based on steelmaking slag. A dataset containing various physical parameters of the heat-transfer materials was obtained from previous research results and Pearson correlation analysis was used to select the optimal input variable. Three machine learning models-support vector regression (SVR), random forest (RF), and multilayer perceptron (MLP)-were assessed to determine the most accurate model for predicting the thermal conductivity of the heat-transfer materials. K-fold cross-validation was applied to each model to prevent overfitting of the results and to generalize the prediction models. All three models predicted the thermal conductivity better than a previous empirical method. The SVR model exhibited the best prediction accuracy across the whole dataset, confirming that this model can provide a simple and practical method for predicting the thermal conductivity of reinforced soil without the need for time-consuming and expensive experiments. Finally, equations based on SVR were proposed that can predict thermal conductivity under different experimental and material conditions.