Chin S. Chu
Graduate Research Assistant,
Kristin L. Wood
Associate Professor.
Ilene J. Busch-Vishniac
Professor.
Department of Mecfianical Engineering,
Ttie University of Texas,
Austin, TX
A Nonlinear Dynamic Model With
Confidence Bounds for
Hydrodynamic Bearings
In conventional rotordynamic modeling, hydrodynamic bearings are often char-
acterized by a set of linear stiffness and damping coefficients obtained from a
first-order Taylor series expansion of bearing reactions. Theoretically, these
coefficients are only valid for small amplitude motion about an equilibrium
position. In this paper, a nonlinear dynamic model that overcomes the small
amplitude assumption in the conventional linear analysis is described. By includ-
ing higher-order terms in the bearing reaction expansion, nonlinearity in the
oil film forces for large amplitude motion can be captured and represented by
a set of nonlinear stiffness and damping coefficients. These coefficients are
functions of static bearing displacement. A finite difference approach is de-
scribed and is used to solve for these coefficients. The stated model is applied
to a conventional slider bearing and a mechanical smart slider bearing that
experiences large variations in load. Error assessment is performed numerically
on the higher-order solutions to determine an acceptable displacement bound
for the higher order coefficients.
1 Introduction
There are generally two approaches to representing fluid film
bearings in rotordynamic modeling: namely, a linearized model
and a time-transient model (Shapiro and Rumbarger, 1971).
The linearized model represents bearing systems by constant
stiffness and damping coefficients evaluated about a nominal
operating position. Dynamic response and stability conditions
can be evaluated using a time-invariant linear dynamic model
with these stiffness and damping coefficients. The time-transient
model, on the other hand, produces a time history of the bearing
displacement in all degrees of freedom without any linearization
assumptions. The latter approach provides the closest simulation
of the actual system performance; however, repetitive solution
of the governing fluid equation is required over a time span,
resulting in costly computational requirements. It is therefore
a common practice to use the linearized approach unless the
linearized assumption deteriorates, i.e., when the bearing motion
amplitude becomes large.
In some cases, the linearized model may be shown to be
valid over a range of bearing parameters and operating condi-
tions. Shapiro and Rumbarger (1971) suggest that displace-
ments over 10 percent of the operating film thickness would
require appropriate modification of linear spring and damping
constants. Lund (1987), who introduced the linearized bear-
ing analysis, states that the linear coefficients maybe valid for
amplitudes up to 40 percent of bearing minimum clearance.
However, there is no justification of his criterion in terms of
coefficient accuracy at such high displacement amplitudes.
Obviously, such a linearized approach would require an evalu-
ation of the linearized error to set forth an acceptable bound
for the linear solutions. Hashi and Sankar (1984) have at-
tempted to estimate the linearized error. Deviations of linear-
ized journal bearing coefficients from the nonlinear systems
are calculated. Error estimation charts for some important
cases of plain journal bearings under unbalance loading are
provided.
Contributed by the Tribology Division for publication in the JOURNAL OF
TRIBOLOGY. Manuscript received by the Tribology Division June 7, 1995; revised
manuscript received June 16, 1997. Associate Technical Editor: M. J. Braun.
When the bearing displacement amplitude exceeds a certain
limit, bearing reactions become nonlinear. There are limited
studies which have been performed on the nonlinearity within
the bearing oil film. Hattori (1993) shows that for a rotary
compressor under large dynamic loads (displacement ampli-
tude varies more than 50 percent of the radial clearance per
revolution), journal bearing stiffness and damping coeffi-
cients can vary by more than one order of magnitude, demon-
strating that oil film nonlinearity seriously influences rotor
motion. The linearized model is therefore not acceptable for
such cases and some form of nonlinear analysis is required.
Choy, et al. (1991) use a nonlinear approach in modefing the
nonlinearity within the oil film forces. They expand the bear-
ing reactions using a power series and retain the higher order
terms. They show that at displacements far away from the
equilibrium, nonlinearity in the oil film is significant and can
be modeled closely by higher order stiffness and damping
coefficients. Their results indicate that the range of accuracy
for the higher-order coefficients can be extended by including
more terms in the expansion.
In this paper, a quasi-static nonlinear dynamic model for
hydrodynamic bearings is described. There are two features to
this model: (j) a nonlinear dynamic model capable of capturing
the nonlinearity in oil film forces under large amplitude excur-
sions, and («) an error scheme that evaluates higher-order trun-
cation error in the oil film force expansion. The first feature
enables the designer to calculate the nonlinear oil film forces.
The second feature provides a useful way for the designer to
set a confidence bound on the nonlinear solutions. Depending
on the application requirements, additional terms can be added
to the bearing load expansion.
A finite difference scheme with successive-over-relaxation
(SOR) is employed to solve the stated nonlinear model. This
model is first applied to a conventional slider bearing to bench-
mark the new model and then to a new Micro-Electro-Mechani-
cal Systems (MEMs) smart bearing concept (Hearn et al.,
1995). The MEMs smart bearing concept involves active con-
trol of the film thickness profile to induce desirable bearing
performance. A typical MEMs smart bearing consists of an
actively deformable surface (actuators) and pressure sensors
(Hearn et al., 1995). Preliminary results from a linear model
Journal of Tribology Copyright © 1998 by ASME JULY 1998, Vol. 120 / 595
Downloaded 24 Aug 2012 to 128.83.63.20. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm