You are on page 1of 60

ESTLJT~202

Copy No C)'

European Space
Tribol~gy Laboratory
:

Technical Memorandum

Ball Bearing Stiffness


Study

S.D. Lewis & S.J. Fullam

European Space Agency


Contract Report

The work described in this report was done under ESA


contract. Responsibility for the contents resides in the
author or organisation that prepared it.

July 1997

AEA AEA Technology


ESA Contract No. 9844/92/NLIBS WP4458 Activity 20
Subject/Title Ball Be:lring Stiffness Study

Contractor ESTL
Report Reference ESTL/TM/~02
Copyright and The copyright in: this document is vested in AEA
reproduction Technology plc. This document may only be
reproduced in whole or in p:lrt, or stored on :l
retriev:ll system, or tr:lnsmined in :lny form, or by
any me:lns electronic, mech:lniC:ll, photocopying or
otherwise, either with the prior permission of AEA
Technology plc or in :lccordance with the terms of
ESA Contract No. 9844/92/NLIBS

Publication Date: Tuly 1997

File reference I h:\paps\brgstiff\stifrep2 & allres1

ESTL (European Space Tribology bboratory)


AEA Technology plc
Risley
Warrington
WA36AT
Telephone +44 (0)1925 252631
Facsimile +44 (0) 1925 252415

AEA Technology is the trading name of


AEA Technology plc which is certified to IS09001

Author (5) Name (s)


Signature (s)

ESTL Approval Name I R.A. Rowntree


Signature R.. A. &.~ '":....

ESTEC Approval Name


(YMM) Signature
D:lte

AEA Technology i
European Space Tribology laboratory
Executive Summary
Though most ball bearing analysis software packages pen11it. prediction of the Hertzian bearing
stiffness, ESTL is not aware of any experimental validation of such predictions-:-Furthermore,
little is known of how the overall stiffness of a bearing can be predicted, since this includes not
only the Hertzian stiffi1ess, but also dfe effycts of engineering inter£lces (fits between bore and
shaft, or housing and bearing outer diameter) which have not been systematically quantified.
From a design viewpoint, clearly such data would clearly be a valuable input.

As part of its 1995 Primary Appliqtions Package, ESTL conducted a series of experiments to
generate data on the (static) stiffness performance of 3 types of typical spacecraft bearings in real
housing and shaft applications. This data was compared with the predictions ofESTL's own
bearing analysis software, CABARET and data from bearing manufacturers.

The programme aims are to compare experimentally obtained Hertzian stiffness perfon1unce
with predictions of ball bearing analytical software (CABARET V1.08); to identitY the extent
to which differences between prediction and experiment are attributable to inter£tce effects and
to investigate whether incorporation of bearing interface compliance's within the CABARET
software could improve overall stiffness predictions.

The conclusions from this work are:

• Ball bearings housed within typical engineering inter£lces do not behave exactly as predicted
by CABARET (or other predictive codes) however it seems this is not due to errors in the
modelling of bearing perfon1unce itselfbut due to omission of interface effects from the
models.

• Overall agreement between CABARET predictions for bearing behaviour and experimental
data is good. Expressed as a percentage of the experimental value, the overall mean
linearised stiffness predicted by CABARET (averaged over all bearing types and loading
conditions) was within 2% of the experimental results (though the overall standard deviation
is 170% of the experimental value).

• On the whole, CABARET predicts higher stiffness characteristics than are measured
experimentally. Agreement between CABARET and experiment for moment and axial
deflections is closer than for radial loading. Both the higher predicted stiffness and closer
agreement for angular and axial deflections are attributable to CABARET omission of
interface stiffnesses.

• For higher stiffness bearings, Hertzian n0n11al approach within clearances could be a
significant source of error for CABARET predictions of bearing radial stiffness. Therefore a
n10del for this effect should be included in future versions of the software.

Further work is proposed to quantitY the effects of inter£lce compliance in isolation of other
£tctors. Due to the need for this additional work, the extent to which inclusion of these
£tctor can generate an improved model of real bearing and inter£lce performance has not
been quantified.

AEA Technology ii
European Space Tribology Laboratory
Contents

1 Introduction 1
,
2 Experimental Approach 2

3 Test Programme 9
3.1 TEST PROCEDURE AND MATRIX 9
3.1.1 Data Processing 10
3.2 DEFLECTION PREDICTIONS 13
3.2.1 Structural Deflections 14
3.2.2 Bearing Deflections 14

4 Experimental Results 20

4.1 AXIAL DEFLECTIONS 20


4.2 RADIAL 26
4.3 AI\JGULAR 31

5 Discussion 36
5.1 SHAFT STIFFNESS 36
5.2 HERTZIAN INTERFACE STIFFNESS 36
5.3 SHAFT ROTATION 39
5.4 INTERFACE HYSTERESIS EFFECTS 39
5.5 INTERFACE FIT EFFECTS 44
5.5.1 Radial Deflections 45
5.5.1.1101H 45
5.5.1.2 ED20 45
5.5.1.3 SEA65 45
5.5.2 Angular Deflections 45
5.6 STIFFNESS VARIATIONS WITH BALL COMPLEMENT 46
5.7 SUMMARY 50

6 Conclusions 51

7 Appendix A 52

AEA Technology iii


European Space Tribology Laboratory
ESTUTM/202

1 Introduction

Due to the need for very low power consumption and accurate component 195iltion, low
torque and high stiffness are frequently critical requirements for spacecraft ball bearings.
Whereas the torque perfom1Jnce for given preload can often be reliably predicted (for example
based on manufacturers data or by ust! of ball bearing analysis software), the stiffness of a bearing
assembly under defined loads is much mote difficult to predict, partly due to the unknown
effects of bearing ring/housing or ring/shaft interface compliance's. These effects tend to
reduce the assembly stiffuess by an amount which is, in general, unknown at the design stage.

As part of its 1995 Primary Applications Package and shaft applications (WP4458, Activity 20),
ESTL conducted a series of experiments to generate data on the (static) stiffuess perfom1Jnce of
3 types of typical spacecraft bearings in real housing and shaft applications. The supplementary
activity, reported in Appendix A was funded under WP7, Activity 4 of the same contract.

Stiffness was evaluated under axial loading (typical of in-flight conditions), and also in radial and
moment loading (as might occur during ground testing or launch). In all cases the ball bearings
used were angular contact, hard preloaded in back-back configuration (a combination which is
frequently selected for maximum stiffuess). The types of bearings selected (manufactured by
SNFA (ED20 and SEA 65) and Barden Corp. (101H)) were selected in order to provide a range
ofPCD (pitch circle diameter), confomlity and contact angle.

The objectives of the programme are:

• To compare experimentally obtained stiffuess perfonnance with predictions of ball bearing


analytical software (CABARET V1.08).

• To identify and quantify the extent to which differences between predictions and
experiment are attributable to interface effects.

• To investigate whether incorporation of structural and bearing inter£'lce compliance's within


software can improve overall stiffuess predictions.

AEA Technology 1
European Space Tribology Laboratory
ESTUTM/202

2 Experimental Approach

As ball bearings are fundamentally high stiffness (radial stiffrless of order 100MN/m) systems,
assessment of their stiffness behaviour requires position and dimensional measurements at the
micron level, repeatable assembly of Geari?gs and reliable assessment of parasitic structural
deflections.

The experimental programme was carried out on a stiffness measurement jig. This comprised:

• A basic structure of high fundamental stiffuess (radial stiffness of order 100MN/m) .


• A set of different sized shafts and housing rings (one set for each type of bearing to be tested)
to provide typical shaft/housing fits
• A loading facility (static loads were applied to the shafts in either axial, radial or angular
senses using a system of pulleys and weights)
• A displacement measurement system (non-contacting inductive position sensors were
originally proposed).

