Key points are not available for this paper at this time.
Fiber-reinforced polymer (FRP) composites are widely used in aerospace, automotive, marine, and construction sectors due to their high strength-to-weight ratio, corrosion resistance, and design versatility. Achieving optimal mechanical performance, however, remains challenging due to the complex interactions among fiber type, filler loading, chemical treatment, orientation, and processing parameters. To address this, advanced modelling approaches such as Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) are increasingly adopted to reduce experimental efforts and enhance prediction reliability. RSM enables efficient experimental design, factor significance assessment, and multi-response optimization, while ANN provides superior nonlinear prediction capability with minimal dependence on underlying data assumptions. Recent research highlights the growing use of hybrid RSM-ANN methodologies that combine the interpretability of RSM with the high prediction accuracy of ANN, especially for multifactor nonlinear composite systems. Comparative studies indicate that although RSM effectively identifies key parameters and visualizes interactions, ANN typically achieves higher accuracy, with reported R 2 values exceeding 0.95. Hybrid models further improve prediction precision, reduce the number of experimental iterations, and capture complex input-output relationships beyond the capacity of standalone models. Remaining challenges include limited standard datasets, interpretability concerns, and generalization across material systems. Future opportunities lie in deep learning, physics-informed neural networks, metaheuristic-ANN coupling, and digital monitoring for next-generation composite design.
Shettar et al. (Tue,) studied this question.