You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/359391354

Thermal Simulation of High-Speed EV Transmission Bearings for Minimum


Lubricant Volume

Article in SAE Technical Papers · September 2022


DOI: 10.4271/2022-01-1120

CITATIONS READS
3 241

4 authors:

Marton Ferenc Toth David Sz. Gangl


AVL LIST GMBH Self-Employed
2 PUBLICATIONS 3 CITATIONS 7 PUBLICATIONS 4 CITATIONS

SEE PROFILE SEE PROFILE

Hemant Bansal Michael Leighton


BET-Motors AVL LIST GMBH
18 PUBLICATIONS 22 CITATIONS 72 PUBLICATIONS 390 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Michael Leighton on 23 March 2023.

The user has requested enhancement of the downloaded file.


20-08-2022

Thermal Simulation of High-Speed EV Transmission Bearings for Minimum Lubricant


Volume
M. Toth, D. Gangl, H. Bansal & M. Leighton
AVL List GmbH

Abstract solution to define oil speed properties inside the bearing, as it will be
discussed in current study.
Minimizing the lubricant volume in a transmission system reduces the
churning losses and overall unit costs. However, lubricant volume It should be noted, applied formulas regarding heat transport contain
reduction is also detrimental to the thermal stability of the system. many uncertainties, and the application has its own limits. As can be
Transmission overheating can result in significant issues in the region observed in [1], where finite element method has been applied for a
of loaded contacts, risking severe surface / sub-surface damage in deep groove ball bearing, temperature results have shown a large
bearings and gears, as well as reduction in the lubricant quality deviation comparing with measurement. After calibration of the heat
through advanced oxidation and shear degradation. transfer coefficient, the analytical relation was able to describe
accurately the thermal conditions at the bearing. In [6] can be found
a presentation about a complex thermal network model for angular
The increasing trend of electrified transmission input speeds raises
contact ball bearings, including effects of cage and thermal expansion
the importance of understanding the thermal limits of the system at
of bearing elements. It has been concluded, errors between
the envelope of the performance to ensure quality and reliability can
calculation and test results were getting bigger with increasing speed,
be maintained, as well as being a key factor in the development,
up to 8.5% at 12,000 rpm. This result is quite great, comparing with
effecting internal housing features for the promotion of lubrication.
another study [8], investigating also angular contact ball bearings
with thermal network model, where the deviation was up to 19% at
A nodal bearing thermal model will be shown which utilizes thermal
the same speed level. So, similar accuracy can be expected in current
resistances and smooth particle based CFD for determining bearing
paper, and although it was not possible to measure directly bearing
lubricant feed rates during operation. Heat transfer is evaluated
temperatures, but some tests have been carried out to compare them
transiently for a range of operating conditions with the bearing
with calculation results. Another applied thermal network system,
contact mechanics solutions applied for determination of the contact
described in [4], where heat effect of ball bearing’s degradation has
thermal resistances and the corresponding lubricant flow across the
been analyzed. Observing the numerical results with experimental
bearing. The nodal bearing model will be shown to have been
data, similar or lower differences have been achieved, compared to
validated against measured thermal data from a high-performance
the mentioned studies, but note, in this case grease lubricant was
vehicle e-axle with a significant focus on the determination of the
considered.
minimum lubricant volume.
It is often quite difficult to directly validate thermal conditions with
Introduction measurements for transmission system. The critical temperatures are
typically at the functional surfaces of interacting mechanical
Determination of bearing thermal conditions has a huge importance components. This reduces the proximity that can be achieved with a
in transmission systems, especially in the case of high-speed systems. sensor without disturbing the performance of the system or inclusion
For the investigation of thermal effects, various approaches can be of telemetry for the measurement. Of course, there are also non-
used, for example finite element method [1], computational fluid contact thermal sensors but the installation and accuracy of these is
dynamic (CFD) [2] or thermal-network system [3, 4, 5, 6, 7, 8], which typically equally challenging. With the positioning of sensors away
has been applied in current study. Thermal-network model offers a from the contact, the value in the data for validation diminishes as
simple approach to estimate the temperature of bearing elements the transfer of the energy and any dissipation between the sensor and
and its environment. The largest benefit of this method is the the point of interest must also be considered.
simplicity of adaptation between applications and the low
computational demand. According to the relevant literature, some As it will be later detailed, the thermal-network model is based on the
papers can be found investigating the bearing thermal conditions in analogue of heat transfer and electrical circuits. Each node or point
case of grease lubrication, as in [1, 4]. Dealing with this condition, represents an isothermal element in the system, and these points are
lubricant has no additional axial speed inside the bearing, which connected through heat resistances, calculated from the geometry
differs from a situation in a high-speed transmission system. In [5, 6, and material properties. Limitations and weaknesses of this modelling
7] oil speed into the bearing is not negligible, but cannot found method will be also highlighted, and the model validation considered.
detailed description, how to take into account this effect in the heat
transfer calculations. SPH based CFD method offers a reasonable
Page 1 of 11

