A rapid calibration method for the discrete element model of straw fodder particle mixtures based on the UVE-PLS-GD algorithm

The accuracy of parameter settings for straw fodder particle mixtures, which include straw rind, pith, bracts, and dust, is crucial for in-depth research on the mechanical behaviours of straw, such as motion, collision, and separation between particles. This study proposes a method for the rapid calibration of the discrete element model for straw fodder particle mixtures, based on the Uninformative Variable Elimination-Partial Least Squares-Gradient Descent (UVE-PLS-GD) algorithm. The UVE-PLS-GD algorithm and traditional experimental methods were used to calibrate the numerical simulation parameters of individual particles in the mixture, followed by comparative validation. The experimental results indicated that both methods identified the factors significantly affecting the repose angle, in the following order: JKR surface energy coefficient between bracts and rind, bracts and pith, pith and rind, pith and pith, rind and rind, and bracts and dust. The relative errors between the repose angles obtained from the numerical simulations using the UVE-PLS-GD algorithm and traditional experimental methods and those from the physical experiments were 0.37 % and 0.45 %, respectively, verifying the accuracy of the calibrated numerical simulation parameters. Among them, the parameters obtained using the UVE-PLS-GD algorithm resulted in the smallest error, and the optimal combination of significant factor parameters were 0.1337, 0.0862, 0.1339, 0.1238, 0.1254, and 0.1186. This study demonstrated that the UVE-PLS-GD algorithm can significantly improve both the accuracy and speed of calibrating the discrete element model for particles, and the calibrated numerical simulation parameters can provide data support for the motion and separation of mixtures.