SOIL PARTICLE MODELING AND PARAMETER CALIBRATION FOR USE WITH DISCRETE ELEMENT METHOD

The discrete element method (DEM) has broad prospects for application in soil-tool simulations. To ensure the reliability of simulations, appropriate simulation parameters and particle modeling are essential. Therefore, in this article, a method combining simulation and actual tests is proposed to calibrate the critical soil parameters. First, the effect of soil particle shape on particle contact was considered. Soil particle shapes were statistically analyzed using an improved GrabCut algorithm and k-means algorithm. Four representative soil particles were obtained. Second, a soil particle model was established by microscope and three-dimensional non-contact surface topography. Finally, taking the angle of repose as the response value, the three parameters with significant effects on the angle of repose, i.e., soil shear modulus, HertzMindlin with Johnson-Kendall-Roberts contact model (JKR), and soil-soil restitution coefficient, were obtained via a Plackett-Burman experiment. The optimal value intervals of the significant parameters were determined by the steepest climbing test. A polynomial regression model between the angle of repose and the three significant parameters was established with a Box-Behnken experiment using three factors and three levels. The interactions between the three significant parameters were not significant, as revealed by response surface analysis. The optimal values of the significant parameters were obtained by taking the actual angle of repose as the target and resulted in a soil shear module of 9.8 MPa, JKR of 0.063, and soil-soil restitution coefficient of 0.478. To verify the reliability of the calibrated parameters, the soil angles of repose from the simulation and from actual tests were compared and analyzed. For a simulated angle of repose of 38.5 degrees, the actual angle of repose was 38.6 degrees, and the relative error was 0.26%. DEM was also used to simulate a rotary tiller with the calibrated parameters. The maximum error of the simulated soil throwing angle was less than 10% when compared with the actual throwing angle. The experimental results showed that the calibrated parameters were accurate and can provide a reference for the selection of soil discrete element parameters.