Details of the bearings tested are given in Table 1 below. All are angular contact, matched pairs
for hard preload in back-back configuration. Preload values were the lowest standard values
available from the respective manufacturers. Low values were chosen to give minimum stiffuess
(so that rneasurable deflections were generated). The preload of the SEA65 bearings was
confimled to be 71.3N, it was more difficult to measure the preload of the 101H and ED20.
bearings, but these values were measured as 4.8N for the ED20 bearings and 12.9N for the
101H bearings. The nominal value for the ED20 was 15N and the 101H was 17.8N. All
bearings were 15 degree contact angle, but conformity values were all different.
Table 1 Details of Bearings Tested
101H ED20 SEA65
Bore (mm) 12 20 65
a.D. (mm) 28 42 85
Ring Width (mm) 8 12 10
Contact Angle (0) 15 15 15
LR. Conformity 1.08 1.14 1.04
a .R. Conformity 1.08 1.14 1.04
Ball Complement 10 10 29
Ball Diameter (mm) 4.76 7.14 5.55
Actual Preload (N) 12.9 4.8 71.3
Nominal Preload (N) 17.8 15 71

A general assembly of the jig is shown in Figure 1. The bearings are located by their outer
rings in the housing rings which are in tum clamped into the upright member. Loads are then

AEA Technology 2
European Space Tribology Laboratory
ESTUTM/202

applied via the central shaft using weights, see Figures 2 - 4, and shaft deflections measured as
shown using the displacement probes. These were originally intended to be non-contacting
inductive probes, but due to late supply, initial tests were carried out using LVDTs to measure
deflections. Based on these initial measurements, it was revealed that the assumed magnitude of
structural effects had been under-estimated and therefore the non-contacting probes could only
be used for the lowest loads. For this reason, the use of an t VDT became the..preferred solution
as this could measure across a much wider working range of deflections than the non-contacting
probes. Note also that deflection measurements for radial and angular tests are referenced from
the upright member itself, thus measJrem~nts are not affected by the moment stiffuess of this
member. An in-line load cell was tempo;arily used in order to confirm the applied loads were
not influenced by pulley friction.

AEA Technology 3
European Space Tribology Laboratory
ESTLfTM/202

~G
.

I
i [i',-
, j W '=:Jte~:::r:01TI I

Sholl Sid, B
Shofl Sid, A 86aring~ In PloeG In The Hou~ing

.~t_______GJI
~./.L - - - -

\<

Figure 1 G A OfTestJig

AEA Technology 4
European Space Tribology Laboratory
ESTUTM/202

Cable
-"'"

Holder

-~ v, /! ~ - - - - - 11/ / / I

~HTsl
~
<

Figure 2 Configuration for Application of Axial Loads

AEA Technology S
European Space Tribology Laboratory
ESTUTM/202

Loading Wire
I
I
Clevis Pin
I

.",>

_ _ _ _ _ _J
~EIGHTSI .
\

Figure 3 Configuration for Application of Radial Loads

AEA Technology 6
European Space Tribology Laboratory
ESTLfTM/202
J _. .._ I

=~~~~,1LVDT Pcob,:7 I I ;p ~~ I
8:1 ITjDjL ---0
j I .
_LjC~~_
--1- rr= __~
J+'__
CG L+J_J
Loading Wire

Load Cell - ~-=]


( -

Pin

lli ll

.",>

------It- . _ - - - - - - L / '\1

I ]
I
I WEIGHTS]
,
<

Figure 4 Configuration for Application of Moments

AEA Technology 7
European Space Tribology Labor atory
ESTUTM/202

The parasitic deflections of the structure and interf:1ces were assessed using "dummy rings", steel
annular rings which were substituted in place of the test bearing pairs during structural
assessments. The dummy rings were manuf.1ctured to tolerances representative of the bearing
rings themselves. Dimensional checks were carried out on these and all other critical
components in order to pemlit assessment of operating fits. The operating fits are summarised
in Table 2 below (by convention negative interference values imply c1earanc~·.

Table U Measured Fits


Mean Interference (microns) Mean Interference (microns)
Bore Housing Bore Housing
A A B B
101H* Dummy ring -30.5 :~ -7.0 -32.3 -11.2
Bearing -38.3 1.2 -38.8 -0.2
ED20 Dummy ring -29.1 -13.6 -26.2 -9.3
Bearing -21.4 -5.2 -21.8 -1.5
SEA65 Dummy ring -45.6 -8.8 -55.8 -5.9
Bearing -52.5 2.7 -52.7 -3.7

* Due to some material pick-up which occurred on initial assembly, the 101H shaft was re-
worked manually to permit use. This resulted in a tapered shaft of around 95 nlicron clearance
at tip and 5 nlicron at root (the mean values were as indicated). The possible effects of this are
discussed in section 5.5.

AEA Technology 8
European Space Tribology Laboratory
ESTUTM/202

3 Test Programme

3.1 TEST PROCEDURE AND MATRIX

The tests were carried out according to the procedure documented in a Job Instruction Sheet
(28346041/1 Issue 1) to a test matrL'{ ~s d<:,fined in Table 3 below. The matrix was developed in
order to provide comparisons between bearings at full, (approximately) one third and half ball
complement for different confom1ity values at three values of bearing PCD. In almost all cases,
bearings were hard preloaded in the back-back configuration (the exception being Test 8.3, a
single bearing test). Assembly was~arried out as described below (see Figure 5):

Bearing pairs (and by definition the durrimy rings) were assembled into the housing (outer
sur£lces were greased and the housing heated to around 55°C in order to pem1it this) and the
outer ring retaining plates clamped in place (using the upper and lower clamping plates). The
shaft was then slid through the bearings and bolted loosely in place whilst the
housing/bearing/shaft assembly was installed in the fixture. Once installed, the shafts were
aligned and the preload applied by clamping the bearing inner rings (the clamping screw was
greased and a preload of lONm applied each time (this was calculated as sufficient to avoid
separation of the shaft/bearing ring interface).

Loads (or moments) were incrementally applied up to the maximum values shown in the test
matrL,{ (in some cases these values were marginally exceeded due to available mass increments).
Once maximum values were achieved the system was unloaded back to zero. Table 3 below
shows CABARET predictions for maximum contact stresses under maximum applied test loads
or moments for the minimum (nominal one third) ball complements tested.

Table 3 CABARET Predicted Maximum Ball/Raceway Contact Stresses For Test


Bearing Ball Applied Ma...x. Inner Ring
Load/
Type Complemen Moment Contact Stress (MPa)
t
101H 3 400 N 1765
Radial
6Nm 2363
300 N Axial 1824
ED20 3 400 N 1552
Radial
30Nm 3037
300 N Axial 1665
SEA65 10 400 N 1119
Radial
30Nm 1831
300 N Axial 1178

AEA Technology 9
European Space Tribology Laboratory
ESTUTM/202

3.1.1 Data Processing

On average tests were repeated 3 times for each configuration and load case to ensure
repeatability of results.

Figure 6 shows an example of the un-processed data from three repetitions ottest 1.3. The
results of the 3 tests were then averaged together to generate a single load versus deflection
curve for each test. In order to remope the effects of rig stiffness, the corresponding dummy
ring deflection versus load curves were subtracted from the curves generated for each test
resulting in the curves shown in this report.

Errors in load measurement were shown to be 1 to S(X> at worst. Deflections were measured
using ESTL's Novibra LVDT's. These are believed absolutely accurate to around O.lllm,
though only a spot step-height reading at lOllm could be made ( this showed no difference
between indicated slip-gauge step height and measured value). On this basis experimental
accuracy was assumed adequate.

AEA Technology 10
European Space Tribology Laboratory
ESTUTM/202

Bearing Housing --I


Lower Clamping Plate

Shaft Side B
.",-
Shaft Side A

Bearings In Place In The Housing

Upper Clamping Plate

,1

Figure 5 Bearing Assembly

AEA Technology 11
European Space Tribology Laboratory
ESTLlTM/202

Comparison of Unprocessed Axial Load v Deflection Data for 3 Test Runs (Test1.3)

60

'2
...o
1 50

----
bn
;:;
~
r 40

o... • Run 1

<B • Run 2
] 30 ~ Run 3
.....
U
III - - Mean Curve
t::
o
u
.....
o
~ 20
;:;
o
'nu
III
<:;:l
III

o
Cil
10

o ...-=- • , \,
o 50 100 150 20() 250 300

Applied Axial Load (N)

Figure 6 Example ofDn-Processed Load/Deflection Data (Test 1.3)

AEA Technology 12
European Space Tribology Laboratory
Table 4 Test Matrix for Stiffness Tests

Max. .~

Bearing Type Test 1.0. Load Config. Ball Com pI. Load/Mom.
ED20 (BB pair) 1.1 , Radial 10 400N
1.2 jMoment 10 30Nm
1.3 Axial 10 300N
SEA65 (BB pair) 2.1 Radial 29 400N
2.2 Moment 29 30Nm
2.3 Axial 29 300N
ED20 (BB pair) 3.1 Radial 5 400N
3.2 Moment 5 30Nm
3,3 Axial 5 300N
SEA65 (BB pair) 4.1 Radial 15 400N
4.2 Moment 15 30Nm
4.3 Axial 15 300N
ED20 (BB pair) 5.1 Radial 3 400N
5.2 Moment 3 30Nm
5.3 Axial 3 300N
SEA65 (BB pair) 6.1 Radial 10 400N
6.2 Moment 10 30Nm
6.3 Axial 10 300N
SEA65 (single bearing 7.1 Axial 29 300N
7.2 Axial 15 300N
7.3 Axial 10 300N
101 H (BB pair) 8.1 Radial 10 400N
8.2 Moment 10 30Nm
8.3 Axial 10 300N
101 H (BB pair) 9.1 Radial 5 400N
9.2 Moment 5 30Nm
9.3 Axial 5 300N
101H(BBpair) 10.1 Radial 3 400N
10.2 Moment 3 6Nm
10.3 Axial 3 300N

AEA Technology 13
European Space Tribology Laboratory
ESTUTM/202

3.2 DEFLECTION PREDICTIONS

3.2.1. Shaft & Upright Deflections

Predictions of test shaft deflections versus load were made using statics (classical equations) 011
the assumption that the test shaft itself behaves as a rigid cantilever (its section''Was modelled as
unifom1 over the entire length since accurate modelling of the reduction in section adjacent to
the bearing would require Finite ElerJilent analysis not available for this task), These predictions
are summarised in Figure 7. The deflection predicted in Figure 7 has been corrected for by
subtracting the deflection with dummy rings from the experimental data.

