Академический Документы
Профессиональный Документы
Культура Документы
& Design
Materials and Design 26 (2005) 281–289
www.elsevier.com/locate/matdes
Abstract
The stresses and deformations in the shrink-fitted hub–shaft joint for various fit forms have been analysed using finite element
method. First, the results of finite element method (FEM) were compared with the results of LameÕs equations along central radial
line of one shrink fit type for accuracy of model and it was proved by these results. Then, the most appropriate fit type was inves-
tigated. It was observed that the most effective parameter to determine appropriate fit type is to be stepped or not for shaft and the
geometries of hub of which edges were wasted have been found as more appropriate to reduce the stresses and deformations.
Finally, it was also observed that the each point of the shaft or the hub at the moment of the first meeting to its partner during
fit can enter the plastic region, and this situation must be taken into consideration as design parameter.
2004 Elsevier Ltd. All rights reserved.
0261-3069/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2004.06.014
282 A. Özel et al. / Materials and Design 26 (2005) 281–289
Nomenclature
Prasad et al. [6] analysed the stresses in hollow shaft that the geometry has a significant effect on the stresses
using finite element method (FEM) for hub–shaft sys- and deformations at the beginning and at the ends of the
tem with different combinations l/d (fit-length/nominal fit zone.
diameter), D/d (hub outer diameter/nominal diameter)
and interferences. Their results are in good agreement
with LameÕs equation results at the centre point of the 2. Computation of stresses and deformations at shrink fit
fit [6]. based on lameÕs equations
Zang et al. [7] applied LameÕs equations and three-di-
mensional FE stress analysis to interference fits in ring In order to compute the stresses and deformations at
gear-wheel connections. They found that LameÕs equa- shrink fits, the results of elasticity theory (LameÕs equa-
tions do not give good results for the interference stres- tions) for thick-wall cylinder subjected to both internal
ses and deformations of complex geometry. They also and external pressure are used. Although the shape of
found that three-dimensional finite element analysis machine components used in technical are very different
gives more complete and accurate results. (for example ring gear, wheel) from the shape used in
The finite element analysis for a tubular alumina LameÕs analysis and the results obtained from LameÕs
liner, which was shrink-fitted into a heat-treated high- equations are approximate calculation, due to its practi-
speed steel (HSS) sleeve and subjected to high internal cal applicability, it has been used widely.
pressure and high temperature, was performed. Based
on the analysis a prototype alumina liner-steel sleeve 2.1. Stresses
was manufactured, and it was found that the alumina
might be used as the structural liner for high pressures In thick-wall cylinder subjected to both internal and
and temperatures when it was shrink-fitted into a heat- external pressure, tangential (rt) and radial (rr) stresses
treated HSS sleeve [8]. occur on condition that deformations are in elastic re-
The stresses and deformations in thin rings and disc gion. Assuming the cylinder subjected to internal pres-
after being interference-fitted as thermally or not are sure as hub and the cylinder subjected to external
analysed. The fit-length which the plastic deformation pressure also as shaft, the stresses can be calculate by
occurs is also determined [9,10]. The behaviour of car- LameÕs equations as follow. The stresses in shaft:
bon face seals, used in aircraft engines and assembled !
via a shrink-fit procedure which generates residual stres- d2 d 2Si
rtS ¼ p 2 1þ 2 ; ð2:1Þ
ses in both components were researched using a two-di- d d 2Si di
mensional axisymmetric finite element analysis [11].
!
In this work, the stress and deformation analysis of d2 d 2Si
shrink fitted shaft (without hollow) with six different rrS ¼ p 2 1 2 ; ð2:2Þ
fit forms which have the same size fit features was per- d d 2Si di
formed by FEM and the effects of geometry on stress where p is contact pressure existed between the two
and deformation were researched. It was determined parts. d, dSi and di are nominal diameter, internal diam-
A. Özel et al. / Materials and Design 26 (2005) 281–289 283
eter in hollow shaft and any diameter of shaft, respec- and decrease in external diameter of solid round shaft,
tively. Stress distribution at solid round shaft does not dS can be calculated as
depend on d and is rtS = rrS = p. Depending on these
stresses, equivalent stress is rSeq = p according to von- pd
dS ¼ ½1 mS ; ð2:8Þ
Mises criteria. The stresses in hub: ES
d2 D2H where EH, mH and ES, mS are modulus of elasticity and
rtH ¼ p 2 1þ 2 ; ð2:3Þ
DH d 2 Di PoissionÕs ratio of hub and shaft, respectively. Total
deformation is calculated by the equation
d2 D2H
rrH ¼ p 1 ; ð2:4Þ
D2H d 2 D2i d ¼ dH dS : ð2:9Þ
where DH is external diameter of hub and Di is any
diameter of hub. To simplify the equation, assuming When Eq. (2.7) and Eq. (2.8) are substituted into Eq.
