Artificial intelligence-based lunar soil granular material mesoscopic parameter calibration method considering real shape

The invention discloses a lunar soil granular material mesoscopic parameter calibration method considering a real shape based on artificial intelligence. The method comprises the steps that discrete element mesoscopic parameters to be calibrated of a contact model and the range of the discrete element mesoscopic parameters are determined; performing discrete element biaxial test simulation; processing macro-micro parameters, and constructing a sample group; training a machine learning model; and calibrating discrete element mesoscopic parameters. According to the method, the real shape of the granular material is considered in the parameter calibration process, the accuracy of discrete element simulation is improved, and more accurate analysis on mechanical characteristics and engineering characteristics of the material is facilitated. In addition, the method also has certain extrapolation, can be popularized to the microscopic parameter calibration process of various other materials and other contact models,