Envelope Signature Extraction Of Ship-radiated Noise Based On HOS

Ship-radiated noise(SRN)is non-Gaussian nonlinear signal,and higher-order statistics(HOS)are powerful tools for the research of non-Gaussian nonlinear signal.In this work,signatures of shaft and blade frequency are extracted from the envelope of SRN by using the properties of HOS which can characterize the quadratic phase coupling property and suppress Gaussian noise in signals.The basic theory of HOS leading this paper.The facts that higher-order cumulants and spectra can suppress all symmetrical distribution noises which include Gaussian noise and characterize the nonlinear phase coupling in signals,power spectrum is phase-blind,and can not characterize the nonlinear phase coupling in signals are demonstrated.Then the physical mechanisms of the envelope modulation of SRN are introduced to demonstrate that the envelope modulation character of SRN is caused by the fact that the propeller is working in a circumferential non-uniform wake.Then the signal model is given from the view of statistics.The comparison of the performance of conventional DEMON spectrum and 3/2-D DEMON spectrum is analyzed in this paper.The properties of 3/2-D DEMON spectrum are introduced and proved through simulation.Then the 3/2-D DEMON spectra of simulation SRN signals and data recorded at a sea trial are analyzed,the optimal demodulation frequency band is chosen by filter bank when the actual data signal is analyzed.The bispectrum signature extraction of the envelope of SRN is analyzed in the end,and its diagonal slice spectrum is also analyzed.The performance of all these feature extraction method to extract the ship propeller shaft and blade frequency proves well.