Abstract: High resolution imaging of the fault zone structure is crucial to understanding the characteristics of strong earthquake activity and the deep seismogenic environment. In seismological studies, the fault zone is generally considered to be a low velocity zone with host rock on both sides. In order to determine the main parameters of fault zone, such as physical properties and interface characteristics, many efforts have been made. However, many key fault parameters still lack constraints, such as the depth extent, width and dip angle of the low velocity zone. With the advancement of the large-N array techniques in recent years, seismologists have collected high-quality data with larger apertures and denser arrays for better analysis of fault zone structures. These array data have also facilitated the development of new seismic imaging techniques. In this paper, a new waveform inversion method for fault zone parameters based on generalized teleseismic waveforms is proposed. Generalized teleseismic event is defined as the local seismic signal whose epicentral distance is greater than 7−10 times the aperture of the array. In order to efficiently simulate high frequency wavefield propagation from long distance local earthquakes, a hybrid modeling approach is proposed, which greatly reduces the computational cost for teleseismic waveform inversion. We apply the proposed new inversion method to a dense array data across an inactive fault in the Qilian Mountains, Gansu Province. As an active-source analogues of generalized teleseismic, the recording waveforms of a 270-meter-long linear array are very clear excited by an airgun source 1.8 km away. Setting cross-correlation travel time of direct P wave as the misfit function, we perform waveform inversion for the main structural parameters of the fault zone through grid search strategy. The new method is very suitable for geophysical survey of fault zone lack of local seismicity.