Abstract: Rotor is one of the important components of centrifugal pump, which is of great significance to the detection and diagnosis of rotor running state. The characteristics of faults caused by unbalanced and misaligned rotor faults are similar. In order to effectively identify the two rotor faults of centrifugal pump rotation, the influence of location selection of different measuring points on distinguishing faults is effectively understood. Sensors are arranged on the inlet flange of centrifugal pump for signal acquisition, and the time-frequency domain characteristics of the original signal are extracted. The random forest algorithm is used to screen out 6 features with high importance, and the classification results obtained by the random forest are used as the input of PSO-SVM, so as to distinguish the faults of normal, misaligned rotor and unbalanced rotor. At the same time, the fault recognition rate of this method is compared with that of traditional PSO-SVM. The results show that the PSO model has fewer iterations and higher recognition rate, and the recognition rate of faults reaches 99.36%.