AVL List GmbH


Materials and Methods the current study it has been neglected, due to its minor effects
compared with drag losses.
Thermal-network model A calculation method for this can be found in [5], where the forces
acting on the rolling elements have been considered as drag forces on
In the following sections a simple thermal-network model will be a moving sphere in a fluid medium. The drag force and power loss can
presented for a deep groove ball bearing. It will be shown, how a be expressed as:
thermal system of this model can be constructed and the calculation
method for heat resistances and heat sources can be seen.
𝜌𝑒𝑓𝑓 𝐶𝐷 𝜔𝑐 𝑑𝑚 2 𝜋𝐷𝑏𝑎𝑙𝑙
𝐹𝑑𝑟𝑎𝑔 = ( ) 𝐷𝑏𝑎𝑙𝑙 ( − 𝑡𝑐 ) (5)
2 2 4
Heat sources
𝐹𝑑𝑟𝑎𝑔 𝜔𝑐 𝑑𝑚
The sources of the generated heat around the bearing can be divided 𝑃𝑑𝑟𝑎𝑔 = 𝑛𝑏𝑎𝑙𝑙 ∙ (6)
2
into two main components.
where 𝜔𝑐 is the angular velocity of the cage, 𝑑𝑚 denotes the pitch
1. The power loss due to the friction can be considered at the diameter, 𝐷𝑏𝑎𝑙𝑙 is the diameter of ball, 𝑛𝑏𝑎𝑙𝑙 represents the number
rolling elements and inner, outer rings. Later, these losses of rolling elements and 𝑡𝑐 is the thickness of the cage. In this equation
will be mentioned as mechanical or friction heat sources. 𝐶𝐷 is the drag coefficient between the lubricant and rolling elements,
2. Power loss will also be produced by the drag forces, during it can be estimated from empirical formulas for different Reynolds-
the interaction between the lubricant medium and rolling number (Re) ranges for a sphere. Additionally, 𝜌𝑒𝑓𝑓 expresses the
elements, this effect will be also included in the presented effective density of the medium, that the rolling elements are moving
thermal model. through.

Power loss due to the interaction between cage/balls and cage/rings In case of a multiphase fluid, e.g., an oil-air mixture, an oil-air fraction
can be also generated additional heat in the system, but in the current should be set up and the equivalent properties given. These
study, these sources have been neglected. relationships have been used also in [5] and [10]. The equivalent
physical properties for an oil-air mist can be expressed as:
To determine the friction losses, a simple empirical formula can be
used according to [1] for the calculation of moments coming from the 𝜇𝑜𝑖𝑙
𝜇𝑒𝑞 = 𝜇 (7)
frictional power losses: 𝑜𝑖𝑙 (1
∙ − 𝑋) + 𝑋
𝜇𝑎𝑖𝑟
𝑀𝑓𝑟𝑖𝑐 = 𝑓1 𝑃1 𝑑𝑚 (1)
𝜌𝑒𝑞 = 𝑋 ∙ 𝜌𝑜𝑖𝑙 + (1 − 𝑋) ∙ 𝜌𝑎𝑖𝑟 (8)
Where 𝑓1 can be expressed for ball bearings as:
Where 𝜇, and 𝜌 are the dynamic viscosity and density of the
0.55 corresponding medium (oil or air). X denotes the oil-air fraction.
𝑃0
𝑓1 = 0.0009 ( ) (2) These equivalent physical parameters will be applied later for other
𝐶0 relations regarding convective heat transfer and fluid mechanics.

𝑃0 is the equivalent static load and 𝐶0 denotes the rated static load. Unfortunately, the oil-air ratio is a quite questionable value and
𝑃1 is another calculation factor, that can be calculated as: strongly influences the properties of the lubricant medium. Smooth
Particle Hydrodynamic (SPH) CFD could be a method offering
𝑃1 = 3𝐹𝑎 − 0.1𝐹𝑟 (3) reasonable estimation of this parameter.

Where 𝐹𝑎 and 𝐹𝑟 are the axial and radial loads on the bearing. 𝑑𝑚 is Heat resistances
the pitch diameter of the corresponding bearing. For the calculation
of power loss (𝑄̇𝑓𝑟𝑖𝑐 ) the rotational speed of the shaft should be used The thermal network model of the investigated bearing consists of
(𝑛𝑟𝑜𝑡 ): elements that are considered as isothermal parts, such that a nodal
system is formed. Based on the analogue of electrical circuits, these
𝑄̇𝑓𝑟𝑖𝑐 = 𝑀𝑓𝑟𝑖𝑐 ∙ 2𝜋𝑛𝑟𝑜𝑡 (4) parts are represented with points, and each point represents the local
temperature. Additionally, connections between isothermal parts are
Many other approaches and relationships regarding the described with heat resistances, so equations can be constructed for
determination of mechanical power losses in the bearings have been two neighbor points as can be seen in the following relation:
reported in literature, for example [9]. For initial results the simplistic
analytical estimation was considered sufficient. Δ𝑇
𝑅= (9)
𝑄̇
The second part of the power loss is generated by drag forces
between the rolling elements and the lubricant. Based on [9], Where R, and Δ𝑇 represent the heat resistance and the thermal
churning losses could be considered in addition to drag losses, but in potential difference between the nodes, and 𝑄̇ is the heat flow. Using
this formula for the whole network, a numerical solution of the
Page 2 of 11

AVL List GmbH


equation system can give results about the thermal conditions of the was considered. Regarding the heat transfer, inner race surface was
corresponding system (analogous of Kirchhoff’s law). The numerical set up as the contact area. Another calculation method can be used
resolution of this system will be discussed later. based on [5] and applying formula for concentric rotating cylinders.
In further investigations, it could be also beneficial to compare
Conductive resistances resulted thermal conditions using different approaches.

