Use of Artificial Neural Network to Determine the Pavement Layer Properties Based on Automated Plate Load Test

Nowadays, automated plate load tests (APLTs) are used to evaluate the performance in terms of deflection data obtained from different magnitudes of loading. Several back-calculation methods are currently available to determine the pavement layer modulus based on the deflection bowl data obtained from the regular falling weight deflectometer (FWD). However, the configuration and number of sensors used for the APLT slightly differ from the routine FWD test. To determine the pavement layer properties from the APLT, there is a need to develop a back-calculation approach. In this study, a series of pavement analyses have been performed to simulate the loading condition of APLT with a multi-layered elastic analyses approach. The elastic deformations obtained from the pavement analyses were correlated with the pavement layer thickness and moduli based on the artificial neural network (ANN) approach. The feedforward approach of ANN was selected for this study. The developed ANN model was further used to predict the base layer modulus of different pavement sections.