Prediction of moisture damage properties of asphalt mixtures using machine learning models

This study investigates the application of machine learning techniques to predict moisture damage indicators in asphalt mixtures, specifically focusing on dry indirect tensile strength (ITS), wet ITS, and tensile strength ratio (TSR). Six machine learning models were evaluated: Linear Regression, Lasso Regression, Ridge Regression, Decision Tree Regressor, Random Forest Regressor, and Multi-Layer Perceptron. A comprehensive dataset of asphalt mixture properties was used to train and test the models. The Random Forest and Decision Tree Regressors consistently demonstrated superior performance across all three target variables, achieving R-squared values of 0.98 and the lowest RMSE values. Linear and Ridge Regression models also showed good performance, while Lasso Regression consistently underperformed. The results highlight the potential of machine learning, particularly tree-based models, in accurately predicting moisture damage susceptibility in asphalt mixtures. This approach could offer a more efficient and cost-effective alternative to traditional laboratory testing methods, potentially enhancing asphalt mixture design and evaluation processes.