Sunteți pe pagina 1din 16

Engineering Failure Analysis 14 (2007) 557572

www.elsevier.com/locate/engfailanal

Analysis of a shrink-t failure on a gear hub/shaft assembly


C.E. Truman *, J.D. Booker
Department of Mechanical Engineering, University of Bristol, Queens Building, University Walk, Bristol BS8 1TR, UK

Received 21 March 2006; accepted 23 March 2006


Available online 15 May 2006

Abstract

The paper presents results and conclusions arising from an investigation into a shrink-t failure on a gear hub/shaft
assembly. The work formed one element of a full failure investigation and set out specically to determine whether or
not fretting or micro-slipping was the cause of failure. The basic rationale of the study was to start with assumed operating
conditions and progressively rene the model in order to accurately take account of the true operating conditions. This
approach permitted conclusions to be drawn after every simulation concerning the propensity to slip along the gear
hub/shaft interface. Two modelling techniques were used in parallel: a novel analytical approach and a nite element
study. Results from the two approaches were consistent and in relatively good agreement. The principal conclusion made
was that the observed slip and subsequent fretting damage could only have occurred as a result of another, principal, fail-
ure mechanism which was responsible for increasing the internal diameter of the gear hub.
2006 Elsevier Ltd. All rights reserved.

Keywords: Shrink-t; Micro-slip; Residual stress; Fretting failure

1. Introduction

A shrink-t is a semi-permanent assembly system that can resist the relative movement or transmit torque
between two components through the creation of high radial pressures at the interface of its constituent parts.
It provides a low cost joining method and is widely used in industry, with applications to cutting tool holders,
wheels and bands for railway stock, turbine disks, rotors for electric motors and for locating ball and roller
bearings. Shrink-ts are also an eective way of assembling machine elements such as a gear to a shaft to
transmit torque. The underlying principle involves establishing a pressure between the inside diameter of
the gear hub and the outside diameter of a shaft through interference in dimensions at their radial interface.
Commonly, expansion of the external part by heating, or cooling of the shaft is employed, the part located and
then the whole assembly returned to operating temperature where upon the pressure maintains part location
to allow transmission of a torque [1,2].

*
Corresponding author. Tel.: +44 1179289737; fax: +44 117 9294423.
E-mail address: c.e.truman@bris.ac.uk (C.E. Truman).

1350-6307/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2006.03.008
558 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

Shrink-ts must be properly designed and produced in order to achieve the required functionality in a con-
sistent manner, particularly with reference to the following:

Interference: Reference to standards [3] and/or design guidelines [4] is required in order to determine the
optimum interference between the shaft and inner diameter of the gear hub for a given nominal dimension
and resultant radial pressure at the interface.
Component dimensions: Very low dimensional variation or precise dimensional control through inspection
of the component parts is required.
Surface nish: Surface roughness values should range from 0.4 to 1.6 lm rad in order to provide adequate
frictional adhesion between the shaft and hub bore [5,6].
Working stresses: Stresses due to the shrink-t pressure and additional stresses when the assembly is in use
must not exceed strength of the parts.
Stress concentrations: These are created in the shaft and hub due to the abrupt transition from uncom-
pressed to compressed material [1,7].
Fretting: Present where the contacting surfaces of mechanical parts are subjected to rubbing and an alter-
nating stress. A shrink-t assembly carries a torque by virtue of the radial pressure between the two con-
tacting component surfaces, so self-evidently, fretting may occur if the interface pressure is not sucient to
prevent relative circumferential slip between the shaft and the hub. Depending on the magnitude of the slip,
both fretting wear and fatigue may be potential causes of premature failure.
Assembly requirements: Component surfaces must be cleaned thoroughly and rapid assembly achieved after
component heating/cooling, with the avoidance of misalignment.

The failure under consideration in this paper concerns several of the above, interrelated, items. The inu-
ences of interference pressure, component dimensions and applied loads on an apparent fretting failure of a
gear hub/shaft shrink-t assembly is investigated. A mechanistic approach is adopted which, in essence, deter-
mines which combinations of the above parameters lead to conditions where subsequent fretting damage is
likely, and correlates these predictions to observations made on the failed gear hub/shaft.

2. Failure details, scope of investigation and strategy

2.1. Failure details

The research presented in this paper was initiated by a need to determine the cause of a shrink-t failure on
a gear hub/shaft assembly. The gear hub/shaft shrink-t was contained within a gearbox employed in a large
industrial application. After a relatively short working period, the gear proved incapable of transmitting the
applied operating torque and had to be removed from service, causing signicant nancial loss due to pro-
longed downtime and failure investigation costs. The need to determine the cause of the failure was heightened
by the fact that only one replacement assembly was available, and it was imperative a similar failure did not
occur on the replacement. After failure and removal of the gear hub/shaft assembly, the shaft was bored free
from the gear hub. Signicant fretting damage was observed along the contact interface, emanating from the
shaft entrance (the driving end) and extending to a depth of 0.75a, where a was the radius of the contact
interface. The gear hub internal diameter was measured and was found to have increased by an amount
0.00115a compared to the original design dimensions. In a separate metallurgical examination of the failed
assembly, circumferential slip in the form of scratches of angular length 0.000030 rad was observed on the
surface of the shaft at the non-fretted end and 0.000500 rad at the fretted end.