For the determination of heat resistances inside the inner and outer In order to estimate the average oil flow speed, the following relation
ring of the bearing, the general formula for cylinders has been has been assumed:
applied, as used also in [4]:
̇
𝑉𝑓𝑙𝑜𝑤
1 𝑑𝑒𝑥𝑡 𝑢̅𝑓𝑙𝑜𝑤 ≈ (16)
𝑅𝑐𝑜𝑛𝑑 = ln ( ) (10) (𝑑𝑜2 − 𝑑𝑖2 )𝜋
2𝜋𝐾𝐿 𝑑𝑖𝑛 − 𝑠𝑐
4

Where K denotes the thermal conductivity of the ring, or the ̇


Where 𝑉𝑓𝑙𝑜𝑤 denotes the lubricant flow rate into the bearing, 𝑑𝑜 , 𝑑𝑖
corresponding component, L is the length of the races, and 𝑑𝑒𝑥𝑡 , 𝑑𝑖𝑛 are the guide diameters of outer and inner ring and 𝑠𝑐 represents the
mean the outer and inner diameter of the cylinder where heat cage cross-sectional area.
transfer is considered. This approach has also been used in case of
heat transfer between shaft/inner ring and the housing/outer ring. Next, the forced convection is the heat transfer between rolling
elements and lubricant will be considered. The calculation method is
Convective resistances similar to the previous section, and also used in [4], but the
characteristic length is taken as the ball diameter and for the
The cases for convective heat resistances can be divided into two characteristic speed is taken as the Root Mean Square (RMS) of cage
types: forced and free convection. In general, the convective heat velocity and the average axial oil flow velocity. Another approach can
resistances can be expressed with this formula, as used in [4]: be found in [11] for moving balls in a fluid medium but the resulted
thermal conditions show negligible difference using the alternative
1 𝐿𝑥 relation.
𝑅𝑐𝑜𝑛𝑣 = ( ) (11)
𝑠 𝐾𝑁𝑢
Convective heat transfer was considered between the outer (static)
Where s is the contact surface for heat exchange, 𝐿𝑥 denotes the ring and the oil, with calculation of thermal resistance based on
characteristic length, K is the thermal conductivity of the fluid equation (11). In this case, characteristic length is taken as the race
medium, and Nu is the Nusselt number. Most of the calculation length, and the characteristic velocity is taken as the Root Mean
methods regarding convective heat transfer are based on this Square (RMS) of cage velocity and the average axial oil flow velocity.
equation. As a result, the following sections present these methods Additionally, the outer race surface was applied for the contact area
for the forced convection. of heat exchange.

First, a convective heat resistance was applied between the rotating The investigated ball bearing is rotating on a hollow shaft, therefore
inner ring and the lubricant. The calculation method was the same as thermal resistance can be determined between the hollow shaft
in equation (11), including the inner race surface (s), length of the surface and the oil-air mixture inside the shaft. Assuming forced heat
bearing (L) and the thermal conductivity of the lubricant (K). The convection, based on [6, 12], the following relations can be found for
Nusselt number is a questionable parameter. According to [4] Nu this case:
number can be determined roughly with the calculation method for
fluid flow around a plate: 𝑁𝑢 = 30.5 𝑅𝑒 −0.0042 𝑅𝑒 ≥ 9600 (17)
103 < 5∙105
𝑁𝑢 = 0.332𝑅𝑒1/2 𝑃𝑟 1/3 𝑓𝑜𝑟 { 𝑃𝑟 ><0.5
𝑅𝑒 (12) 𝑁𝑢 = 𝑅𝑒 0.37 7300 ≤ 𝑅𝑒 < 9600 (18)
5
𝑁𝑢 = 0.028𝑅𝑒 4/5 𝑃𝑟 1/3 𝑓𝑜𝑟 { 𝑅𝑒 > 5∙10
𝑃𝑟 > 0.5 (13) 𝑁𝑢 = 0.00308𝑅𝑒 + 4.432 7300 > 𝑅𝑒 (19)

Where Re and Pr are the Reynolds number and the Prandtl number, Lastly, the free convection between housing surface and ambient air
dimensionless values: should be also considered as convective thermal resistance. For this
purpose, the relation of the horizontal cylinder heat exchange was
𝐷𝑥 𝑈 used, according to [13]:
𝑅𝑒 = (14)
𝑣
1/6
𝐺𝑟𝑃𝑟
𝑣 𝑁𝑢1/2 = 0.6 + 0.387 { } (20)
𝑃𝑟 = (15) [1 + (0.559/𝑃𝑟)9/16 ]16/9
𝛼
Where Pr is the Prandtl number, and Gr represents the Grashof
The 𝐷𝑥 represents the characteristic length, and 𝑣 is the kinematic
number:
viscosity, and 𝛼 the thermal diffusivity. U denotes the characteristic
speed, in the current case of inner ring-lubricant interaction, the
superposition of race velocity and the average lubricant flow velocity
Page 3 of 11

AVL List GmbH


𝑔𝛽(𝑇𝑤 − 𝑇∞ )𝑥 3
𝐺𝑟 = (21)
𝜈2

Where g and 𝛽 are the gravity factor and the thermal expansion
coefficients, 𝑇𝑤 and 𝑇∞ denote the wall temperature and the ambient
(air) temperature respectively. In this equation x is the characteristic
length and 𝜈 is the kinematic viscosity of the air. The housing around
the bearing, of course, contains some ribs and other geometrical
deviations, but as a simplification, the housing has been considered
as a horizontal cylinder with an average diameter, as a characteristic
length.

