Transverse Failure Mechanisms in UD ‐ FRP Composites: A Parametric Micromechanical Study Spanning Broad Modulus Regimes via Interepretable ML

ABSTRACT The transverse mechanical integrity of unidirectional fiber‐reinforced polymer (UD‐FRP) composites is critical for structural performance yet remains difficult to predict due to complex nonlinear couplings among microstructural features, fiber transverse modulus, and interfacial properties. This study establishes a unified, data‐driven micromechanical framework to systematically investigate transverse failure mechanisms across a broad parametric space spanning these critical variables. A high‐fidelity database comprising 648 Finite Element Analysis (FEA) simulations on Representative Volume Elements (RVEs) was generated. The interpretable tree‐based Machine Learning (ML) architecture achieved a prediction accuracy of R ² > 0.91, and Symbolic Regression (SR) subsequently extracted explicit algebraic equations (R ² > 0.81). SHapley Additive exPlanations (SHAP) analysis quantified distinct failure regimes governed by modulus mismatch: (i) under transverse tension, a “modulus‐mismatch penalty” was identified: fiber transverse modulus beyond 60 GPa causes high interfacial strength to compromise global capacity by intensifying matrix stress; and (ii) under transverse compression, a “rigid obstacle effect” was revealed, where high‐modulus fibers ( E _f > 50 GPa) deflect shear bands to enhance strength, effectively compensating for weaker interfaces. These findings culminate in quantitative design maps and explicit analytical formulas, demonstrating how integrating interpretable ML with Symbolic Regression accelerates composite optimization beyond traditional trial‐and‐error.