You are on page 1of 14

1

Normalization of the stress concentrations at the


rounded edges of a shaft–hub interference fit
A Strozzi*, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi
Engineering Faculty, University of Modena and Reggio Emilia, Modena, Italy

The manuscript was received on 23 December 2010 and was accepted after revision for publication on 23 February 2011.

DOI: 10.1177/0309324711403845

Abstract: The elastic stress concentrations developed from the keyless frictionless static press-
fit of a shaft into a hub are addressed. Two configurations are examined, namely (a) an infi-
nitely long solid shaft press-fitted into a hollow hub with bore rounded edges, and (b) a shaft
with filleted extremity, partially inserted into a hub. Derived from an analytical approach, a
normalizing parameter is proposed that accounts for the combined effects on the stress con-
centrations of the fillet radius, the shaft radius, the interference, and Young’s modulus. With
the aid of finite elements, various design charts are compiled that report the elastic stress con-
centrations within the hub versus the proposed normalizing parameter. Each curve is valid
for a fixed ratio of inner to outer hub radii.

Keywords: interference-fit, contact stresses, stress concentration factor, normalizing


parameter, finite elements

1 INTRODUCTION In this paper the elastic maximum contact stresses


are analysed for the two configurations illustrated in
Interference fits are widely employed to semi-perma- Figs 1(a) and (b). A detailed knowledge of the contact
nently connect gears, pulleys, flanges, wheels, discs, stresses is relevant, since it assists the designer in fore-
rotors, and similar mechanical components, to a casting the mechanical strength of the coupling. In
shaft, and for locating ball and roller bearings. The Fig. 1(a) an infinitely long, solid cylindrical shaft is
range of tolerances on the fitted members must be press-fitted into a keyless cylindrical hub whose bore
selected so that a minimum interference allowance edges are rounded; the indentation of the filleted edges
gives a sufficiently tight assembly, whereas the maxi- of the hub bore into the shaft surface produces loca-
mum interference will not produce excessive stres- lized stress peaks, examined in this paper. In Fig. 1(b)
ses, see for example references [1, 2]. To confidently a solid shaft with filleted extremities is partially
evaluate the contact stresses, it is generally unneces- inserted into a keyless hub; the indentation of the shaft
sary to account for the whole geometry of the com- rounded extremity on the web bore surface promotes
ponent press-fitted onto the shaft; most times, only local stress concentrations, analysed in this paper. The
the part surrounding the shaft has to be considered, configuration of Fig. 1(a) is considered first.
see for example references [3]; often this part may be The contact stresses that take place in an axisym-
modelled as cylindrical, that is, it is essentially axi- metric shaft–hub press-fit illustrated in Fig. 1(a)
symmetric, and it exhibits an approximately rectan- have been addressed in references [1–29], whereas
gular cross-section; it is referred to as the hub. the parallel plane situation of a rectangular com-
Figure 1(a) illustrates the hub and the shaft. ponent with rounded extremities indenting a half-
plane has been considered in reference [30], p. 49,
and in references [31–34]. The available salient
*Corresponding author: Engineering Faculty, University of information on the contact stresses in a shaft–hub
Modena and Reggio Emilia, Via Vignolese 905, Modena 41100, press-fit is summarized in the following.
Italy When a cylindrical shaft of infinite length is
email:antonio.strozzi@unimore.it press-fitted into a keyless cylindrical hub of finite

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
2 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

