A micromechanics-based artificial neural networks model for rapid prediction of mechanical response in short fiber reinforced rubber composites

Abstract

The complex microstructural characteristics inherent in short fiber reinforced rubber composites (SFRRCs) impose considerable computational burdens in predicting the mechanical behavior of such composite materials. To address this challenge, this research extends the applicability of the homogeneous model predicated on the orientation averaging method to encompass composite materials featuring hyperelastic matrices. Combined with finite element method, a comprehensive mechanical response database encompassing various volume fractions and fiber orientation distributions is established. Leveraging this database, a micromechanics-based artificial neural network (ANN) model is meticulously designed to rapidly predict the mechanical response of SFRRCs across varying volume fractions and fiber orientation distributions, utilizing a fixed strain step strategy. To ascertain the efficacy and precision of the developed ANN model, representative volume elements portraying both planar and three-dimensional random distributions of composites are constructed and subjected to finite element analysis. Results indicate that the predicted outcomes from the ANN model align closely with finite element calculations within a certain strain range, while significantly reducing computational costs.