Neural Networks with Radial Basis Function and NARX Structure for Material Lifetime Assessment Application

In the present paper, neural networks (NN) with radial basis function and non-linear auto-regressive exogenous inputs (NARX) structure is introduced and first applied for predicting fatigue lives of composite materials. Fatigue life assessment of multivariable amplitude loading linked to the concept of constant life diagrams (CLD), the well known concept in fatigue of material analysis and design, was investigated. With this respect, fatigue life assessment using the RBFNN-NARX model was realized as one-step ahead prediction with respect to each stress level-S corresponding to stress ratio values-R arranged in such a way that transition took place from a fatigue region to another one in the CLD. As a result, composite materials lifetime assessment can be fashioned for a wide spectrum of loading in an efficient manner. In addition, the produced mean squared error (MSE) values of fatigue life prediction results of the RBFNN-NARX model competed favorably, even better, with those of the MLP-NARX model previously obtained. The simulation results for different multidirectional laminates of polymeric-based composites and loading situations were presented and discussed.