A improved way for fault diagnose based on EEMD and SVM

The change of the vibration signal can reflect the mechanical state of the HV circuit breaker. An efficient method of feature extraction of the vibration signal usually pays a key role in the validity of the fault diagnosis and also lays the foundation for the fault classification in the subsequent stage. The paper presented a feature extraction method which is based on average empirical mode decomposition (EEMD) and GP algorithm. The EEMD algorithm is used to decompose the non-stationary signal into several the conqueror functions (IMFs). However, the first four orders usually have great influence on the classification of signal, so we select the first four orders to calculate the correlation dimension. Finally, the SVM (Support Vector Machine) is used to classify different types of fault signals after decomposing (EEMD) and calculating correlation dimension. The test indicates that the represented method have a high recognition rate. To a certain extent, the method is likely to provide a new method for fault diagnosis.