A study on Artificial Neural Networks – Genetic Algorithm model and its application on back-calculation of road pavement moduli

Abstract

Recently, Falling Weight Deflectometer (FWD) is one of the most significant testing used to measure the surface deflections under impact load subjected to circle plate which was used to back-calculating elastic moduli of the road pavement layer. Several back-calculation programs are useful for back-calculating road pavement layer moduli. A genetic algorithm (GA) was used successfully in this problem but it requires more computation time to a variation of computed deflection using optimized moduli value of road pavement layer based on the GA. There is a few research adopted to Artificial Neural Network (ANN) for computing deflection of the road pavement system. This article aimed to develop ANN for computing surface deflection of pavement using layer moduli and its thicknesses as input parameters. We have also discussed the solution techniques and algorithms for use in developing the program, including Burmister theory for determining deflection based on a cylindrical coordinate system, the GA optimization for back-calculating road pavement moduli, and the development of the ANN-GA model. The evaluation shows that the predicted deflections using the ANN compare well with computed deflections from the hypothetical model. Backcalculated layer moduli based on the GA-ANN model are well with a hypothetical model based on FWD test.