You are on page 1of 13

12th International Conference on Vibrations in Rotating Machinery ­

Institution of Mechanical Engineers, ISBN 978-0-367-67742-8

Influence of thrust bearings in lateral


vibrations of turbochargers under axial
harmonic excitation
T.F. Peixoto, K.L. Cavalca
School of Mechanical Engineering, University of Campinas, Brazil

ABSTRACT
Turbocharger rotordynamic modelling usually neglects thrust bearing coupling effects
on lateral oscillations. However, using thrust bearings in turbochargers is mandatory,
due to the compressor and turbine gas flows axial force imbalances. Furthermore,
approximating the bearings dynamic characteristics by linear coefficients is not suit­
able in this rotor-bearing system, because of the high rotational speeds achieved by
a typical automotive turbocharger. This paper brings a model of the entire turbochar­
ger, considering the nonlinear behaviour of both thrust and floating ring bearings, with
thermal effects, analysing the turbocharger response to an axial harmonic excitation
and the thrust bearing influence on lateral vibrations.

1 INTRODUCTION
The current practice of engine downsizing with a turbocharger (TC) requires a great
knowledge of both the engine and turbocharger behaviours. A reliable model of the
turbocharger is of primal importance for the development and optimization of newer
designs. Given the low mass and extremely high rotational speeds achieved by
a typical automotive turbocharger, high nonlinear phenomena can be observed in
turbocharger oscillations. To support these high vibrations, the common practice is to
utilize a floating ring bearing (FRB) to support the rotating shaft. The FRB has proved
to increase the damping characteristics of radial bearings and it is an optimal choice to
sustain a high-speed TC [1]. However, because of the two oil films arranged in series,
with the floating ring working as a mass pedestal between both films, the phenomenon
of fluid-induced instability is particularly strong in this type of bearing [2]. Either the
inner film, the outer film, or even both films simultaneously, can induce lateral oscilla­
tions [3] and recent researches have focused on minimizing these sub-synchronous
oscillations [4].
Several key components may influence the FRB instability, such as the inner or
outer film clearances, the inner-to-outer film clearances ratio, the bearing lengths,
the bearing length-to-diameter ratios, the ring size, the mass unbalance level, the
injection pressure, among others [5]. Study the effect of each parameter in the
system response is crucial to understand the overall TC dynamic characteristics.
Recent research has suggested increasing the outer film clearance may reduce, or
even totally suppress, some sub-synchronous components [6]. However, this is not
an optimal solution, as other sub-synchronous components may become higher.
Unfortunately, most works on turbocharger focus in the lateral vibrations, entirely
neglecting the axial dynamics. Previous works completely neglect axial loads and
the thrust bearing (TB) effect on the overall turbocharger behaviour [1]–[6]. Just
recently, a few works have begun to include the TB on the complete turbocharger
dynamic simulations. Novotný et al. approximated the TB by a linear spring with
a constant stiffness [7], but improved the TB model to account for nonlinear stiff­
ness and damping characteristics in a time-transient analysis [8]. The key factor to
account for the TB nonlinear axial behaviour was the creation of a lookup table,

