Critical-Scale Lunar Soil Particle Identification Based on Ensemble Learning

To address the impact of critical-scale lunar soil particles on the drilling process, a parameter identification model for critical-scale lunar soil particles based on ensemble learning algorithms was proposed. Discrete element method (DEM) simulation software was employed for simulation analysis, and a central composite design was used to conduct experiments under various operating conditions. The load characteristics of lunar drilling tools under different conditions were obtained. The model took critical-scale lunar soil particle size and offset position as main variables. Torque data from the lunar drilling process were collected and analyzed. By extracting the features of the experimental data, an ensemble learning algorithm was applied to establish regression models for identifying particle size and offset position under two working conditions, with drilling torque as the input data. Experimental results show that the ensemble learning algorithm achieved high accuracy in predicting particle positions, with a mean squared error (MSE) of 0, while the prediction error for particle size was larger, with an MSE of 1.61. These findings can provide a reference for developing identification model technologies for autonomous drilling and sampling.