Contact resistances for the constriction heat transfer

At the rolling element and ring contacts contact heat resistances


should be applied to complete the thermal network model. The
estimation of this resistance is possible using the properties of the
Hertzian contact and the contact speed, so the following equation has
been created to determine this resistance, used in [4] and referred in Figure 1 Applied thermal network model for the ball bearing.
[14]:
Although the cage geometry was included in the calculation of
0.918 average flow speed and drag force, its effect on the heat exchange
𝑅𝑐𝑜𝑛 = (22)
2𝑏𝜒√2𝑎𝑉 has been assumed negligible. Thermal expansion of the bearing
elements has not been included in this model, due to the minor
Where a and b are the two axes of the Hertzian contact area, V is the effects on the thermal conditions and with the expectation that a
relative speed of the contact surfaces and 𝜒 denotes the thermal reasonable temperature rise will be considered. Additionally, heat
effusivity, that can be expressed as: exchange between ambient air and the side of bearing races was also
neglected, because of the small contact area and negligible thermal
𝜒 = √𝐾𝜌𝑐 (23) effects comparing with forced convection between oil and rolling
elements.
Where K is the thermal conductivity, 𝜌 is the density and c represents
Table 1. Elements of the thermal network model
the specific heat of the corresponding material. Using these formulas,
contact heat resistances have to be taken into account and added to
Node Description
the thermal network system.
1 Oil-air in hollow shaft

Boundary conditions 2 Inner ring, inner surface


3 Inner ring mass
For the applied thermal network model boundary conditions have 4 Inner ring-rolling element contact
been considered in some points of model, as will be detailed in the
5 Rolling element mass
next chapter. The ambient air temperature was set up as 25°C out of
the housing and the oil-air mixture temperature inside the hollow 6 Outer ring-rolling element contact

shaft was considered as a constant value, as 80°C. The oil inlet 7 Oil-air mixture (lubricant)
temperature into the bearing was set up also as 80°C. Heat source 8 Outer ring mass
from frictional losses has been equally distributed and added to the 9 Outer ring-housing contact
inner ring-ball and outer ring-ball contact. The other heat source from
10 Housing surface
drag forces was added directly to the lubricant inside the bearing.
11 Ambient air

Nodal distribution and model conditions

As can be seen from Figure 1, the thermal network model has been In the model of the corresponding bearing, nodes represent the
generated for a deep groove ball bearing and comprises 11 nodes. In thermal property of the given point. Each node and its role can be
the current model the contact resistances between the inner ring and seen in Table 1. As mentioned before, in some points of network
shaft surfaces and the outer ring-housing contact have been system have been applied boundary conditions. So, the temperature
neglected, assuming a well-fitted connection with minimal thermal of oil-air mixture inside the hollow shaft was considered at node #1,
resistance. (‘N1’), constant oil inlet temperature was assumed at ‘Oil-Inlet’ point,
connecting to node #7 (‘N7’), and air temperature as set up outside
the transmission (node #11, ‘N11’). ‘Q4’ and ‘Q6’ are representing the
heat source from friction losses, have been added to the node #4 and
#6 (‘N4’, N6’). ‘Q7’ denotes the heat source from drag forces, it was
prescribed at node #7 (‘N7’). Furthermore, apart from the mentioned
Page 4 of 11

AVL List GmbH


relations and conditions, heat exchange through the flowing lubricant
and the mass transport should be taken into account. Considering an
average mass flow (𝑚̇), the following equation can be constructed:

𝑄̇ = 𝑚̇𝑐𝑝 (𝑇𝑜𝑖𝑙,𝑜𝑢𝑡 − 𝑇𝑜𝑖𝑙,𝑖𝑛 ) (24)

Where 𝑇𝑜𝑖𝑙,𝑜𝑢𝑡 and 𝑇𝑜𝑖𝑙,𝑖𝑛 are the average oil outlet and inlet
temperatures, 𝑐𝑝 is the specific heat of oil and 𝑄̇ represents the rate
of thermal energy transfer to the lubricant. Assuming the average
value of inlet and outlet temperatures for the oil temperature inside
the bearing, it can be expressed in the following relation, similar to
[5]:

𝑇𝑜𝑖𝑙,𝑜𝑢𝑡 + 𝑇𝑜𝑖𝑙,𝑖𝑛
𝑇7 = (25)
2

Using these two additional equations, the thermal network model can
be completed, and a thermal equilibrium condition can be
determined for a given set of boundary conditions.

