Abstract: In order to provide an effective numerical simulation method to reveal the damage mechanism study of surrounding rock under the combination of the change of principal stress value and the rotation of the stress principal axis from the fine-scale level, based on the uniaxial compression and conventional triaxial compression test characteristics of hollow cylindrical gray sandstone, the sensitivity of the macro fine-scale parameters of the parallel bonding model in the discrete element software PFC^3D was studied by using the one-factor analysis method, the response surface analysis method, and the regression analysis method, respectively. characteristics of the parallel bonding model in the discrete element software PFC^3D. The results show that: the reliability of the macroscopic parameters determined by the one-factor analysis method is extremely low, and the macroscopic parameters obtained based on the response surface analysis and regression analysis method show a good fit with the laboratory results at low perimeter pressure; the numerical model compressive strength \sigma _\mathrmc is positively correlated with cohesion \barc , and the elastic modulus E is positively correlated with the quadratic of the effective modulus of adhesion \barE_\mathrmc and stiffness ratio \bark_\mathrmn/\bark_\mathrms , and positively correlated with stiffness ratio \bark_\mathrmn/\bark_\mathrms positively correlated with the stiffness ratio \bark_\mathrmn/\bark_\mathrms and negatively correlated with the quadratic of the stiffness ratio \bark_\mathrmn/\bark_\mathrms , Poisson's ratio v is positively correlated with the stiffness ratio \bark_\mathrmn/\bark_\mathrms , and the shear cracking ratio α is linearly correlated with the quadratic of the stiffness ratio \bark_\mathrmn/\bark_\mathrms . Finally, the fine-scale parameters obtained by response surface analysis and regression analysis were optimized using the iterative method, and the optimized numerical simulation results were in good agreement with the laboratory results in terms of the peak strength, deformation parameters and damage morphology of the specimens under different circumferential pressures. The iterative method provides an effective method for the calibration of the fine apparent parameters of the parallel bonding model in PFC^3D, and this optimization can simplify the calibration steps, as well as enhance the applicability of the calibration results, and improve the accuracy and reliability of the model.