Documente Academic
Documente Profesional
Documente Cultură
=
= . If rigid ball is used, which means K2 , so: 0
1
2
K
,
then
total
K
1
( +
1
1
K
3
1
K
) which shows that in the whole stiffness, there is no contribution
from the stiffness of the ball. Thats not physical.
(a) (b)
Fig.1-7(b)_1 simplification of model
So the flexible rings and flexible ball have to be used for the simplification.
* Simplification step 2: part model with ball to part model with other element
In order to investigate the possibilities of further simplification of ball, since flexible rings
and flexible components have to be used, there are two possible solutions:
The first idea: use a spring between the inner and outer ring which is shown in
Fig.1-7(b)_2.
Fig.1-7(b)_2 possible simplification
- 6 -
The disadvantage of this model is obvious:
- No large deformation
- No axial force transmission
- No nonlinearity of the reaction force
The second idea: use beam elements between the inner and outer ring which is shown in
Fig.1-7(b)_3.
Fig.1-7(b)_3 possible simplification
The advantages of this model are:
+ Axial and radial force transmission
+ Nonlinear reaction force and displacement
+ Large deformation
Fig.1-7_(c) simplification of the part model
The static structure analysis of the model is the interest.
As shown in Fig.1-8, it is the free body diagram for all the components of the bearing. The
boundary condition is shown in Fig.1-7(b)_2.
- 7 -
The force R_y will be the reaction force which one wants to investigate and find some
reasonable value for it because it can provide a reference force to the design of shaft or
housing. Thats why it is chosen to evaluate all the numerical results.
R_y2
Fig. 1-8 free body diagram
Conclusion:
To do the simulation with the part model of bearing, flexible-flexible model has to be used
for the rings and ball. The model with beams between the inner ring and outer ring will be
used to simplify the angular contact ball bearing. The model is shown in Fig.1-7(c). Finally
all the results will be evaluated by comparing the reaction force from the housing to the
outer ring.
- 8 -
1.4 Overview of Each Chapter
In chapter 2, simulation with exact geometry of ball bearing through basic ANSYS work
flow is performed. Firstly the procedures to create the model are described. Secondly some
basics settings for contact are given. Thirdly the influencing factors which affect the
reaction force are investigated and finally the reasonable reference reaction force for this
angular contact ball bearing is attained.
In chapter 3, the contact analysis is discussed. The meaning of some important contact
items, the behavior of contact and the principle of different contact algorithms are explained.
Some simple but important examples are also given in order to get better understanding
about contact problem.
In chapter 4, the simulation of simplified model is carried out. As shown in Fig.1-7(c), the
ball is simplified with beams between the inner ring and outer ring. Firstly possible or
reasonable solutions to simplify the model are discussed (Even if beam elements are used
between the inner ring and outer ring, there are still several possible solutions). Secondly
solving the simplified model under certain boundary condition is performed. Finally results
of model with beam and with ball are compared. The possible number of beams and the
radius of each beam are figured out as well.
- 9 -
2 Static Analysis of Angular Contact Ball Bearing
2.1 Introductions
In this chapter, the procedures to create the model* will be introduced and some other
basics about angular contact ball bearing will also be discussed. The CAD model of the
bearing, generation of the mesh and solution of the model are done by ANSYS Workbench.
2
2.2 Construction of the FE Model
In order to determine the geometry of this angular contact ball bearing, the following
parameters firstly are necessary to be figured out.
D. diameter ball
r radius groove outer
r radius groove inner
d diameter raceway outer the
d diameter raceway inner the
o
i
o
i
deep groove ball bearing angular contact ball bearing
Fig.2- 1 geometry of rolling bearing
] 1 [
* Because of lacking of the geometry data from the manufacturer, the geometry data of the model here used is from the
reference book [1].
2
- 10 -
According to the geometry, the following equations hold:
The pitch diameter =1/2*( + ) *
m
d
i
d
o
d
3
The distance between the center of inner groove ( ) and outer groove ( ) A= + -D
i
r
o
r
i
r
o
r
Diametrical clearance = - -2*D
d
P
o
d
i
d
4
* angle contact free the o = ) 2 / 1 ( cos
1
A P
d
<
>
sticking Fn, Ft
sliding Fn, Ft
3.5 Contact Algorithm
There are several contact algorithms for contact problem:
Pure Lagrange multiplier (on contact normal and tangential direction)
Pure Penalty method
Augmented Lagrange
Multipoint constraint (MPC)
In order to understand how these algorithms work, the following points are necessary.
The weak form for all the FEM formulation is reading as below:
0
T
= = H
} } }
udA t udV f dV
A V V
o o oc o o (3-1)
Principally the internal energy minus external energy should be zero then the solution can
find the equilibrium point and converges. For the contact problem within the contact area,
just an additional term is added at the end of the equation, which reads as follows:
0 ' '
T
= + = H
} } }
m contactter udA t udV f dV
A V V
o o oc o o (3-2)
Different contact terms in equation (3-2) result in different contact algorithms.
3.5.1 Pure Lagrange Multiplier
For the Pure Lagrange method, the contact term in equ.(3-2) will be:
Contact term (3-3) dA g g
T T
A
N N
) ( o o + =
}
Where,
N
,
T
are the Lagrange multiplier in the normal and tangential direction
respectively and they are denoted as contact force in the contact analysis.
