Developing a prediction model for rutting depth of asphalt mixtures using gene expression programming

•A new machine learning-based model for prediction of rut depth of asphalt mixtures.•Formulating the model in terms of typical mixture components and test conditions.•Sensitivity analysis has a good agreement with laboratory results.•A viable rut depth model for different climates and traffic levels. This study presents a new model for the prediction of rutting depth of asphalt mixtures using a machine learning technique called gene expression programming (GEP). A database containing a comprehensive collection of Hamburg test results is used to develop a GEP-based prediction model. The database includes 96 tests results for various asphalt mixtures. The model is formulated in terms of typical influencing variables such as asphalt binder high temperature performance grade (PG), mixture type, aggregate size, aggregate gradation, asphalt content, and total asphalt binder recycling content. The model accuracy was assessed through a rigorous validation process. A sensitivity analysis was performed to evaluate the effect of the variables on the rutting depth in the GEP model. A comparative study has been conducted to benchmark the prediction performance of the GEP model against an artificial neural network model. The model is capable of capturing the differences in mixture properties and test conditions. The model is recommended for pre-design purposes or as a tool to determine rut depth in asphalt mixtures when laboratory testing is not feasible.