Vehicle-pavement Interaction Under Thermal and Vehicle Speed Actions Using Deep Learning Method

Abstract

Finite element analysis (FEA) is usually used to understand the behavior of the structure such as pavement structure. Besides, the pavement roughness, temperature and vehicle speed can be considered as the main factors which may influence the vehicle riding comfort and serviceability of the asphalt layer. FEA process is generally used to study the vehicle running along the pavement structure. Unfortunately, FEA is time consuming when the process should be iterating thousand time. Therefore, this study introduces a deep learning (DL) model. By taking the elasticity modulus, temperature and vehicle velocity as main inputs to predict the strain, stress, deflection, and heat flux responses of the pavement structure. FEA provided a bunch of images recorded during the simulation to train the DL model. The asphalt pavement of Babadjou-Bamenda road located in Cameroon’s western region has been selected for case study. The results shown that The Deep learning model can predict the results with accuracy and effectiveness, the NMAE error is less than 2%. The foundation layer thickness decreases with the traffic speed. Moreover, results shown that the mean of the vertical displacement corresponding to pavement roughness with Class B profile is higher than the pavement roughness with Class A.