2.2. Scope of investigation

A full failure investigation was commissioned, of which the results of this paper comprised one element.
Specically, the research set out to determine whether fretting had been the primary failure mode, or had
arisen as a result of another, principal failure mode which had caused the gear hub internal diameter to
increase. To provide an answer to this question, the investigation considered three scenarios. First, the origi-
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 559

nal, or nominal, dimensions of the gear hub and shaft were assumed to be those that were present when the
assembly entered service, and the operating conditions as supplied by the end-user were assumed to be correct.
Analysis would then be undertaken which would determine the extent and magnitude of any micro-slip pres-
ent along the hub/shaft interface and compare the predictions with the post-failure observations. Second, the
original, or nominal, dimensions of the gear hub and shaft were assumed to be those that were present when
the assembly entered service, but the possibility of an in-service torque overload was admitted. Analysis would
then be undertaken which would determine the size of the torque overload required to produce damage con-
sistent with that observed. Third, if some other failure mechanism had caused the internal diameter of the gear
hub to increase either before service commenced or during service, would the assumed in-service torque be
sucient to cause damage consistent with the observations. Once these three scenarios had been considered,
it would be possible to conclude whether fretting was the primary cause of assembly failure.

2.3. Strategy

Formulae for the stresses which arise in a hub and shaft shrink-tted together have been available in the
literature for many years, dating back to the work of Lame, and are readily available in standard textbooks,
e.g. [8]. However, these models are not readily extendable to incorporate the inuence of an applied torque.
The traditional approach to determine the failure torque of a shrink-t is based on determining the torque
required to cause slip along the total contact interface. The value of torque required to cause this gross slip
is usually called the holding torque of the assembly. The holding torque for a shrink-t assembly is therefore
a function of the friction, the radial pressure between the hub and shaft, their diameters and the length of the
contact area. This approach, however, fails to account for the likelihood of localised circumferential slip,
which is not constant in magnitude, along the interface at signicantly lower applied torques than the tradi-
tional holding torque. The point of rst slip is usually at the entrance of the shaft into the hub, directly adja-
cent to the applied torque. This localised slip is a key driving parameter for fretting damage, so it is imperative
that more realistic models of the stresses in shrink-ts caused by applied torsional loads are used in integrity
assessments such as that under consideration in this paper. Recent work [912] has provided a framework for
the solution of this type of problem, and the results presented in this paper represent the rst application of
this solution methodology to a real industrial failure.
A schematic diagram of the gear hub/shaft shrink-t assembly is shown in Fig. 1. The shaft had been fab-
ricated from a tempered steel and the gear hub from a carburized steel. After assembly of the shrink-t, the
radius of the contact interface was a = Di/2, the length of the contact interface was L = 3.10a and the gear hub
outer diameter was Do = 3.84a. There existed an interference pressure rrr (r = a, z) along the hub/shaft inter-
face. For convenience, an interference pressure r0 is dened as the constant value of rrr (r = a, z) assumed to
exist in the assembly according to Lame theory. This will be fully detailed in a later section. All dimensions will
subsequently be normalised with respect to a as the absolute dimensions are commercially sensitive. Likewise,

Fig. 1. Schematic diagram showing the gear hub/shaft, the overall dimensions, and dening coordinate system.
560 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

stresses will be normalised with respect to r0. The assumed in-service operating torque of the gear was T/
a3r0 = 0.14. It was also assumed that during operation the shaft increased in temperature from 30 C to
60 C, and the gear hub increased in temperature from 30 C to 100 C. The coecient of friction, f, was ini-
tially assumed to be 0.15 following advice contained in a number of sources [2,7,12,13], although this assump-
tion was also investigated.
The solution strategy was to initially employ the assumed operating conditions and progressively rene the
model in order to more accurately take account of the true operating conditions. This approach permitted
conclusions to be drawn after every simulation concerning the propensity to slip along the gear hub/shaft
interface, and thus gain a quantitative knowledge of the likelihood of fretting. So, for each model, it was
required to determine whether there would be any likely relative circumferential slip between hub and shaft,
how far along the contact interface would this slip propagate, and the slip magnitude itself. These results could
then be correlated against the post-failure observations.
It is emphasised that although normalised quantities are used throughout the paper, the solution procedure
detailed in the next section is entirely linear elastic, so the results are completely general in nature and may be
suitably scaled for other, similar, problems. The next section summarises the analysis procedures and models
employed in the investigation. Section 4 then presents a series of micro-slip analyses of increasing complexity
using the nominal geometrical, material and load parameters. Section 5 presents the results of friction mea-
surements made on coupons extracted from the failed gear hub and shaft, and then uses these values in the
same models as Section 4 (which assume a friction coecient of 0.15). Section 6 contains a reverse analysis,
where the measured post-failure data is used to infer the in-service torque required to cause the observed dam-
age. The nal section of the paper presents conclusions and discusses the consistency of the results presented.

