By studying the amplitude-frequency and distribution characteristics of the surface vibration signal of the high-voltage shunt reactor, combined with the Spearman correlation analysis, the characteristic parameters that can characterize the mechanical fault state of the high-voltage shunt reactor are found and verified. The segmented discrete power spectrum of the vibration signal, the principal component coefficients and other parameters are used to form the feature vector, and the KNN,SVM, neural network and other machine learning methods are used to diagnose the mechanical fault of the high voltage shunt reactor. Then, based on this, an online monitoring system has been developed, which has functions such as signal acquisition and analysis, fault diagnosis and early warning, data intelligent sampling and storage, and feature observation and analysis. The experimental results show that the system has accurate diagnosis, stable performance, convenient and intelligent, and has certain practical value.