Composilator_Code

This repository contains the implementation of ComposiLator, a hybrid AI–mechanistic surrogate model for stiffness prediction in heterogeneous polymer composites. The framework combines classical Voigt–Reuss micromechanical bounds with a Random Forest correction factor trained on structured morphological descriptors.Unlike purely empirical regressors, the model is physically constrained: predictions are restricted within Voigt–Reuss limits and satisfy endpoint consistency (matrix recovery at zero reinforcement). The machine-learning component operates at the descriptor level, capturing morphology- and interphase-dependent deviations from mixture-rule scaling while preserving mechanistic interpretability.The code enables full reproduction of the calibration and independent validation results reported in the associated study and serves as a reusable template for physics-informed machine learning in composite materials design and heterogeneous system modeling.