High-purity (HP) copper targets with grain sizes of 50, 130 and 200 μm are constructed by using the Voronoi method. Damage nucleation points are randomly prefabricated at the grain boundaries. A two-dimensional axisymmetric finite element model is established to simulate the spallation experiment of HP copper target. The effects of grain size and loading stress on the macro- mechanical response and meso-damage evolution of HP copper spallation are studied and compared with the relevant experimental results. Based on the analysis of free surface velocity profiles, the effects of grain size on the location of pull back velocity rebound point, velocity rebound slope and velocity rebound amplitude are revealed. It is demonstrated that the spalling strength corresponds to the peak value of tensile stress in the damage zone, which essentially represents the critical stress of micro damage nucleation or early growth. Based on the characteristic analysis of damage evolution nephogram, the evolution process of localized plastic strain field around the micro-voids in the growth and coalescence process is reproduced, and the strong dependence of micro-void coalescence behavior on grain size is clarified. The loading stress amplitude has little effect on the location of pull back velocity rebound point, but has a significant effect on the growth and coalescence behavior of micro-voids. The slope and amplitude of pull back velocity rebound increase with loading stress increasing, which is consistent with the relevant experimental result. With the increase of the loading stress, the micro-voids grow from independent growth to coalescence, thus forming spalling surface. The physical process of damage evolution determines the wave oscillation characteristics after the pull-back rebound point. The numerical simulation results reproduce the physical process of damage evolution and its influence on the macroscopic mechanical response, which is of great significance for further understanding spall damage evolution mechanism and theoretical model construction.