3. Solution methodology

This section describes the analytical and nite element models used to determine the likelihood of fretting
along the gear hub/shaft assembly. The key mechanistic variable required to assess the likelihood of fretting
damage is the magnitude and extent of relative circumferential slip between the hub and shaft, or micro-slip.
Predictions of micro-slip are principally made using a novel analytical approach [10,11], with validations being
provided by nite element analysis. The basic rationale employed is to determine the shear stresses along the
gear hub/shaft interface, rrh (r = a, z), and determine whether they are of sucient magnitude to cause relative
circumferential movement of hub and shaft and if so, to what extent. The analytical approach adopted has the
advantage that it accurately models steep stress gradients, expected in this problem near the shaft/hub
entrance, whereas nite element predictions traditionally struggle to model steep stress gradients, requiring
very ne meshes. The rst point of micro-slip was expected to be close to the shaft/hub entrance, where
the stress gradient was at its most severe. However, the analytical approach assumes that the hub and shaft
are elastic half-spaces, i.e., that the contact interface is much longer than the shrink-t radius. In this problem
the length of the contact interface was 3.1a, suciently long for the validity of analytical results to be accept-
able. By undertaking analytical and nite element predictions in parallel, extra condence in the results would
be gained.

3.1. Analytical approach

Making use of Lames well-known equations for stresses in thick-walled cylinders, the changes in diameter
due to pressure r0 at the inner surface of the wheel and the outer surface of the shaft are respectively [8],
d 0 r0
cshaft  1  ms 1
Es
 
Di r0 D20 D2i
chub m h 2
Eh D20  D2i
where d0 is the outside diameter of the embedded shaft, Es,h are the Youngs modulae of the shaft and hub,
respectively, ms,h are the Poissons ratios of the shaft and hub, Do is the outside diameter of the hub and Di is
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 561

the inner diameter of the hub. The diametral interference c = chub  cshaft is taken up by deformation of both
the shaft and hub, so Eqs. (1) and (2) may be re-arranged to nd r0 in terms of c,

c Ec
r0   D2  d 20 if Es Eh : 3
Di D20 D2i
mh dE0s 1  ms 2d 0 D20 0
Eh D20 D2i

Once the interface pressure is known, the radial and circumferential stresses in the shaft and hub may be
found. For the shaft these stresses are independent of radial position, r,

rrr rhh r0 4

while for the hub


D2i r2  D20
frrr ; rhh g r0 : 5
r2 D20  D2i
The traditional holding torque for the assembly is then found as
p
T H f r0 d 20 L; 6
2
where L is the length of the contact interface and f is the coecient of friction.
The stress state arising from the application of a torque T to a non-slipping shaft of radius a embedded in
an elastic half-space may be modelled using two alternative assumptions regarding the boundary conditions at
the entrance, z = 0, of the shaft, which act as bounds on the true situation, Fig. 1 [10]. If the shaft is assumed
to be extremely rigid, circumferential displacements at this position will be proportional to radial position, up
to the shaft radius a. This situation takes the form of the so-called ReissnerSagoci (RS) problem, for which
the shear stresses, rrh, at radius r and axial position z are given as
3T q2 n
rrh r; z 2
: 7
4a3 pn2 g2 1 n2
Alternatively, if the shaft is assumed to be extremely exible, the shear stresses, rrh, on the shaft at its point of
entry into the half-space are prescribed to be proportional to the radial position, and the shear stresses at ra-
dius r and axial position z are given by a dierent expression,
2T 4 2 2
rrh r; z p 3d  16d 16Kd 8d  16Ed 8
3a3 p2 d 3 q
where
4q
d2 : 9
1 q2 x2
In the above expressions, K() and E() are, respectively, the complete elliptic integrals of the rst and second
kind, x and q are dimensionless cylindrical coordinates and n and g are dimensionless oblate spheroidal coor-
dinates dened as
qp
z r
x ng; q 1 n2 1  g2 : 10
a a
The stress eld due to the slip of the shrink-t assembly is based on the solution for the stresses in an innite
elastic space of shear modulus l due to a circular dislocation of radius a and Burgers vector bh, calculated at
radial position r and depth z relative to the dislocation,
 
D lbh 0 2 k 4  16k 02
rrh a; z; bh 8k 2  k Kk  Ek 11
4pk 2 k0
where
562 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

k 02 1  k 2 ; 12
4
k2 : 13
4 z=a2
To determine the shear stresses and slip penetration along the interface of the gear hub/shaft assembly, an
integral equation is formed, embodying the fact that the interface shear stress within the slip region is limited
to frrr (r = a, z) where f is the coecient of friction. This integral equation must be satised along a region of
unknown length c of the shrink-tted interface within which slip is occurring. This is a singular integral equa-
tion with a Cauchy kernel, and is evaluated numerically using a GaussChebychev quadrature scheme de-
scribed in detail in [10].
The magnitude of the relative circumferential slip between the shaft and hub at any axial position, z, may
then be determined using
Z c
Dhz Bh ndn; 14
z
dbh
where Bh dn
is the dislocation density. The surface relative slip displacement is thus Dh(0).

