Recognition of Incipient Defects in the Units of Ship Machinery by Vibrodiagnostics Based on Optimum Decision Rules

Abstract

We developed a method for revealing incipient defects in ship machinery during vibration-based inspection, based on studying the statistical properties of diagnostic features in the vibroacoustic signals of ship mechanisms and on the principles of the pattern recognition theory. The following applied solutions of the problem are obtained; an algorithm for constructing references for multidimensional feature spaces allocated in vibroacoustic signals and characterizing incipient defects in mechanisms for various modes (states) of their operation in the form of conditional multidimensional probability densities; an optimum decision rule for recognizing faults-incipient defects-and intact states of mechanisms, taking changes in the dimensionality of the feature spaces into account. The developed method has been verified experimentally in bench vibroacoustic and field tests of ship machinery.