The moment stiffness of the upright section of the test jig was also predicted (though this
stiffness was largely irrelevant to tht; d.<xial test results as the test methodology removed the effects
of deflection of this member from assessment of the bearing deflections). The predicted stiffuess
of this member was calculated as 48MN/ m (however because of the bore which houses the
bearing housings, again only calculable using FE modelling, we consider this an upper bound
on stiffuess).

3.2.2 Bearing Deflections

Predictions of the deflections to be measured as a function of applied loads or moments were


made using ESTL's CABARET (Vl.08) bearing analysis software. These predictions are
summarised for full ball complement bearings in axial, radial and moment loading in Figures 8,
9 and 10. These predictions have been included in future graphs showing the corresponding
results obtained experimentally.

The software predictions assume:

• Infinite shaft and housing stiffuesses.


• Deflections are due to the (non-linear) deflection of the ball-race contacts alone.
• No interface compliance effects are included.

Over the relatively small range ofloads and moments, the stiffuesses can be approximated as
linear to enable numeric comparisons between experiment and CABARET predictions. These
"linearised" stiffuess values are shown together with experimental results in Table 5 later.

Manufacturers were also asked to supply stiffuess data for the bearings used on test. This data is
summarised below:

For the 101H data, Barden provided axial load/deflection curve and tabulations of "Yield
Rate" for axial and radial loading. This data was derived from a Bearing Compliance computer
program. No experimental data was received.

For the ED20 bearings no data has yet been received from the bearing manufacturer at the time
of issuing this report.

For the SEA 65 bearing, manufacturers quote axial "offset" stiffuess values of 16. 9MN/m for
bearings preloaded at 71N. As will be shown later the axial stiffuess obtained by

AEA Technology 14
European Space Tribology Laboratory
ESTUTM/202

experimentation is 94.7MN/m while that estimated by Cabaret is 52.6MN/m. These values


differ considerably from the values quoted by the manu£.1cturers.

Figure 10 shows a comparison of the predictions of bearing manu£.1cturers with those from the
CABARET code, for axial deflection. It can be seen from the one set of data received from
manu(lcturer of Barden 101H bearings, that their results are very similar to the'CABARET
predictions

AEA Technology 15
European Space Tribology Laboratory
ESTLlTM1202

Calculated Shaft Deflection v Applied Radial Load (Assuming unifonn section Cantilever)

(,

,-...
'"s::
0
t
]. 4
s::
0
OJ::
U

~III
Q
....
3
III

III
13
s::
Cd
U
"Cl
III ')
_
....u
:BIII
...
0..

oV '.
o 50 100 150 200 250 300 350 400
Applied Radial Load (N)

Figure 7 Predicted Cantilever Deflection of Shaft Under Radial Load

AEA Technology 16
European Space Tribology Laboratory
ESTLlTM/202

CABARET Predictions For Angular Deflection of Type 101H, ED20 and SEA65 Bearing Pairs (full ball
complement)

SOOE-OJ

4.S0E-03
...---'_~---'
..-........ ~.-
,........ ..-.-
4.00E-03 ~.~

....,."....
~ ......,./
,., ..... /
~

~
J.SOE-03
..../
/'./ ......
:a"'
.......
..,./
J.OOE-03
~ "' ,.........
----IOIH
I'l
0 , .........
..0
u 2.S0E-03 .".,.~'~,. - - - - - - EDS20
'"
<;:l
, ..,-,., - - - SEA6S
Cl'" ;".
'"
"'
"3/)/)
2.00E-03

~ I.SOE-OJ

IOOE-OJ

S.00 E-04

---------~----------------------------------------------------------------------------------
O.OOE+OO
0 S 10 IS 20 2S 30 3S 40 ~
Applied Moment (Nm)

Figure 8 CABARET Predicted Angular Deflection v Applied Moment for type 101H, ED20 and SEA65 Bearing Pairs

AEA Technology 17
European Space Tribology Laboratory
ESTLlTM/202

CABARET Predictions For Radial Deflection of Type 101H, ED20 and SEA65 Bearing Pairs (full ball
complement)
7

()

.... -_ _..-_ .

~
...uo -"'>
g4 - - - - )OIH
;::
.g . - - - .. ED20
u
~ - - - SEA6S
Q3
"3
:a
~
2

o so 100 1S0 200 2S0 300 3S0 40(i


Applied Radial Load (N)

Figure 9 CABARET Predicted Radial Deflection v Applied Moment for type lOlH, ED20 and SEA65 Bearing Pairs

AEA Technology 18
European Space Tribology Laboratory
ESTLlTM1202

Cabaret Predictions For Axial Deflection Of Type 101H, ED20 and SEA65 Bearing Pairs Compared With
Manufacturer's Data (full Ball Complement)

30

25 I

~
c:: 20
o ~ ...... ... ....
....
(j ........ .. ;.~-- .- .""
g
c::
o
15 ..
~ /
,..".s-"" - . - •
-- -
~
101H
ED20

-' - - - -
t> ,. . ~

ll)
/:/'" - - - - - - SEA65
c;:::;
ll)

Q .. / .... ..- - - - - 101 H Manufacturers data


'til
-<
x 10 ...-
...-
..-
- -
...-
--
--
...-,...-' ,/

" .
,/ ,/
.. ,/ ...-...-"'-
5
...-,," / ,/

o .<", "", "", "", "", """""""", "", "", ""'''''''''''''''''' "",
" " /
/ "/ ...- "
.., ' /,"

o 50 100 150 200 250 300


\,
Applied Axial Load (N)

Figure 10 CABARET Predicted Axial Deflection v Applied Load For Type 101 H, ED20 And SEA65 Bearing Pairs
Compaired With Manu£1Cturers Data

AEA Technology 19
European Space Tribology Laboratory
ESTUTM/202

As expected CABARET predictions agree closely with data obtained from the manu(1cturers.
The most likely reason for this is that manu(1cturers use similar predictive codes to generate the
data provided.

,
4 Experimental 'Results

The experimentally derived data is sUr11marised below for axial, radial and angular deflections
and compared to CABARET predictions.

4.1 AXIAL DEFLECTIONS


The axial load-deflection characteristics of the test jig with the dummy ring sets installed are
shown in Figure 11 (for dummy ring types 101H, ED20 and SEA65 respectively). Note that all
3 types of dummy ring follow the same load v deflection characteristic (approximate effective
stiffuess of 1OMN/ m) which is due to the stiffuess of the upright member of the test fixture in
which the bearings are housed and the stiffuess of the lower clamping plate (which is resisting
the loads in the test fixture). This value is in reasonable agreement with the earlier stiffness
prediction for the upright alone which was itself based on a slightly over-stiff assumption for the
structure (48MN/m).

A'Cialload-deflection plots are given in Figures 12-15 for type 101H, ED20 and SEA65 bearings
respectively. These plots have been corrected for the deflection of the test facility by
subtracting the measured deflections using the dummy rings from the values obtained with the
test bearings. In this orientation bearing deflections thus obtained are mainly thought to be due
to the Hertzian compliance of the ball-raceway contacts (which is the component predicted by
CABARET).

For the type 101H bearings, agreement with CABARET predictions is reasonable up to 150N.
above this value there is divergence (agreement is only within about 50 tYc'J at 300N axial load).
For the ED20 and SEA65 bearings, agreement between predicted and experimentally measured
deflection curves as a function ofload is generally reasonable for full- and ha1£- ball
complement, though again divergence is noted for one third ball complement, these deflection
values being about 50% higher than predicted at high axial load.

