Investigation on Thermal Conductivity of Soil Under Freeze–Thaw Action Based on Machine Learning Models

Thermal conductivity is a crucial factor for the soil, which is significantly affected by environmental conditions. Based on the variation in the thermal conductivity and the influencing factors of silty clay obtained by the freeze–thaw cycle test, this paper adopted four machine learning models optimized by particle swarm optimization (PSO), including the artificial neural network model (ANN), random forest model (RF), support vector machine model (SVM), and extreme gradient boosting model (XGBoost) to predict the thermal conductivity of the soil. Meanwhile, mean absolute error (MAE), root mean square error (RMSE), and correlation coefficient(R2) were used to evaluate the accuracy of the models. The accuracy of the machine learning model and empirical model were also compared. Then, the Monte Carlo simulation was used to analyze the stability of the models. The research results showed that the predicted performance of the machine learning models is significantly better than the empirical models. Among all the machine learning models, the R2 of the PSO-ANN model is above 0.95, while both RMSE and MAE values are below 0.02 (W·m⁻¹·K⁻¹). In addition, the stability order of the machine learning models is PSO-XGBoost, PSO-ANN, PSO-RF, and PSO-SVM. Therefore, comprehensively considering the accuracy and stability of the four machine learning models, the PSO-ANN model is recommended to predict soil’s thermal conductivity under freeze–thaw action.

This creative work has been published under the Creative Commons Attribution 4.0 International (CC-BY 4.0) license which allows copying, and redistribution as well as adaptation of the original work provided appropriate credit is given to the original author and the conditions of the license are met.