Physical parameters of the lubricant were set as equivalent properties


applying a given oil-air fraction, which can in this case be called the
density ratio. As mentioned before, it is a questionable parameter of
the study, and describes how much the bearing cavity volume is filled
with lubricant. Based on another study [5] and according to
parameter analysis, 0.1-0.2 seemed to be a reasonable value for this
property.
Figure 2. Scheme about the calculation flow of thermal network model.
Numerical resolution
Temperature measurement set up
Constructing thermal equations, based on equation (9), a non-linear
equation system can be created, based on the characteristic equation
The potential value of a thermal network model for application to
for thermodynamic systems, as applied in [4]:
design development or future design validations is dependent on the
𝑑𝑢 validation of the thermal results with experimental data. A validated
𝑄̇ = 𝑊 + (26) model could be used to support development by showing how an
𝑑𝑡
optimal lubricant volume could be set up to achieve better thermal
performance, reliability and efficiency.
Where 𝑄̇ is the heat flow and u represents the energy variation. This
relation can be expressed for the current system, based on the
To evaluate the model, temperature measurements have been
electrical circuit analogue, as used in [4]:
carried out on a high-speed Electric Drive Unit (EDU) A diagram of the
𝑛 testbed installation as can be seen on Figure 3. One temperature
𝑇𝑖 − 𝑇𝑗 𝑑𝑇𝑖 sensor was placed on the housing surface, close to the investigated
𝑄̇𝑖 = ∑ + 𝑚𝑖 𝑐𝑝𝑖 (27)
𝑅𝑖𝑗 𝑑𝑡 deep groove ball bearing. The tests have been performed at two input
𝑗,𝑗≠1
speeds and with two rotational directions.
Where 𝑇𝑖 , 𝑇𝑗 are the temperatures of the neighboring nodes, 𝑅𝑖𝑗
denotes the thermal resistance between these points, 𝑚𝑖 and 𝑐𝑝𝑖 Additionally, SPH lubrication simulations have been also performed
for this transmission system and bearing flow rate values were
represent the mass and specific heat capacity at each node.
evaluated for the mentioned operation points. These flow rates have
Considering a steady state investigation, equation (27) will be
been used in the thermal network model as input parameters.
simplified, applying this condition:

𝑑𝑇𝑖
=0 (28)
𝑑𝑡

The resulted equation system has been linearized and solved with
iteration method, using python environment. The scheme of the
calculation methodology can be seen on Figure 2.

Page 5 of 11

AVL List GmbH


Steady-state temperature devations between
test and calculation results
40

30

Deviation [%]
20

10

0
Figure 3. Diagram of EDU testbed set-up.
-10
Results
-20
4000 [rpm] 8000 [rpm]
The measured temperature data has been compared with results of
the presented nodal thermal model, assuming a steady-state
Density ratio: 0.15 Density ratio: 0.2 Density ratio: 0.25
operating condition. The sensor has been placed on the housing and
in thermal network model at this point and its temperature are
represented with node #10 (‘N10’). Assuming an isothermal housing Figure 4. Comparison of model and measurement results for housing surface
surface temperature, the value of this node was considered and temperature at different density ratio values. Implemented lubricant flow
compared with measurement results, as can be seen in Figure 4. In rates have been determined from lubrication simulation.
order to show the difference between test and model results,
deviation can be calculated as: It can be also seen that, for the greater speed rate, a lower density
ratio provides a better fit between the calculated results and
𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 = (𝑇𝑐𝑎𝑙𝑐 − 𝑇𝑡𝑒𝑠𝑡 )/𝑇𝑡𝑒𝑠𝑡 ∙ 100 [%] (29) measurements. This is likely to be due to the higher rotational speeds
increasing the aeration within the bearing and a higher rate of
discharge from the bearing surfaces moving faster. In case of lower
Where 𝑻𝒕𝒆𝒔𝒕 and 𝑻𝒄𝒂𝒍𝒄 are the measured and the calculated
speed, the trend of this effect is contrary, so smaller discharge results
temperature on the housing surface (node ‘10’).
smaller amount of oil in the bearing and a lower density ratio has
been considered for the right calculation. It can be also remarked,
It was earlier discussed that the oil-air fraction, or density ratio, is one
with higher density ratio lubricant’s cooling effect is too much, so
of the most questionable parameters in the proposed model.
model result differs from the test and middle value of the investigated
According to the further analysis, the model is quiet sensitive to this
density ratios was chosen.
property.
According to the parameter analysis, choosing even smaller density
Figure 4. shows that assuming a higher density ratio causes the
ratio also results in higher temperature values, due to the insufficient
underestimation of housing temperatures comparing with test
lubrication condition in the model. Based on this analysis, density
results. The reason for this is that the lubricant higher specific heat
ratio has been set up as 0.2 and 0.15 for the two speed rates, as these
capacity and density than the air and increasing the ratio increase the
values allowed calibration of the model.
presence of oil in the oil-air mixture.

Figure 5. Transient evaluation, comparison test and model results.

Page 6 of 11

AVL List GmbH


Temperature results from thermal model at different flow
rates (8000 rpm, density ratio: 0.15)
After the evaluation of steady-state calculation and calibration with 150
test results, calculation has been performed also for transient case
based on equation (27), presented on Figure 5. There can be seen the

Temperature [°C]
comparison of measured and calculated temperatures at the housing 130
surface for each time step. It can be observed, the middle range of
temperature values shows a low correlation with test results, 110
especially at high-speed level. To improve model accuracy regarding
transient results, should be analyzed the effects of several input
parameters, so that could be the aim of further investigations. In the 90
next chapters, steady-state conditions have been assumed and its
results are shown regarding the influence of oil flow rate and oil-air
70
fraction.
0 2 4 6 8 10 12 14
Flow rate [l/min]
Figure 6. presents the effects of flow rate on the simulated bearing
temperature condition. Nodes ‘3’, ‘5’, and ‘8’ are representing the Inner ring mass ('N3') Outer ring mass ('N8')
Balls ('N5') Oil outlet
inner ring mass, rolling elements and outer ring mass.

