DISCRETE ELEMENT PARAMETER CALIBRATION AND OPTIMIZATION OF LAMINARIA JAPONICA BASED ON GP-PSO-XGBOOST MODEL

During the mechanized harvesting of Laminaria japonica, it is prone to breakage and damage, resulting in an increased loss rate. To accelerate the optimization of harvesting equipment for Laminaria japonica, this study established a simulation model based on the discrete element method. The Hertz-Mindlin with Bonding contact model was used, and parameters of Laminaria japonica were calibrated through shear tests. Using the maximum shear force (Fmax) as the test indicator, the optimal parameters were obtained through Plackett-Burman test, the steepest climb test, Box-Behnken test, and the GP-PSO-XGBoost regression prediction model. The results indicated that when the coefficient of restitution of Laminaria japonica-steel was 0.45, the normal stiffness per unit area was 303 MN/m3, the shear stiffness per unit area was 378 MN/m3, and the bonding radius was 0.70 mm, the relative error of Fmax was 0.75%. The average error of the Fmax for samples at different thicknesses was 3.09%, and the relative error of the maximum puncture force in puncture test was 5.59%. Finally, a discrete element model of the whole Laminaria japonica was established. This study offers theoretical support for the development and optimization of harvesting equipment for Laminaria japonica.