When the flexible thinwalled bearing is mounted on the harmonic reducer,the inner and outer rings are forced to form an ellipse,and a periodic shock signal is generated during operation,which makes the extraction of fault features of flexible thinwalled bearings more complicated.In order to realize the effective extraction of fault feature signals of flexible thinwalled bearings,a fault feature extraction and diagnosis method based on multiresolution SVD package and MED was proposed.The multiresolution SVD package was used to decompose the collected vibration signal.After decomposing,the MED noise reduction was performed using the approximation signal with the highest signaltonoise ratio in the first 4 layers.The signal reconstruction and spectrum analysis was performed between the signal after noise reduction and the detail signal of the first 4 layers,and the fault characteristic frequency was obtained.Finally the obtained fault characteristic frequency was compared with the theoretical calculated value to determine the fault type.Compared with the experimental results of multiresolution SVD package processing alone,this method has better effect on the fault feature extraction of flexible thinwalled bearings.