For the single bearing (Test 7.1) the experimentally derived axial load v deflection behaviour
showed a similar trend to the bearing pair data, being less stiff than predicted by CABARET.
No reduced ball complement tests with single bearings could be completed as the stimless of the
system was too low resulting in shaft/inner ring tilt and bearing offloaded

20
ESTLlTM1202

Axial Deflection v Applied Axial Load for Bearing Type 101H, ED20 and SEA65 Dummy Ring Pairs

35

30

25

'""
Ul
;::
...u0
g 20 '- ~101H
;::
.$l
..... ------ ED20
u
<ll ----.- SEA6S
o
t;:
15
";;J

~
10

0...... '!
o 50 100 150 200 250 300
Applied Axial Load (N)

Figure 11 Axial Deflection v Axial Load For Bearing Type 101 H, ED20 and SEA65 Dummy Ring Pairs

AEA Technology 21
European Space Tribology Laboratory
ESTLlTM1202

Axial Deflection (lllean of3 tests) v Applied Axial Load for 101H Bearings COlllpared to CABARET
30
Prediction

25

II

1
'"J::
...o
u

J::
-')0

....- .--'"
.-- .
.--
:...-
.",,<

.g
.
..
15 .--'-- 1 - ... - CABARET Prediction (10 ball) .
u
<l>
.--" ---Test 8.3 *(10 ball) :
t;l .-- ......... 1

.
.--
..... .-- .--
<l>
o
Cd
~ ]() -/ .--'--
~.--'--
-~~
""
?

5
::?

0 ..
. ~.,...
?

L--~~;--------;;,;--~
?

,,
o 50 JOO 150 200 250 30n
Applied Axial Load (N)

Figure 12 Axial Load v Deflection for 101H Bearing Pair (Mean of:3 tests corrected for rig deflection effects)

AEA Technology 22
European Space Tribology Laboratory
ESTLlTM/202

Axial Deflection (mean of3 tests) v Applied Axial Load for ED20 Bearings Compared to CABARET
Prediction
70

(,()

50
~
__ k''- .- ..... .-
Iio
.
u ....... j-_. .~

g 40 - -.- CABARET Prediction (10 bJIl)


- .. - CA1}bR~T Prediction (5 bJIl)
l::
.g - .. - CABARET Prediction (3 bJII)

'---
U
III
<;:l Test 1.3 *(10 bJII)
Q 30
l
Test 3.3 *(5 bJII)
~
---Test 5.3 *(3 bJIl)
~
:w - - . - _.- - -+- - ~- _. - _ ..- -
_ ..- _ __ . .... x.--x .. ·x ..
• x x x---- x
-- ---------- --
10 x
._x-- ... x
_.. --x·-··
x
o
o 50 100 150 200 250 300 ,,
Applied Axial Load (N)

Figure 13 Axial Delfection v Load for Type ED20 Bearing Pair

AEA Technology 23
European Space Tribology Laboratory
ESTLlTM/202

Axial Deflection (mean of3 tests) v Applied Axial Load for SEA65 Bearing Pair Compared to CABARET
Predictions
14

12

10

E
• . ",'"

.
o
u
",'"
.. '"
"':,¥
- .. -
.-.,.
CABARET Prediction (29 b311)

.'"
:.[ H ",.J(' - .. - CAllAl~ET Prediction (15 b311)
;::
o ",*' '" '" - .. - CABARET Prediction (10 b311)

.'"
'.0
u • ",'" ~
OJ

o ()
t;:l ",A
___B' '"
---Test 2.3 *(29b311)

_.. -
---¥ --- ---Test 4.3 *(15 b311)
-;;
.--- --- _. - ....... - - - Test 6.3 *( 1() b311)

_.
~ ___D' '"

_.- - _.-
4
--- .....- -
v"~ ",- ---
....------- . . ---

- ...-
.

2
",'" _.--_.-
() -
~ .
I
() so 100 150 200 250 300 \,
Applied Axial Load (N)

Figure 14 Axial Load v Deflection for SEA65 Bearing Pair

AEA Technology 24
European Space Tribology Laboratory
ESTLlTM/202

Axial Deflection v Applied Load for SINGLE SEA65 Bearing Compared to CABARET Prediction

30

2S

~ 20
E
eu • i"" • RUIl I
:[
l::i
.gu IS
-------------------------------- A RUI12
- -~. CABARET :
'"
<:;:I
.,-~,---
.......----~ 1 - MC311 Dclfcctiol1 (of2 runs) I
Q
"@ _..--,,,,,,~"'-
~ 10 .---.--
.--'--
~,,.,.,.,

~---.".
",,,
.".""
S

o ir I
o SO 100 lS0 200 2S0 300
~
Applied Axial Load (N)

Figure 15 Axial Deflection v Applied Axial Load for SINGLE SEA65 Bearing Compared to CABARET Prediction

AEA Technology 25
European Space Tribology Laboratory
ESTUTM/202

4.2 RADIAL

For tests under pure radial load, again the stiffness of the test shaft and dummy rings is initially
assessed. These results are summarised in Figure 16 (note the relatively high radial deflection
obtained from the 101 H bearings).

Test bearings were then substituted in place of the dummy nngs and loading t~sts repeated. For
type ED20 and SEA65 bearings, deflections using the dummy rings were subtracted from the
measured deflections using the test be,ttringS in order to generate deflection/load plots
sununarised in Figures 17 & 18.

For the ED20 and SEA65 bearings the corrected data showed less good agreement with the
CABARET predictions than did the axial load data. Perhaps because in this plane, phenomena
not modelled in CABARET are significant.

For the type 101H bearings however it was not possible to carry out subtraction of the dummy
ring data as the deflection with the test bearings was actually less thal) that measured with the
dummy rings. Figure 19 shows radial deflection as a function ofload for CABARET
predictions, test results and dummy rings plotted separately.

Note that though the one third-ball complement tests were attempted for type 101H and ED20
bearings, no repeatable results could be obtained, because the ring movements were highly
non-linear with only three balls present. These results have not therefore been included in this
report.

AEA Technology 26
European Space Tribology Laboratory
ESTLlTM1202

Radial Deflection v Applied Radial Load for type 101H, ED20 and SEA65 Dummy Ring Pairs
70

60

50
--.
Vl
;::
...u0
I JP .""
:[ 40
-+--101H

/
;::
0
.~ ___ ED20
u
<ll
~
<ll
----.- SEA65
Q 30
-;
:a
~
20

10

~
()~i· • : • • •
: .... ~ J
. : •

o 50 100 150 2()() 250 30() 350 400


Applied Radial Load (N)

Figure 16 Radial Deflection v Load For Type 101 H, ED20 and SEA65 Dummy Rings

AEA Technology 27
European Space Tribology Laboratory
ESTLlTM/202

Radial Deflection (ITlean of3 tests) v Applied Radial Load for £D20 Bearing Pair COITlpared to CABARET
Prediction
(,()

50

-;;;- -10
l:l
...uo
!
."'<,

- .. - CABARET Prediccion (10 ball)


l:l
o - .. - CABl\-RET Prediccion (5 ball)
°B 30
Qj
<;:l
Tesc L1 *(10 ball)
Qj

Q ---Tesc3.1 *(5 ball)


til
:a
~ 20
..--------c--~
..------ ..~
. .. -----
--------

.~.~--
.--~

~::j..::.~---A~-·- ~--·----
]0
___ .. ..----- --
.... - . _..- - . - ..- - - . - .... - . - ... - . -..- - ... - ..- -
n _.~ 4' =c,.:..- - .. - -+- - . - ..- - . - .. - - . -..- - . - -+- - . - ..- -+- ..
!
o
50
I
100 ISO 200

Applied Radial Load (N)


250 300 350 400
,
,

Figure 17 Radial Deflection v Applied Radial Load for ED20 Bearing Pair

AEA Technology 28
European Space Tribology Laboratory
ESTLlTM1202

Radial Deflection (mean of3 tests) v Applied Radial Load for SEA65 Bearing Pair Compared to CABARET
Prediction
3.5

~
2.5 . "'J;:~
E
...u
0
..
.-¥' '" . /
'" /'" /.//
"
.",<

- -+- CABARET Prediction (29 ball)


..--~'" ~/
2
!r: ...--"-- -'-11- CABARET Prediction (15 ball)
..--..--6> e// ..-- • ..-- i
"'''7// •..-
- -A- CABARET Prediction (10 ball)
0
.p Test 2.1 *(29 ball)
u

-
<Il 1.5
/.-¥' -----

.
<;:l - - - T e s t -1.1 *(15 ball)
<Il ..--it: ,.
..... .....---

-~- . .-
Q / Test6.1 *(10 ball)

~~ -- e...-/
---
... -

--:''!>~.
.,-

//
/
~/

.-----//. ---_.--- _.---_.-


_ .. -- - . - - - _.-
~
0.5
..-- ... ~...
.. . .-<:::::.- - .. - .-...- - •
. ,..-.:~-

o ..,---
o 50 1(j0 150 200 250 300 350 -100 ~,

Applied Radial Load (N)

Figure 18 Radial Deflection v Applied Radial Load for SEA65 Bearing Pair

AEA Technology 29
European Space Tribology Laboratory
ESTLlTM/202

Radial Deflection (mean of3 tests) v Applied Radial Load for 101H Bearing Pair Compared to CABARET
Predictions & Dummy Ring Behaviour
HO

70

(,0

----
'"
~ .,,<
8u 50

!~ - .. - CABARET Predictions
o
''8 40 ________ 101 H Dummy Rings
Gl
<:;:l -----.-- Test 8.1
Gl
Q
(;
:a 30
~

20

10

(). . . . ..= . =- • -+-


I
.. ... -+- .. -+- .. ... .. .. -+- .. -+- ... ...
I
\,

o 50 100 150 200 250 300 350 400


Applied Radial Load (N)

Figure 19 Radial Deflection v Applied Radial Load for Type 101H Bearing Pair

AEA Technology 30
European Space Tribology Laboratory
ESTUTM/202

4.3 ANGULAR
For angular deflection tests, again the behaviour of the dummy rings was first established. This
is summarised in Figure 20. Note that the deflections of the 101 H dunm1y rings seem much
greater than the corresponding data for either ED20 or SEA65 rings.

