An identification method of spindle bearing fault based on rough sets theory is proposed. By collecting bearing′s typical fault signals and using signal information processing techniques, vibration fault data are obtained. Then, equidistant clustering analysis method is introduced into discretization of experimental data of continuous attributes. In this way, vibration fault data table meets the requirement of rough sets data analysis. Besides, discernibility matrix algorithm is used to realize the reduction of condition attribute in the decision table. Thus, fault information hidden in huge signal data is extracted. Therefore, simple and clear fault pattern rules are acquired. The result indicates that the method can realize fault pattern identification of spindle’s bearings and it is of great application value in practical fault pattern identification.