- 36 -
Inserting equation. (3-3) into (3-2), then the system of equation is got as follows:
.
0(
(
(
T
g
g K
=
(
u
(
0 g
f
(3-4)
Where:
From equation (3-4) the contact condition can be determined as follows:
g=
T
g . u +g0 (3-5)
=
>
<
happens contact 0, g
contact no 0, g
happens n penetratio 0, g
In order to have a better understanding about contact, as shown in Fig.3-5, there is an
example and see what contact actually means. These two springs represents two isotropic
material bodies.
f1
k1 u1 u2 k2
go=1
Fig.3-5 contact example
The conditions given are: the stiffness of both springs is k1=k2=2,
The force acting on spring 1 is f1=5, initial gab between two springs is g0=1.
Question: what will be the contact status for two springs?
Solution:
From Hooks Law: f=k*x
yields: u1=f1/k1=2.5
since the initial distance between the two springs are 1, so penetration will happen. The
amount of it is: pene=u1-go=2.5-1=1.5
- 37 -
in the matrix notation, reads :
Stiffness matrix: K = = , force vector = =
(
2 0
0 1
k
k
(
2 0
0 2
f
(
0
1 f
(
0
5
Since , so
=
= u K f
= f K u
1
= , which means u1=2.5, u2=0.
(
0
5 . 2
So the penetration is : pene=2.5-1=1.5
Using Pure Lagrange method and determine under what conditions the contact will
happen?
Insertion K and g =| and | 1 , 1 f into equation (5-4) and let
g=
T
g . u +g0 =0
yields the following results :
u = , =3.5,
(
75 . 0
75 . 1
int
F =
(
5 . 1
5 . 3
Which means that k1 is tension and k2 is compression at amount of u1=1.75 and
u2=0.71then the exact contact between two bodies will happen.
From above, some basic idea about how the Lagrange algorithm works is gained. Now the
application characteristics of normal Lagrange multiplier method are given as follows.
Advantages:
Good accuracy, almost no penetration generated
. No problems with ill conditioned matrices
. No contact stiffness needed, no trial and error
Disadvantage
. Extra degree of freedom for each contact element added
. System of equations blows up when the model has too many degrees of freedom
. The fast iterative PCG solver not available because the diagonal elements are zero
. Over constraint may happen
3.5.2Pure Penalty Method
For the penalty method, Lagrange multiplier is not needed but the contact stiffness in
normal direction
N
c and tangential direction
T
c will be used. The contact term in equation
(3-2) followed as following term:
- 38 -
Contact term (3-6) dA g g g g
T T T
A
N N N
) ( o c o c + =
}
Insertion (3-6) into (3-2) yields the system of equation in the following matrix form:
[ K +c g
T
g ] u = f (3-7)
Solve the equation (3-7) in the normal direction yields:
u1=
N
N
c
c
4 4
5 10
+
+
, u2=
N
N
c
c
4 4
5
+
, g=
N
c 4 4
5
+
+g0 (3-8)
From equation (3-8), it can be seen that the final results depends on the parameter (contact
stiffness)
N
c , which is a kind of ill-conditioned solution.
iteration n iteration n+1
Fig.3-6 numerical iteration
Fig.3-6 shows how pure penalty method works. The normal contact stiffness factor FKN
plays an important role in the solution. The spring represents FKN. When the stiffness of
spring is two large, after the iteration n, the upper part moves upwards because of the large
reaction force between these two bodies. Then both parts can not touch each other and the
program can not find the equilibrium position and contact points. So finally it doesnt
converge. On the other hand, when the contact stiffness is too small, the upper body can
move downwards easily and even penetrate into the target body. Then too much penetration
will be generated.
Application characteristics for pure penalty method:
+ System of equation doesnt blow up
+ Fastest method in ANSYS
+ Large model (>300.000 Dofs) can be handled well
+ All solvers are available
- Convergence behavior and accuracy depend on the contact stiffness
- Large contact stiffness may result in ill conditioned matrices
- 39 -
Conclusion:
The contact stiffness plays an important role in the penalty algorithm. It is necessary to find
an appropriate value of contact stiffness with which the solution not only converges well
but also the penetration will be within the criterion user defined.
3.5.3 Augmented Lagrange Method
In the last two sections, basic principles for Normal Lagrange multiplier method and
penalty method are discussed. Both algorithms have their own advantages and
disadvantages. Now a new algorithm which combines both namely Augmented Lagrange
will be presented.
The contact term in equation (3-2) reads as following:
Contact term (3-9) dA g g g g
T T T T
A
N N N N
) ) ( ) (( o c o c + + + =
}
Where:
N
,
T
are Lagrange multiplier in normal and tangential direction
N
c ,
T
c are the contact stiffness in normal and tangential direction
Augmented Lagrange augments the Lagrange part by penalty terms. When the program is
processed with Augmented Lagrange, the program first uses penalty method to find the
equilibrium point and then use Lagrange part to minimize the penetration. The user can also
specify the penetration tolerance for the final solution. ( One thing to mention is that after
each equilibrium point the solver eliminates penetration 3 times but no matter how much it
will be at the end).Also the system doesnt blow up and all the solvers can be used.
