Characteristic study of typical faults of rotor systems combining wavelet packet with fractal
Sheng Zhong · 2007
Based on the natural consistence of wavelet packet and fractal theory on the multi scales analysis and self-similarity, this paper presents a method of fault characteristics extraction combining the wavelet packet decomposition with the correlation dimension for several typical fault (such as friction, mass imbalance and oil film whirling) of rotor system. Firstly, the wavelet packet is applied to decompose the vibration signal, and the correlation dimensions of decomposition coefficient obtained are regarded as characteristic vector of fault, then the characteristic vector of fault is identified by neural network. The simulation result verifies that the method presented by this paper is very effective for several typical fault diagnoses of rotor system.