A discrete element modeling of rock and soil material based on the machine learning

The discrete element modeling of rock and soil mass usually relies on tedious parameter adjustment mechanical property testing operations to obtain appropriate parameters the element. In this paper, based linear elastic contact theory, a model samples was established software MatDEM with high-performance numerical analysis, actual properties were predicted using machine learning method, XGBoost algorithm. A certain amount sample data sets generated, through experiments, training, validation test operation (compressive strength, Young's modulus, Poisson's ratio tensile strength), quantitative relationship between input explored, showing following results: (1) Based characteristics 12000 correlation tag shows that are positively correlated properties;(2) With increase number in training set, average error becomes smaller smaller, distribution more stable.(3) results converge set threshold, prediction errors all less than 4%, is stable, which proves method can quickly accurately establish specific properties.