Autoregressive Modelling and Synthetic Generation of High-Fidelity, Statistically Equivalent 3D Microstructures for As-Manufactured Misalignments in Fiber-Reinforced Composites

This study presents an integrated framework for processing, modelling, and generating statistically representative three-dimensional fiber microstructures from experimental X-ray μCT observations. The framework introduces an analytical slice-segment ellipse-intersection method to extract per-fiber misalignment profiles, constructs a stochastic autoregressive model with copula-based dependence, and employs Bayesian optimization for calibration. The optimized statistical model is coupled with a physical generation strategy using Delaunay-based neighbourhood construction and ellipse-based contact resolution. The framework successfully generates approximately 2,400 synthetic fibers while preserving strong statistical fidelity to the original X-ray μCT data, providing a scalable route for generating simulation-ready fiber composite microstructures for virtual testing and analysis.

Mohamad A. Raja, Clemens Dransfeld, Boyang Chen. arXiv:2606.20117 [cs.CE], 2026.
Reference | Download Paper