Feature extraction method of bearing AE signal based on improved FAST-ICA and wavelet packet energy

• The Fast-ICA algorithm is improved by third-order Newton iteration method.
• The simulation results show that the improved Fast-ICA algorithm is superior to the classic Fast-ICA algorithm in convergence speed and the accuracy of separation.
• The improved Fast-ICA algorithm can effectively separate acoustic emission signal which contains two kinds of damage and fracture failure.
• The feature extraction of mixed fault signal was realized by the Improved Fast-ICA algorithm which is combined with the wavelet packet energy spectrum.

In order to accomplish the feature extraction from a mixed fault signal of bearings, this paper proposes a feature extraction method based on the improved Fast-ICA algorithm and the wavelet packet energy spectrum. The conventional fast-ICA algorithm can only separate the mixed signals, while the convergence speed is relatively slow and the convergence effect is not sufficient. The method of the third-order Newton iteration is adopted in this paper to improve the Fast-ICA algorithm. Moreover, the improved Fast-ICA algorithm is confirmed to have a faster convergence speed and higher precision than the conventional Fast-ICA algorithm. The improved Fast-ICA algorithm is applied to separate the acoustic emission signal in which two kinds of fault components are comprised. The wavelet packet energy spectrum is used to extract the feature information in the separated samples. In addition, the fault diagnosis is performed based on the SVM algorithm. It is confirmed that the slight damage and fracture of a bearing can accurately be recognized. The results show that the improved FAST-ICA and wavelet packet energy method in feature extraction is sufficiently effective.