Fatigue life prediction of glass fiber reinforced epoxy composites using artificial neural networks

Abstract

In this work, fiber-reinforced composites, consisting of glass fibers with an epoxy matrix, have been considered to assess their fatigue life under combined rotating bending loads. Rotating bending fatigue tests were performed under fully reversed loading cycles at an operating frequency of 10 Hz. Fatigue test results revealed that the fatigue life of glass fiber reinforced with epoxy (GFRE) matrix composite decreased with an increase in stress levels. In addition, an artificial neural network approach was used in this study to demonstrate the impact of stress levels on the fatigue strength and failure modes of GFRE composites. The neural network was trained by the Levenberg-Marquardt algorithm using MATLAB. The feasibility of artificial neural networks for modeling and predicting the fatigue behavior of GFRE composites was verified by the correlation results (R2 = 0.99857) between the experimental data and ANN outputs. Three failure modes viz. matrix cracking, interfacial debonding and interfacial splitting of fibers were observed under low and high cycle fatigue tests. Further, ANN was successful in classifying these failure modes using a conjugate gradient backpropagation algorithm with 100% accuracy.