Artificial Neural Network-based fatigue behavior prediction of metals and composite materials

Abstract

This article presents a study devoted to predicting the fatigue behavior of two different materials: aluminum alloy AL-2024-T6 and glass fiber composite samples. The approach used in the study involves the use of artificial neural networks (ANNs) to develop accurate models for predicting the fatigue life of these materials at various skewness ratios (R). For the first case study, the S-N curve of tensile-tested AL-2024-T6 was predicted for different values of R using a few sets of data for learning. The model was then tested on the same values of R as the learning set, as well as on a different value of R (-0.4), to demonstrate the ability of the model to predict fatigue data under varying conditions. The results showed that the model was capable of accurately predicting the fatigue life of AL-2024-T6 for different values of R. For the second case study, the stiffness degradation of bending-tested glass fiber woven composite samples was predicted for different values of R using ANN. Different layups of composite samples were considered in this study. The model was trained on a few sets of data and tested on the same and different values of R, demonstrating the ability of the model to accurately predict stiffness degradation of composite samples under varying coefficients of asymmetry. The results of both case studies showed that ANN-based models can be effective in predicting the fatigue behavior of different materials tested using different methods under varying coefficients of asymmetry. These findings have practical implications for industries involved in the design and manufacturing of materials, particularly in the aerospace and automotive sectors, where fatigue behavior is critical to the structural integrity of components.