In order to diagnose the typical mechanical looseness and insulation fault of gas insulated switchgear (GIS) in operation, this paper sets up experimental platform to simulate insulator loosenesses, tip discharges and suspended particle discharge faults. Acceleration signals of GIS vibration in different fault patterns are collected. Then the Hilbert marginal spectrum can be obtained by analyzing vibration signal using virtue of Hilbert-Huang transformation (HHT) after the modified ensemble empirical mode decomposition (MEEMD). The mechanical looseness or partial discharge of GIS can be judged through the energy of signals in different frequency bands. The wavelet transform is applied to calculate discharge factors, which reflects the characteristics of different types of partial discharge fault. Results obtained by experimental data analysis indicate that, the above methods can distinguish the faults of GIS and reflect the characteristics of partial discharge faults effectively. Thus providing a useful method for diagnosis of mechanical looseness and insulation fault of GIS.