3.2. Finite element model

In parallel with the analytical modelling of micro-slip along the gear hub/shaft interface, a nite element
model was created. This provided an independent validation of the analytical solutions, however it was
assumed the nite element models would generally provide less accurate results in areas of steep stress gradi-
ent, such as near the surface entrance of the hub/shaft.
The model was created in ABAQUS CAE and executed in ABAQUS 6.4. A linear elastic material model
was assumed, using material properties provided by the manufacturer. The material properties were typical of
those for the general class of materials described in Section 2. The model contained 3,775 elements and 4,042
nodes, with 87 elements being present along the gear hub/shaft interface. The mesh was rened towards the
entrance end of the interface, which resulted in an element size of 0.01a. The element type was set as
CGAX4T, a four-noded generalized axi-symmetric, thermally coupled, quadrilateral, bilinear displacement
and temperature element capable of including twist. This choice of element permitted the most general, tor-
sional analysis of the gear hub and shaft to be undertaken. Friction was modelled using the contact surface
property. A hard contact was assumed in the direction normal to the contact, r, but slip was permitted in
the tangential direction. The coecient of friction was set in the input le.

4. Micro-slip assessment using the nominal gear hub/shaft dimensions and assumed load parameters

4.1. Simulation 1: Standstill analysis

Fig. 2 shows the calculated contact pressure along the gear hub/shaft interface in the standstill condi-
tion, i.e., immediately after assembly but prior to the shaft spinning and neglecting any thermal gradients
which may be present. This formulation was referred to as the baseline case. Three lines are shown. The
line labeled Analytical pressure, constant represents the constant value of interference pressure determined
using Eq. (3) and assuming a constant outer hub diameter of 3.84a. This interference pressure is the value,
r0, with which other stresses are normalized. The line labeled Analytical pressure, piecewise again uses Eq.
(3) but uses outer hub diameters of 3.84a at the two ends and 3.00a in the center, as shown in Fig. 1. The
line labeled FE pressure shows the result obtained from the nite element analysis, detailed in Section
3.2.
The global (or traditional) holding torques, TH and corresponding safety factors for the shrink-t assembly
may be routinely determined from Fig. 2 by employing Eq. (6) on a piecewise basis. The values obtained
assuming f = 0.15 are TH/a3r0 = 2.92 with corresponding safety factor 20.69 for a constant, analytically deter-
mined value of r0, TH/a3r0 = 2.79 with corresponding safety factor 19.7 for a piecewise constant analytical rrr
(r = a, z), and TH/a3r0 = 2.84 with corresponding safety factor 20.13 for the nite element determined FE rrr
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 563

1.4

1.2
Normalised contact pressure, rr/0

0.8

0.6

0.4

0.2 Analytical pressure, constant


Analytical pressure, piecewise
FE pressure
0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 2. Lame and nite element interference pressure, standstill condition.

(r = a, z). Thus, there was a high margin of safety for this case, and any micro-slip present would not have
caused slip along the complete shaft/hub interface.
Predictions of the shear stress, rrh (r = a, z), along the gear hub/shaft interface are shown in Fig. 3. The lines
labeled Limit analytical and Limit FE are the maximum values of shear stress that may be supported along
the interface before slipping commences. These lines are plots of rrh (r = a, z) = frrr (r = a, z), where the inter-

0.2
Normalised shear stress, r/0

0.15

0.1

0.05
Limit analytical
Limit FE
Traction bc
RS bc
FE rough contact
FE slip
0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 3. Predicted shear stresses, rrh, along the shaft/hub interface with an applied torque T/a3r0 = 0.14 and f = 0.15, standstill condition.
564 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

face pressure is taken from Fig. 2. The lines labeled Traction bc and RS bc are the calculated values of rrh
(r = a, z) using the analytical approach detailed in Section 3.1, and the lines labeled FE rough contact and
FE slip are the nite element predictions detailed in Section 3.2. Two analytical results are shown corre-
sponding to the two bounding solutions, ReissnerSagoci and Traction, explained in Section 3. The two nite
element predictions dier in the way factional behavior is modeled in the nite element model.
The two bounding solutions from the analytical approach are almost identical. There is, generally, good
agreement between nite element and analytical predictions of the shear stress along the interface. The most
severe shear stress is predicted by the analytical model. This is expected as the nite element analysis struggled
to model the steep stress gradients present. All of the predictions showed negligible relative circumferential slip
along the interface and the conclusion from simulation 1 was that conditions along the hub/shaft interface
would not lead to subsequent fretting.

