Abstract: The uncertainty in the grain-based model, such as grain shape, size and spatial distribution, has a significant effect on the reliability and reasonability of numerical simulation for rocks. Based on the Voronoi model in UDEC, two modified Trigon models were proposed and realised through the built-in programming language FISH. Meanwhile, the effects of grain shape, iteration and random seed on the macroscopic response and failure mode of the coal sample were analysed in uniaxial compression test. The results showed that the effects of iteration and random seed on uniaxial compression strength and deformation of coal samples depended on the grain shape. For the two Trigon models, the influences of iteration and random seed were similar. However, for the Voronoi model, the influence of random seed was greater than that of iteration. Furthermore, the type, number and development of cracks were mainly affected by the grain shape. The failure modes of the coal sample were determined by the grain shape, iteration and random seed. Hence, the reliability of numerical simulation in the UDEC grain-based model can be improved by considering the real grain shape, grain size and spatial distribution of rocks.

Key words: discrete element, grain-based model, micromechanics, iteration, random seed