cH = d/DH and maximum tangential and radial stresses (2.9), we obtain
at internal diameter of hub are calculated as
d ¼ pdðk H þ k S Þ; ð2:10Þ
1 þ c2H
rtH max ¼p and rrH max ¼ p: ð2:5Þ 1 1 þ c2H 1
1 c2H kH ¼ þ mH ; kS ¼ ½1 mS :
EH 1 c2H ES
Depending on these stresses, equivalent stress according
to von-Mises criteria is
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 þ c4H 3. Finite element model
rHeq ¼ p : ð2:6Þ
1 c2H
In order to take advantage of the geometric symme-
2.2. Deformations try of the problem, a two-dimensional axisymmetric
model was carried out. ANSYS code version 5.7 was
By using Hooke law for two-dimensional plane stress, used in the analysis. For six different types shrink fit
increase in internal diameter of hub, dH can be calcu- form, models were formed separately and smaller ele-
lated as ments in the critical regions were used for each model.
The shaft and hub were modelled using PLANE42
pd 1 þ c2H and four node quadratic elements. The contact between
dH ¼ þ m H ð2:7Þ
EH 1 c2H hub and shaft was modelled using CONTAC48 and
Fig. 1. Incremental progression of hub onto shaft and finite element mesh details.
284 A. Özel et al. / Materials and Design 26 (2005) 281–289
Z
two-dimensional surface contact elements with elastic T T
fdug ½B ½D½B dðvolÞfug fdugT
coulomb friction. The axial progression was completed vol
in 150 repetitions by given 0.1 mm incremental Ux dis- Z
placement to hub in axial direction for 15 mm pressing ½BT ½D eth dðvolÞ þ fdugTk
vol
length. Boundary conditions and mesh details are given Z
T
in Fig. 1. y-axis was considered as axial symmetry of ½N n ½N n d areaf fug
shaft (Ux = 0) and shaft was fixed at x = 0. Loading areaf
The principle of virtual work states that a virtual Noting that the {du}T vector is a set of arbitrary
(very small) change of the internal strain energy must virtual displacements common in all of the above
be offset by an identical change in external work due terms, the condition required to satisfy Eq. (3.7) re-
to the applied loads, or duces to
nd
½K þ K fe fug F th
e ¼ ½M e f€ug þ F pr
e þ Fe ;
dU ¼ dV ; ð3:1Þ
ð3:8Þ
where U = strain energy (internal work) = U1 + U2, where:
V = external work = V1 + V2 + V3, d = virtual operator. Z
The virtual strain energy is T
½K ¼ ½B ½D½B dðvolÞ ðElement stiffness matrixÞ;
Z vol
T
dU1 ¼ fdugT ½B ½D½B dðvolÞfug fdugT Z
Z
vol
K fe ¼k ½N n T ½N n d areaf ;
½BT ½D eth dðvolÞ: ð3:2Þ areaf
vol
ðElement foundation stiffness matrixÞ;
Another form of virtual strain energy is when a sur- Z
face moves against a distributed resistance, as in a foun- T
F th
e ¼ ½B ½D eth dðvolÞ
dation stiffness. This may be written as vol
Z ðElement thermal load vectorÞ;
dU2 ¼ fdugTk ½N n T ½N n d areaf fug; ð3:3Þ Z
areaf
½M e ¼ q ½N T ½N dðvolÞ ðElement mass matrixÞ;
vol
where {e}, [B], [D], {u} and [Nn] are strain vector, strain–
displacement matrix, elasticity matrix, nodal displace- d2
ment vector and matrix of shape functions for normal f€ug ¼ fug ðAcceleration vectorÞ;
dt2
motions at the surface, respectively. Next, the external Z
virtual work will be considered. The inertial effects will pr T
Fe ¼ ½N n fP g d areap
be studied first: areap
Z ðElement pressure vectorÞ;
T d2
dV1 ¼ fdugT q ½N ½N dðvolÞ 2 fug: ð3:4Þ
vol dt Eq. (3.8) represents the equilibrium equation on a
The pressure force vector formulation starts with one element basis. Where k, q, t, {P} and fF nd e g are
Z the foundation stiffness in units of force per length
dV2 ¼ fdug T
½N n fP g d areap : ð3:5Þ per unit area, density, time, the applied pressure vector
areap (normally contains only one non-zero component) and
Unless otherwise noted, pressures are applied to the the nodal forces applied to the element, respectively.
