Sedimentary rock is widely distributed on the earth and is closely related to the oil and gas industries as they are the main container of oil and gas. Scientifically understanding and simulating the behaviour of sedimentary rock is of great significance to improving the efficiency of industrial production. In our study, Discrete Element Method (DEM) is used for rock behaviour prediction. We established a DEM simulation framework including particle packing construction, contact model selection and automatic parameter calibration, which can be used for comparing different bonded particle models. By using a conical contact model for particle-particle contact, we improved and developed an improved parallel bonded particle model in the open source software Kratos Multiphysics DEMApplication, which is used for capturing the changes of rock Young's modulus under different confining pressures. Furthermore, a hybrid Genetic Algorithm (GA) and XGBoost (eXtreme Gradient Boosting) algorithm is developed for automatic DEM parameter calibration. The results show that the efficiency is significantly improved by applying the new GA-XGBoost algorithm. The developed DEM models and automatic calibration algorithm are applied to the triaxial compression test. The results can help us deeply understand sedimentary rock behaviour and provide reliable references for industrial production process design.