39
a Database, storing the bearing load-capacity for different oil film thicknesses. This
so called Database Method, term coined by [9], allows for a fast evaluation of the
bearing characteristics. However, the most limiting assumption on [8] was neglect­
ing the TB angular misalignment effects. It has been show [10]–[13] that the angu­
lar misalignment may influence the axial supported load of a typical turbocharger
thrust bearing and induce restoring moments in the shaft, affecting the lateral
dynamics of such systems. The influence of thrust bearings in lateral vibrations has
already been verified in other types of rotating machines [14], [15], but just
recently this concept has been applied to turbochargers. Chatzisavvas et al. [10]
considered the angular misalignment and showed the thrust bearing influence on
lateral oscillations, emphasizing its capacity on suppressing some sub-synchronous
components and changing the onset of oil whirl in the lateral oscillations. This model
was further investigated by Koutsovasilis [11], analysing the influence of the thrust
bearing position along the axis, noticing the bearing size and position may have
positive, negative or neutral impact on lateral oscillations. However, the modelling
of [10], [11], although considering angular misalignments, neglected thermal vari­
ations, which has also been shown by Peixoto and Cavalca [12], [13] to affect the
load-carrying capacity of turbocharger thrust bearings. The simplifications adopted
either by [8], [11] are justifiable, in order to complete a time-transient analysis in
a reasonable amount of computational time, given the high computational cost of
solving the Reynolds equation at each time step altogether with the Energy equa­
tion. To overcome this time problem to analyse transients in turbocharger dynamics,
but still considering shaft angular misalignment and lubricant thermal effects, the
proposed Database Method has a great appeal. [7], [8] mention the creation of
a database to use it in the time integration of the turbocharger dynamic equations,
but the work focus on the development of the Reynolds equation, and some vari­
ations of it, to account for different phenomena on the bearings. The analysis
admits some strong assumptions, that may not be suitable for all turbocharger
applications, and little detail is given to the Database Method itself. Chasalevris and
Louis [9] presented a detailed explanation of the Database Method, solving for the
loads in isothermal floating ring bearings in a turbocharger. The study does not
focus on the bearing modelling itself, using a simplified isothermal approach, but
instead compares the Database Method Solution to a Direct Solution, solving the
Reynolds equation at each time step. The results revealed a great agreement
between both solutions, evidencing the advantages of the Database Method in
terms of computational time and noticing the method could be extended to include
any other desired effect. This work proposes to use the Database Method to model
thrust bearings, including angular misalignment effects and temperature variations
in the oil films. The study presents the thrust bearing modelling and the use of the
Database Method in the analysis. It utilizes the method to study a turbocharger sub­
ject to a harmonic axial excitation and the coupling effect the thrust bearing may
have on lateral oscillations.

2 METHODOLOGY
The turbocharger model consists of a Finite Element (FE) model of the rotating shaft
with the compressor and turbine wheels modelled as rigid discs. The rotating shaft is
supported by floating ring bearings and a double-acting thrust bearing, to support
both the lateral and axial oscillations. Because the thrust collar, initially parallel to the
thrust bearing, is attached to the shaft, any shaft rotation imposes a collar rotation,
inducing restoring moments of the thrust bearing on the shaft, which may influence
lateral oscillations. The next sections present the FE model of the rotating shaft and
the bearing models. This entire model is detailed in [13]. We focus on the description

40
of the thrust bearing database and the differences on the response compared to
a direct solution.

2.1 Turbocharger model


Given the high nonlinear characteristics of a typical automotive turbocharger, approxi­
mating the bearing dynamics by its equivalent linear coefficients is not a suitable
approach [4], [8], [11], [13]. In order to accurately model the hydrodynamic forces,
a nonlinear approach must be sought. In order to accomplish it, the turbocharger
equations of motion are written as:

wherein in the left-hand side of the equation, the mass, damping, gyroscopic and stiff­
ness matrix are obtained through a FE modelling of the rotating shaft, compressor and
turbine wheels [13], [16]. In the right-hand side of the equation, the excitation vector
is separated in three terms. The first term relates to the static forces, the turbocharger
weight. The second term relates to the external excitation forces. The third term rep­
resents the nonlinear hydrodynamic bearing forces. During the time integration of
these equations of motion, the hydrodynamic forces are considered in its full form,
instead of approximating it by equivalent stiffness and damping coefficients. Time
integration is performed utilizing the Newmark integration scheme [13].

Figure 1. (a) TC FE model and (b) FRB variables.

The FE model utilized in this work is shown in Figure 1a. Nodes 1 and 7 comprise the
compressor and turbine wheels, as rigid discs. Nodes 5 and 6 represent the floating
ring bearings at compressor-side (CS) and turbine-side (TS). The double-acting thrust
bearing is discretized in nodes 2 to 4. Each oil film will produce an axial force and
restoring moments in the thrust collars in nodes 2 and 4, so the CS thrust bearing is
located at node 2 and the TS thrust bearing is at node 4.

2.2 Floating ring bearing lubrication model


The floating ring bearing is utilized to increase the damping characteristics of a typical
radial bearing. The outer film, a second layer of lubricant, provides a higher damping
characteristic to this type of journal bearing [3], [4] and is especially useful to support
the lateral oscillations in a high-speed turbocharger. A cross section of the floating ring
bearing, with its main variables, is shown in Figure 1b. The shaft has a radius Rj, the
floating-ring has an inner and outer radius Rri and Rro , respectively, and the bearing
house has a radius Rb . The shaft rotates with an angular speed Ωj , while the ring

41
rotates with an angular speed Ωr induced by fluid shear in the oil films. The inner and
outer clearances are C1 and C2 , respectively. Both oil films are assumed to be infinitely
short, given its dimensions provide a length-to-diameter ratio close to 0.5. The com­
plete derivation of the short bearing equations modelling the isothermal floating ring
bearing is provided by [2], while the thermal effects were included and described by
[13]. Essentially, the inner and outer film forces in both directions and the torques on
the ring are given by:

wherein the dimensionless forces fi;oy;z are obtained from short bearing theory applied
to each film, considering global thermal effects [13].

