[111] Application of a non-standard approach for diagnostic analysis of vibration waveforms

This paper is dedicated to a non-standard vibration diagnostics method to be used as an additional tool to empower the commonly applicable spectral methods. The fact is that any machinery fault changes the nature of the associated vibration from linear (usual for normal conditions) to significantly non-linear. One of the best ways to register these modifications of machinery vibration features is to apply a stochastic approach in analysing vibration waveforms picked up at different measurement points. The results of investigations involving a joint two-dimensional probability distribution of vibration amplitudes for rotating machine diagnostics are presented. The usage of such probabilistic features of shaft-relative vibration in journal bearings for two perpendicular measurement directions, such as cross-regression functions and other probability characteristics (for example, numerical information concerning the stochastic interdependence of vibration signals), permits vibration linearity to be estimated both numerically and graphically. As it is shown, the parameters used for the condition monitoring of a centrifugal compressor are linear under normal compressor conditions and non-linear under a journal bearing rough fault condition and may be used as sensitive diagnostic symptoms of machinery operational damage. Moreover, they can be successfully applied for quality maintenance estimation, for example immediately after repair.