Non-Periodic RVEs for Composite Modeling via Simplified Random Sequential Expansion

Abstract Realistic digital microstructures are increasingly recognized as essential for predicting the mechanical behavior of fiber-reinforced composites within multiscale modeling frameworks. This study presents a computationally efficient Simplified Random Sequential Expansion (sRSE) algorithm for generating threedimensional, statistically random, non-periodic representative volume elements (RVEs) with non-overlapping fiber distributions. The sRSE simplifies the original RSE framework while enhancing its performance, introducing randomized reference fiber selection and spatial hashing for efficient, statistically robust microstructure generation. These enhancements enable rapid generation of digital microstructures that preserve statistical authenticity across a wide range of fiber volume fractions, achieving up to 68% in the present implementation. Validation against experimental benchmarks—using nearest-neighbor distributions, Ripley’s K function, and pair correlation functions—confirms that sRSE produces non-periodic fiber arrangements that exhibit true spatial randomness. When integrated into a finite element workflow with kinematic (symmetric) boundary conditions, the generated microstructures predict transverse elastic properties of glass-fiber/epoxy composites within 3% of experimental results. The proposed approach offers a robust and scalable pathway for producing non-periodic, simulation-ready microstructures, with direct applications in virtual testing, multiphase material design, and machine learning–driven property prediction.