Fiberglass-Reinforced Polyester Composites Fatigue Prediction Using Novel Data-Intelligence Model

Abstract

During the design and inspection of polyester composites reinforced with fiberglass for structural engineering application, the accurate and reliable prediction of the fatigue failure is essential. In this paper, extreme learning machine model (ELM) which is a modern data-intelligence model is developed for predicting fatigue cycle in composite materials. The fatigue of a particular fiberglass-reinforced material was predicted using its specific experimental fatigue data by determining the number of cycles based on the loading conditions that were predetermined. Several modeling input variables such as the geometry dimension, the stress, and the orientations were considered during the modeling, and the results of the ELM modeling approach were ascertained against artificial intelligence-based model called generalized regression neural network (GRNN). The results of the prediction study suggested a better performance of the tested modern model ELM over the classical GRNN. Also, the best attributes to conduct the optimal predictive models were the geometry of the samples and the applied stresses. Quantitatively, the absolute error values (root-mean-square error and mean absolute percentage error) were enhanced by 39 and 38% over the testing phase of the modeling.