Mechanical defects of gas insulated switchgear (GIS) may occur in the process of assembly and long?term operation. Detecting and diagnosing mechanical defects is of great significance to ensure the reliable operation of GIS. At present, scholars have carried out a large number of vibration signal detection. However, the result analysis is mostly time?frequency analysis based on FFT (fast Fourier transform), and the diagnostic methods for vibration?signal feature extraction under different typical mechanical defects is deficient. Therefore, based on the basic principle that frequency characteristics differences exists in the vibration signals of different typical mechanical defects, the method adopting EMD?FFT joint algorithm to extract the characteristics of GIS vibration signals is proposed in this paper. According to the vibration signals detection and analysis of typical mechanical defects in 550 kV entity GIS equipment, the characteristic maps of GIS mechanical vibration signals under different defects are summarized. Thus the effective diagnosis of mechanical defects of on?site GIS equipment can be realized. The results show that the proposed EMD?FFT algorithm can extract the main characteristic frequency points of vibration signals under different typical defects effectively. The constructed characteristic spectrum can reflect the variation of frequency information under different defects directly. Which leads to realize the diagnosis of typical mechanical defects. Based on the proposed diagnosis method, field test is conducted to detect the loosening defects of anchor bolts in a certain GIS equipment, which proves the effectiveness of the proposed method. The research results can provide methods and test results reference for on?site GIS mechanical defect diagnosis.