Abstract: To investigate the influence of critical-sized lunar regolith particles on the drilling process, a parameter identification model based on ensemble learning algorithms is proposed. Discrete Element Method (DEM) simulation software is employed for modeling and analysis. Various working conditions are constructed using Central Composite Design (CCD), enabling the acquisition of load characteristics of lunar drilling tools under different scenarios. Focusing on the size and offset position of critical-sized particles as core variables, drilling torque data are collected throughout the drilling process, from which key feature values are extracted. On this basis, ensemble learning algorithms are introduced for the first time to identify parameters of critical-sized lunar particles, resulting in the development of a bivariate model capable of simultaneously identifying both particle size and position. Experimental results show that the model demonstrates exceptional performance in predicting particle position, achieving a mean squared error MSE of 0, indicating extremely high prediction accuracy. For particle size identification, the model also performs well, with an MSE of 1.61. This model effectively addresses the challenge of parameter identification under conditions where particle size and position cannot be directly observed. The findings provide a valuable reference for the development of identification model techniques in unmanned drilling and sampling applications.