A novel machine learning framework informed by the fractional calculus dynamic model of hybrid glass/jute woven composite

Abstract

This study investigates the influence of glass‐jute fiber hybridization on the dynamic viscoelastic performance of woven natural fiber composites. Experiments elucidated the effects of low‐frequency vibration and glass/jute ratio on the storage modulus and loss factor. Results exhibited nonlinear escalations in storage modulus and loss factor with increasing frequency and glass fiber content. The loss factor also demonstrated nonlinear rises with frequency and jute content. A fractional calculus‐based two‐branch model successfully captured these behaviors by incorporating frequency and hybrid ratio as key variables, showing excellent agreement with measurements. To further improve predictive accuracy, an artificial neural network informed by the fractional calculus dynamic model is implemented by enforcing physical constraints during training. The physics‐informed artificial neural network achieved higher correlation to experiments than unconstrained models, affirming the value of fusing physics knowledge into data‐driven models. This study highlights the promise of physics‐guided machine learning for predicting the intricate dynamics of sustainable natural fiber hybrid composites. The integrated analytical and data‐driven techniques provide a pathway to comprehensively model the mechanical performance of advanced multiphase materials over diverse conditions.