2.3 Thrust bearing lubrication model


The main geometric variables of a typical turbocharger TB pad are shown in Figures 2a
and 2b. A TB pad has an inner and outer radii ri and ro , respectively, and an angular
extent θ0 . The converging gap has an angular length θramp and shoulder height sh . The
minimum oil film thickness is h0 . If the thrust collar is able to rotate around the Y and
Z axes, as illustrated by Figure 2c-f, the oil film profile must consider these rotations.
The TB modelling is the same as the one described by [12], [13], consisting of a full
three-dimensional thermo-hydrodynamic model of the entire TB . Pressure distribu­
tion is governed by the generalized Reynolds Equation (2.3), while temperature is
governed by the 3D Energy Equation (2.4). Pressure and temperature equations are
compactly written using the r operator. The integrals F0 , F1 and F2 in the Reynolds
equation account for the viscosity variation throughout the oil film thickness, accord­
ing to Eq. (2.5), and the oil film shape in the TB and its time derivative are described
by Eqs. (2.6) and (2.7), considering the collar rotations. The pressure and tempera­
ture equations are solved simultaneously by the Finite Volume Method for all pads in
the TB and the axial load and restoring moments of the TB are estimated integrating
the converged pressure over each pad area, Eq. (2.8).

Details on the modelling and the implementation on the actual equations to estimate
the thrust bearing load and restoring moments are given in [12].

42
Figure 2. (a,b) TB variables, (c) allowable displacements, rotations around
(d) Y axis and (e) Z axis and (f) displacements along X axis.

2.4 Thrust bearing database model


The actual lubrication model described by Eqs. (2.3)-(2.8) is computationally expen­
sive to solve, due to the coupling between pressure and temperature equations. An
iterative approach is sought in order to correctly estimate both the pressure and tem­
perature fields in the oil film in each TB pad. Also, the analysis cannot be simplified to
only one pad, because the collar rotation induces different pressure distributions on
each pad. This greatly enhances the computational time to solve the TB governing
equations. Further simplifications can drastically reduce the computational time to
solve it, such as the isothermal approach, neglecting heat generation within the bear­
ing. However, it has been shown this may lead to erroneous predictions [12], [13] and
the best approach to accurately model the dynamical behaviour of turbochargers sup­
ported by thrust bearings must consider thermal effects. In a dynamical analysis,
given the TB high nonlinearities, the Reynolds equation is solved on every step of the
time integration scheme. To avoid the high computational cost of solving Eqs. (2.3)­
(2.8) at every step, the Database Method described by [9] will be applied to thrust
bearings, with correct adaptions. The Database Method consists in creating a lookup
table, with the TB axial forces and restoring moments, for different input variables,
such as the collar displacements and velocities, and, once the bearing loads are
required, an interpolation in this database is performed. Different interpolation
schemes can be utilized. In this work, we propose to utilize the linear interpolation
scheme. Some important observations must be done before further developing the
Database Method:

1. Different thrust bearings will have different geometry, which means


a database should be constructed for each different thrust bearing considered
in the simulations. Given the thrust bearing geometrical parameters ri , ro ,
θramp , θ0 , sh and Npad , one could consider create one database for every desired
combination of these parameters. This, however, is not recommended, as the
database will have some terabytes in size and storage may be problematic.
2. The same observation is valid for different oil films, if thermal effects are con­
sidered. Isothermal analysis reduces the number of variables, solving the iso­
thermal Reynolds equation in dimensionless form. To include thermal
variations, however, it is necessary to specify the oil parameters ρ, cp and k and
the viscosity-temperature relation. Nondimensional analysis can characterize
heat transfer for different lubricant flows, but still three independent param­
eters (the Péclet, Eckert and Reynolds numbers) must be defined [12] to
create the database. Moreover, the dimensionless analysis relies on the defin­
ition of a reference temperature, usually taken equal to the replacement oil
temperature, which raises the number of independent variables to at least
four. The same storage problems may be encountered here if one tries to
create a single database for different lubricants.
3. Given a TB and lubricant, the number of variables in the determination of the
bearing characteristics may be six or seven, depending on whether thermal

