Artificial neural networks—genetic algorithm based model for backcalculation of pavement layer moduli

TLDR

Backcalculation of pavement layer moduli uses surface deflections to evaluate pavement layers, and while genetic algorithms have been successfully applied, they are robust but computationally intensive. The study develops artificial neural network models to compute surface deflections from pavement layer elastic moduli and thicknesses. These ANN models are integrated into the BACKGA program, replacing repeated surface‑deflection calculations with rapid predictions to preserve GA robustness while cutting computational time. The resulting BACKGA–ANN model achieved satisfactory performance in backcalculation tasks.

Abstract

Backcalculation of pavement layer moduli refers to the process of evaluating the pavement layers using pavement surface deflections. The genetic algorithm (GA) technique was successfully used in the past for backcalculation. The BACKGA model developed by the Indian Institute of Technology, Kharagpur is one such program used for backcalculation using the GA technique. Though GA-based backcalculation models are considered to be robust due to the search algorithm adopted in the process, they require more computational time due to the large number of times the surface deflections are computed using different sets of layer moduli. In the present work, artificial neural network (ANN) models have been developed for computing surface deflections using elastic moduli and thicknesses of pavement layers as inputs. The ANN models have been used in BACKGA for forward calculation of surface deflections to combine the computational efficiency of ANNs with the robustness of the GAs. The performance of the resulting model, BACKGA–ANN, has been evaluated and found to be satisfactory.