Fatigue life prediction based on Artificial Neural Network

Since damage tolerance concept has been widely used in the aerospace engineering, the fatigue crack propagation prediction is playing an important role in Structural Prognostic and Health Management (SPHM). The fatigue damage evolution of metallic structures is fundamentally a nonlinear process even under the constant amplitude loading, which makes the fatigue life prediction difficult. The Artificial Neural Network (ANN) is a powerful tool to perform nonlinear multivariate projection. In this paper, a novel fatigue life prediction method based on ANN is proposed to predict the fatigue crack propagation life. The fatigue crack growth rate data with different stress ratios (da/dN−ΔK−R) in various aluminum alloys obtained from the available literatures are used to train the network. And then an algorithm based on the well-trained ANN is developed to calculate the crack propagation cycle by cycle. The results show the good agreements with the experimental data. Forman’s equation is also employed to compute the fatigue life. And ANN works much better than the Forman’s equation for most given failure criterions. The current investigation indicates that the ANN-based model is probably a sophisticated method to describe the fatigue damage evaluation which is a typical nonlinear and multivariate process. Furthermore its good performance implies its great potential in predicting the fatigue life even under the variable amplitude loading spectrum.