In order to solve the problem of gearbox fault diagnosis

a new method based on minimum entropy deconvolution（MED）and support vector machine（SVM）is proposed.MED is used for gearbox vibration acceleration signal under background noise

then feature parameters extracted on breadth domain

frequency domain and energy domain of decreased signal are carried out

and the feature vector is built.Taking the feature vector as input

the multi-classification support vector machine is established

and the model parameters optimized by cross validation method are used to identify gearbox fault types.The fault diagnosis result of practical gearbox vibration signals shows that the proposed method can effectively identify different fault types of gear and bearing

and the optimizing model parameters can evidently improve fault identification accuracy.