In order to investigate the micro-process and inner mechanism of rock failure under impact loading, the laboratory tests were carried out on an improved split Hopkinson pressure bar （SHPB） system with synchronized measurement devices including a high-speed camera and a dynamic strain meter. The experimental results show that the specimens were in the state of good stress equilibrium during the post failure stage even when visible cracks were forming in the specimens. Rock specimens broke into strips but still could bear the external stress and keep force balance. Meanwhile, numerical tests with particle flow code （PFC） revealed that the failure process of rocks can be described by the evolution of micro-fractures. Shear cracks emerged firstly and stopped developing when the external stress was not high enough. Tensile cracks, however, emerged when the rock specimen reached its peak strength and played an important role in controlling the ultimate failure during the post failure stage.
The strength of rock materials is largely affected by water and loading conditions, but there are few studies on mechanical properties of saturated rocks at high strain rates. Through compressive tests on dry and saturated sandstone specimens, it was found that the dynamic compressive strength of both dry and saturated sandstone specimens increased with the increase of strain rates. The saturated rock specimens showed stronger rate dependence than the dry ones. The water affecting factor （WAF）, as the ratio of the strength under dry state to that under saturated state, was introduced to describe the influence of water on the compressive strength at different strain rates. The WAF under static load was close to 1.38, and decreased with the increase of strain rate. When the strain rate reached 190 s^-1, the WAF reduced to 0.98. It indicates that the compressive strength of saturated specimens can be higher than that of dry ones when the strain rate is high enough. Furthermore, the dual effects of water and strain rate on the strength of rock were discussed based on sliding crack model, which provided a good explanation for the experimental results.
