DOI: 10.14489/td.2018.10.pp.044-049

Sokolova A. G., Balitsky F. Yа., Ivanova M. A., Shirman A. R.
REGRESSION FUNCTIONS AND OTHER PROBABILITY CHARACTERISTICS OF VIBRATIONS AS TOOLS TO ENHANCE CONDITION MONITORING SYSTEMS OF CENTRIFUGAL COMPRESSORS. PART 2
(pp. 44-49)

Abstract. The same approach to advanced machinery condition monitoring with probability vibration features application, which was originated in the Part I, is continued and expanded with numerical information of stochastic interdependence of vibration signals in the second part of the paper. It is no secret that any machinery fault manifests itself by changing vibration characteristics from usually linear (or quasi-linear) ones to significantly nonlinear ones. If it were possible to measure simultaneously both the source vibration signal (input) and vibration signal of response on it (output), then the best way to understand all about these changes is to investigate cross-regression function shapes and the numerical information of stochastic interdependence for these signals. The “input” information unfortunately is unapproachable in vibrodiagnostics approach, so one has to use indirect signs, i.e. features of two different “output” signals. The vibration cross-regression functions along with numerical correlation data, directly describing the degree of nonlinear distortion of machine transient characteristics are proposed as helpful additional effective diagnostic tool. As it is shown, these parameters used for the centrifugal compressor condition monitoring are linear ones under normal compressor condition, and nonlinear ones under its journal bearing rough fault condition.

Keywords: vibration diagnostics, rotary machinery, journal bearing, relative shaft vibration, probability distribution, stochastic liaison, cross regression function, correlation coefficient and correlation ratio.