Studying Statistical Properties of Diagnostic Features in Vibroacoustic Signals of Ship Machinery

Abstract—The statistical properties of multidimensional diagnostic features, the locations and amplitudes of discrete components in the power spectra of vibration-acceleration envelopes that characterize incipient flaws in mechanisms, have been studied. Experimental data have been processed. Using a nonparametric rank correlation test, the mutual and combined independence of the one-dimensional values of multidimensional feature quantities has been proved. The distribution laws for these features have been researched using the Kolmogorov–Smirnov nonparametric test. It has been taken that in the general case, the distribution laws for these features can be of any form.