Neural network paradigms for fatigue strength prediction of fiber-reinforced composite materials

The paper presents artificial neural network models to evaluate the fatigue life of unidirectional glass fiber-reinforced epoxy-based composites under tension-tension and tension-compression loading. The fatigue behavior of the composite materials was analyzed using three parameters: fiber orientation angle, stress ratio, and maximum stress. These parameters formed the input vectors, and the number of cycles corresponding to the failure was taken as the output parameter for the assessment of the fatigue life. The architecture of the network was selected based on a detailed parametric study and it was trained and tested with data generated analytically using finite element analysis. The predicted results of the neural network model were compared with the available experimental values and were found to be in good agreement. Three different networks such as feedforward, recurrent, and radial basis function networks were used in the present investigation, and a comparative study was carried out to get the optimum network. The significance of the present work is that the same network could be used for assessing the fatigue strength of unidirectional glass/epoxy composite specimens with different fiber orientation angles tested under different stress ratios.