4.2. Simulation 2: Spinning analysis

Fig. 4 shows the calculated contact pressures along the gear hub/shaft interface when stress relaxation
eects caused by spinning, or centrifugal eects, are considered. The line labeled Analytical pressure shows
the value of 0.49r0. This value is the mean value of the nite element predictions, labeled FE pressure. A
constant value of interference pressure was extracted from the nite element results in order to run the slip
models of Section 3.1. The interface pressure had thus dropped to 48.9% of its standstill value.
The holding torques and corresponding safety factors were again routinely determined from Fig. 4. The
values obtained, assuming f = 0.15 were TH/a3r0 = 1.43 with a corresponding safety factor 10.12 for the con-
stant interference pressure 0.49r0 and TH/a3r0 = 1.48 with a corresponding safety factor 10.51 for the nite
element determined rrr (r = a, z). So, as for the standstill case, when the centrifugal eects of the hub/shaft
spinning were also considered, the shrink-t had a high factor of safety if failure by global slip was considered,
although the factor of safety had reduced by a factor of 2 compared to the standstill case.
Predictions of the shear stress, rrh (r = a, z), along the gear hub/shaft interface are shown in Fig. 5. The lines
labeled Limit analytical and Limit FE are the maximum values of shear stress that may be supported along
the interface before slipping commences. These lines are plots of rrh (r = a, z) = frrr (r = a, z), where the inter-

1.4

1.2
Normalised contact pressure, rr/0

0.8

0.6

0.4

0.2
FE pressure, standstill
Analytical pressure
FE pressure
0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 4. Lame and nite element interference pressure, including centrifugal eects.
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 565

0.12
Limit analytical
Limit FE
Traction bc
FE
Normalised shear stress, r/0 0.1

0.08

0.06

0.04

0.02

0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 5. Predicted shear stresses, rrh, along the shaft/hub interface with an applied torque T/a3r0 = 0.14 and f = 0.15, including centrifugal
eects.

face pressure is taken from Fig. 4. The line labeled Traction bc are the calculated values of rrh (r = a, z) using
the analytical approach, and the remaining line the nite element predictions. Only one analytical result is
shown corresponding to the Traction boundary condition explained in Section 3.1, as both bounding solutions
were essentially identical, as for the standstill case.
It can be seen that under spinning conditions a small amount of relative slip between shaft and hub occurs,
due to the decreased interference pressure. The location of the stick/slip boundary was determined to be at a
depth of 0.087a along the hub/shaft interface. The predicted relative circumferential slip between the shaft and
hub at the shaft entrance was calculated to be 0.000025 rad using Eq. (14). It was considered that this mag-
nitude and extent of slip would not be sucient to cause the observed damage.

4.3. Simulation 3: Combined spinning and thermal analysis

To include thermal eects in the analysis, the thermal boundary conditions described in Section 2 were
incorporated into the nite element model. The periphery of the shaft was permitted to rise in temperature
from 30 C to 60 C, and the periphery of the hub was permitted to rise in temperature from 30 C to
100 C. The gear hub/shaft assembly was then allowed to reach steady state conditions. A specic heat capac-
ity of 586 J kg1 C1 and a thermal conductivity of 52 J m1 s1 C1, representative of steel, were assumed
in the analysis. Fig. 6 shows the steady state nite element predicted temperature distribution in the gear. The
applied thermal boundary conditions are clearly visible. Fig. 7 shows the calculated contact pressures along
the gear hub/shaft interface when thermal eects are considered in addition to the stress relaxation eects
caused by spinning. Results are also shown for the standstill analysis detailed in Section 4.1 and the spinning
analysis detailed in Section 4.2. This is to permit ready comparisons to be drawn.
It can be seen from Fig. 7 that thermal eects do not reduce the interface pressure compared to the spinning
model. Compared to the standstill model, the interface pressure reduces by 20%. By allowing steady state
conditions to be reached, a holding torque of TH/a3r0 = 2.35 was calculated with an associated safety factor
of 16.64, assuming f = 0.15. These values are higher than the corresponding values obtained from the spinning
analysis of Section 4.2.
566 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

Fig. 6. Steady state temperature distribution in the gear hub and shaft.

1.4

1.2
Normalised contact pressure, rr/0

0.8

0.6

0.4

0.2
FE pressure, standstill
FE pressure, spinning
FE pressure, thermal+spinning
0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 7. Contact pressure distributions for simulations 13.

