An advanced numerical method is proposed to generate more realistic 3D meso-scale models for concrete-like particle-reinforced composites with high volume fraction of particles. In traditional methods, it usually consumes an excessive amount of computational time, and even worse, it cannot be realized when the volume fraction exceeds the computational limitation. A novel idea is presented in this paper in which free fall acceleration is applied to simulate the falling process of randomly distributed particles. Firstly, certain numbers of convex particles are generated based on a grading curve, and then the particles are randomly placed into a higher space than the required specimen size in height. The time for the intersection detection is significantly saved as the intersection possibility among particles in a larger space is decreased. Secondly, the surfaces of the generated particles are meshed with shell elements, and then the particles fall freely into a certain container. Finally, the matrix and particles are meshed with solid elements for further numerical analysis. Comparing with other modeling methods, this method can establish models with higher particle volume fraction. Besides, the minimum gap between adjacent particles can be flexibly controlled by changing the thickness of the shell elements. So numerical models with different volume fractions can be obtained easily. Moreover, computational cost and time can be effectively reduced. Based on this method, some numerical simulations have been carried out, and the results of numerical simulations in different loading cases all have good consistency with experimental data, which implies the accuracy of the method and the isotropy of the phase arrangements. This method is of great importance in performance analysis, structural optimization, and material design for particle-reinforced composites.
Read full abstract