However, it needs more calculation time and ill-conditioned matrices may result in large
contact stiffness.
3.5.4 Multipoint Constraints(MPC)
The internal multipoint constraint (MPC) approach is combined with certain bonded and no
separation contact definitions to define various contact assemblies and kinematics
constraints.
The internal MPC approach can overcome the drawbacks of the traditional contact
algorithms and other multipoint constraint tools available in ANSYS. For example:
- Degrees of freedom of the contact surface nodes are eliminated
- No contact stiffness is required as input.
- Both translational and rotational degrees of freedom can be constrained.
- Generation of internal MPC is simple because it uses contact pair definitions.
- 40 -
Application:
* Connection of dissimilar meshes, like solid-shell (shown in Fig.3-7), or shell-shell, or
solid-solid.
* apply remote displacement, shown in Fig.3-8, load can be applied on pilot node (one
single node), the rotation can be applied as well.
Fig.3-7 connection of shell-solid element
Fig. 3-8 load applied on pilot node
3.6 Conclusions
In this chapter, the basics about contact problem are given: contact algorithm, contact
behavior, contact types and so on. Of course there are some other quantities like contact
status, contact pressure to evaluate the result of contact problem. Relevant information can
be found from some reference books and here are not discussed.
- 41 -
4 Simulation of Angular Contact Ball Bearing with Simplified Model Using
Beam Element
4.1 Introduction
In chapter 2 simulation using balls between inner and outer ring has be analyzed and the
reaction force can be got from the result, which is our interest because it gives a reference
for the design of housing. Now in this chapter the simulation using simplified modelwith
beam between the inner ring and outer ring will be performed. Finally the comparison
between the simulation results with ball and beam will be also discussed.
4.2 The Whole Model of Angular Contact Ball Bearing
Fig. 4-1 shows the model of inner ring and outer ring which can be done in ANSYS
Workbench simply suppressing all the balls. The simplified model will also use this
geometry to do the calculation, so the influence from the geometry of inner ring and outer
ring is the same as that to the exact model. All the effort will be put on the influences of
beams to the structure in order to finally make the difference between the solution with ball
and beam as small as possible.
Fig.4-1 inner and outer ring for the simplified model
It has been explained in chapter 1 that a part of the whole model can be used for the
calculation, which is shown in Fig. 4-2.
To make the model in Fig.4-2 converge is not so easy because of the initial geometrical gab
between balls and inner ring and outer ring and the rotation of the two balls, which makes
the program not so easy to find the contact points and also the rotation of the two balls. The
trick to these is to apply some constraints on both balls to prevent the translation in Z
direction shown in Fig. 4-2. In order to solve the initial gab, at first step, it is possible to
solve only the contact gab and then from second step on load will be applied. This is a very
convenience way to solve convergence problem because of initial gab or penetration.
- 42 -
Fig. 4-2 part model for the calculation and mesh
As the solid element does not affect the reaction force too much, one can see that in Fig.4-2,
the element for the inner and outer ring is relatively large compared to the elements in
contact region. The material model and parameter is the same to that in chapter 2.3.1
4.3 Result of Target Model
Solving the model shown in Fig.4-2, the solution can be attained.
r_y(ball)
0
20000
40000
60000
80000
100000
120000
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
r
_
y
[
N
]
r_y(ball)
Fig.4-3 reaction force of the model with 2 balls
Fig.4-4 contact status
- 43 -
The force reaction is plotted in Fig.4-3 and contact status and penetration are shown in
Fig.4-4 and 4-5 respectively.
Fig. 4-5 contact penetration
This penetration is small enough and this reaction force plot will be used as the target plot
and finally the solution of simplified model with beam between the inner and outer ring will
be compared with this plot.
4.4 Possible Ways to Simplify Angular Contact Ball Bearing
a 5 *
Now the simplified model with beam is going to be created.
Procedures to generate the simplified model:
Firstly, the ball in the model shown in Fig.4-2 will be suppressed. The mesh of the rings
will not be changed at all.
Secondly, the macro file which will generate the beam element between the inner and outer
ring will be inserted into the Workbench operating tree and ANSYS Workbench will read it
automatically.
Finally all the boundary conditions for the model with ball are exactly the same to the
model with beam elements. The analysis settings are also the same. The outline comparison
between these two is shown in Fig.4-5-1.
a 5 *
Here beam element will be used between the inner and outer ring to simplify this angular contact ball bearing.
- 44 -
Fig.4-5-1 outline comparison
There may be several possible ways to generate beam between inner and outer ring. First
lets discuss the possible solutions using beam elements and compare which way is
reasonable to simplify the bearing.
4.4.1 Possible Solution (1)
As shown in Fig.4-6, this is one possible simplified model to simulate the ball bearing.
Two parts of beams are used to replace one ball. Between these two parts, point-point
contact will be used. This model is already applied by the customer.
Advantages and Disadvantages:
+ Easy to use
+ Short calculation time
- No large deformation (no rotation available)
- Limited meshing (fit for only regular geometry )
Beam elements Outer ring
inner ring point- point contact
Fig:4-6 possible solution_1
- 45 -
Conclusion:
This solution is not a reasonable one because when large deformation is applied this model
will crack. The main reason is the point to point contact between two beams. It is very
difficult to keep two points in contact when large deformation is applied and also it has high
requirement for the geometry of inner and outer ring.
