EffectiLator: Morphology-dependent model selection for elastic moduli prediction in polymer composites

Accurate prediction of elastic moduli in polymer composites remains challenging because reinforcement participation depends on morphology and interphase characteristics that cannot be represented by a single micromechanical relation. This work introduces EffectiLator , a physics-constrained framework that selects between alternative effective-volume formulations within strict Voigt–Reuss bounds, enabling morphology-dependent representation of reinforcement participation. The approach does not regress elastic properties directly; instead, it identifies the participation regime governing stiffness evolution and activates the corresponding mechanistic pathway. The framework is trained on 73 composite systems and validated on 33 independent systems (234 datapoints) spanning diverse matrices, reinforcement types, and interphase conditions. On validation data, EffectiLator achieves a mean absolute error of 0.215 GPa and RMSE of 0.420 GPa while preserving strict micromechanical admissibility across heterogeneous composite systems. The results indicate that logarithmic-type participation scaling provides a robust baseline description for many polymer composites, suggesting a dominant reinforcement mechanism associated with gradual load-transfer development. However, systematic deviations from this behavior are observed in specific morphology classes, indicating the presence of distinct participation regimes that cannot be captured by a single model form. By extending effective-volume concepts from single-series calibration to cross-family regime identification, the proposed framework enables descriptor-conditioned selection between alternative effective-volume formulations while retaining interpretability and micromechanical consistency. More importantly, it establishes a systematic methodology for the objective comparison and evaluation of alternative effective-volume formulations within a common descriptor-informed framework, thereby providing a transferable basis for future developments in morphology-dependent composite modeling.