Figures 21, 22 and 23 show the corre;.ronding behaviour (corrected for the effects of the
dunm1y rings by subtraction of dunm1y ring angular deflections) of the test bearings themselves
for full, one third- and half- ball complen;ents. For the 101H and ED20 bearings, experimental
results were in close agreement with predictions. For type SEA65 bearings however, the
agreement is less good. However it should be noted that the deflections are much smaller
(typically 10(% of the deflections fot the 101H and ED20 bearings).

AEA Technology 31
European Space Tribology Laboratory
ESTLlTM1202

Angular Deflection v Applied Moment for Type lOlH,ED20 and SEA65 Dummy Ring Pairs

2.50E-03

2.()()E-03

'"t:
os
:a~ ""
1.50E-03
t: ---.....- 101 H
o
'p _ _ ED20
u
Q)
<;:l
Q)
----A- SEA65
o
~ 1.00E-03
::l
b1J

5.00E-04

O.OOE+OO 1~ : .. . . : . • ~ I

o 5 10 15 20 25 30 35 40
Applied Moment (Nm)

Figure 20 Angular Delfection v Load for Type 101H, ED20 and SEA65 Dummy Ring Pairs

AEA Technology 32
European Space Tribology Laboratory
ESTLlTM/202

Angular Deflection (mean of3 tests) v Applied Moment for 101H Bearings Compared to CABARET
Prediction
5.00 E-O]

~
4.50E-O] .-- .--
.--'-
.-- .--
.-- •
4.o0E-O] .. '-- .--
,..--
,/
,/

]50E-OJ
, ,/
,/

,
~
Vl
:::
~ ,/
.""
:a
~ 3.()(IE-03
,/

~ ,/
,/
::: - .. - CABARET Prediction (lO ball)
0
'!l
u
~
2.50E-O] ,;-
,;- -* • Test 8.2 *(10 ball)
<:;:l ,;-
---Poly. (Test 8.2 *(10 ball))
~
~
Q
/
,;- ............
~ 2.00E-03 / * Accounting for effects of
"3bJ)
/ • dummy rings
~ /

.
/
1.50E-O]
If
/
/
1.00E-O]
,.
/
/

I
I
5.00E-O 4
I
I \,
I
O.()(IE +00
0 5 10 15 20 25 30 35 40
Applied Moment (Nm)

Figure 21 Angular Deflection v Applied Moment for Type 10tH Bearing Pair

AEA Technology 33
European Space Tribology Laboratory
ESTLlTM/202

Angular Deflection (mean of3 tests) v Applied Moment for ED20 Bearing Pair Compared to CABARET
Predictions
7.()()E-03

...
x/
//
600E-03
/~
.1(/
".
"./
/It
5.00E-03 //
"..&
~ ",.."'"

:a'" "...&

~ '" 4.00E-03
",..'"

r:l /"..&
o /" ---Test 3.2 *(5 ball)
°au
.&/"
OJ
<;:I
OJ ---Test 5.2 *(3 ball)
Cl 3.00E-03 /It///
"
'"
~ /
/
.-..- _ _ .. _ _ A
.ij .Jf
2.00E-03
/
/ .-..---"-- * Accounting for ,effects of
/
/ -..---"-- dummy rings

/

1.00E-03 /
/
/
/
£
O.OOE+OO rE-= I
~,
o 5 10 15 20 25 30 35 40

Applied Moment (Nm)

Figure 22 Angular Deflection v Applied Moment for Type ED20 Bearing Pair

AEA Technology 34
European Space Tribology Laboratory
ESTLlTM/202

Angular Deflection (Mean of3 tests) v Applied Moment for SEA 65 Bearing Pair Compared to CABARET
Predictions
7.00E-04
........
........... ---
~ ....
(dlO E-04

.... "
... --- ---
" ... " ..... --- ---
5.00E-04
... .... " ,.-.- --- ,.-.-
~
.....
CABAR.ET Prediction (29 ball)
Vl
l:i ,," /--- x
' - ' -

."" I
:a'"
-_ ..
1

,J(.... //----
'" 4(10E-04
~ ,," .... /11'
- .. - CABAR.ET Prediction (15 ball!
l:i
.g )i" 11'---
...... '- '. I

U
<lI " ............
- .. - CABARET Prediction (10 ball)

~ ,,-' ,-11'
---Test 2.2 *(29 Ball)
Q
... 3.00E-04 ~ ,/ I
'" / "II'
~ /
....
,- - - T e s t 4.2 *(15 ball) I

....
,/ i
.;j If /.

... ---/
I
2.00E-04 / / ... /" ---Test 6.2 *(10 ball)
,i
",
/
./ ....... ""'"
---
~ / .......
-'./ --- .... * Accounling for eft\:cts of
1.(lOE-04 1/ / ........
.!..." / ....... dummy rings

(I.OOE+OO~ ~
o 5 10 15 20 25 30 35
I
40
,,
Applied MOlnent (Nm)

Figure 23 Angular Deflection v Applied Moment for Type SEA65 Bearing Pair

AEA Technology 35
European Space Tribology Laboratory
ESTUTM/202

5 Discussion

Though in general agreement between predictions and bearing behaviour was found to be
reasonable, there are differences which need to be further investigated. Table 5 summarises the
experimentally derived "Linearised" stiffness, CABARET predictions and the"i'~ltio of
experimental results to CABARET predictions In general radial stiffness values are at first sight
over-predicted by CABARET, thereris closer agreement for moment stiffness (but CABARET
tends to under predict this) and for axial sti.ffi1ess (for which there is no consistent trend across all
bearings) .

There are a number of reasons why there might be differences between experimentally derived
and predicted bearing stiffuess behaviours. These relate to the non-inclusion of some
phenomena in the current CABARET model which assumes rigid shafts, that the only
compliance in the system is that of the Hertzian ball-raceway contacts. Other sources of
compliance and experimental error which may be relevant are reviewed below:

5.1 SHAFT STIFFNESS

In reality, the shaft is not infinitely stiff as assumed in CABARET (already shown in Figure 7
are the predicted shaft radial deflections assuming the shaft behaves as a simple encastre
cantilever). Deflections of order 5 microns maximum would be expected. However these
deflections would apply equally to dummy rings and test bearings. Therefore it is concluded
that this effect should have negligible influence on experimental results.

5.2 HERTZIAN INTERFACE STIFFNESS

Under radial loading (and to an extent under angular loading too), the Hertzian normal
approach of the clearances (between shaft/bearing inner ring and between housing/bearing
outer ring) should be considered in a stiffness approximation. ESTL has calculated the Hertzian
nonnal approach based on measurements of the actual clearances measured for test bearings and
dummy rings. The calculations assume mathematically smooth and perfectly cylindrical
components in contact, ESTL considers that in reality fom1 errors and roughness effects will
serve to reduce the scale of these deflections, by interfering with possible movements through
the clearances. However results of these calculations, which represent an upper bound, are
summarised in Figure 24. Typical predicted normal approach at maximum loads are of order 2-
3 microns (note that there were insignificant differences between predictions for test bearing
rings and dummy rings therefore the fom1er are shown). Whilst Hertzian normal approach
effects are not modelled in CABARET, these effects would (in theory at least) apply equally to
test bearing rings and dummy rings. Therefore the practice of subtracting dummy ring
measurements from experimental results would only fail to compensate these effects if there was
a difference in either dummy ring or bearing ring behaviour compared to Hertzian theory. Such
a difference is not anticipated.

AEA Technology 36
European Space Tribology Laboratory
ESTLITM/202

Predicted Total Hertzian Normal Approach for Bearing Rings and Shaft/Housing Contacts (based on
measured fits)
4

3.5

;:
g 2.5
1

~~
..<:: -+-IOIH
u
"...0 2
...- ---ED20 I
!