4.4.2 Possible Solution (2)
As shown in Fig.4-7, it is another possible solution to simplify this model: use one beam
between the inner ring and outer ring. The process to generate beam elements is: first the
nodes on the surfaces of outer ring and inner ring are generated. Then two nodes from the
inner and outer ring will be connected by beams of beam elements. Not any contact is
defined in this model.
Pros and cons :
+ beam element, easy to use
+ short calculation time
- no nonlinearity
- limited meshing
Beam element
Fig. 4-7 possible solution_2
Conclusions
This model is not a reasonable one because:
At one hand it can not simulate the nonlinearity of the angular contact ball bearing, which is
fatal. At the other hand, it does not have the contact between the beam and the raceway,
which makes it difficult to investigate the contact pressure and the penetration between the
beam and the raceways. So this model is also useless.
- 46 -
4.4.3 Possible Solution (3)
Combing the possible solution (1) and (2), another possible simplified model can be gained.
As shown in Fig. 4-8, node-surface contact is defined. At one side of the beam, bonded
contact is used for the node and outer ring surface and at the other side frictional contact is
applied. The generation of the beam element is similar to that in possible solution (2).
Pros and Cons
+ Easy to use
+ Short calculation time
+ Nonlinearity of the force and displacement curve
+ large deformation
+ no meshing limitation
frictional contact
Bonded
Contact Fig. 4-8 possible solution_3
Conclusion
This model is a possible model to simplify the ball bearing. In the following part, the
simulation using this model will be carried out.
- 47 -
4.5 Test of Final Chosen Way
In section 4.4, all the possible solutions to simplify the model of the ball bearing have been
discussed. Now the verification of this idea will be investigated using roller bearing and see
whether this model works or not.
Fig.4-9 original model of roller bearing
The model of roller bearing is shown in Fig.4-9. What will be done is to use only the
geometry of the inner ring and outer ring and beam element in between to test this model.
As shown in Fig.4-10, 9 paralleled beams are used to simplify one roller and the number of
beams is a parameter and can be changed as well.
Fig.4-10 simplified model of roller bearing
Generation of the beam is done by APDL, which is a short term of ANSYS Parametric
Design Language and actually it is macro file. This macro file will be added in the
accessory part of the thesis.
- 48 -
Relevant Parameters:
* Node-surface contact CONT175 and target 170 are used for the contact pair
* MPC bonded contact between the beam and surface of outer ring
* Frictional contact for beam and surface of inner ring
* Beam element 188, circular solid
The contact photos are shown in Fig. 4-12.
4.5.1 Applied Displacement to the Model
In order to test whether this model can support large deformation and rotation, in this
section the displacement with different direction and magnitude will be applied. In section
4.5.2, both rotation and displacement will be applied so that the further test can be
investigated.
One thing to mention is that here only the simulation of the model with beam is done and
not any comparison between the real roller bearing and FEM results with beam. The
purpose of this section is just to see whether the idea explained in section 6.5 works or not.
The displacement will be applied on the inner ring. From 0 to 90, totally 5 steps will be
applied on the structure and the angle between each step load vector is 22.5. The diagram
of load vector is shown in Fig.4-11.
0
0,05
0,1
0,15
0,2
0,25
0 1 2 3 4 5 6
step
m
a
g
n
i
t
u
d
e
o
f
d
i
s
p
l
a
c
e
m
e
n
t
[
m
m
]
u_x
u_y
Fig.4-11 load diagram
- 49 -
Fig.4-12 contact
- 50 -
Fig. 4-14 minimum principle stress of the outer surface of inner ring
The minimum principle stress which means the least positive principle stress can be
interpreted here as the contact pressure between the beam and the surface of the inner ring
and it is plotted in the picture 4-14. From the animation of the min. principle stress (shown
in presentation) it can be seen that with the direction change of the load vector, the
maximum pressure also moves. It means when the load is applied, the force can be
transmitted through the beam to the outer ring and the model does not crack.
] 3 [
4.5.2 Apply Combined Load to the Model
Now the combined loadsdisplacement plus rotation shown in chart 4-1 will be applied
onto the inner ring and see whether this model still works.
step u_y Rotation z
1 0.2 0
2 0 5
3 0 10
4 0 15
5 0 20
Chart 4-1 mixed load magnitude
As shown in Fig. 4-15, the minimum principle stress on the surface of the inner ring can
also rotate as the rotation of the load, which means that the model can simplify the roller
bearing and can support large deformation and rotation. Further job using this idea will be
done to simplify the angular contact ball bearing and make further comparison between the
simplified model with ball and the model with beam between.
- 51 -
Fig.4-15 minimum principle stress of outer surface of inner ring under combined load
- 52 -
4.6 Simulation of Angular Contact Ball Baring with Beam
4.6.1 Introductions
Now the way mentioned in section 4.5 will be used to simplify the ball bearing. A little bit
changes will be made in the macro file in order to generate different styles of the beam.
4.6.2 Original Simplified Model with Beam
For the roller bearing, several parallel beams shown in Fig.4-10 can be used to replace one
roller, which is possible and reasonable because the roller bearing can only support the load
in the normal direction and no load in the axial direction can be applied. The parallel beams
can also support normal direction load. So the simplified model works well.