It can be observed from Figure 6. that, after flow rate reaches Figure 7. Investigation of lubricant flow rate’s effect regarding temperature
approximately 2.5 l/min, the temperature of bearing elements seem condition.
to be stabilized. It should be noted that the average lubricant flow
velocity has been calculated with an estimation of the flow rate. Figures 8 and 9 show the sensitivity in bearing temperature results to
Considering oil splashing and other irregular flowing effect around the the applied density ratio.
bearing, it is difficult to model the flow conditions.
It has been shown that, increasing the density ratio, the thermal load
Although, it is possible to outline a trend for the thermal influence of of bearing elements is reduced. This effect can be understood,
lubricant flow rate and determine a given oil volume inside the considering that a higher oil-fill rate can increase the coolant effect.
transmission system, when oil flow rate and its coolant effect reach a It can be also observed that for the case of low-speed operation
critical value. In Figure 7, it can be seen than the same trend and (Figure 8), the calculated outer ring temperature produced extremely
similar results are found for 8000 rpm as were found for 4000 rpm high results when a low-density ratio was applied (<0.15).
(Figure 6). Temperature differences are significantly high between
outer and inner ring for the higher rotational speed.

Temperature results from thermal model at different flow Temperature results from thermal model at differnt density
rates (4000 rpm, density ratio: 0.2) ratios (4000 rpm, flow rate: 4 l/min)

150
150
Temperature [°C]
Temperature [°C]

130 130

110 110

90 90

70 70
0 2 4 6 8 10 12 14 0.1 0.2 0.3 0.4 0.5 0.6
Flow rate [l/min] Density ratio [-]
Inner ring mass ('N3') Outer ring mass ('N8') Inner ring mass ('N3') Outer ring mass ('N8')
Balls ('N5') Oil outlet
Balls ('N5') Oil outlet

Figure 8. Investigation of density ratio (oil-air fraction)’s effect regarding


Figure 6. Investigation of lubricant flow rate’s effect regarding temperature temperature condition.
condition.
Equations regarding heat transfer usually contain empirical relations
The higher inner to outer race temperature variation can be explained
with specific limitation on their application, such that at the operation
by the different heat exchange conditions, as the rotating inner ring
point with low speed and low density ratio, the calculation of the Re-
has a more intensive interaction with lubricant then the outer race.
number is likely to be invalid.

Page 7 of 11

AVL List GmbH


Temperature results from thermal model at density ratios
(8000 rpm, flow rate: 4 l/min)

150
Temperature [°C]

130

110

90

70
Figure 10. Investigated bearing in case of oil-bath lubrication.
0.1 0.2 0.3 0.4 0.5 0.6
Density ratio [-] Calculations were performed only for the current bearing, but these
Inner ring mass ('N3') Outer ring mass ('N8') could be extended to a transmission system. For the thermal model,
Balls ('N5') Oil outlet all input parameters have been set up as in the earlier investigations,
but the density ratio (oil-air fraction) was estimated with the
Figure 9. Investigation of density ratio (oil-air fraction)’s effect regarding following formula:
temperature condition.
𝐴𝑖𝑚
𝑋≔ (31)
Bearing speed, density ratio and oil feed rate are not independent (𝑑𝑜2 − 𝑑𝑖2 )𝜋/4
properties, so it is difficult to determine the effect of these
parameters in isolation. To this end, CFD can offer a useful support Where 𝑑𝑜 and 𝑑𝑖 are the guide diameters of outer and inner races
regarding these critical values. Using a 3D SPH lubrication simulation respectively, and 𝐴𝑖𝑚 represents the considered immersion-area, (as
it is possible to evaluate oil flow rate and the density ratio for a given shown in Figure 10).
oil volume in a transmission system. Therefore, these properties can
be implemented in the thermal calculation to find an optimal Implementing the calculated density ratios depending on immersion
lubrication condition and oil volume. This methodology can be depth of oil, drag and churning losses and steady state bearing
presented through the investigated ball bearing, in case of oil-bath temperature values can be determined with different flow rates, as
lubrication. As can be seen on Figure 10, if the corresponding bearing shown in Figure 11.
is submerged in oil with a given oil-level, depending on the oil volume
in the transmission.

Then, a relative immerge rate can be defined as:

𝐼𝑚𝑚𝑒𝑟𝑠𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 = ℎ𝑖𝑚 /𝐷 ∙ 100 [%] (30)

Where ℎ𝑖𝑚 is the oil immersion level, and D the bearing diameter,
(as shown in Figure 10 Assuming different immersion levels for
different oil volumes, drag and churning losses can be calculated
quiet accurately using an in-house program, based on calculation
methods from [15, 16].

Figure 11. Bearing outer ring temperature at different oil immersion rates
and its churning and drag losses (shaft speed: 8000 rpm).

It can be seen that, in case of an oil-bath lubrication, increasing


lubricant presence, the bearing produced higher drag and churning
losses, but the thermal load of bearing does not show remarkable
decrease after a certain point.

