Interpreting acoustic emission signals by artificial neural networks to predict the residual strength of pre-fatigued GFRP laminates

Abstract Artificial neural networks (ANNs) were used to predict the residual strength of glass fibre-reinforced plastic beams pre-fatigued in flexure up to different portions of their fatigue life. To this aim, the acoustic emission signals recorded during the tests for the measurement of the residual strength, and the associated applied stress, were provided as input. An optimisation of the network configuration was carried out, using the root-mean-square error calculated in the training stage as the optimisation parameter. The predictive accuracy of the optimised ANN, consisting of two nodes in the input layer, four nodes in the hidden layer, and a single node in the output layer, was tested by the “leave-k-out” method. From the results obtained, ANN provided quite reliable predictions when the applied load was sufficiently far from the failure load, performing better than a previous theoretical model, relying on fracture mechanics concepts. Therefore, ANN was shown to be a valid tool in the non-destructive evaluation of composite materials employed in fatigue-sensitive applications.