However, for the angular contact ball bearing, it has a different geometry and should be able
to support the load in both axial and radial directions. If paralleled beams are generated to
replace ball, these beams have no stiffness in the axial direction. That is the problem. What
can be done is to use a beam star to replace one ball. The model is shown in Fig.4-16,
where all the contact nodes are connected to the centre node. This model can support axial
and radial load.
The parameters for the simplified model:
- Beam element188, circular solid, Timoshenko beam
- Node-surface contact , frictional and bonded contact
- Initially number of beam is 9 and radius of beam is 10 mm for each beam
Fig.4-16 beam star
- 53 -
The simplified model is shown in Fig.4-17. One beam star represents one ball. For one
beam star, the number of beams is a parameter and can be changed by the input of the
customer. The radius of the beam is also changeable.
Fig.4-17 original simplified model
There is one problem with this model. As shown in Fig.4-18 when the displacement in the
x(axial) direction is applied, there are large deflections of these beams in the
y(circumferential) direction.
Large displacement of beam
Fig.4-18 deflection of the beam in the y direction
It can be seen that the model can not support any more load because the beams are flying
- 54 -
away from their initial positions. Fig.4-19 can be an indirect proof to the above conclusion.
It plots the reaction force v.s. displacement. The total displacement is 0.4mm, but after
0.1mm, the model has almost cracked and can not support any more axial load.
0
5000
10000
15000
20000
25000
30000
35000
40000
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4 0,45
u_x[mm]
r
_
x
[
N
]
r_x
Fig.4-19 reaction force
4.6.3 Improved Model
As described in section 4.6.2, the beam stars have too much deflection in the y direction,
which makes the model can not support too much load. The reason is that the structure does
not have enough stiffness in the circumferential direction. The measure that all the central
nodes of the beam star are connected by using beams can be taken, which can prevent too
much deflection in the circumferential direction. The improved model is shown in Fig.4-20.
Fig.4-20 improved model
After the change it works well. The displacement of the model in the direction of y
direction is shown in Fig.4-21, which is much less compared to the model shown in Fig.
- 55 -
4-18. This model may be reasonable to simplify the angular contact ball bearing.
Small displacement in y direction
Fig.4-21 displacement after updating
4.7 Comparison between the Simulation Results with Ball and Beam
6
*
For the first calculation, 9 beams are used and radius of every beam is 10mm. The reaction
force in y direction is plotted in Fig.4-22.
reaction force comparison
0
20000
40000
60000
80000
100000
120000
140000
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x[mm]
r
_
y
[
N
]
r_y(ball)
r_y 9beams
Fig.4-22 reaction force comparison
6
* The material model of every beam is the same. All the beam parameters here used are the same to that
used in roller bearing.
- 56 -
From 4-22 one can clearly see that the reaction force of simplified model with beam
element is similar to that using ball between the raceways. In order to compare the result in
quantity, the error is plotted in Fig.4-23. From Fig.4-24, it can be seen that at the beginning
for the displacement between [0 0.2], the difference between the model with ball and beam
is very larger. After this region, the difference decreases dramatically until finally the error
is approx.13%, which is a little bit large. The value of error in each step is shown in
chart4-2
error %
0
50
100
150
200
250
300
350
400
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
e
r
r
o
r
%
error %
error %
0
10
20
30
40
50
60
70
80
90
100
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
e
r
r
o
r
%
error %
Fig.4-23 error vs displacement
7
*
u_x [mm] Difference %
0
5.00E-02 352.031516
0.1 106.873368
0.15 58.7101677
0.2 40.9098718
0.25 30.7559374
0.3 24.6700333
0.35 20.4422123
0.4 17.7326835
0.45 15.6337555
0.5 13.0811126
Chart 4-2 simulation result difference between the ball and beam solution
When the displacement is in the region [0, 0.2], as is explained in chapter 2.4.2 that the
contact stiffness plays a dominant role in the whole stiffness of the structure. From chart 4-2,
one can also find that it is also in this region that the simplified model has big error which
maybe means the simplified model can not simulate the contact properly.
7
* The number of beams is 9 and the radius of one beam is 10mm
- 57 -
In Fig.4-24, the ellipse represents the contact region of angular contact ball bearing with
ball schematically. One point in this picture represents one beam contact point. For the
simplified model with beam, point-surface contact is used. The contact area for point is so
small that the contact stiffness is huge at the beginning which results in the higher reaction
force than the model with ball. That may be the reason for the big error at the beginning.
Fig.4-24 contact region
With more displacement applied, the contact stiffness does not play an important role in the
whole stiffness of the structure but the stiffness of the beam and the solid element of inner
and outer ring. As the beam has similar capability to the ball (when slice is done to the ball,
each slice can be thought of a beam), the error decreases and the finally error is 13%.
This error is large and it is not so satisfying. Further work should be done in order to
decrease it. What can be changed for the beam stars are:
- the radius of each beam
- the number of beams in one beam star
- the Youngs modulus for one beam
Because here all the materials of bearing, no matter the inner and outer ring and rolling
element, are structure steel and it is generally a constant which is 2.0* MPa. So here
the change of Youngs modulus will not be considered. The focus will be on the change of
radius and the number of beam.
