Parameter variation and data mining of oil-film bearings: a stochastic study on the Reynolds’s equation of lubrication
The modeling and simulation process of oil-film bearing dynamics constitutes a rather essential task integrated in the workflow of various mechanical products. Specifically, in the turbo charger industry, the correct capture and understanding of the associated nonlinear rotating dynamics is of utmos...
Gespeichert in:
Veröffentlicht in: | Archive of applied mechanics (1991) 2014-05, Vol.84 (5), p.671-692 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The modeling and simulation process of oil-film bearing dynamics constitutes a rather essential task integrated in the workflow of various mechanical products. Specifically, in the turbo charger industry, the correct capture and understanding of the associated nonlinear rotating dynamics is of utmost importance, since the system’s efficiency and lifetime span depends on it. The root cause of the nonlinear rotordynamic effects is the oil-film concentrated in the rotor’s journal bearings. Its behavior is highly coupled with both the system’s geometric and dynamic configuration. The dynamics of the oil-film are described by the well-known Navier–Stokes equation, which under a series of assumptions and simplifications results to the, so-called, Reynolds equation. In this paper, the Reynolds equation is numerically solved based on a finite difference scheme and several parameter variation studies are conducted in an effort to pinpoint the most influential parameters—journal bearing geometric dimensions, oil-film properties and rotor-velocity-driven inputs—with respect to designated responses—friction, oil-film pressure force, minimum oil-film thickness and boundary oil-flow—all of which are regarded as important in terms of the aforementioned system’s efficiency and lifetime span. Based on multivariate analysis algorithms, correlation outcomes and global sensitivity results are presented. In an effort to capture possible nonlinear phenomena, which might not be possible via linear data mining tools, the Spearman rank-order coefficient and self-organizing maps methodology are applied. |
---|---|
ISSN: | 0939-1533 1432-0681 |
DOI: | 10.1007/s00419-014-0824-3 |