0..
0.. -.-SEA6S!
-0::
C;
E
... 1.5
0
Z

o......-lL~~~~;;;-----;;--------:
0.5

o 50 100 150 200 250 300 350 400


Applied Radial Load (N)
\,

Figure 24 Predicted Total Hertzian Normal Approach For Bearing Rings And Shafts/ housings contacts ( Based On Measured
Fits)

AEA Technology 37
European Space Tribology Laboratory
ESTUTM/202

Table 5 Comparison ofExperimnetally Derived Linearised Stiffness Values With CABARET Predictions

Bearing Type Ball Complement . "Linearised" Stiffness


(Test No.) Radial (MN/m) Moment (KNm/radian) Axial (MN/m)
Expt or Cabaret Expt/CABARET Expt or Cabaret Expt/CABARET Expt or Cabaret Expt/CABARET
Predictions Predictions Predictions
ED20 10 (Tests 1.1, 1.2, 1.3) 19.6 0.3 17.4 1.2 13.9 1.0
5 (Tests 3.1, 3.2, 3.3) 8.1 0.2 9.2 1.0 7.5 0.8
3 (Test 5.1, 5.2, 5.3) - - 6.2 1.1 4.8 0.7
SEA65 29 (Tests 2.1,2.2, 2.3) 117.1 0.4 135.4 1.5 94.7 1.8
15 (Tests 4.1, 4.2, 4.3) 123.1 0.6 79.8 60.0 1.6
10 (Tests 6.1, 6.2, 6.3) 133.3 0.9 129.4 2.5 . ""
25.0 0.8
101H 10 (Tests 8.1, 8.2, 8.3) 5.5 0.1 11.0 1.4 re.5 0.7
E020 Dummy Rings 17.5 - 81.6 - 10.34
SEA65 Dummy Rings 55.6 - 218 - 10.34
101H Dummy Rings 4.3 - 15.8 - 9.87
E020 10 (CABARET) 75.3 14.7 13.5
5 (CABARET) 44.1 9.4 9.0
3 (CABARET) - 5.6 6.8
SEA65 29 (CABARET) 300.8 92.1 52.6
15 (CABARET) 190.5 64.4 37.4
10 (CABARET) 142.9 52.2 30.6
101H 10 (CABARET) 76.6 7.9 16.1
ED20 (Hertzian Approach) 151.1 ,
\

SEA65 (Hertzian Approach) 117.8


101H (Hertzian Approach) 107.6

AEA Technology 38
European Space Tribology Laboratory
ESTUTM/202

5.3 SHAFT ROTATION

During this test programme, the test bearings were not constrained to prevent rotation (since
any constraint might modifY the measured bearing stiffness behaviour). In consequence, some
tests were aborted and repeated due to obvious rotation of the test bearing during loading.

Though this phenomenon would be oflittle consequence in axial tests, in radi.al or angular tests,
rotation would influence the apparent deflection obtained. The influence of small rotations
cannot be discounted in considering the re;mlts of the tests presented. Such rotations are likely
to be of increasing significance as loads inqease and bearing rotating moments (which at small
loads are insufficient to overcome friction;l torque) become larger. In Figures 25 and 26 we
show the calculated apparent radial and angular deflections due to shaft rotation (assuming
LVDT offsets of 1,3 and 5 mm from the shaft axis at zero rotation). As can be seen, from these
graphs, even small combinations of offsets and shaft rotation (e.g. 1mm offset and 1 degree shaft
rotation) can be responsible for the' otherwise unexplainable apparent shaft deflections seen in
the experimental results previously given for radial and angular deflections.

5.4 INTERFACE HYSTERESIS EFFECTS

One difficulty with the test setup is that (for radial and angular loading) the true "zero
deflection position of the shaft or housing within its interface clearance is uncertain at the start
of a test. The assumption made in the initial analysis done above is that zero interface position is
the position resulting on assembly. However, it was frequently found that on unloading a
hysteresis loop was forn1ed with deflection not returning fully to zero. In order to verifY this,
the experimental setup was modified as shown in Figure 27 for a single test (using ED20
bearings with full ball complement). The modified test setup permitted a complete radial
loading cycle fully negative through zero to fully positive and back to fully negative, passing
through true zero twice in order to observe the hysteresis loop so formed.

An example of a hysteresis plot so obtained is shown in Figure 28. Note at the zero load
position, the difference in deflections amounts to around 1 micron, not the 26.6 microns
pern1itted by the full radial clearance in this case. This deflection of 1 micron at the zero
position is probably due to the effect of clearances on stiffi1ess and so can be subtracted from the
measured deflection for the 3 types of bearing shown in Figures 17, 18 and 19. This would
bring the measured deflections closer in line with the CABARET predictions for the bearings.

Using this method, radial stiffi1ess values for ED20 bearings were found to be of the order 12-
14MN/m (approximately confirnung previous results which gave a value of 19MN/m). This
indicated that the zero-load deflections could not account for the difference between
CABARET predictions and measured stiffi1ess values.

AEA Technology 39
European Space Tribology Laboratory
ESTLlTM/202

Comparison of Effects of Shaft Rotation on Radial Delfection Measurements for Diferent Assumed Offsets of
LVDT from Shaft Centreline
400

3S0

~ 300
~
...uo
-,"",
]. 2S0
.:
o
'C
u --+- t mm Oilier Ii

"
"il 200
_ _ 3mm Oilier,
P - . - Smm Oilier]
';ij
:.a
~.: tSO
"...co
0..
0..
~ too

so

()~ I ~
o 2 3 4 S (, ,
Shaft Rotation (degrees)

Figure 25 Apparent Radial Deflection Due to LVDT Offsets of1, 3, and 5 mm

AEA Technology 40
. European Space Tribology Laboratory
ESTLlTM/202

Comparison of Effects of Shaft Rotation on Angular Deflection Measurements for Different Assumed Offsets
ofLVDT from Shaft Centreline
(,.OOE-03 , r - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ,

a :>.OOE-03
:a'"
~'"
...
";;
0..
tlll 4.00E-03
l:i
l:i
"C

~''"" ---+- JUlIll Ot1Set

/ /
'-
0
l:i
_____ 3Illlll Offset
0
3 (HIE-03
"P ---.- :>Illlll Offset
u
'"
t;l

Cl'"
...
"3'" 200E-03
tlll

~
~
...'"
'"
0..
0.. t.OOE-03
<t:

O.OOE+OO'~
o 2 3 4 :> ~
<
Assumed Shaft Rotation (degrees)

Figure 26 Apparent Angular Deflections due to Shaft Rotation for LVDT Offsets of1,3, and 5 mm

AEA Technology 41
European Space Tribology Laboratory
ESTUTM/202

CRANE

T-1~-Laod Cell

LVDT Probe

O~D!I=I~-I=r~
jf ! i
I ---.,.
I

(8
lr~ Loading Wire

L~
Clevis

~~--:==JL ~----J[_
_ _8
,~_
-~L-_- __- 1
\,
WGHTS]

Figure 27 Modified Test Setup To Pertuit Hysteresis Loops to be Generated

AEA Technology 42
European Space Tribology Laboratory
ESTUTM/202

k'* /7mm- o,gyn1 .",<

D..ef'{ Q..Gt-ion
\,

Figure 28 Example of Load v Deflection Hysteresis Loop for ED20 Bearing Pair

AEA Technology 43
European Space Tribology Laboratory
ESTLfTM/202

5.5 INTERFACE FIT EFFECTS

Finally, in ESTL's view the most likely cause of experimental variations from the CABARET
predictions is considered to be the role of the inter£Lce fits.

Clearances between shaft/inner ring and housing/ outer ring can give rise to additional
deflection or tilt under radial or moment loading (this differs from the interfaceTHysteresis as the
fit defines the maximum variation, whereas the hysteresis measures the variation at zero load).
In principal, the additional radial defltj!ction achievable may be up to the value of total diametral
clearance and the angular deflection can 1* calculated by assuming the shaft or outer ring tilts as
a free body within the available clearances.

In Table 6 below we summarise these possible tilts and deflections based on component
nleasurenlents.

