In allusion to the difficulty to obtain fault samples of multi-stage centrifugal pumps in practical engineering

three typical faults containing rubbing

misalignment and unbalance were simulated through the fault simulation test-bed of multi-stage centrifugal pumps. And a fault diagnosis model based on support vector machine (SVM) was established to realize the classification of three types of faults. High dimensional feature samples were constructed by extracting time-frequeney domain characteristies of vibration signal with ensemble empirical mode decomposition(EEMD). combined with characteristies of time domain

frequeney domain and information entropy. The efficient fault classification was achieved by optimizing the quality of input samples with principal component analysis (PCA). In addition

by comparing the classification effects of SVM and back propagation (BP) neural network

it shows that the SVM model has better classification effect and high applicability in fault diagnosis of multi-stage centrifugal pump.