A set of suitable parameter values is crucial in discrete element method (DEM) simulations. As a non-spherical particle, the coefficients of rolling friction between maize seed particles and between the seed particles and the boundary are hard to measure directly and must be obtained by calibration. In this paper, taking three representative maize varieties as research objects, the necessity for the accurate calibration of the coefficients of rolling friction between seed particles and between seed particles and boundaries is investigated. Subsequently, by studying the sensitivity of the coefficients of rolling friction between seed particles and between seed particles and the boundary to the angle of repose test, the methods used to calibrate the coefficients of rolling friction between seed particles and between the particles and the boundary are determined. It can be seen from the results that the coefficients of rolling friction between the seed particles and between the seed particles and the boundary have a significant influence on the test results, so these two parameters must be accurately calibrated. Additionally, the coefficient of rolling friction between seed particles and the boundary has no effect on the piling angle, but the piling angle is highly sensitive to the coefficient of rolling friction between seed particles. By comparing the simulation results and the experimental results in the lifting cylinder and “self-flow screening” tests, the calibrated seed parameters were found to be accurate and valid.

To solve the poor universality in the existing modelling approaches of soybean particles, we proposed a soybean particle modelling approach by combining five, nine, and 13 balls. The soybean seeds from three varieties (Suinong42, Jidou17, and Zhongdou39 with a sphericity of 94.78%, 86.86%, and 80.6%, respectively) are chosen as the study objects. By the comparisons between the simulation results and the test results in the “self-flow screening” and “piling angle” tests, it is concluded that the soybean particle modelling approach we presented in this paper is a universal modelling approach appropriate for soybean particles with different sphericities. The five-ball model is appropriate for the soybean particles with high sphericity, and the nine- or 13-ball models are applicable to those with low sphericity. The soybean particle modelling approach we presented is also compared with the ellipsoidal equation modelling approach for soybean particles and with the modelling approaches presented by other researchers. From an overall perspective, the soybean particle modelling approach we presented is better than the ellipsoidal equation modelling approach and those modelling approaches presented by other researchers. Additionally, it is shown that the multiple contacts issue in the multi-ball model has a little influence on the simulation results of soybean particle models. The study in this paper provides a new modelling approach for soybean particles in the DEM simulation of the contacts between soybean particles and the related machines.