Artificial neural networks for fatigue life estimation of glass-carbon/epoxy hybrid composites

Hybrid composites enable to ingeniously design typical machine components by reinforcing much stiffer and stronger carbon fibers and inexpensive glass fibers within the same resin matrix to achieve better mechanical advantage to suit specific requirements, particularly where properties are required in the direction of loading. In this study, the influence of applied stress levels and operating frequencies on the fatigue strength of hybrid [HC]₄ and non-hybrid composite [NHC]₄ specimens were evaluated by modeling the fatigue strength of composites using artificial neural networks (ANN). With increasing alternating stress levels and test frequencies, the fatigue strength of [HC]₄ and [NHC]₄ composites decreased significantly. Further, the Levenberg–Marquardt (LM) approach was utilized to train and evaluate the fatigue strength of [HC]₄ and [NHC]₄ composites. Plotting the S–N curve on the linear σ_a-N plane shows that the artificial neural network (ANN) derived curve agrees more closely with the experimental fatigue life. The cumulative correlation, R² value was found to be 0.99133 and 0.99514 for [HC]₄ and [NHC]₄ composites respectively. As a result, the ANN model aligns closely with the experimental results.