outside surface of each element and are normal to For in this work, the following equation from Eq.
curved surfaces, if applicable. Nodal forces applied to (3.8) is derived:
the element can be accounted for by
½K þ K fe fug ¼ 0; ð3:9Þ
dV3 ¼ fdugT F nd
e : ð3:6Þ
Finally, Eqs. (3.1)–(3.6) may be combined to give frg ¼ ½D½Bfug: ð3:10Þ
A. Özel et al. / Materials and Design 26 (2005) 281–289 285
4. Results and discussion of shrink fit for I type are given in Fig. 3. rtS and
rrS of shaft along central radial line are approximately
In present work, six different shrink fit forms were equal to contact pressure p, 4.6666 MPa. When the
modelled separately, as given in Fig. 2. In all models, stresses at the midpoint of the hub along radial line
dimensions were taken as l/d = 0.75 and d/D = 0.333. are examined, it is seen that the maximum tangential
For theoretical shrinkage d = 1 lm, p pressure at inter- stress (rtHmax) occurs at internal surface of hub and
face, from Eq. (10), was solved and found as 4.6666 its value is 1.24p and the maximum radial stress
MPa. In the FE analysis, this contact pressure was taken (rrHmax) also occurs at internal surface of hub and
into consideration. The all stress results were normalised its value is p. The results are compatible with the re-
by divided to p value. sults of LameÕs equations and with results given in ref-
The rt and rr stress distributions obtained from erence [6]. The accuracy of models was proved by
FEM and LameÕs equations along central radial line these results.
Fig. 2. Shrink fit types and geometries. (a) I Type, (b) II Type, (c) III Type, (d) IV Type, (e) V Type, (f) VI Type.
286 A. Özel et al. / Materials and Design 26 (2005) 281–289
1.5 1.5
σr FEM
I Type
σt FEM 1.2
1 II Type
σt analytic
σr analytic 0.9 III Type
σeqv/p
0.5
IV Type
0.6 V Type
σ /p
0
VI Type
0 0.25 0.5 0.75 1
0.3
-0.5
0
-1
0 0.25 0.5 0.75 1
y/l
-1.5 Fig. 6. The reqv stresses at external surface of the shaft for all types
d/DH along fit-length.
Fig. 3. Comparison of rt and rr stress distributions obtained from
analytic and FEM for I type shrink fit model along the central radial
line. fit-length (y/l = 0.5). However, the maximum radial
stress at y/l = 1 occurs in I type and is 3.907 times of
stress at the centre of the fit-length. These indicate that
-0.1 the important stress concentrations happen at the ends
0 0.25 0.5 0.75 1
of the fit-length. Therefore, the stresses occurred in the
-1.1 I Type ends of fit length determine the most appropriate fit-
II Type
type. The fit forms according to the maximum radial
III Type
σr/p
-2.1
IV Type stresses occurred at y/l = 1 are I, V, II, III, VI and IV
V Type type, respectively. At the end points, it is observed that
-3.1 VI Type the tangential stresses are lower than the radial stresses.
At the y/l = 1, the arrangement of fit forms according to
-4.1 the maximum tangential stresses are same as those of the
y/l maximum radial stresses.
Fig. 4. The rr stresses at external surface of the shaft for all types
The von-Mises equivalent stresses (reqv) at outer sur-
along fit-length. face of shaft along fit-length for six different fit forms are
shown in Fig. 6. At the y/l = 1, for determining the
appropriate fit-type, the important parameter is the
-0.4
shaft to be stepped or not. The maximum stress concen-
0 0.25 0.5 0.75 1
-0.8 trations occur in the II and IV shaft types which the step
I Type is at the centre and is along fit-length, while the mini-
-1.2 II Type mum stress concentrations occurs in the V and VI shaft
III Type types which the step is at the end edge of fit-length. At
σt/p
-1.6
IV Type the shaft type without the step (for example at the I
-2 V Type and III types), the stress concentrations happen lower
VI Type than those of stepped shaft types at the centre and great-
-2.4
er than those of stepped shaft types at the end. Thus, for
-2.8 the shaft, it is concluded that the best design types to re-
y/l duce stress concentrations are V and VI types, which the
step is at the end edge of fit-length.
Fig. 5. The rt stresses at external surface of the shaft for all types
along fit-length.