5
10
It is obvious that the more beams in one beam star, the stiffer the structure will be. Also the
large radius for one beam, the stiffer the structure will be. In the following 2 sections, the se
two parameters will be changed and the comparison will be done between the results with
ball and with beam solution:
* The radius constant, the number will be the changed
* Number of beams constant, radius will be changed
- 58 -
4.7.1 Number of Beams and Reaction Force
In this section, the radius for all the beams is 10mm and the number of beams will be
changed. The reaction force plot is shown in Fig. 4-25.
reaction force comparison
0
20000
40000
60000
80000
100000
120000
140000
160000
180000
200000
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
r
_
y
[
N
]
r_y(ball)
r_y(9 beams)
r_y(11beams)
r_y(21beams)
Fig.4-25 reaction force with different beams
0
100
200
300
400
500
600
700
0 0.2 0.4 0.6
u_x [mm]
e
r
r
o
r
%
diff. 9beams
diff. 11beams
diff. 21beams
0
20
40
60
80
100
120
140
160
180
200
0 0.2 0.4 0.6
u_x [mm]
e
r
r
o
r
%
diff. 9beams
diff. 11beams
diff. 21beams
Fig.4-26 number of beams and error
From Fig.4-25, it can be seen that the number of beams in one beam star played an
important role in the reaction force. When there are more beams in one beam star, the
structure will be stiffer and the reaction will be larger as well which is in accordance with
the engineers general knowledge. This phenomenon can also be seen from Fig.4-26, which
plots the difference between the ball solution and the beam solutions with different number
of beams. It is obvious that decreasing the number of beams can minimize the relative
difference between the ball solution and beam solution.
- 59 -
4.7.2 Cross Section and Reaction Force
It can be seen in Fig. 4-25 that all these models with the radius of each beam 10mm have
larger reaction force at every displacement which means that they are all stiffer than the
model with ball.
From Fig.4-24, one already finds that the node-surface contact can not simulate the
surface-surface contact properly, however it is true that more points(beams) will have better
results than less point to simulate the contact.
Considering it is not so easy for the beam element to converge with small cross section area,
what will done will be: a relatively large number of beam will be used and the radius of
each beam is not too small.
In this part the number of beams as constant (12 beams) and the radius of the beam will be
changed.
reaction force with different beam radius
0
20000
40000
60000
80000
100000
120000
140000
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
r
_
y
[
N
]
r_y(ball)
r_y(r_4)
r_y(r_4,5)
r_y(r_8)
Fig. 4-27 reaction force comparison with different beam radius
8
*
From Fig.4-27 it can be found that decreasing the radius of beam, i.e. decrease the area of
cross section of beam, can decrease the stiffness of the structure which complies with the
general knowledge.
It is obvious from Fig.4-27 also that with the radius of 4.5mm for each beam can generate a
better result than the simulation with some other radius.
8
* the notation`r_4`means the radius for each beam is 4mm.
- 60 -
It can be seen that the final error is approx.5%, which is small enough for the simulation. In
reality, maybe the accumulated error for the real bearing is approx.2%. For the bearing,
there exist imperfections in every procedure: manufacture deviation of each part of the
bearing, assembly difference from different staffs, manufacture deviation from other parts
which are assembled together with bearing to work as one mechanism and so on. So the
result is acceptable.
difference between beam and ball solution
0
50
100
150
200
250
300
350
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
d
i
f
f
e
r
e
n
c
e
%
difference
difference between beam and ball solution
0
5
10
15
20
25
30
35
40
45
50
0 0,1 0,2 0,3 0,4 0,5 0,6
u_x [mm]
d
i
f
f
e
r
e
n
c
e
%
difference
Fig.4-28 difference plot
u_x [mm] Difference(9_10)% Difference(12_4.5)%
0
5,00E-0 352.031516 325,183748 2
0,1 106.873368 90,3454613
0,15 58.7101677 50,5781445
0,2 40.9098718 33,0206508
0,25 30.7559374 23,2410013
0,3 24.6700333 17,4547611
0,35 20.4422123 13,0440191
0,4 17.7326835 10,7825587
0,45 15.6337555 8,5251597
0,5 13.0811126 5,0184789
Chart 4-3 data difference comparison
- 61 -
- 62 -
Chart 4-3 lists the differe umber and radii of beam
.8 Conclusion
he final conclusion for this chapter may be drawn: it is possible to use beam to simplify
owever, if one is interested in investigating the contact issue inside the bearing, for
nce of two solutions with different n
element in contrast to the solution with ball between the inner and outer ring. It can be seen
that at each displacement, the model with 12 beams and radius of each beam is 4.5 mm has
small error than the solution with 9 beams and radius of beam is 10mm.
4
T
the roller bearing. But for the angular contact ball bearing, the ball is possible to be replaced
by beam if and only if the interest is to compare or investigate the final reaction force. For
this bearing, 12 beams in one beam star and radius of each beam 4.5 mm can have a similar
stiffness of one ball.
H
example, contact penetration or contact pressure, then it is not a good idea to use beam to
represent the ball between the inner and outer ring. Perhaps it is very necessary to use the
ball itself to compare the simulation results with the experiment data.
5 Summary and Prospects
In this study the static analysis of rolling bearing especially one specific angular
contact ball bearing using FEM has been performed.