There is negligible relative circumferential slip between shaft and wheel when spinning and thermal eects
are considered, and fretting problems were not expected to occur. The most severe conditions considered (and
hence the most likely to promote fretting) are those pertaining to the spinning analysis of 4.2. It was thus
decided to use these conditions for worst case predictions of the likelihood of fretting damage.

4.4. Comparison with observations

None of the analysis conducted up to his point, i.e., simulations 13, demonstrated that signicant micro-slip
along the hub/shaft interface was likely, with no model predicting signicant circumferential slip. The maxi-
mum extent of slip penetration was 0.087a, which was an order of magnitude lower than the 0.75a observed.
It was interesting to note that the maximum circumferential slip predicted (under spinning conditions) was
0.000025 rad. This was in excellent agreement with the value of 0.000030 rad measured in the separate
metallurgical examination of the failed gear hub and shaft at the non-failed end of the shaft/hub interface.
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 567

In-service, it was possible for both ends of the shaft to be the driving end (although mainly the failed end
would be the end where the torque was applied) and thus this close agreement between the predicted and mea-
sured relative circumferential slip provided a high degree of condence in the numerical models.

5. Micro-slip assessment using post-failure dimensions, assumed load parameters and measured coecient of
friction

The coecient of friction is one of the key parameters which inuence the extent and magnitude of slip in
the gear hub/shaft assembly. The uncertainty of using an assumed friction coecient was removed by under-
taking friction measurements on samples extracted from the failed gear hub and shaft. The value of friction
measured was then used in the most severe micro-slip model, i.e., the spinning model. After this analysis, the
measured, post-failure internal gear hub diameter (0.00115a larger than the nominal value) was used in con-
junction with the measured friction coecient. The rationale of this analysis was that if signicant micro-slip
was predicted using the measured, post-failure, internal gear hub diameter, but was not predicted with any of
the simulations based on the design values, then it could be justiably claimed that the fretting occurred after
some other principal failure mechanism had already occurred.

5.1. Measurements of friction coecient

A friction test rig was designed around samples provided from the failed gear hub and shaft. A simple
clamp system was devised in order to press the samples together using a at steel plate and steel convex block
machined with a radius replicating the actual shaft diameter, and manufactured using a wire-EDM process.
Two bolts were torqued so as to provide 89% of the standstill interference pressure, r0. This value was
deemed representative of the real contact pressure in the gear hub/shaft assembly. This arrangement permitted
the coecient of friction to be determined for the sample surface combinations received in a laboratory test.
In order to determine the coecient of friction f, the hub sample was pulled in a tensile test machine whilst
being clamped to the coupon of shaft material. The coecient of friction at the non-sample surfaces was made
very low by grounding and polishing with grease. It was therefore assumed the highest coecient of friction
was that of the sample surfaces.
The force was applied to the hub sample by an Instron 250 kN tensioncompression test machine pulling a
rod located through the hub sample. The sample shaft sections (0.002a thickness) were located using three cap
head screws and clamped precisely on the convex block in order that the hub inner diameter sample surface
mated with the shaft specimen at the correct diameter. All other surfaces were cleaned with acetone. A test rate
of 1 mm/minute was used to represent quasi-static conditions for measuring the response of the hub on shaft.
Fig. 8 shows the friction response for the non-damaged sample surfaces and the damaged sample surfaces.
The highest value of applied tangential stress reached before slip (also audible at test) for each sample surfaces
are highlighted on the gures with precise values used in calculations below,
0:236r0
fnon-damaged 0:139 15
0:89r0
0:2230r0
fdamaged 0:251: 16
0:89r0
These tests revealed that a slightly lower value of friction coecient than assumed in Section 3 (f = 0.15)
should be used in the micro-slip calculations. The tests also showed the increased value of friction at the dam-
aged end.

5.2. Simulation 4: Inuence of as-measured friction coecient

The lowest value of friction measured, f = 0.139, was then used in the same model as Section 4.2, i.e., the spin-
ning model. As described earlier, this model was employed as it was found that the largest magnitude of micro-
slip would pertain under these conditions. In other words, this may be considered as the worst case model.
568 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

Fig. 8. Friction test results for the two samples provided.

It was found that using the slightly reduced value of friction coecient produced negligible dierences from
the results of Section 4.2. In other words, the same conclusions were reached as Section 4.2, and in particular,
the extent of micro-slip would not be sucient to cause the damage observed. Fig. 9 demonstrates the negli-
gible eect of lowering the friction coecient by showing the shear stresses along the shaft/hub interface for an

0.04
FE, f=0.15
FE, f=0.139
0.035
Normalised shear stress, r/0

0.03

0.025

0.02

0.015

0.01

0.005

0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 9. Comparison between predicted shear stresses, rrh, along the shaft/hub interface with an applied torque T/a3r0 = 0.14 and f = 0.15,
0.139, including centrifugal eects.
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 569

applied torque T/a3r0 = 0.14, and frictional coecients of 0.15 and 0.139. It can be seen that the lines are
almost identical.