4.2. Stress distributions in the hub
4.1. Stress distributions in the shaft For all fit form types, the radial and tangential stres-
ses in the hub along fit-length are given in Figs. 7 and 8,
The radial and tangential stresses (rr and rt) at outer respectively. Radial stresses in the hub are with the same
surface of shaft are given in Figs. 4 and 5, respectively. character with radial stresses in the shaft, but at y/l = 1,
The all stresses occurred in the shaft are with compres- the maximum stresses happen in V type instead of I
sive character. The stresses in the different fit forms type. The stresses are approximately equal p along fit-
are not noteworthy different, except for the ends of fit- length in the hub as in the shaft. While the radial stresses
length. The maximum radial stress at y/l = 0 occurs in are with the compressive character, the tangential stres-
V type and is 2.24 times of stress at the centre of the ses in the hub are different and generally with positive
A. Özel et al. / Materials and Design 26 (2005) 281–289 287
I Type ses in the shaft. The reason of this is that the radial and
-0.5 0 0.25 0.5 0.75 1
II Type tangential stresses are with different character (tension-
III Type
compression). The normalized von-Mises equivalent
-1.5 stress at the centre point of the hub obtained from Eq.
IV Type
(6) is 1.952 and the normalized von-Mises equivalent
σr/p
V Type
-2.5
VI Type
stresses of the same point obtained from FEM change
from 1.924 to 2.027 for various fit form types. This also
-3.5 indicates that the models are accurate. There are not
clearly differences among fit form types according to
-4.5 the stresses occurred along overall fit-length as much
y/l as in the shaft. For the hub, the important determining
factor for the appropriate fit-type is the von-Mises
Fig. 7. The rr stresses at internal surface of the hub for all types along equivalent stress concentrations occurred at the fit ends.
fit-length.
When the von-Mises equivalent stresses at the y/l = 1 are
analysed, it is seen that the stress concentrations of the
1,6
I Type
hub types of which the edges wasted (III, IV and VI
1,4
II Type
types) are lesser than those of the hub types of which
1,2 the edges filled (I, II and V types). According to decrease
III Type
1 the stress concentrations, the arrangement of three fit
IV Type
0,8
types of both hub types are the same and the type of
σt/p
V Type
0,6
VI Type stepped shaft at the centre (IV type for hub type with
0,4
edge wasted and II type for the hub type with edge
0,2
0
not wasted), the type of shaft without stepped (III and
-0,2 0 0,25 0,5 0,75 1 I types) and the type of shaft by stepped at the end
-0,4 (VI and V types), respectively. Therefore, for the hub,
y/l the most appropriate fit type is IV type. According to
the stresses occurred along overall fit-length, the maxi-
Fig. 8. The rt stresses at internal surface of the hub for all types along mum stress occurs in V type and the minimum stress
fit-length.
also occurs in III type.
3,5
4.3. Displacement in the shaft and the hub
I Type
II Type
For six types, the Ux displacements of shaft (shrink-
III Type
ing) and hub (extending) along fit-length are shown in
σeqv/p
IV Type
2,5 Figs. 10 and 11, respectively. To be wasted of the hub
V Type
edges changes the displacement distributions at the shaft
VI Type and the hub. When the hub types of which edges are not
wasted are used as I, II and V type, it is observed that
1,5
the displacements in shaft are greater, as for the hub,
0 0,25 0,5 0,75 1 displacements are smaller. Moreover, as closed from
y/l
Fig. 9. The reqv stresses at internal surface of the hub for all types
along fit-length. -7.50E-05
0 0.25 0.5 0.75 1 I Type
II Type
character. Therefore, the maximum tangential stresses in
-1.00E-04 III Type
the hub occur at outside of ends, while the maximum
IV Type
tangential stresses in the shaft occur at ends. When the
Ur
V Type
tangential stresses in the hub are examined, it is seen
-1.25E-04 VI Type
that the stresses in III, IV and VI types are clearly greater.
In addition, the stresses in the other three types change
the sign at y/l = 1.
-1.50E-04
When the von-Misses equivalent stresses of the hub
given in Fig. 9 are examined, it is observed that the y/l
equivalent stresses in the hub along fit-length are more Fig. 10. The Ur displacement at external surface of the shaft for all
regular, but three times greater than the equivalent stres- types along fit-length.
288 A. Özel et al. / Materials and Design 26 (2005) 281–289
4.00E-04 16
I Type 12
3.90E-04
II Type 8
3.80E-04 III Type 4
σ/p
IV Type
Ur
3.70E-04 0
V Type 0 30 60 90 120 150
-4
3.60E-04
VI Type σr
-8
3.50E-04 σt
-12
σeqv
3.40E-04 -16
0 0.25 0.5 0.75 1 Time
y/l
Fig. 13. rr, rt and reqv stresses at the first end point of the hub fit-
Fig. 11. The Ur displacement at internal surface of the hub for all length for I type during axial progression.
types along fit-length.