43
effects or not will be accounted. If thermal effects will be considered, the rota­
tional speed is an important parameter related to the shear dissipation in the
oil film, so it should be considered as an independent input parameter. This
dependency does not exist under the isothermal hypothesis, as the dimen­
sionless Reynolds equation is solved.
Considering the three aforementioned observations, the strategy to develop
a database accounting for thermal effects considers the fixed-geometry thrust bear­
ings with known parameters and just one type of lubricant. With that in sight, observ­
ing the expressions for the bearing load, Eq. (2.8), and its dependence on the
Reynolds and Energy Equations, (2.3) and (2.4), the oil film shape (2.6) and its
derivative (2.7), the axial force and restoring moments can be written as a function of
seven variables:

After creating a 7D grid for different values of the rotational speed Ω, the minimum oil
film thickness h0 , the rotations y and z , the squeezing velocity h_ 0 and the rotation vel­
ocities _ y and _ z , a database can be stored for each interested parameter and further
evaluations of the axial load and restoring moments can be quickly performed by a 7D
interpolation on the database. The database can be extended to store any desired par­
ameter during the calculations, such as the average oil film temperature on the bear­
_ _ _

gations on thermal effects. The creation of the database should be carefully planned.
The evaluation of the thrust bearing thermal characteristics is highly cost, and the
right choice for the input parameters is mandatory to create the database in
a reasonable amount of time. The displacement variables h0 , y and z are restricted to
geometrical limitations. Given that a double-acting thrust bearing supports the turbo­
charger, the maximum oil film thickness is the sum of both thrust bearing clearances,
while, clearly, the minimum value should be zero. The zero value makes no physical
sense, however, one can discretize this variable  starting in a small, but nonzero,
value. The rotations are limited to the restriction y;z ro =h0 50:4, indicating that, with
the rotation, the film thickness is always positive. The specified limit is smaller than
1.0 to assure that, in the case of simultaneous rotation y and z , the film thickness is
always positive [12], [15]. The presented simulations suggest it is sufficient to con­
sider the values 10  μm  h0  40  μm and 3  mrad  y;z  3  mrad in the database cre­
ation. The geometrical limitations cannot be extended to restrict the velocities. Even
with the nondimensionalization proposed by [12], whether  
the variables are the
dimensional set h_ 0 , _ y and _ z or the nondimensional set h, y and z , both sets of vari­
ables have the domain ð∞; ∞Þ, which makes the variable discretization unfeasible.
Chasalevris and Louis [9] propose a second change of variables to the dimensionless
Reynolds equation applied to radial bearings, dividing it by a factor Q and utilizing
a second variable x1;2 to account for the squeeze term. The proposed change of vari­
ables, however, is not suitable for the thrust bearing, nor for thermo-hydrodynamic
analysis. Novotný et al. [8] restricts the approximation  velocity
 of both thrust bearing
surfaces to a positive, and yet arbitrary, limit, h_ 0 2 0; h_ max0 , but does not mention any
appropriate value for setting this limit. This approach also neglects strong cavitation
effects that may appear in the separation of both thrust bearing surfaces (for negative
velocities). In the absence of any reference  
value, we propose to create the database

for the set of variables ðΩ; h0 ; y ; z ; H ; ’y ; ’z Þ, wherein the nondimensional variables
related to the velocity terms are defined as:

44
     
H ¼ h =ro O; ’y ¼ y =O; ’z ¼ z =O ð2:10Þ

and the range


   of  each nondimensional
 variable
   is restricted to a maximum arbi­
        
trary valueH  5 H max, ’y  5 ’y max and ’z  5 ’z max. The choice of this arbitrary
   
value, however, is not completely random. If one knows, or at least has a good
estimate, of the maximum operating velocities of the turbocharger, this values are
promptly ready to use. However, if this information is not known beforehand, we
recommend the adoption of a maximum estimated velocity following a (simplified)
isothermal analysis. Given that the isothermal Reynolds equation is much faster to
evaluate than the generalized Reynolds equation, several techniques can be
employed to solve the equations-of-motion of the turbocharger. Aside from the
already established Finite Difference solution, Koutsovasilis approach [11] uses
a Galerkin form to get a fast approximated solution of the isothermal Reynolds
equation by means of a truncated series of sine terms. Analytical equations are
also available to some thrust bearing configurations [17]. Either way, after a time
transient analysis with an isothermal thrust bearing, a good estimate of the max­
imum velocities will be known. Given that the load-carrying capacity of
a turbocharger thrust bearing decreases with the inclusion of thermal effects [12],
[13], these maximum velocities estimated by the isothermal modelling may be
multiplied by a safety factor SF in order to get the range within the variables
  