.. A nguI ar an dRdilDfl
T a bl e 6 Summary 0 fP arasltlc a a t rf:ace Ft
e ec fIons WI'thi n I ne I s

Bearing/Dummy Ring Parasitic Angular Parasitic Radial Defin. (microns)


(DR) Defln. (radians)
Type Locatio Calculated Total Both Calculated Total Both
n
Locations Locations
ED20 a.D. 8.98E-04 1.24E-03 5.2 26.6
Bore 3.40E-04 21.4
ED20DR a.D. 1.19E-03 1.67E-03 13.6 39.7
Bore 4.85E-04 26.2
101H a.D. 2.50E-03 8.64E-03 0.2 38.6
Bore 6.14E-03 38.3
101H DR a.D. 2.03E-03 7.85E-03 11.2 41.7
Bore 5.82E-03 30.5
SEA65 a.D. 2.72E-03 2.75E-03 3.7 56.1
Bore 2.58E-05 52.5
SEA65 DR a.D. 2.63E-03 3.01E-03 8.8 54.4
Bore 3.77E-04 45.6

In the case of the 101H bearings, the shaft pick-up problems (see section 2 above) resulted in a
tapered shaft from 95 to 5 microns clearance along its length. The allowable tilt for both 10lH
dummy rings and bearings at the bore is in consequence a factor of around 10 times larger than
for the ED20 or SEA65 bearings. At low loads, this would not be expected to influence results,
however at high load, it is possible that this additional tilt might influence results (depending on
the sliding friction of the clamping rings and the normal load achieved in relation to the applied
load or moment).

The ED20 shaft was parallel, though the SEA65 shaft was tapered to a very slight extent (due to
manu£'1cturing errors from 73 microns at tip to 22 microns clearance at root).

Inter£Lce fit effects are relevant only for Radial and Angular tests. The results of these tests are
reviewed below:

AEA Technology 44
European Space Tribology Laboratory
ESTUTM/202

5.5.1 Radial Deflections

In the else of radial deflections all 3 bearing types show un-predicted behaviour which may be
attributed to either interface behaviour or shaft rotation during test However for the dummy
rings, the measured radial deflections are larger than predicted solely on the basis of the
cantilever theory (and in this case no shaft rotation is possible) implying that differences are due
to inter£1Ce effects. These differences are of order 17.51, 3.89, 0.51 microns at,100N load for
101H, ED20 and SEA65 dummy rings respectively. Ifit is temporarily assumed that no shaft
rotation occurred in any test, the observed radial deflection behaviour can be explained as
follows: I' J

5.5.1.1 101H

In the case of the 101 H bearings, ~eflections with the dummy rings were (surprisingly) higher
than for the test bearings. As this bearing type is the only one for which both bearings have
larger mean clearances than the dummvrings (and line-line fits at outer ring/housing contacts)
we would tend to expect the opposite therefore if inter£1Ce effects are relevant we would expect
the deflection of the bearings to be to be larger than the dummy rings by on the order of 8
microns (ifloads are sufficiently high and friction sufficiently low to allow the shaft to displace
within the clearances). Close to 20 microns of extra deflection is observed for the dummy rings
compared to the bearing. If shaft rotation effects are discounted, this can only be explained
assuming that the (tapered) shaft is able to tilt within the clearances. As 20 microns of radial
3
deflection is equivalent to around 0.2x10- radians of angular deflection of shaft relative to
3
bearing bores (well within the calculated values of around 6x10- radians) this is a possible
explanation for the 101H results.

5.5.1.2 ED20

In this case (at for example 300N radial load) the radial deflections are larger than CABARET
predictions by 20 microns for the 5 ball and around 10 microns for the 10 ball test. If this
deflection were expressed in terms of angular tilt, then this is Gust) achievable, there being
3
around 0.3-0.5x10- radians of available tilt within the shaft/bearing bore or shaft/clearances.

5.5.1.3 SEA65

In this case, predicted deflections are of order 1-2 microns. Experimentally determined
deflections exceed predictions by of order 1-2 microns. Such deflections can easily be
accounted for in tem1S of the difference in shaft/bore clearance between bearing and dummy
ring bores.

5.5.2 Angular Deflections

In general experimental results for angular deflections agree well with the CABARET
predictions. Possible reasons for differences are experimental errors at low load or moment. To
demonstrate this experimental data has been re-plotted, assuming exact agreement with the
corresponding CABARET prediction at maximum applied moment, in Figures 29-31. For the
101 Hand ED20 bearings it can be seen that agreement between the CABARET predictions
and the measurements is almost exact except at very low applied moments (implying that the
assumption of small offsets causing apparently large errors may well be correct). If the SEA65

AEA Technology 45
European Space Tribology Laboratory
ESTUTM/202

bearing data is treated in the same way, the agreement appears less good, but it should be noted
that the scale is a (,ctor 10 smaller. This small offset seen as a result of clearances could be partly
corrected for using the -1 micron of measured clearance shown in Figure 28.

5.6 STIFFNESS VARIATIONS WITH BALL COMPLEMENT

CABARET predicts reduced stiffness with reduced ball complement. Though in the case of
the angubr deflection, the ratio of stiffuess versus ball complement variation predicted by
CABARET seems to agree almost exactly with experiment, for other loading cases agreement
seems less good.

AEA Technology 46
European Space Tribology Laboratory
ESTLlTM1202

Angular Deflection (Mean of3 tests) v Applied Moment for 101H Bearing Pair Compared to CABARET
Predictions (assuming prediction is correct at maximum applied moment)
O.OOS

O.o04S "?

,/,/

,/~
OJ)04 ... ,/

",'"
... ,/
,/,/

O.003S
j ",
", ... ",

:aco
..!:-
0.003 ",
... '" "'"
l'l
",'"
0 - ... - C1mARET Prediction (10 ball)
"0 ./ ,,¥
,,/ "
~ O.OO2S
<;:1
---Test 8.2 *(10 ball)
'"

. ."
Q
~ 0.002 ""
~
/ /
~
0.001 S //
/
/
/
O.OO!
~/
/
/
O.OOOS
/
/
/
0
0 S 10 1S 20 2S 30 3S 40 \<
Applied Mom.ent (Nm.)

Figure 29 Comparison of Experimental CABARET Predicted Angular Deflections Assuming Agreement at Maximum Applied Moment

AEA Technology 47
European Space Tribology Laboratory
ESTLlTM/202

Angular Deflection (Mean of3 tests) v Applied Moment for ED20 Bearing Pair Compared to CABARET
Predictions (assuming prediction is correct at maximum applied moment)
7.00E-03

6.00E-03

'"
,,"
It"
500E-O,3 /'"
.... '"
~ ,," "
.
:a'" ." .... - .. - CABARET Prediction (10 balll
i
'"
-; 4.00E-03
y"
- .. - CABAH':E'r Prediction (5 ball) I
~
0
"au - ... - CABf,.L<..E-r, Prediction (3 ball) I
III
<;I
- - - Test 1.2 *(10 ball)
III
~ - - - Test 3.2 *(5 ball)
... 3.00E-0.3
- - - Test 5.2 *(3 ball)
'" Q
~1:1 Q

~
I A
fi

2.00E-O,3

IOOE-03

O.OOE+OO II' !
o 5 10 15 20 25 30 35 40

Applied moment (Nm)

Figure 30 Comparison of Experimental and CABARET Predicted Angular Deflection for ED20 Bearing Pair Asuming
Agreement at Maximum Moment

AEA Technology 48
European Space Tribology Laboratory
ESTLlTM1202

Angular Deflection (Mean of3 tests) v Applied Moment for SEA 65 Bearing Pair Compared to CABARET
Predictions (assuming prediction is correct at maximum applied moment)
7.00E-04 - I

/
,Ji'/
//

("()OE-04 /,Ji'

5.00E-04 //
~/
/
"K/
-
.:'"
til
:.a .&
,," "
,£"

//
.... /
- .. -
- .._---.-
CABARET Prediction (29 ball)
!.: ...
..
4.00E-04 /" - .. - CABi\ll..ET Prediction (15 ball)
"
/

