Abstract: Particle size is an important factor affecting macroscopic mechanical properties and computational efficiency of particle discrete element model. To account for the uncertainty of numerical results caused by the stochastic model, we study the effect of particle size using statistical methods. The result of the global test shows that the distribution position of mechanical parameters (i.e. peak strength, elastic modulus, Poisson's ratio and peak strain) and the failure characteristic parameter (bond failure rate) is significantly affected by the characteristic length ratio L/R. The coefficient of variation (CV) of these parameters increases as L/R decreases. The further multiple comparisons show that L/R has no significant effect on the distribution of these parameters as L/R≥125. When L/R≥79, the L/R has no significant effect on the distribution position of bond failure rate within the range of three adjacent ball radius levels. With the decrease of L/R, the damage degree of the model increases gradually, and the failure mode is changed from the whole shear failure to the unstable failure caused by the local damage, which makes the model lose the simulation effectiveness to rock materials. Finally, by combining the statistical results of mechanical parameters, the failure mode of the model and the computational efficiency, the L/R=200 should be suitable for the simulation of rock materials.

Key words: macroscopic mechanical properties, rock model, statistical test, particle size effect, characteristic length ratio