H ; ’y ; ’z can be discretized. Our simulations suggest the values of
 
2  104  H  2  104 and 5  104  ’y;z  5  104 .
Creating the database requires solving the governing equations (2.3)-(2.8) for all
combinations of the input parameters. This is the highly cost operation on the method.
Its advantage is that the database is evaluated only once. Every further evaluation of
the thrust bearing performance parameters is done interpolating the desired output
from the database. Because 7 different input variables are necessary to create the
database, a 7D multivariate interpolation is necessary. This study restricts to linear
interpolation, but observe any interpolation scheme can be employed. For multilinear
interpolation, because we have a function of 7 variables, the 7-linear interpolation will
be the weighted average of 27 ¼ 128 neighbour values. Essentially, each parameter
(the axial force Fx , the restoring moments My;z and the fluid film mean temperature
Tavg ) will be evaluated as
X128
θ¼ i¼1
Ni θi ð2:11Þ

wherein each value θi is the calculated force/moment/temperature for an input value


of the 7D grid and the coefficient Ni is the weight coefficient of the linear interpolation.
This is simply an extension of the linear interpolation presented by [9] to a 7D inter­
polation. The great benefit of the Database Method is the low computational time to
evaluate the interpolation. It is stated in [9] that this evaluation time is comparable to
analytical solution evaluations. This greatly reduces the computational evaluation of
the thrust bearing thermal performance. On average, the implemented routine to
evaluate the pressure and temperature distributions of a thrust bearing takes 30 to 60
s to converge, for a known thrust collar rotational speed, displacements and velocities.
If an excessive number of points on the fluid mesh discretization is utilized, this time
can greatly increase. The interpolation scheme proposed is independent of the mesh
size, as the evaluation relies simply on the database interpolation, and this interpol­
ation time reduces to a few milliseconds, which reduces the total computational time
to do a turbocharger time transient analysis.

45
2.5 External excitation
The external axial excitation considered in the turbocharger is a harmonic excitation
on the turbine node. Lüddecke et al. [18] provided some experimental results in
dimensionless forms of the thrust force in a turbocharger due to engine operation,
indicating the thrust force from the engine is periodic. They observed that the thrust
force has a cycle average (static) value directed from the turbine towards the com­
pressor and the cycle resolved (dynamic) thrust load may vary depending on the oper­
ation point of turbocharger. In the absence of a better description of the harmonic
excitation, and noticing that any periodic function can be expanded in a Fourier series,
the external axial force, applied on node 7, considers only the first Fourier term in the
series expansion, as described by Eq. (2.12).

The negative sign in the equation defines the force direction from the turbine to the
compressor. The cosine term is also neglected as it introduces only a phase difference
in the external excitation. Lüddecke et al. [18] notices that if the static average value
is much greater than the dynamic thrust force, almost no axial displacement is
observed in the rotor. On the other hand, if the amplitude of the oscillating force is
high, the axial rotor displacement is mainly influenced by the engine load conditions.
They also provided all results in dimensionless form, dividing the axial thrust forces by
the turbocharger own weight. Observing these points, in order to observe the axial
displacement of the rotor, but to also consider the more realistic case of a nonzero
cycle average thrust force, we admit the terms in Eq. (2.12) to be equal to ten times
the rotor weight, i.e., A0 ¼ A1 ¼ 10WTC ¼ 51:3N. The frequency of excitation is composed
of two pulses at a complete engine revolution [18], so that the excitation frequency is
approximately ω1 ¼ 628 rad/s; it is twice the engine speed, admitted for an engine run­
ning at 50 Hz (3,000 rpm).