5.3. Simulation 5: Spinning analysis using post-failure measurements

Fig. 10 shows the calculated contact pressures along the gear hub/shaft interface when the measured, post-
failure, dimensions of the gear hub were used in the simulations. A gear hub inner diameter 0.00115a larger
than the nominal dimension was modelled. The simulations used the measured friction coecient value of
f = 0.139 and also considered stress relaxation eects caused by spinning. The line labeled Analytical pressure
shows the value of 0.0050r0 This value is the mean value of the nite element predictions, labeled FE pres-
sure. It was necessary to extract a constant value of interference pressure in order to be able to run the slip
models of Section 3.1. Also shown, for comparison purposes, are the equivalent results obtained using the as-
assembled dimensions. The reduction in interference pressure may clearly be seen.
The holding torques and corresponding safety factors may again be routinely determined from Fig. 10. The
values obtained assuming f = 0.139 are TH/a3r0 = 0.136 with corresponding safety factor 0.96 for the con-
stant value of interference pressure and TH/a3r0 = 0.166 with corresponding safety factor 1.17 for the nite
element predicted interference pressure. The interface pressure drops to under 5% of r0 and there is no safety
margin. Failure would be expected under these conditions.
Predictions of the shear stress, rrh (r = a, z), along the gear hub/shaft interface are shown in Fig. 11. The
lines labelled Limit analytical and Limit FE are the maximum values of shear stress that may be supported
along the interface before slip commences. These lines are again plots of rrh (r = a, z) = frrr (r = a, z), where
the interface pressure is taken from Fig. 10. The line labelled Traction bc shows the results obtained using the
analytical approach, and the line labelled FE shows the results obtained using nite element analysis.
It can be seen that excessive slip occurs when the post-failure gear hub dimensions and the eects of spin-
ning are incorporated in the micro-slip model. The location of the stick/slip boundary was determined to be at

0.8
FE pressure, nominal+spinning
Analytical pressure
FE pressure
0.7
Normalised contact pressure, rr/0

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 10. Lame and nite element interference pressure, including centrifugal eects, using as-measured post failure parameters. Also
shown is the interference pressure for the equivalent analysis using nominal dimensions.
570 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

0.02
Limit analytical
Limit FE
Traction bc
FE
Normalised shear stress, r/0

0.015

0.01

0.005

0
0 0.5 1 1.5 2 2.5 3
Normalised distance along shaft, z/a

Fig. 11. Predicted shear stresses, rrh, along the shaft/hub interface with an applied torque T/a3r0 = 0.14, f = 0.139, including centrifugal
eects obtained using as-measured, post-failure, dimensions.

a depth of 2.96a along the shaft/hub interface. In other words, slip over almost the complete interface was
expected to occur. The predicted relative circumferential slip between the shaft and hub at the shaft entrance
was calculated using Eq. (14) to be 0.000300 rad. This amount was consistent with the size of the scratches
observed on the failed shaft surface, 0.000500 rad.

6. Estimation of torque required to produce observed damage

Finally, the assumption that the shaft was loaded in torsion to a maximum value of T/a3r0 = 0.14 was
relaxed and estimates of the torque required to produce the observed damage were made. The nal set of
analyses was thus conducted in order to determine the applied torque necessary to produce the extent of
damage observed (0.75a), and to check for consistency in the earlier results. Fig. 12 shows a graph of
predicted slip length versus applied torque for four of the models previously considered. On the graph,
the applied torque has been normalised with respect to the nominal, or assumed, in-service torque T/
a3r0 = 0.14. Two of the curves show the slip versus applied torque characteristics for a hub/shaft assem-
bly assumed to be operating under design parameter values. Thus, the lines labelled nominal standstill
and nominal spinning show the compliance performance for the gear hub/shaft assuming a friction coef-
cient of 0.15 under standstill and spinning conditions. The curve labelled post standstill shows the
compliance characteristics of the assembly using the post-failure dimensions and a friction coecient
of 0.139 in the standstill condition, and the curve labelled post spinning shows the compliance charac-
teristics of the assembly using the post-failure dimensions and a friction coecient of 0.139 under spin-
ning conditions.
It can be seen that in order to generate damage to the observed extent (line labelled Observed slip), an
applied torque of at least 3.29 larger than the assumed operating torque would be required if no change in
the internal dimensions of the gear hub had occurred. Conversely, if the internal dimensions of the gear
hub changed according to the measured values, an applied torque of only 0.32 would be required to produce
the observed damage. This is only 32% of the nominal service torque.
C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572 571

0.8

0.7

0.6
Normalised slip length, c/a

0.5

0.4

0.3

0.2
f=0.15, nominal standstill
f=0.15, nominal spinning
0.1 f=0.139, post standstill
f=0.139, post spinning
Observed slip
Lines A and B
0
0 1 2 3 4 5 6 7 8
Applied torque/nominal torque

Fig. 12. Predictions of slip length versus applied torque. Cases considered are standstill and spinning with f = 0.15 using nominal input
parameters and standstill and spinning with f = 0.139 using post-failure measured dimensions.

