A review of recent machine learning advances for predicting and optimizing mechanical and tribological properties of fiber-reinforced thermoset polymer nanocomposites

Abstract This review article comprehensively examines how machine learning (ML) methods can be used to predict the mechanical and tribological properties of synthetic fiber-reinforced thermoset composites, with and without nanofillers. In the beginning, the basic ideas of supervised and unsupervised machine learning techniques, such as artificial neural networks, support vector machines, random forests, and Gaussian process regression, are described, with their applications in materials science. Subsequently, recent research published between 2020 and 2025 is thoroughly examined with respect to composite systems, input parameters, dataset features, and anticipated outputs, including wear rate, tensile strength, flexural strength, and stress-strain behavior. A serious discussion is held regarding the impact of nanofillers (such as silica, graphene, carbon nanotubes, and nanoclay) on composite performance and model accuracy. Furthermore, techniques like multi-fidelity modeling and physics-informed ML are discussed, and issues with small datasets, feature selection, and model generalization are explored. Lastly, suggestions for future study are made, with a focus on sustainable composite design, open databases, and data standardization. This paper shows that machine learning techniques can enable accurate prediction and optimization of fiber-reinforced epoxy/polyester/vinyl ester nanocomposites while drastically reducing experimental effort.