Important physical and numerical affecting factors of the mechanical system of
rolling bearing have been analyzed. Two modeling ways were constructed: full FE
model with ball and FE simplified model where balls are replaced by a kind of
specific beam structure beam star. The factors like curvature of ball and raceways,
the boundary condition, mesh density and penetration are investigated. The most
important parameter which affects the reaction force is the geometrical curvature of
the ball and the raceway of inner and outer ring. A very small change of the ball
diameter (increased by 1.3%) can lead to large variation of the final reaction force
(increased by 27%). One can find relevant investigation in Chapter 2. So if the FE
model of angular contact ball bearing is constructed, the radii of ball and raceways
should be as accurate as possible.
If the numerical aspects are considered, mesh density and normal contact stiffness
factor FKN (if Augmented Lagrange is used) are the two important parameters. The
contact stiffness and the reaction force are sensitive to the mesh density mainly in the
region where the structure behaves most nonlinearly. In other words, beyond that
region, the mesh density affects the reaction force only a little. For instance when the
number of nodes increases by approx. 6 times, the final reaction force increases only
13.8%. It is suggested that for the single element size it should be less than approx.2%
of ball diameter (here ball diameter is approx.22mm).
For the penetration further investigation shows that the radial reaction force is not so
sensitive to it. For example, when the penetration increases from 0 (Pure Lagrange
method) to 0.0056 mm (Pure Penalty method), the final reaction force decreases only
approx.1%, but in the most nonlinear behavior region, the reaction force decreases
maximally 8.5%. It is advised that under the conditions of good convergence behavior
and not so long computation time, the customer can choose the FKN as large as
possible (if Augmented Lagrange or Pure Penalty method is used). One can find this
information in Chapter 2.
With respect to the simplification of rolling bearing with beam elements, it is possible
to use beams between the inner and outer ring to simplify the roller bearing.
Through the investigation of the beam star, one finds that the stiffness of the structure
is sensitive to the number of beam and the radius of beam. For this specific angular
contact ball bearing, twelve beams in one beam star can have a similar stiffness to the
ball. However, for the contact part, the beam star can not simulate the real situation of
contact inside the angular contact ball at all. So for the angular contact ball bearing, it
- 63 -
is not an ideal plan to replace the ball by beam elements.
All the mentioned results are based on the specific angular contact ball bearing with a
ball diameter around 22mm and the inner and outer raceways diameter around 11mm.
For other geometrical sizes of this kind of angular contact ball bearing, the number of
beams or the radius of the beam in one beam star is still unknown and further
investigation must be carried out.
These two reasons lead to the final conclusion that for the angular contact ball bearing
the beam star is not a good simplification model.
For the static analysis this FE model of angular contact ball bearing is from the
literature and the static real bearing testing results about it is still missing. As soon as
one gets relevant experimental data, it is possible to verify these numerical results
further.
For the future study it is also very important to simulate the dynamic behavior of
rolling bearing using FEM. Generally many rolling bearings are used in high speed
machines. For the high speed machinery, the vibration and noise are two challenging
problems which are normally from the bearing. Hence in the future it is of great
importance to investigate the dynamic behavior of rolling bearing so that one can
predict the system vibration responses and minimize the noise level.
- 64 -
Bibliography
[1]Tedric A. Harris; Rlling Baring Aalysis, John Wiley& Sons, Inc.1967
[2] Tedric A. Harris; Rlling Baring Aalysis, John Wiley& Sons, Inc., fourth
edition,2001
[3]Vince Adams and Abraham Askenazi; Bilding Btter Poducts with Finite Element
Analysis; on world press,1998
[4]T.Stolarski Y.Nakasone. S.Yoshimoto; Engineering analysis with ANSYS software;
Elsevier Butterworth-Heinemann, Oxford, 2006
[5]http://www.kmlbearing.cn/common/catalog_data/01/0101/010101/01010101/image
_01010101_b.gif
[6] Release 11.0 Documentation for ANSYS
[7]Dr.-Ing.GnterMller/Dipl.-Ing.Clemens Groth; FEM fr Praktiker, 4.aktualisierte
Auflage 1999; expert-Verlag
[8] http://www.promshop.info/cataloguespdf/a005-011.pdf
[9]http://www.skf.com/skf/productcatalogue/jsp/viewers/productTableViewer.jsp?&la
ng=de&tableName=1_3_1&presentationType=3&startnum=14
[10]Lothar Issler Hans Ruo Peter Hfele; Festigkeitslehre Grundlagen,
2.Auflage; Springer
[11] Nadellager Zylinderrollenlager Katalog 307; INA Wlzlager Schaeffler
Ohg;2000
[12]http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20040016066_2004008347.pd
f
- 65 -
Appendix (1)
APDL generating simplified model of the angular contact ball bearing
!!!!!!!!!!!!!!!!! user input file!!!!!!!!!!!!!!!!!!!!!!
