The methods based on coherent and variance are used to determine faults, which take advantage of the strong signals in seismic data; nevertheless they are limited by seismic resolution and hence fail to accurately identify small faults. In the seismic data, the boundary of large-sized geological body has obvious reflection characteristics. However, in the case of smaller geological bodies such as pinchouts and small faults, their weak reflection energy exists in the form of diffraction or residual diffraction, which is overwhelmed by the strong reflection. In this paper, plane wave destruction method is used to extract diffraction. The reflected wave angle body is calculated using regularization iterative algorithm. And then the mean filter method is used to extract strong energy reflection, which is subtracted from seismic data to get the diffraction wave information. Using diffraction wave information, we can decide the development extent and combination style of the faults. Theoretical model analysis and actual data application show that this method is practicable and effective in identifying faults. This method has a certain value for industrial application.