An ANN model to correlate roughness and structural performance in asphalt pavements

- Proposal of a neural network method to correlate roughness and structural capacity.
- Adoption of an LTPP-based database, including many influencing parameters.
- Comparison of various networks to analyse the method potential.
- The method can reduce traditional deflection tests (HWD, FWD) frequency for PMSs.

In this paper, using a large database from the Long Term Pavement Performance program, the authors developed an Artificial Neural Network (ANN) to estimate the structural performance of asphalt pavements from roughness data. Considering advantages of modern high-performance survey devices in the acquisition of road pavement functional parameters, it would be of practical significance if the structural state of a pavement could be estimated from its functional conditions. To differentiate various road section conditions, several significant input parameters, related to traffic, weather, and structural aspects, have been included in the analysis. The results are very interesting and prove that the ANN represents an adequate model to evidence this relation. The papers shows the effectiveness of the adoption of a large database for the analysis of the correlation. ANN provides also better results in comparison with Linear Regression. Further, the authors trained three different ANNs to analyse the effects of modified datasets and different variables. The numerical outcomes confirm that, by using this approach, it is possible to correlate with good accuracy roughness and structural performance, allowing road agencies to actually reduce the deflection test frequency, since they are generally more costly, time consuming, and disruptive to traffic than functional surveys.