Abstract: In allusion to the low accuracy of rubbing fault diagnosis of multi-stage centrifugal pump in engineering, a rubbing fault diagnosis method for multi-stage centrifugal pump based on optimized support vector machine（SVM） is proposed. The method can complete diagnosis with a small amount of sample data by extracting the most effective fault features from vibration signal to optimize the input sample quality of the model and optimizing the parameters of the model. The rubbing signal of the actual product is obtained through the fault simulation testbed of multistage centrifugal pump, and an SVM model is constructed by extracting time-frequency domain characteristics and information entropy characteristics of vibration signal with ensemble empirical mode decomposition（EEMD）, features are selected with sequential feature selection（SFS） algorithm, and the parameters of SVM with grid search（GS） algorithm and particle swarm optimization（PSO） algorithm are optimized. The experimental results show that the proposed model has high recognition accuracy for rubbing fault of multi-stage centrifugal pump in engineering and has good practicality.It can provide a certain reference for the rubbing fault diagnosis of the multistage centrifugal pump.