Pavement moduli back-calculation using artificial neural network and genetic algorithms

An Artificial Neural Network (ANN)-based back-calculating program combined with a Genetic Algorithm (GA) optimization algorithm was developed for the back-calculation of flexible pavement layer moduli. Deflections measured using geophones on a model pavement structure under accelerated pavement testing (APT) with the MLS30 were utilized for back-calculating the pavement layer moduli. As theoretically expected of visco-elastic materials that are temperature dependent, the back-calculated moduli of the asphaltic layers, namely the surfacing and base, exhibited a decreasing trend with an increase in the temperature and number of APT load cycles. By contrast, the moduli of the unbound base layer and subgrade exhited insensitivity to temperature changes and did not decay significantly as a function of the APT loading. Overall, the integrated GABP algorithm (based ANN formulation) exhibited potential in satisfactorily back-calculating the pavement layer moduli form geophone measured deflections with acceptable accuracy. The splitting fatigue test results show that the fatigue life of asphalt mixture decreases with the increase of loading cycles, which can verify the feasibility of back-calcultion model. (C) 2021 Elsevier Ltd. All rights reserved.

Automated summary

An ANN-based back-calculating program combined with a GA optimization algorithm was developed to back-calculate flexible pavement layer moduli. The moduli of asphaltic layers decreased with temperature and number of APT load cycles, while the unbound base layer and subgrade moduli were insensitive to temperature changes. The integrated GABP algorithm showed potential in accurately back-calculating pavement layer moduli from geophone measured deflections.