Time-variation and frequency-variation analysis of wavelet packet-characteristic entropy for vibration signals

In this paper a new method to extract characteristic parameters from vibration signals of HV CB (circuit breakers) is introduced based on wavelet packet. For one normal signal, the method splits each reconstructed-signal envelope into 15 segments with the same energy, at the third layer of the wp-decomposition, and the testing signal is also split with the same breakpoint as the normal signal. The characteristic entropy is extracted from the segmental energy of reconstructed signals, and an 8-D vector can be get. At last, many simulated signals with time-domain and frequency-domain changes from the normal signal are used to test the character of the new method. The simulation results indicate the characteristic vector extracted from HVCB vibration signal with the new method can reflect the changes of the mechanical states visually and sharply, and can be used to diagnosing the fault of HVCB.