In order to accurately identify the fault type of wind turbines, this paper considers the correlation between vibration and current signal sources, presents a fault diagnosis method of information fusion and a new relevance vector machine. Got the data from wind power generator test bench, extracting data with higher sensitivity characteristic parameters of diagnosis as samples, then built an improved vibration and current detection based on relevance vector machine model of fault diagnosis. The model of multi signal source fault diagnosis is established by means of information fusion, and finally the fault diagnosis result of wind turbine is obtained. The experimental results show that the proposed method is more accurate and can identify the fault types of wind turbines with electromechanical coupling characteristics compared with the single signal based fault diagnosis method.