Optimized Mori-Tanaka Model with FEM-RVE Corrections and Machine Learning for High-Fidelity Composite Mechanical Property Estimation

The accurate prediction of composite mechanical properties remains a critical challenge in materials science, particularly in bridging the gap between analytical micromechanics and computationally expensive numerical simulations. Traditional homogenization techniques, such as the Mori-Tanaka model, often fail to capture microstructural interactions at high fiber volume fractions, leading to significant deviations from experimental results. To address this gap, our research introduces an optimized Mori-Tanaka model, enhanced by Representative Volume Element (RVE) corrections and machine learning-driven optimization techniques. By leveraging higher-order polynomial regression, least squares optimization, and AI-based predictive models, we refine the Mori-Tanaka estimations, achieving near-experimental accuracy while maintaining computational efficiency. Our innovative approach integrates FEM-based RVE simulations, particularly using the UD Square configuration, to systematically calibrate correction factors, enabling robust predictions of longitudinal modulus (E1), transverse modulus (E2), and shear modulus (G12). This hybrid methodology bridges the gap between analytical and numerical techniques, significantly improving stiffness estimations across various fiber volume fractions. The optimized solver enhances predictive accuracy and ensures a scalable and cost-effective solution for composite material characterization. By combining micromechanics, FEM, and AI-driven optimizations, our research provides a next-generation framework for efficient composite property predictions, enabling advanced material design in aerospace, automotive, and structural applications. This breakthrough methodology lays the foundation for future research in nonlinear analysis, failure modeling, and adaptive AI-driven material optimization, ensuring superior performance and efficiency in composite engineering.