Evaluation of a Machine Learning Approach for Temperature Prediction in Pavement Base and Subgrade Layers in Alberta, Canada

Abstract

The performance of flexible pavement is influenced by pavement material properties and the strength or the stiffness of the pavement layers. Pavement temperature significantly impacts the material properties of flexible pavements. However, to date there has not been much research that investigates the prediction of the pavement temperature in unbound material (base and subgrade layers). The goal of this research is to apply a new approach, machine learning, to predict pavement temperature in unbound material. Pavement temperature recordings collected at the Integrated Road Research Facility (IRRF) test road in Alberta from January 2013 to February 2020 were used to train and validate machine learning models. Finally, high-performance machine learning models with two parameters (air temperature and day of the year) were developed to predict the average daily pavement temperature at 0.5–2.7 m below the road surface. The accuracy of the temperature in the base and subgrade layers predicted using the machine learning models was found to be higher than for an existing model.