A new method of fault diagnosis for high voltage circuit breakers based on vibration signals

A new method of fault diagnosis based on wavelet packet and support vector machine is presented. Wavelet packet decompresses vibration signals, and wavelet packet nodes sensitive to the change of state, as eigen-nodes, form a decomposition tree; the signal restructured at all eigen-nodes can be used as the fingerprint in many breaker diagnostic systems. Support vector machine with the "most important" factors at eigen-nodes as input vector classifies the states at "one to others" strategy. The experimentation without loads indicates the method can easily and accurately diagnose mostly breaker faults.