3 RESULTS
3.1 Turbocharger model
The turbocharger model is the same one described by [13]. The FE model of the turbo­
charger shown in Figure 1, whose parameters are listed in Table 1, is composed of six
beam elements, two rigid discs representing the compressor and turbine wheels, two
equal rotating floating ring bearings and a double-acting thrust bearing. The oil circu­
lating in the bearings is the Essolube X2 20W, whose temperature-viscosity relation is
described by Eq. (3.1).

46
Table 1. Finite element model parameters.

Shaft parameters
Young
Density (kg/ modulus Poisson’s Length
m3) (GPa) ratio (–) Diameter (mm) (mm)
15.5-1.00­
7860 200 0.30 11.0 1.00- 15.5­
39.0-33.0
Rigid discs parameters
Transverse
Mass (kg) Polar MoI (10-6 kg.m2) MoI (10-6
kg.m2)
Compressor
0.118 44.0 32.7
Wheel
Turbine
0.326 81.0 77.0
wheel
Floating ring bearing geometrical parameters
Radial Ring Ring polar
Bore radius Length
clearance mass MoI (10-7
(mm) (mm)
(μm) (g) kg.m2)
Inner film 35.0 5.535 6.5
7.5 3.504
Outer film 75.0 8.000 9.0
Thrust bearing geometrical parameters
Pad Ramp
Outer
Inner radius Shoulder angular angular Number of
radius
(mm) height (μm) extent extent pads (–)
(mm)
(°) (°)
5.5 10.0 20.0 100 75 3
Oil physical parameters
Specific
Density (kg/ heat cap­ Inlet oil tem­
Thermal conductivity (W/m.°C)
m3) acity (J/ perature (°C)
kg.°C)
880 1950 0.130 100

3.2 Database method in time transient analysis


Given the rotor-bearing system model described in the previous section and the exter­
nal axial harmonic excitation force discussed on Eq. (2.12), it is possible to simulate
the system time response. In order to observe only the bearing damping characteris­
tics, the damping matrix of the rotating shaft is assumed to be zero. In addition, in
order to correctly ascertain the different phenomena in this rotating system, the
unbalance forces will also be neglected. Finally, the engine excitations due to base
motion will also be neglected. This assumption corresponds to the investigations of
Tian et al. [2], who noticed that for rotor speeds higher than 45 krpm, the engine

47
induced vibrations are suppressed by the dominant sub-synchronous oscillations due
to oil whirl and whip in the FRBs. The simulations are performed for a rotational speed
of 100 krpm, so it is expect little influence of the engine excitations on the lateral oscil­
lations. With these assumptions, the work aims to investigate the fluid-induced
instability presented in the FRBs and the TB coupling effect in lateral oscillations.

Figure 3. – TB axial displacement. (a) direct solution (DS), (b) database solu­
tion (DB), (c-f) comparison of both solutions at selected time spans.

Figure 3 presents the axial displacement of the thrust bearing midpoint (node 3 in the
FE discretization), when the system is subjected to the axial harmonic excitation. Fig­
ures 3a and 3b present the response utilizing the Direct Solution (DS), solving the
Reynolds equation at each time step, and the Database Method solution (DB), respect­
ively. Figures 3c-f highlight different time intervals in order to observe the differences
utilizing each approach. Figure 3c presents the solution at the very beginning of the
simulation, Figure 3d, at the instability transition, further explained, Figure 3e, at
some time after the instability transition and, finally, Figure 3f at the end of the simu­
lation. Clearly, both approaches do not predict the exact same response, but they
have an excellent agreement. The biggest differences are observed only during the
instability transition, in Figure 3d.

Figure 4. – TS FRB: (a) orbit, (b) vertical displacement and (c) eccentricities.

48
The initial conditions admitted for the simulations are the static equilibrium position of
the rotor structure. Provided no external loads or perturbations interfere in the float­
ing ring bearings, the system would remain in this static equilibrium position forever
[13]. This static equilibrium position, however, is an unstable position. Given that the
TB may induce bending moments on the shaft, the perturbation from the TB induces
the FRB instability, as shown in Figure 4a. Figure 4a presents the journal orbit of the
TS FRB. It is possible to identify the journal rapidly diverges from its initial position,
but reaches a limit cycle smaller than the bearing clearance and remains within this
cycle. This can also be observed on the eccentricities of the inner and outer films on
Figure 4c. Figure 4b presents the vertical displacement of the journal as a function of
time, during the first half of total time, in order to show the negligible differences of
both models. A great agreement is observed between both curves. The small differ­
ences in the lateral response are due to the TB restoring moments estimated by the
two different methods.

