A novel demodulation method for rotating machinery based on time-frequency analysis and principal component analysis

Demodulation is a highly effective method for feature extraction of rotating machinery. Vibration/acoustic signal of rotating machinery carries substantial information that represents the mechanical equipment conditions. Therefore, the demodulation of vibration/ acoustic signals is a key issue in fault diagnosis and target recognition. However, background noise enhances the difficulty of signal demodulation. This study proposes a novel demodulation method based on time-frequency analysis and principal component analysis (DPCA) to enhance demodulation performance under low signal-to-noise ratio (SNR). The periodic modulation wave signal is extracted by combining time-frequency analysis and principal component analysis (PCA). The effectiveness of the proposed method is verified by mean of simulation analysis and two real application cases. Firstly, monocomponent and multicomponent modulated simulation signals are used to test the performance of the proposed method. Secondly, the proposed method is applied to the vibration signals of the pump and acoustic signals of the propeller. Finally, the superiority of the proposed demodulation method is demonstrated by comparing the demodulation results with the kurtogram and cyclostationary analysis method under low SNR condition. (C) 2018 Elsevier Ltd. All rights reserved.