In order to apply the technique of acoustic emission to monitor and identify the damage condition of structures on line, a damage monitoring method is presented using wavelet packet analysis. The wavelet packet energy spectrum identified under the wavelet packet analysis from the measurement of acoustic emission are fused using D-S evidence theory to obtain improved wavelet packet energy spectrum, on this basis the damage condition discrimination index DCDI is proposed using mahalanobis distance. Then a hypothesis test involving t-test method is carried out to investigate the change of the index DCDI of different damage conditions. The analysis results of the acoustic emission signals of swivel bearings reveal that: the effective information can be extracted from the testing samples of wavelet packet energy spectrum from the measurement of acoustic emission using D-S evidence theory, the damage condition discrimination index DCDI has preferable capacity of damage identifying and can identify the different damage condition of structures accurately.