7. Conclusions

A series of conclusions concerning the failure of the gear hub/shaft shrink-t assembly were subsequently
drawn after considering the results presented in the previous sections:

1. The gear hub/shaft assembly would not experience signicant micro-slip (and subsequent fretting damage)
if the original on-assembly dimensions were maintained and the maximum operating torque was restricted
to T/a3r0 = 0.14.
2. The predicted magnitude of relative circumferential slip at the gear hub/shaft interface entrance obtained
using the as-assembled parameters (about 0.000025 rad) was consistent with the 0.000030 rad measured in a
separate metallurgical examination of the failed assembly at the non-failed end of the shaft/hub interface.
3. Reducing the friction coecient from 0.15 to 0.139, in accordance with measurements, did not alter con-
clusion point 1.
4. The gear hub/shaft assembly would experience signicant micro-slip (and hence fretting damage)
if the inner diameter of the gear hub had increased in accordance with the post-failure
measurements.
5. The results of the analysis employing the post-failure, measured, dimensions were consistent with the
observed damage. In particular, slip was predicted to occur along the majority of the hub/shaft interface.
Using an increased value of friction coecient (measured at 0.251 at the failed end of the shaft) would have
produced closer agreement between predicted slip length and observed slip length (0.75a). Consistency was
also provided by the predictions of relative circumferential slip present. A predicted value of 0.000300 rad
at the failed end compared favourably with the observed circumferential slip of 0.000500 rad measured
independently.
6. If the diameter of the gear had not changed, an applied torque at least 3.2 times larger than the nominal
operating torque would be required to produce the observed damage. This would not be consistent, how-
ever, with the post-failure measurements of the inner hub diameter.
572 C.E. Truman, J.D. Booker / Engineering Failure Analysis 14 (2007) 557572

7. The observed slip and fretting damage could only have occurred as a result of another, principal, failure
mechanism which was responsible for increasing the internal diameter of the gear hub. One conjecture is
that the dimensional increase of the gear hub bore may be due to the carburizing process. The temperature
for this surface heat treatment process can be up to 1090 C, well above the recrystallisation temperature
for low alloy steel causing a phase change and volumetric increase from ferrite (body centered cubic) to
austenite (face centered cubic). If further subsequent heat treatment processes or secondary machining were
not adequately conducted to specication, this would account for the decrease in interference pressure and
premature failure at the service load.

More generally, the paper applied a novel, analytically based, solution technique to a real industrial prob-
lem. The approach was validated at every step by a parallel nite element implementation of the problem.
Agreement between the two approaches was good, with both approaches being complementary to each other.
The analytical approach naturally captured the steep stress gradients present at the entrance of the shaft into
the gear hub, and the nite element approach included the complex boundary conditions present. As the solu-
tion technique was based on linear elasticity, the solution framework is equally valid to any other similar
problems.

References

[1] Kutz M. Mechanical engineers handbook. New York: Wiley; 1986.


[2] Wick C, Veilleux RF. Quality control and assembly tool and manufacturing engineers handbook, vol. 4. Dearborn, (MI): Society of
Manufacturing Engineers; 1987.
[3] BSI, BS4500A: Specication for ISO limits and ts. Data sheet: selected ISO ts hole basis. London; 1970.
[4] Avallone EA, Baumeister T, editors. Marks standard handbook for mechanical engineering. Boston: McGraw-Hill; 1996.
[5] Bolz RW, editor. Production processes: the productivity handbook. New York: Industrial Press Inc.; 1981.
[6] Norton RL. Machine design: an integrated approach. New Jersey: Prentice hall; 2000.
[7] Beitz W, Kuttner KH, editors. Dubbel handbook of mechanical engineering. London: Springer-Verlag; 1993.
[8] Timoshenko SP. Strength of materials part II: advanced theory and problems. 3rd ed. Krieger Pub Co; 1956. p. 20513.
[9] Hills DA, Barber JR, Sackeld A, Truman CE. A shrink-t shaft subject to torsion. Eur J Mech A-Solids 2002;21(1):7384.
[10] Sackeld A, Truman CE, Hills DA. A stepped shrink-tted shaft subject to torsion. IMechE C J Mech Eng Sci 2002;216:9971004.
[11] Truman CE, Sackeld A, Hills DA. Torsional loading of a nite shrink-t shaft. IMechE C J Mech Eng Sci 2002;216:110915.
[12] Booker JD, Truman CE, Wittig S, Mohammed Z. A comparison of shrink-t holding torque using probabilistic, micro-mechanical
and experimental approaches. IMechE B J Eng Manufact 2004;218:17587.
[13] Rothbart HA. Mechanical design handbook. New York: McGraw-Hill; 1996.

S-ar putea să vă placă și