关键词: 振动与波, 混沌信号, EMD, 小波分析, 连续均方误差

Abstract:

For de-noising from low SNR rotor vibration signal, a method based on EMD and wavelet analysis is proposed. Firstly, the noisy signal is decomposed by EMD, and the IMF components are divided to high and low frequency by using consecutive mean square error(CMSE). Then, different thresholds are determined and the useful signal details of high and low frequency IMF components are extracted based on wavelet analysis and combined with the low-frequency IMF components so that the noise of IMF is removed efficiently. Numerical simulation and rotor vibration signal were carried out to evaluate the performance of the proposed method. The results showed that the proposed method outperformed the EMD and wavelet threshold de-noising algorithm.