Abstract: As the parameters of discrete element model for yam harvesting are difficult to obtain directly from literatures or experiments, this study proposes a parameter calibration method that combines experiments with numerical simulations. Firstly, a physical test was carried out. Compression, shear and bending tests were carried out on yam with the moisture content of 80%. The axial and radial elastic modulus of yam were obtained, which were 4.48 and 4.03 MPa, respectively. The results showed that the axial and radial elastic modulus were close, and the structure of yam was homogeneous. At the same time, the density, length and radial dimension of yam were measured. Based on the discrete element method, the bimodal distribution model of yam was established by the particle filling method. According to the research theories of domestic and foreign scholars, the bonding parameters of yam were calibrated. Through the compression and shear simulation tests in the discrete element software, the calibration parameters were checked and verified to make the calibration results more accurate. The experimental results showed that there were 9 887 non-equal size particles and 111 723 effective bonds in the bimodal distribution model of yam, the coordination number was about 11, the porosity was low, which could better characterize the mechanical properties of yam. The normal stiffness, shear stiffness, critical normal stress, critical shear stress and bonding radius were 9.3×105 N/m, 3.0×106 N/m, 0.58 MPa, 0.14 MPa, 3.5 mm. Secondly, the contact parameters of sand loam particles were calibrated by the soil accumulation angle test. The particle size of sandy loam was less than 4 mm, and the moisture content was 14.15%. After screening, the particles of sandy loam met the requirement of uniformity. The method of funnel test was used to form a certain stacking angle of sandy loam and the stacking body was photographed on the side. The acquired image was grayed and binarized by MATLAB and the edge was detected by Canny operator. Then the image was imported into Origin software to obtain the contour point coordinates. The scatter map was fitted with Gaussian distribution to find the inflexion point of the image and the sharp angle between the tangent of inflection point and X axis as the accumulation angle of soil was took. Then, several tests were carried out to determine the accumulation angle of sandy loam. Combined with the Box-Behnken response surface analysis test, the regression model between the sand soil accumulation angle and coefficient of restitution, coefficient of static friction, coefficient of rolling friction and surface energy was established and optimized. The optimal parameter combination was 0.42, 0.20, 0.30 and 0.40 J/m3. The results showed that the average value of simulation soil accumulation angle was 41.25°, the average value of actual soil accumulation angle was 41.87°, and the error between them was only 1.48%. Finally, the simulation sample of yam, steel plate and particle bed model were established in EDEM software. Through the test of coefficient of restitution, static friction coefficient and coefficient of rolling friction, the coefficient of restitution, static friction coefficient and coefficient of rolling friction between yam-steel plate and yam-sandy loam were 0.34, 0.26, 0.049, 0.21, 0.38 and 0.075, the calibration results was reliable and accurate compared with the actual experimental errors of less than 5%. It provides accurate contact parameters for the subsequent establishment of composite model of yam and sandy loam, thereby we can visualize and parameterize the harvest process of yam; meanwhile, the calibration results can provide theoretical basis for the relevant simulation test of yam post-production processing and the promotion of the whole mechanization process of yam industry.