Changing oil flow rate could be another option to reduce bearing


temperature and the effect can be seen also on Figure 11. Using these
relations and diagrams, it could be possible to find an appropriate
Page 8 of 11

AVL List GmbH


lubrication condition keeping a balance between thermal load of the Summary and Conclusions
bearing and power losses.
A thermal network model has been successfully created for a high-
Evaluation of drag and churning losses could be achieved also for speed transmission bearing, and reasonable temperature results
other components in a transmission system, making it possible to could be achieved when compared with the steady state conditions
analyze the system effects of the higher oil volume. However, as measured at the testbed. On the one hand, this model contains many
mentioned before, in a full transmission system more complicated analytical and empirical formulas, giving a simplified approach to
situations can be experienced, where the oil-feed, cavity fill of estimating the temperature conditions of the bearings. On the other
bearing, oil volume and operating conditions are changing and not hand, the developed methodology provides a practical tool to predict
independent. Figure 12. presents a recommended methodology to bearing thermal load to support the development process. Thermal
determine an optimal lubrication condition. First, thermal conditions finite element methods and CFD could prove more accurate and
should be evaluated at different oil volumes, taking into account the sophisticated techniques for this purpose, but the time for model set
uncertainties of model and SPH CFD simulations also need to be up and the computational demand are much higher, especially in a
performed to determine oil feed data. complex flow-thermal coupled simulation.

The results of the applied thermal network model have shown the
thermal effects of lubrication conditions, such as oil flow rate or oil-
air fraction. It has been concluded that, despite the simplified
analytical approaches, the thermal model is able to predict
temperature values that can be calibrated and compared with
measurement results.

Applying flow results from lubrication simulation and implementing


power loss values, it’s possible to determine optimal lubrication
parameters, such as oil volume, which is of value in determining the
minimum level of sufficient lubrication.

Nomenclature

𝑎 Small axis of Hertzian contact [m]

Figure 12. Scheme of suggested methodology for the determination of an


𝐴𝑖𝑚 Oil immersion area [m2]
optimal lubricant volume. 𝑏 Large axis of Hertzian contact [m]

These parameters can be implemented in the bearing thermal model 𝑐 Specific heat [J/kg/K]
with other input data regarding bearing and housing geometry, and 𝐶𝐷 Drag coefficient [-]
other operating condition. Then, according to the bearing material
properties, a critical temperature can be determined. Finally, and 𝐶0 Rated static load [N]
considering temperature limit of bearing elements and power losses
𝑑𝑖 Inner guide diameter of bearing [m]
in transmission system and also based on engineering judgement, it’s
possible to estimate an optimal oil volume. 𝑑𝑜 Outer guide diameter of bearing [m]
𝑑𝑒𝑥𝑡 External diameter [m]
𝑑𝑖𝑛𝑡 Internal diameter [m]
𝑑𝑚 Bearing pitch diameter [m]
𝐷 Bearing diameter [m]
𝐷𝑏𝑎𝑙𝑙 Diameter of rolling element (ball) [m]
𝐷𝑥 Characteristic length [m]
𝑓1 Dimensionless factor
𝐹𝑎 Axial load [N]
𝐹𝑟 Radial load [N]
𝐹𝑑𝑟𝑎𝑔 Drag force [N]
Figure 13. Suggested workflow for the evaluation of bearing thermal 𝑔 Gravity factor [m/s2]
conditions
𝐺𝑟 Grashof number [-]
Page 9 of 11

AVL List GmbH


ℎ𝑖𝑚 Oil immersion level [m] 𝑉𝜇𝑒𝑞 Relative
Equivalent
speed
dynamic
of contact
viscosity
surfaces
[Pas][m/s]
𝐾 Thermal conductivity [W/K/m] ̇𝜇𝑜𝑖𝑙
𝑉𝑓𝑙𝑜𝑤 Lubricant
Dynamicflow
viscosity
rate of 3/s][Pas]
[moil
𝐿 Length of bearing [m] 𝑋𝜇𝑎𝑖𝑟 Oil-air
Dynamicfraction
viscosity
[-] of air [Pas]
𝐿𝑥 Characteristic length [m] 𝛼𝑣 Thermal
Kinematic
diffusivity 2/s]
viscosity[m[m 2/s]

𝑚 Mass [kg] 𝛽𝜌𝑒𝑓𝑓 Thermal


Effectiveexpansion
density [kg/m 3]
coefficient [1/K]
𝑚̇ Average mass flow [kg/s] 𝜇𝜌
𝑒𝑞𝑜𝑖𝑙 Equivalent
Density ofdynamic 3]
oil [kg/mviscosity [Pas]
𝑀𝑓𝑟𝑖𝑐 Torque of friction loss [Nm] 𝜇𝜌
𝑜𝑖𝑙
𝑎𝑖𝑟 Dynamic
Density viscosity
of air [kg/m 3] [Pas]
of oil
𝑛𝑏𝑎𝑙𝑙 Number of rolling elements (balls) 𝜇𝜔
𝑎𝑖𝑟
𝑐 Dynamic
Angular viscosity
velocity of
of the
air [Pas]
bearing cage [1/rad]
𝑛𝑟𝑜𝑡 Shaft speed [1/s] 𝑣 Kinematic viscosity [m2/s]
𝑁𝑢 Nusselt number [-] 𝜌𝑒𝑓𝑓 Effective density [kg/m3]
𝑃0 Equivalent static load [N] 𝜌𝑜𝑖𝑙 Density of oil [kg/m3]
𝑃1 Dimensionless factor 𝜌𝑎𝑖𝑟 Density of air [kg/m3]
𝑃𝑑𝑟𝑎𝑔 Power from drag losses [W] 𝜔𝑐 Angular velocity of the bearing cage [1/rad]
𝑃𝑟 Prandtl number [-]

