01629175

Component

Precise performance evaluation of flexible pavements is a challenging matter when deciding the rehabilitation and maintenance strategies. A commonly performed method for evaluation of a pavement’s condition is to use Falling Weight Deflectometer (FWD) device, which records deflections of a pavement’s surface when exposed to simulated traffic loading. The structural properties of pavements are backcalculated using FWD data, and used to decide the current conditions of pavement. In this paper, a backcalculation algorithm, named LCA-ANN, based on the combined use of metaheuristic optimization method called League Championship Algorithm (LCA) and finite element method based Artificial Neural Network (ANN) is proposed for backcalculating layer properties of conventional flexible pavements. The proposed algorithm uses the reliable finite element based pavement structural models that take nonlinear material characterization into account. Surrogate ANN models are used to provide efficiency and speed without losing the accuracy of finite element methods. LCA-ANN algorithm has the extensive search capability of LCA which can successfully discover the multimodal search spaces formed by the layer properties. To prove the success of LCA-ANN algorithm, a synthetic dataset is used first, and then the data obtained from Long Term Pavement Performance (LTPP) database are attempted as field study. One of the widely used backcalculation software, EVERCALC, is also used for comparison purposes. In addition, the sensitivity analyses of the algorithm are given with respect to its parameters. Comparison with other optimal search methods are provided for robustness. The results indicate that proposed approach provides a reliable performance when backcalculating layer properties fast enough. Within this form, LCA-ANN can be used as a robust backcalculation method.

This paper was sponsored by TRB committee ABJ70 Standing Committee on Artificial Intelligence and Advanced Computing Applications.

01618707

17-05043

English

2p

2017

Transportation Research Board 96th Annual Meeting

Digital/other

Backcalculation; Finite element method; Flexible pavements; Heuristic methods; Neural networks; Pavement performance; Programming (Mathematics)

Data and Information Technology; Highways; Pavements

Transportation Research Board Annual Meeting 2017 Paper #17-05043

TRIS, TRB, ATRI

Dec 8 2016 11:56AM