As gearbox play an important role in the mechanical drive system of wind turbine,Its main function is that the generated mechanical Power transfer to the generator and the corresponding speed under the action of wind turbines in the wind,  thus ensuring the normal operation of wind turbine gearbox is extremely important.And the type of large-scale wind turbine gear box is the subject in this paper.Based on signal analysis.Expert system about state monitoring and fault diagnosis of the wind turbine gear box is developed successfully.The data acquisition system is composed of vibration acceleration transducers ,charge amplifier and NlUSB9215A-DAQ Card and other instruments.And the knowledge base of expert system is established by the characters of signal when faults happen in the wind turbine gearbox.Based on a large number of experimental data，the character of state monitoring is gained from second analysis of kurtosis value and the character of fault diagnosis is consist of energy proportion spectrum of the data.By using BP neural network,the gray relational theory and D-S evidence theory， the inference engine of expert system is designed successfully .The validity and correctness of these principles applied to fault diagnosis are established by simulation·  In this paper,the structure of the expert system is devised by the tools of UML, computer Assistant Software Engineering,MATLAB7.0and VisualBasic6.0.Finally,the experiment shows validity and correctness of the expert system applied to the fault diagnosis of large-scale wind turbines gearbox.