Faults in Gas insulated switch-gear（GIS） account for a large proportion of all kinds of operation accidents in substations. It is urgent to effectively improve the success rate of diagnosis of insulation defects in GIS. In this paper

the defect simulation device for high voltage SF6 gas insulated composite electrical apparatus is developed. The device consists of the SF6 gas chamber

solid insulator

defect simulation device

and observation and measurement device. The defect device can effectively simulate four typical insulation defects including free metal particle discharge

tip discharge

suspension discharge and air gap discharge. Furthermore

an artificial intelligence classification method combining fuzzy iterative selforganizing data analysis techniques algorithm（ISODATA） and ant colony algorithm is proposed

and the structure parameters of the above algorithms are optimized by particle swarm optimization. The field application results of high-voltage SF6 gas insulated composite electrical appliances show that the proposed fault diagnosis method can be adopted to intelligently determine the fault modes according to the types and contents of typical micro-decomposition gases in SF6. The hardware platform of the defect simulator can effectively collect the original classification data and form the defect sample database. The artificial intelligence classification algorithm proposed in this paper is easy to be programmed and implemented. It can be used for the diagnosis and evaluation of insulation defect types in practical engineering. The success rate of insulation defect type diagnosis is 93.3%. The research results are helpful for the detection of the early latent faults for SF6 insulation equipment

and have theoretical guiding value for fault diagnosis and mode recognition of high voltage SF6 gas insulated combination electrical equipment.