σ/p
result of FE, for I, II and V types is Ur/d = 1.46 · 104, 0
for III, IV and VI types is Ur/d = 1.35 · 104, the result -1.5
0 30 60 90 120 150
The stress distributions of the centre point of the shaft Fig. 14. rr, rt and reqv stresses at the centre point of the hub fit-length
fit-length, the first end point of the hub fit-length and the for I type during axial progression.
centre point of the hub fit-length at the first instant of
meeting to its partner during axial progression procedure the hub for I type is seen as 3.31 in Fig. 9 at y/l = 1
for I type are given in Figs. 12–14, respectively. During but, really, the maximum equivalent stress occurs as
axial progression, when a point meets to its mating part 11.12 when meets at first instant to its partner as seen
at the first instant, its stresses rise suddenly and at the in Fig. 13. This is the most interesting and dangerous sit-
afterwards progression steps reduce to its value of final uation because, depending mostly on the interface pres-
progression time. For example, the maximum equivalent sure p, the stresses of a point at the moment of the first
stress at the shaft for I type is seen as 1.286 in Fig. 6 at meeting to its partner can increase excessively, exceed
y/l = 1 but, really, the maximum equivalent stress occurs the yield stress and enter the plastic region. However,
as 2.06 when meets at first instant to its partner as seen in the stresses of this point decrease and its values at the fi-
Fig. 12. Similarly, the maximum equivalent stress at nal progression step are taken into considerations. Con-
3
2
1
0
-1 0 30 60 90 120 150
σ/p
-2
-3 σr
σt
-4
σ eqv
-5
Time
Fig. 12. rr, rt and reqv stresses at the centre point of the shaft fit-length for I type during axial progression.
A. Özel et al. / Materials and Design 26 (2005) 281–289 289
sequently, the point entered the plastic zone is taken into References
considerations as a point of elastic zone. This case is a
very dangerous for design for the reasons provided ear- [1] Ugural AC, Fenster SK. Advanced strength and applied elastic-
ity. UK: Edvard Arnold; 1981.
lier. Although this case was given in Figs. 12–14 for three
[2] Seely FB, Smith JO. Advanced mechanics of materials. 2nd ed.
points, this case is valid for all point of shaft and hub for London: Wiley; 1963. p. 321–327.
all fit types during axial progression. [3] Shigley JE, Mischke CR. Standard handbook of machine
design. New York: McGraw-Hill; 1988.
[4] Attia MH, Kops L. A method for generating desired contact
pressure distributions in experimental interfacial studies. J Eng
5. Conclusion Industry 1985;107:241–246.
[5] Prasad NS, Sashikanth P, Ramamurti V. Stress distribution in
The FE method gives more accurate results in com- interference joints. Computers Struct 1994;51(5):535–540.
plex geometry which the traditional method can not [6] Zhang Y, McClain B, Fang XD. Design of interference fits
via finite element method. Int J Mech Sci 2000;42:
be applied. 1835–1850.
For the shaft types, the most important determining [7] Oh JH, Lee SJ, Lee DG. Manufacturing of an alumina liner-steel
parameter is to be stepped or not. sleeve under high pressure and temperature. Int J Mech Tools
The hub geometries of which edges are wasted are Manufact 1997;37(10):1511–1523.
more appropriate to decrease stress concentrations [8] Gamer U, Lance RH. Residual stress in a shrink fit. Int J Mech
Sci 1983;25(7):465–470.
and displacements. [9] Mack W, Bengeri M. Thermal assembly of an elastic–plastic
In shrink fits, the end zones are more dangerous than shrink fit with solid inclusion. Int J Mech Sci 1994;36(8):699–705.
the central zones. [10] Bengeri M, Mack W. The influence of the temperature
The differences among fit forms appear at end points. dependence of the yield stress on the stress distribution in a
The each point of the shaft and hub at the moment of thermally assembled elastic–plastic shrink fit. Acta Mech
1994;103:243–257.
the first meeting to its partner during fit can enter the [11] Samant RN, Phelan PE, Ullah MR. Finite element analysis of
plastic region. For this reason, it must take into con- residual-stress-induced flatness deviation in banded carbon seals.
sideration as a design parameter. Finite Elem Anal Des 2002;38:785–801.