o
'J::
u

'E'" /
./ / ,/
~
./
/ /. //
/
//

~~~~/~
~~tc/
~A'~/
~
- .. - CABARET Prediction (10 ball)
------- Tdr-2.2(29 ball)

'2
~
til

.;j
3.00E-04

..&" /
/

;.'
/
/ /

>£.---~
~;:-/ - ..

X
~
~K"""---
---Test 4.2 (15 ball)
---Test 6.2 (10 ball)

~;; ...
2.00E-04
/: _ --.~-------x~ . --
/

J( ,//
/

.../-

,/ ,,'
/,/ //

1.00E-04

/.
~//
//-'
" ,:-"
~
///'"

//

O.OOE+OO itt I
o 5 10 15 20 25 30 35 40

Applied MOrrIent (NrrI) \,

Figure 31 Comparison ofExperimentla and CABARET Predicted Angular Delfections for Type SEA65 Bearings Assuming
Agreement at Maximum Applied Moment

AEA Technology 49
European Space Tribology Laboratory
ESTUTM/202

5.7 SUMMARY

Agreement between CABARET predictions and experiment are generally good as shown
earlier in Table 5. Agreement is closest in the axial loading case where interface effects and shaft
rotation are not significant. The first phase of this work could not discriminate the effects of
interfaces and rotation. However the supplementary work reported in Apperi'ffix A
demonstrates that these deviations can be greatly reduced if results are corrected for effects of
small shaft rotations during test. Wht!n this is done, CABARET predictions are shown to
closely match the experimentally observed behaviour in radial and angular loading as is the case
with a..xialloading. Small residual deviations from predicted behaviour are attributable to
interface effects, such as interfaces stiffuess hysteresis.

These interface effects may be modelled and include:

• Hertzian Normal Approach

Hertzian normal approach effects have been calculated to be of order several microns at high
loads. Therefore these effects should be included in a future version of CABARET. Though
relatively small in magnitude, these deflections may be a significant source of prediction error
for radially loaded stiffbearings where bearing Hertzian deflections are of the same magnitude as
Hertzian normal approaches. For low stiffuess bearings, such effects are relatively less
important.

• Interface Stiffuess Effects

A follow-up study is proposed in order to quantify interface stiffuess and hysteresis effects for
different interface fits (examination of the use of different fits for each bearing or dummy ring
was outside the scope of the present study and its extension to accommodate shaft rotation).
The proposed follow-up would remove the possibility of shaft rotation by using dummy rings
of different sizes and fits to generate a set of interface stiffuess coefficients for incorporation into
CABARET.

The proposed programme would:

(i) Manufacture further sets of dummy rings with a range of different interface clearances
(say 4 clearances per bearing size). Adapt the existing fixture in order to permit both
higher loads and effects of hysteresis losses to be investigated. This would be achieved
by providing a capability to both push and pull the shaft so that radial loads and
moments in both positive and negative senses can be applied in each test, thus forcing
the interfaces through the clearances.
(ii) Carry out a series ofload v deflection tests on the modified fixture using dummy rings
in order to quantify the stiffuess effects of clearances and examine the effects of
inter£tce behaviour (radial and angular orientations).
(iii) Carry out a very brief series of supplementary (radial and angular) tests using ball
bearings and monitoring shaft rotation so that direct comparisons using the same test
setup can be made ..
(iv) Where possible, input interface stiffuess data to re-analyse data from the present
programme and so demonstrate agreement with CABARET

AEA Technology 50
European Space Tribology Laboratory
ESTUTM/202

The current programme has shown that in the absence of inter£1Ce effects, CABARET
predictions are in close agreement with experiment. That is, CABARET's predictions of ball
motion and Hertzian ball-race compliance are accurate.

The primary aim of the proposed programme will look in isolation at ring/shaft and
ring/housing interface effects (but without the added compiication of the Henzian contacts, ball
complement, PCD or conformity variations). As a secondary issue, the supplementary bearing
tests will provide confidence in extraffiolatfng results to bearings, which would then permit
a
interface stiffness data derived from such programme to be employed within a future version
of CABARET.

6 Conclusions

• Comparisons of experimental and predicted stiffuess performance (CABARET V1.08) have


been made. It is concluded that ball bearings housed within typical engineering interfaces do
not behave exactly as predicted by CABARET (or other predictive codes) however it seems
this is not due to errors in the modelling of bearing performance itself but due to omission of
interface effects from the models.

• Nevertheless, overall agreement between CABARET predictions for bearing behaviour and
experimental data obtained during this programme is good. Expressed as a percentage of the
experimental value, the overall mean linearised stiffuess predicted by CABARET (averaged
over all bearing types and loading conditions) was within 2% of the experimental results
(though the overall standard deviation is 170% of the experimental value).

• On the whole, CABARET predicts higher stiffuess characteristics than are measured
experimentally. Agreement between CABARET and experiment for moment and axial
deflections is closer than for radial loading. Both the higher predicted stiffuess and closer
agreement for angular and axial deflections are attributable to CABARET omission of
interface stiffuesses.

• For higher stiffuess bearings, Hertzian normal approach within clearances could be a
significant source of error for CABARET predictions of bearing radial stiffuess. Therefore a
model for this effect should be included in future versions of the software.

• The effects of interface compliance for each bearing type (in isolation from other factors)
cannot be derived from this test programme. However the supplementary work described in
Appendix A suggests that shaft rotation may have a larger effect on original radial and angular
test results than the interface compliance effects. The proposed additional work is necessary
in order to verify this and until it is completed it would not be appropriate to modify
CABARET to account for interface effects. Thus the extent to which improved prediction
accuracy can be obtained by incorporating interface effects in the CABARET model cannot
yet be proved.

AEA Technology 51
European Space Tribology Laboratory
ESTUTM/202

7 Appendix A
As a supplementary activity, ESTL was requested to carry out radial and angular deflection tests
on 1 bearing type whilst simultaneously measuring shaft rot;J.tion using inclinometers.
...,T

The test setup for the SEA65 bearings was. therefore modified to interface with the inclinometer
and counterbalance its mass as shown'in F.igure A 1. Full ball complement bearing radial and
angular loading bearing tests 2.1 and 2.2 were then repeated several times. The measured shaft
rotations were very small, of order 0.1-0.2 degrees, however these generated small parasitic
deflection measurements which could be easily calculated once the tilt angle was known using
the principal of similar triangles as ~hown in the example below:
r-- ___
D' I Dr---------_
x ,i

\ ~------
)
D'= (Z tan dQ cos dQ)+X x ( Z - Z)
Z tan dQ cos dQ cos dQ

e.g.
IfX=lmm, Z=22.7, Q= 1°, then D'=0.012mm = 12.186 microns

As the parasitic shaft rotations were repeatable, the data was used to "correct" the original data
from test 2.1 and 2.2 by subtracting the measured parasitic deflections at each load value. Radial
load v deflection and moment v angular deflection results are shown in Figures A2 and A3
respectively. In each case, we compare the original results (uncorrected for the effects of shaft
rotation) with the predictions of CABARET and results corrected for parasitic shaft rotation.
As can be seen, the agreement between CABARET predictions and corrected experimental test
data from Test 2.1 (radial loading) is greatly improved by this means. For the angular loading,
there is also improved agreement. However it should be noted that there remain some small
differences between the cOlTected data and the predictions, which we currently attribute to
interface effects.

AEA Technology 52
European Space Tribology Laboratory
ESTLlTM1202

[41--
-It~ Irr-~
H--

~~I~ ~
r'lyJ r'l LVDT Probe r'l LlJr'l

I
Inclinometer ~ rr-rr+r- ~
LLj-L-l-fJ ----Y-
- lir- I I I
-1'1- - -J-P--LIJ_L-jl_l+LI
I I 1----rT] G
o:::IIllJ
I-~

Counter f3oloncinq

/~
Weiqhls

Clevis Pin Loodinq Wire

/11 'r--------
II
OL

_~I - -------

IWEIGHTS ,-<

Figure A1 Configuration for the Measurement of Shaft Rotation During Radial Loading

AEA Technology 53
European Space Tribology Laboratory
ESTLlTM1202

Comparison Of Predicted And Experimenal Radial Deflection Of SEA65 Bearing Pairs- 29 Ball
Complement ( Results Are Corrected For Structural Deflections And Shaft Rotation)

2 , I

1.8

1.6

en 1.4
l:
e
.~ 1.2
"'"
E
l: - - - - Cabaret Predictions i
o

~~;~e2~t~d*~:p~~:~ental Test ?:!J


:;:;
o
Ql

o 0.8
:c'"
&. 0.6

0.4

0.2

o~ I
,,
o 50 100 150 200 250
Radial Load (N)

Figure A2 Corrected Radial Deflection v Applied Moment For Type SEA65 Bearing Pair

AEA Technology 54
European Space Tribology Laboratory
ESTLlTM1202

Comparison Of Predicted And Experimental Angular Deflection Of SEA65 Bearing Pairs- 29 Ball
Complement ( Results Are Corrected For Structural Deflections And Shaft Rotation)

0.0004

0.00035 •
//

/
0.0003
Ul
c:
III
"C 0.00025 .""
III
=:
c:
0
:;;
/ -----.- CABAR_ET. Prediction (29 ball)
0 0.0002 - - . - Experimental (Test 2.2)
<ll
I;:
<ll - - - - - Corrected Experimental (Test 2.2)
---- .. - ---
0
--~------- ---------~-

...
.!!!
::J
0.00015
Cl
c:
«
0.0001

0.00005

0
0 5 10 15 20 25 30 35 40 ,
<

Applied Moment (Nm)

Figure A3 Corrected Angular Deflection v Applied Moment For Type SEA65 Bearing Pair

AEA Technology 55
European Space Tribology Laboratory
ESTUTM/ZOZ

Circulation List
ESTEC

M. Eiden YMM 5
P. Arrabal ECT 1
Ms L. Valentin AOC-DOC 1

ESTL

S.D. Lewis 1
DCO 2
File 1
Data File 1

AEA Technology 52
European Space Tribology Laboratory

You might also like