𝑄̇ Heat flow [W]


𝑄̇𝑓𝑟𝑖𝑐 Heat power from friction [W]
𝑅 Heat resistance [K/W]
𝑅𝑒 Reynolds number [-]
𝑅𝑐𝑜𝑛 Heat resistance of constriction [K/W]
𝑅𝑐𝑜𝑛𝑑 Conductive heat resistance [K/W]
𝑅𝑐𝑜𝑛𝑣 Convective heat resistance [K/W]
𝑠 Contact surface [m2]
𝑠𝑐 Section area of the cage [m2]
𝑡 Time [s]
𝑡𝑐 Cage thickness [m]
𝑇 Temperature [°C]
𝑇𝑜𝑖𝑙,𝑖𝑛 Oil inlet temperature into the bearing [°C]
𝑇𝑜𝑖𝑙,𝑜𝑢𝑡 Oil outlet temperature from the bearing [°C]
𝑇𝑐𝑎𝑙𝑐 Temperature value from thermal model [°C]
𝑇𝑡𝑒𝑠𝑡 Temperature value from measurement [°C]
𝑇𝑤 Wall temperature [°C]
𝑇∞ Ambient temperature [°C]
𝑢̅𝑓𝑙𝑜𝑤 Average lubricant flow speed [m/s]
𝑈 Characteristic speed [m/s]
𝑉 Relative speed of contact surfaces [m/s]
̇
𝑉𝑓𝑙𝑜𝑤 Lubricant flow rate [m3/s]
𝑋 Oil-air fraction [-]
𝛼 Thermal diffusivity [m2/s]
𝛽 Thermal expansion coefficient [1/K]

Page 10 of 11

AVL List GmbH


References 8. Dexing, Z., Weifang, C., “Thermal performances on angular
contact ball bearing of high-speed spindle considering
structural constraints under oil-air lubrication,” Tribology
1. Wang, Y., Cao, J., Tong, Q., An, G., “Study on the Thermal
International 109, 2017. Doi: 10.1016/j.triboint.2017.01.035
Performance and Temperature Distribution of Ball Bearings in
9. Gupta, P. K. “Advanced dynamics of rolling elements,” 1984,
the Traction Motor of a High-Speed EMU,” Appl. Sci. 2020, 10,
p.295 (Springer-Verlag, New York). doi: 10.1007/978-1-4612-
4373. Doi: 10.3390/app10124373
5276-4
2. Tauviqirrahman, M., “CFD Analysis of Journal Bearing with a
10. Isbin, H. S. Moy, J. E., and Da Cruze, A. J. R. “Tow-phase stream-
Heterogeneous Rough/Smooth Surface,” Lubricants 2021, 9, 88.
water critical flow,” A.I.Ch.E., 1957, 3, (3), 361–365.
Doi:10.3390/lubricants9090088
11. J.P. Holman, Heat Transfer, 7th ed., McGraw-Hill, New York,
3. Morris, S., Morris N. and Leighton, M., “A Time Efficient
1989.
Thermal and Hydrodynamic Model for Multi Disc Wet
12. E.C. Cobb, O.A. Saunders, “Heat transfer from a rotating disk,”
Clutches”, SAE Technical Papers, 2022. doi:10.4271/2022-01-
Proc. R. Soc. 236 (1956) 343–351
0647
13. J.P. Holman, Heat Transfer, 10th ed., McGraw-Hill, New York,
4. Belmiloud, D., Lachi, M., Pron, H., Bolaers, F., “Thermo-
2010. ISBN 978–0–07–352936–3
dynamical modelisation of the degradation of a ball bearing in
14. Muzychka, Y., Yovanovitch, M., “Thermal resistance models for
variables use conditions,” Mechanics & Industry 21, 608 (2020)
non circular moving heat sources on a half space,” Transaction
doi:10.1051/meca/2020083
of the ASME, Journal of Heat Transfer 3,624–632 (2001) doi:
5. Pouly, F., Changenet, C., Ville, F., Damlens, B., “Investigations
10.1115/1.1370516
on the power losses and thermal behaviour of rolling element
15. SKF Group “Rolling Bearings Catalogue” available at;
bearings,” Proc. of the Inst. Of Mech. Eng., Part J: J. of
https://www.skf.com/binaries/pub12/Images/0901d19680280
Engineering Tribology, 224(9): 925-933, 2009. doi:
9de-Rolling-bearings---17000_1-EN_tcm_12-121486.pdf,
10.1177/1350650117750806
accessed 10/04/2022
6. Zheng, D., Weifang, C., Miaomiao, L., “An optimized thermal
16. Harris, T. and Kotzalas, M. “Essential Concepts of Bearing
network model to estimate thermal performances on a pair of
Technology Rolling Bearing Analysis”, Taylor Francis, 2006,
angular contact ball bearings under oil-air lubrication,” Applied
ISBN: 9780849371837
Thermal Engineering 131 (2018) 328–339 doi:
10.1016/j.applthermaleng.2017.12.019
7. Liu, Y., “Improved thermal resistance network model of Contact Information
motorized spindle system considering temperature variation of
cooling system,” Advances in Manufacturing 6:384–400, 2018. Michael Leighton; michael.leighton@avl
Doi: 10.1007/s40436-018-0239-4

Page 11 of 11

AVL List GmbH

View publication stats

You might also like