Efficiency of genetic algorithms and artificial neural networks for evaluating delamination in composite structures using fibre Bragg grating sensors

The efficiency of genetic algorithms (GAs) and artificial neural networks (ANNs) in the quantitative assessment of delamination in glass fibre-reinforced epoxy (GF/EP) composite laminates was evaluated comparatively. For GA-based identification, a theoretical model and a vibration-based objective function were established to relate the delamination parameters to the shift in structural eigenvalues. For the ANN-based approach, feedforward artificial neural networks were configured and trained using the structural eigenvalues obtained from different damage groups, under the supervision of an error-backpropagation neural algorithm. By way of validation, dynamic responses of selected GF/EP laminate beams containing various delaminations were captured using embedded fibre Bragg grating sensors, from which the structural eigenvalues were extracted and used inversely to implement the damage assessment via the GA and the ANN. The performances of the two algorithms were addressed as regards the prediction precision and computational cost.