A Voronoi-based gaussian smoothing algorithm for efficiently generating RVEs of multi-phase composites with graded aggregates and random pores

This paper presents an innovative numerical modeling framework capable of generating highly realistic 3D mesoscale multi-phase concrete models with unprecedented efficiency and accuracy. Addressing a significant wide range in aggregate volume fraction (0 to 80%) and featuring rapid model generation capabilities, our framework marks a breakthrough in reducing computational time—achieving simulations of 1 million elements within mere 30 s on readily available hardware. By analyzing randomly generated concrete samples across varying compositions, our algorithm can provide deep mesoscopic insights into their compressive properties, validated against a wide array of experimental results. Additionally, our comprehensive analytical model sheds light on the intricate roles of aggregate volume fraction and porosity, enhancing understanding of the meso-scale compressive behavior. We also make the source code available on GitHub, offering a valuable tool for the engineering and research community to optimize concrete material design and performance.