Prediction of Asphalt Pavement Responses from FWD Surface Deflections Using Soft Computing Methods

Abstract

This study predicts asphalt pavement responses from surface deflections under falling weight deflectometer (FWD) loading using soft computing methods. Finite-element (FE) models are developed and validated considering viscoelastic properties of the asphalt layer and nonlinearity of unbound layers. The synthetic database of surface deflections and strain responses in asphalt layer are developed for different combinations of pavement structures, material properties, temperature profiles, and loadings levels. An artificial neural network (ANN)-based program combined with genetic algorithm (GA) optimization is trained and verified using the synthetic database. The soft computing model shows better predictive accuracy than the traditional approach of multivariable regression. The model is validated using a pavement section selected from the long-term pavement performance (LTPP) database and pavement instrumentation measurements reported in the literature. The ANN-GA program is proved to be an efficient approach for predicting tensile and shear strains in asphalt layer under FWD loading. The proposed prediction approach provides an efficient way to assess existing pavement condition without layer moduli backcalculation.