Fault Diagnosis of Reactor Based on Vibration Signal Information Entropy

By collecting and studying the time-domain characteristics of the vibration signal on the surface of shunt reactor, it is found that the vibration signal in each period fluctuates violently when the reactor has mechanical failure. The moving average sequence information entropy of vibration signal is extracted as the feature vector, and a One-Class Support Vector Machine (OCSVM) mechanical fault diagnosis model is constructed to realize the health state evaluation of shunt reactor with 99.2% accuracy. Furthermore, a fast fault detection method is proposed. This method only uses four random sampling points, which reduces the difficulty of field operation on the premise of ensuring the average fault diagnosis rate of 98.5%. Therefore, the information entropy feature of moving average sequence is an important feature of fault diagnosis of reactor mechanical equipment, which has strong practical engineering significance for reactor health diagnosis.