Feature extraction methods of vibration signal in automobile main reducer based on morphological un-decimated wavelet

In order to explore methods that can process effectively nonlinear signals, the nonlinear wavelet, morphological wavelet (MW) was introduced into the field of vibration signal processing. Because of the decline of decomposition signal layer by layer, one kind of morphological un-decimated wavelet construction method based on the cascade of morphological opening and morphological closing was proposed. According to the general structure of morphological un-decimated wavelet (MUDW), the filtered signal, filtering by the cascade of morphological opening and morphological closing, constructed the approximate signal, and detail signal was equal to the original signal subtract the approximate signal. Obviously through the process of decomposing, the approximate signal or detailed signal of current layer and low level layer had the same data length to avoid information leaking and provide enough information for signal processing. The method was used in the feature extraction of vibration signal in automobile main reducer. Results showed that the mentioned MUDW had better filtering effect than the existing MUDW and linear wavelet (sym8 wavelet), and it also had better demodulation effect than Hilbert envelope analysis. The mentioned MUDW can extract the feature from nonlinear vibration signal effectively and have good application value.