Aiming at problems of low recognition rate

low reliability of diagnosis system and difficulty in extracting sensitive characteristics of single-sensor in rolling bearing fault diagnosis

a fault diagnosis method adopting feature and decision-making level information fusion of multi-sensor is presented.Vibration signals are processed with the methods of empirical mode decomposition（EMD）

wavelet packet

and local mean decomposition（LMD）method and extract feature vector

establish support vector machine（SVM）classifier.On feature level

the preliminary diagnosis results is gained through crossing diagnosis.On decision-making level

data fusion is carried out with the D-S theory.Tests show that fault recognition rate of rolling bearing can be enhanced with the methods.