关键词: 多尺度时域平均分解, 模糊熵, 船舶旋转机械, 故障诊断, 特征提取, 特征选择

Abstract: Ship rotating machinery often operated in a multi-vibration source environment. How to effectively extract features and reduce noise processing of the vibration signals was a hot topic for scholars. The time-domain synchronous averaging method had good effect iveness on the noise suppression of vibration signals. However， the key phase signals required by this method were difficult to obtain， and fault information other than the specific frequency would be lost， and the waveforms of the frequency multiplication signal would be mixed with each other. The above problems limited the applicability of this method. A MTAD method was proposed to overcome the problems of the traditional time-domain averaging method effectively， and combined fuzzy entropy feature selection to perform fault diagnosis on ship wind turbines. The accuracy and calculation speed are better than that of EMD， EEMD and VMD method. The effectiveness of the proposed method is demonstrated through simulation data analysis and fault simulation experiments. 

Key words: multiscale time-domain averaging decomposition（MTAD）, fuzzy entropy, ship rotating machinery, fault diagnosis, feature extraction, feature selection