In the light of the problem that the early fault feature signal of the hydraulic pump is weak and the fault feature is difficult to extract, a fault diagnosis method of hydraulic pump based on symbolic dynamics entropy and support vector machine(SVM) was proposed. Nine kinds of fault states of hydraulic pump were simulated, and the sample values of the vibration signal of the multi measurement points were measured then, the symbolic dynamics entropy Hk of each vibration signal were calculated using time series symbolic dynamics entropy to determine the corresponding symbol dynamics entropy feature vector of each state. Finally, the training set of feature vectors in different status was established for SVM to diagnose and identify the fault state of hydraulic pump, the accuracy of the test was 98.71%.The comparison of the diagnostic results with the improved BP neural network showed that the method has a higher recognition rate of 98.71% and cost less time which helps in online diagnosis．