Since the implementation of a series of major ecological construction projects, vegetation has been restored on the Loess Plateau. Soil structure and physical characteristics changed by vegetation restoration, such as soil infiltration characteristics and soil water holding capacity and so on, have significant impacts on the surface hydrological process and mechanism of runoff generation and convergence. The change of environmental conditions in the underlying surface is necessarily change the surface hydrological process, the soil hydrological characteristics and mechanism of runoff generation and convergence on the Loess Plateau. The slope-gully system, as an important geomorphic unit of watershed on the Loess Plateau, is an integral part of the process of runoff generation and convergence and sediment transportation. Therefore, study of the rainfall-infiltration process is of great significance to comprehensively understand the rainfall-infiltration process of slope-gully system driven by "Grain for Green" projects on the Loess Plateau and reveal mechanism of runoff generation and convergence. In this paper, the dynamic response of different rainfall types and infiltration processes in the slope-gully system was studied by long-term dynamic monitoring of soil moisture in multiple depths. Based on clustering analysis, we divided the rainfall events into four different types. Afterward, the characteristics of soil moisture infiltration process under different rainfall types were analyzed. Results show that (1) soil moisture is sensitive to the response to rainfall events in the 0-60 cm soil layer, and the deepest level affected by the rainfall process is 60 cm. (2) The soil moisture infiltration amount of slope-gully system is inversely proportional to the depth, and the infiltration amount of downhill is about 9.75% higher than that of the uphill. (3) The depth of the wetting front is deeper with the extreme rainfall and long-duration light intensity type, followed by the short-duration medium intensity type, the shallowest by the short-duration light intensity type. (4) The response time of soil moisture under different rainfall types show different degrees of hysteresis, and there are significant differences. On the whole, the response of downhill is faster than that of uphill. Compare with the uphill, wetting front depth of downhill of each rainfall type is about 5-10 cm deeper than the uphill on average. (5) The Mezencev infiltration model has high accuracy in simulating the infiltration amount of four different rainfall types (Adj-R²>0.96; NSE>0.92). The Horton model can be used to simulate extreme rainfall types (NSE>0.98), while the Kostiakov model and the USDA-NRCS model are suitable for simulating short duration medium intensity rainfall and short duration light intensity rainfall.