Mechanical tillage before cotton sowing is a crucial process in cotton production. Numerical simulations of soil cutting and energy consumption predictions, along with optimization methods, are very important for understanding the interaction between tillage tools and soil, as well as guiding energy-efficient cultivation practices. The focus of this study is on the problem of cutting sandy silt in Xinjiang cotton fields. Sandy silt can be characterized by its low cohesion and large, loose particles. Starting from the macroscopic physical and mechanical properties of the soil, a soil contact mechanics model considering soil plastic deformation and bonding forces between soil particles is established. By optimizing the cotton field soil discrete element model and parameter calibration methods, the accuracy of the soil cutting simulation is improved. The principles and modelling steps of discrete element method (DEM) simulations for cutting soil are explained in detail, enabling the analysis and evaluation of the complex dynamic behaviour of soil under large deformation conditions and the mechanical properties of the cutting tool. The average error between the energy consumption measured in field rotary tillage experiments and simulation results is 7.04%. By utilizing the simulation results as a dataset, an extreme learning machine (ELM) without a physical model is employed to replace traditional polynomial regression for rapid energy consumption prediction based on the cutting parameters. The average error between the prediction results and simulation results is 4.34%. By using response surface methodology based on the predicted energy consumption, optimal working parameters are determined, resulting in a 10.02% reduction in the power consumption compared to the initial parameter settings. This effectively achieves energy savings in rotary tillage and further validates the accuracy of the simulation method and prediction model.