Figure 5. – Average oil temperature on CS TB (a) and TS TB (b).

Thermal effects can also be addressed by the database method. The direct solution is
extremely cost, as it requires to solve the generalized Reynolds equation altogether
with the energy equation at every time step during the time integration scheme, while
the database method just interpolates the values from the lookup table. During the
time integration scheme, the interpolation of the database method can also be used to
analyse any other interested parameter. Storing the average temperature of the fluid
film in each TB allows for the comparison of the temperature evolution of both
schemes. Figure 5 illustrates this behaviour. In Figure 5a, the global average film tem­
perature of CS TB is shown, as predicted by both methods, while Figure 5b shows the
TS TB average film temperature. The oscillating predicted behaviour is captured by
the Database solution. The biggest difference is observed on the TS TB, as the tem­
perature amplitude oscillation predicted by the Database Method is a little higher than
the direct solution. Both solutions are capable to show the temperature increase
during the instability transition and the temperature oscillations due to axial and rota­
tional motion of the thrust collars. Finally, another important aspect of the Database
method not revealed in any results is the computational time to evaluate this time
transient solution. The code for both solutions was written and compiled with Intel
Visual Fortran 11.1.038 and both solutions were run in a personal computer with an
Intel Core i7-5500 CPU @ 2.40 GHz processor and 8.00 GB of RAM. The direct solution
took 3,368 minutes (56 hours) to run, while the database solution took only 2 min­
utes. That is an impressive gain in terms of computational time and shows the suitabil­
ity of the Database Method to model thrust bearings and perform thermal transient
analysis on turbochargers.

49
4 CONCLUSIONS
This work compared two methods for a time transient analysis of a turbocharger
supported by floating ring and thrust bearings. The hydrodynamic thrust bearing
loads and moments are estimated by the generalized Reynolds equation, as ther­
mal effects are also considered. The two approaches in the time integration were
to solve the Reynolds equation at each time step (the Direct Solution) or to inter­
polate it from a lookup table (the Database Method). Both approaches produce
results in very good agreement, and the suitability of the Database Method over
the Direct Solution concerns the computational time to evaluate both models. The
Database Method is an extremely fast model and should pose as an excellent
alternative to the reliable, but slower, Direct method, solving the governing equa­
tions every time step. The dynamic simulations of the turbocharger revealed the
unstable characteristics of the floating ring bearings, as a small perturbation on
the shaft lateral motion induces a diverging movement of the rotating shaft from
its static equilibrium position, reaching a limit cycle (smaller than the bearing
clearances) and remaining within this cycle. When the shaft is axially excited, the
thrust collar rotations induce restoring moments on the shaft, that are great
enough to perturb the lateral motion and induce the oil whirl/whip on the floating
ring bearings. The axial thrust was modelled as a harmonic force, with a mean
value directed from the turbine to the compressor. The axial response is
a harmonic motion, and a perturbation on axial displacements is observed in the
lateral oscillations instability transition, revealing the natural coupling of the axial
and lateral dynamics. Thermal effects were also addressed and both methods
could capture the overall behaviour. The temperature variations resemble the axial
displacement: a harmonic variation of the average film temperature.

ACKNOWLEDGEMENTS
The authors would like to thank BorgWarner Brasil Ltda., SAE-Unicamp, CAPES, CNPq
and FAPESP grant #2015/20363-6 for supporting this research.

REFERENCES
[1] C.-H. Li, “Dynamics of Rotor Bearing Systems Supported by Floating Ring
Bearings,” Journal of Lubrication Technology, vol. 104, no. 4, pp. 469–476, Oct.
1982, doi: 10.1115/1.3253258.
[2] L. Tian, W. J. Wang, and Z. J. Peng, “Dynamic behaviours of a full floating ring
bearing supported turbocharger rotor with engine excitation,” Journal of Sound
and Vibration, vol. 330, no. 20, pp. 4851–4874, Sep. 2011, doi: 10.1016/j.
jsv.2011.04.031.
[3] B. Schweizer, “Total instability of turbocharger rotors—Physical explanation of
the dynamic failure of rotors with full-floating ring bearings,” Journal of Sound
and Vibration, vol. 328, no. 1–2, pp. 156–190, Nov. 2009, doi: 10.1016/j.
jsv.2009.03.028.
[4] L. Tian, W. J. Wang, and Z. J. Peng, “Effects of bearing outer clearance on the
dynamic behaviours of the full floating ring bearing supported turbocharger
rotor,” Mechanical Systems and Signal Processing, vol. 31, pp. 155–175, Aug.
2012, doi: 10.1016/j.ymssp.2012.03.017.
[5] R. J. Trippett and D. F. Li, “High-Speed Floating-Ring Bearing Test and Analysis,”
A S L E Transactions, vol. 27, no. 1, pp. 73–81, Jan. 1984, doi: 10.1080/
05698198408981547.
[6] P. Koutsovasilis, N. Driot, D. Lu, and B. Schweizer, “Quantification of
sub-synchronous vibrations for turbocharger rotors with full-floating ring