number_balls = 6
diameter = 22.225
rb = diameter/2
distance = 10
radius_beam =4
sec_max=1
angle_degree_int=21
/prep7
*GET,et_max,ETYP,0, num, max
*GET, r_max, RCON, 0, num, max
*get,mat_max,mat,0,num,max
et_beam_id = et_max+1
tagetid = et_max+2
cidb= et_max+3 !bonded contact
cidf= et_max+4 !frictional contact
et,et_beam_id,188 ! beam 188
et,tagetid,170 ! target 170
et,cidb,175 ! node-surface contact 175
et,cidf,175 ! node-surface contact 175
!!!!!!!!!!!!!!!!!!!!!!key options for bonded contact!!!!!!!!!!!!!!!!!!!
keyopt,cidb,2,2 !mpc keyopt,5,2,2
keyopt,cidb,12,5 !bonded contact
keyopt,cidb,10,5 !update stiffness each iteration
keyopt,cidb,5,4
!!!!!!!!!!!!!!!!!!!!!!key options for friction contact!!!!!!!!!!!!!!!!!!!
keyopt,cidf,12,0 !standard contact
keyopt,cidf,2,4 !Pure Lagrange
keyopt,cidf,10,5 !!stiffness update
- 66 -
!!!!!!!!!!!!!!!real constant setting!!!!!!!!!!!!!!!!!!!!!!!!!!!!
r,r_max+9
r,r_max+7 !r,5
r,r_max+8 !r,6
r,cidf,,,0.2 ! FKN=0.2
!!!!!!!!!parameters for beam element!!!!!!!!!!!!!!!!!!!!!!
sectype,sec_max+1,beam,csolid !circular solid beam
secdata,radius_beam
mp,ex,et_beam_id,200000 !Youngs Modulus 2e5MPa
mp,prxy,et_beam_id,0.3 !Poisons ratio 0.3
mp,mu,cidf,0.05 !!!!!!! friction coefficient!!!!!!!
!*********************generate beam element*******************!
csys,20
*AFUN, deg
mat,cidb
type,tagetid
real,r_max+7
cmsel,s,inner_outerring
esurf
allsel
mat,cidf
type,tagetid
real,r_max+8
cmsel,s,outer_innerring
esurf
allsel
!!!!!!!!define array for storage of centre node!!!!!!!!!!!!!
n=1
*dim,centre_node,array,5,1,1
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
*do,j,90,111,angle_degree_int
*GET, max_node, NODE, 0, num, maxd
!!!!!!!generate local coordinate at each ball centre!!!!!!!!!!!!!!!
csys,20
- 67 -
k,100+50*j,62.6364,j,15.816 !the keypoints defining centre
!of the local coordinate !sys.
k,100+50*j+1, 62.6364,j, 15.816+22.225/2 !define the keypoint of positive x-axis
k,100+50*j+2, 62.6364+(22.225/2)*sin (50),j, 15.816+(22.225/2)* cos (50)
cskp,1000+50*j,1,100+50*j, 100+50*j+1,100+50*j+2
!!!!!!!!generate beams at outer ring !!!!!!!!!!!!!!!!
csys,1000+50*j
n, max_node+1 , 0,0,0
centre_node(n,1,1)=max_node+1
*do,i,30, 90, distance !i is the number of beams
mat,mat_max+1
type,et_beam_id
secnum,sec_max+1
real,r_max+9
!csys,1000+50*j
n, max_node+i+j, rb,i,0
e, max_node+i+j, max_node+1
*enddo
!!!!!!!generate contact at outer ring!!!!!!!!
*do,i,30, 90, distance !i is the number of beams
type,cidb
real,r_max+7
e, max_node+i+j
*enddo
!!!!!!!!!!generate beams at inner ring!!!!!!!!!!!!!!!!!!
*do,l,210, 270, distance
mat,mat_max+1
type,et_beam_id
secnum,sec_max+1
real,r_max+9
n, max_node+l+j, rb,l,0
e, max_node+l+j, max_node+1
*enddo
!!!!!!!!!!generate contact at inner ring!!!!!!!!!!!!!!!
*do,l,210, 270, distance !i is the number of beams
type,cidf
real,r_max+8
e, max_node+l+j
*enddo
- 68 -
n=n+1
*enddo
!!!!!!!!!!!!generate beam between centre nodes!!!!!!!!!!!!!!!
n=1
mat,mat_max+1
type,et_beam_id
secnum,sec_max+1
real,r_max+9
e,centre_node(n,1,1),centre_node(n+1,1,1)
!***********************generate beam finished****************!
- 69 -
Appendix(2)
The proof of equation ) (
2
1
o i m
d d d in angular contact ball bearing
Geometry of angular contact ball bearing
From the geometry one can get:
2
1
=
m
d
2
1
x d
i
2
1
A Cos()
i.e. =
m
d x d
i
2 A Cos()
=
2
1
( + +4x+ 2A Cos()) (1)
i
d
i
d
In order to prove ) (
2
1
o i m
d d d (2)
Comparing (1) and (2), one has to prove:
+4x+ 2A Cos() (3)
o
d
!
i
d
From the geometry one can get directly the following equation:
2
1
=
o
d
2
1
o i
r x d
i.e. =
o
d
o i
r x d 2 2 (4)
comparing (3) and (4), one has to prove:
- 70 -
4x+ 2A Cos()
o
r x 2 2
!
o
r
From the geometry, one can get :
x -A Cos() (6)
i
r
Comparing (5) and (6), if and only if = , then the equation (2) holds.
i
r
o
r
- 71 -