Dynamic Back-Calculation Approach of Deflections Obtained from the Rolling Dynamic Deflectometer: Application and Validation

Abstract

Pavement response depends not only on loading magnitude and pavement material properties but also on the pavement’s dynamic parameters such as inertia, resonance, and damping. In a previous study, it was found that by using rolling dynamic deflectometer (RDD) free vibration testing, the resonant natural frequency and the damping ratio of the pavement could be determined, which is essential in determining the pavement stiffness, k. In this study, a back-calculation approach using RDD-measured deflections considering the natural frequency and loading frequency was proposed. A three-dimensional (3D) finite-element (FE) model was established simulating RDD loading on a three-layered pavement system consisting of asphalt, subbase, and subgrade. Using the FE model, a synthetic database composed of different pavement conditions and deflection responses was developed. The synthetic database was trained to predict natural frequency and deflections using deep-learning neural networks (DLNN). A back-calculation algorithm was then established determining the pavement modulus and thickness using the pavement’s natural frequency, deflection response, and RDD loading frequency. The proposed approach was validated by comparing the RDD and falling-weight deflectometer (FWD) back-calculated modulus using static and dynamic analysis. The RDD back-calculated modulus at 25 Hz was found to have good correlation with the FWD back-calculated modulus with an assumed hitting frequency of 33 Hz. In addition, modulus values of field cored specimens were compared with the RDD back-calculated modulus and were found to have good correlation.