50
bearings,” Arch Appl Mech, vol. 85, no. 4, pp. 481–502, Apr. 2015, doi: 10.1007/
s00419-014-0924-0.
[7] P. Novotný, P. Škara, and J. Hliník, “The effective computational model of
the hydrodynamics journal floating ring bearing for simulations of long tran­
sient regimes of turbocharger rotor dynamics,” International Journal of
Mechanical Sciences, vol. 148, pp. 611–619, Nov. 2018, doi: 10.1016/j.
ijmecsci.2018.09.025.
[8] P. Novotný, J. Hrabovský, J. Juračka, J. Klíma, and V. Hort, “Effective thrust bear­
ing model for simulations of transient rotor dynamics,” International Journal of
Mechanical Sciences, vol. 157–158, pp. 374–383, Jul. 2019, doi: 10.1016/j.
ijmecsci.2019.04.057.
[9] Chasalevris and Louis, “Evaluation of Transient Response of Turbochargers and
Turbines Using Database Method for the Nonlinear Forces of Journal Bearings,”
Lubricants, vol. 7, no. 9, p. 78, Sep. 2019, doi: 10.3390/lubricants7090078.
[10] I. Chatzisavvas, A. Boyaci, P. Koutsovasilis, and B. Schweizer, “Influence of
hydrodynamic thrust bearings on the nonlinear oscillations of high-speed
rotors,” Journal of Sound and Vibration, vol. 380, pp. 224–241, Oct. 2016, doi:
10.1016/j.jsv.2016.05.026.
[11] P. Koutsovasilis, “Automotive turbocharger rotordynamics: Interaction of thrust
and radial bearings in shaft motion simulation,” Journal of Sound and Vibration,
vol. 455, pp. 413–429, Sep. 2019, doi: 10.1016/j.jsv.2019.05.016.
[12] T. F. Peixoto and K. L. Cavalca, “Investigation on the angular displacements influ­
ence and nonlinear effects on thrust bearing dynamics,” Tribology International,
vol. 131, pp. 554–566, Mar. 2019, doi: 10.1016/j.triboint.2018.11.019.
[13] T. F. Peixoto and K. L. Cavalca, “Thrust bearing coupling effects on the lateral
dynamics of turbochargers,” Tribology International, p. 106166, Jan. 2020, doi:
10.1016/j.triboint.2020.106166.
[14] N. Mittwollen, T. Hegel, and J. Glienicke, “Effect of Hydrodynamic Thrust Bear­
ings on Lateral Shaft Vibrations,” Journal of Tribology, vol. 113, no. 4, p. 811,
1991, doi: 10.1115/1.2920697.
[15] L. San Andrés, “Effects of Misalignment on Turbulent Flow Hybrid Thrust
Bearings,” Journal of Tribology, vol. 124, no. 1, p. 212, 2002, doi: 10.1115/
1.1400997.
[16] H. D. Nelson, “A Finite Rotating Shaft Element Using Timoshenko Beam Theory,”
J. Mech. Des., vol. 102, no. 4, p. 793, 1980, doi: 10.1115/1.3254824.
[17] S. Liu and L. Mou, “Hydrodynamic Lubrication of Thrust Bearings with Rectangu­
lar Fixed-Incline-Pads,” Journal of Tribology, vol. 134, no. 2, p. 024503, Apr.
2012, doi: 10.1115/1.4006022.
[18] B. Luddecke, P. Nitschke, M. Dietrich, D. Filsinger, and M. Bargende, “Unsteady
Thrust Force Loading of a Turbocharger Rotor During Engine Operation,” p. 10,
2015.

51

You might also like