A novel artificial neural network model for wide-band random fatigue life prediction

Abstract

In this paper, a novel artificial neural network (ANN) model is developed to achieve fatigue lifetime prediction under wideband random loads. Many power spectral density (PSD) samples with different spectral shapes and a wide range of material parameters related to fatigue life are used to improve the generalization ability of the model. The bandwidth parameters are used as the input neurons for the first time, compensating for the defect that the existing neural network models have too many input neurons.Numerous numeric simulations demonstrate that the developed model is more accurate in life prediction than the existing frequency domain models.