opposed to polar), detailed, specific analysis is gener-


ally needed to evaluate the local stress peaks and to
assess the hub strength.
The configuration of Fig. 1(b) is considered in the
following. This time the solid shaft is only partially
inserted into the keyless hub, so that the filleted extre-
Fig. 1 Complete insertion (a) and partial insertion (b) mity of the shaft produces stress concentrations that
of the shaft into the hub depend on the fillet radius. There is little information
at the disposal of the engineer on these stress concen-
trations. To the authors’ best knowledge, the only
contributions dealing with the problem of Fig. 1(b)
axial length, stress concentrations take place at the
are reference [11], which specifically deals with the
hub–shaft contact extremities, whereas the contact
shaft insertion mechanism, and reference [37], where
stresses remain reasonably constant along the cen- Fig. 8 addresses an incomplete insertion of the shaft.
tral portion of the contact, see for example refer- For both the configurations (a) and (b) of Fig. 1
ences [5, 8, 21, 22, 28]. If the edges of the hub bore and for a specific shaft–hub press-fit geometry, the
are sharp, infinite elastic pressure peaks occur at stress peaks may be accurately evaluated with a
the contact extremities reference [10]. Conversely, finite element (FE) analysis. It is however difficult to
pressure lateral bumps of finite intensity take place numerically retrieve stress forecasts of ample valid-
when the edges are rounded, and the contact pres- ity, since the level of the contact pressure peaks
sure becomes null at the contact extremities refer- depends on the shaft radius, the hub outer radius,
ence [22]; the Hertzian-type lateral bumps and the the fillet radius, the interference, and the hub and
central flattish zone confer to the contact pressure shaft Young’s moduli. In addition, the title contact
distribution a camel-backed profile. Consistent with problem is progressive, since the hub–shaft contact
the Hertzian nature of the pressure bumps, the length in the shaft axial direction increases, albeit
maximum equivalent stress within the hub does not moderately, with the interference imposed. Con-
occur at the hub bore surface, but at an internal sequently, this contact problem is non-linear, see
point. To introduce fillets at the hub bore edges is for example [38], and this aspect jeopardizes the
therefore a practically valuable means for weaken- feasibility of a normalization procedure with respect
ing the stress peaks. (In reference [7] the contrasting to the imposed interference.
photoelastic result is obtained that ‘the radial and In this paper a viable way to numerically prepare
tangential stresses are maximum . not at the edge design charts addressing the contact stress peaks
of the thick-walled cylinder as commonly assumed.’ in a hub with rounded edges press-fitted onto a
In reference [24] the contrasting opinion is stated solid shaft, is presented for the two configurations
that ‘the corner radius has a negative effect on con- detailed in Figs 1(a) and (b). In fact, it is shown in
tact stress at edges.’) Additional methods adopted to this study that, for a prescribed ratio between the
relieve the pressure peaks consist in introducing a hub inner and outer radii, the stress concentrations
shoulder in the shaft or in the hub, or in in the presence of radiused edges of the hub bore
employing a grooved hub references [2, 12, 16, 21, (see Fig. 1(a)) and their analogues due to the
35], or in imposing an interference that varies in the incomplete indentation of the rounded shaft extre-
axial direction [36]. mity on the hub bore surface (see Fig. 1(b)) depend
The stress state along the hub central portion may upon a single normalizing parameter that com-
be thoroughly predicted by modelling the press-fit prises and summarises the simultaneous effects
problem as plane and by employing the Lamé equa- of a reference radius for the shaft–hub press-fit (for
tions for thick-walled cylinders, see for example example, the shaft radius), the fillet radius, the
references [16, 21, 27]; conversely, with regard to the interference, and Young’s modulus of the two
evaluation of the localized pressure peaks, the pres- press-fitted components.
ent authors are unaware of manageable analytical The above normalizing parameter has here been
formulae of ample validity or of easily accessible dia- determined by following routes analogous to those
grams in the literature. (The analytical studies of [10, successfully adopted in references [39–42], referring
31, 33] attest to the complexity of analytical solu- to pin-in-plate contact problems, and to a press-fit
tions.) Consequently, the transmissible torque may problem. To this aim, the contact problem expres-
be confidently estimated by resorting to the Lamé- sed in the title is formulated in terms of an integral
based plane solution, that is valid along most of the equation, and the normalization of its right- and
contact axial length; conversely, a cylindrical (as left-hand sides evidences the combination of the

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
Normalization of the stress concentrations 3

key variables that constitutes this normalizing para- 2 SIMPLIFYING ASSUMPTIONS


meter. It is additionally noted that the above-refer-
enced examples of definitions of a normalizing It is assumed that the rounded edges of the hub
parameter deal with problems cast in polar coordi- bore, see Fig. 1(a), as well as the filleted shaft extre-
nates, whereas the title problem refers to cylindrical mity, see Fig. 1(b), are described by a quarter of cir-
coordinates. cumference. Consequently, the fillet radius fully
The formulation of the title problem in terms of describes the geometry of the mating surfaces at the
an integral equation serves to define the normaliz- hub–shaft contact.
ing parameter; however, since the integral equation With reference to the configuration of Fig. 1(a), the
approach is quite complex, the stress field is more shaft is assumed to protrude from the hub by a size-
conveniently evaluated with the FE method, by able length, so that the effect on the shaft–hub contact
varying the normalizing parameter within an inter- stresses of the shaft extremities is insignificant with
val covering practically significant keyless shaft–hub respect to an infinitely long shaft, see for example [5].
press-fit configurations. By extending the results of [35], addressing a plane
When a hub is press-fitted onto a shaft, the hub indentation problem, to an analogous cylindrical situ-
expands radially, whereas the shaft contracts radially, ation, it may be deduced that a shaft long twice the
see for example reference [20]. Consequently, the hub axial thickness provides essentially the same
hoop stresses are (generally) tensile in the hub and results as an infinitely long shaft. Comparable conclu-
compressive in the shaft. Since most materials emplo- sions have been derived in reference [28]. In the
yed in mechanical elements exhibit higher strength in photoelastic study of reference [7], a shaft of four
compression, the strength of the press-fitted compo- times the hub axial length has been considered.
nents is nor-mally assessed by examining the stresses Analogously, with reference to Fig. 1(b), the shaft end
within the hub rather than within the shaft. However, is supposed to be sufficiently far from the two hub
the above considerations are strictly valid when the walls perpendicular to the shaft axis, such that the
shaft–hub contact problem may be modelled as plane, stress concentrations caused by the indentation of the
i.e. when no contact pressure bumps take place at the filleted shaft extremity on the hub bore surface are
shaft–hub contact extremities. This situation occurs in essentially independent of
practice when the shaft and the hub possess identical the above distance.
axial lengths, see for example reference [6]. If, instead, Again with reference to Fig. 1(a), the hub is
the shaft protrudes from the hub cavity, this contact assumed to be sufficiently long to produce, in its
problem becomes cylindrical (as opposed to plane), central part, a uniform stress state in the axial direc-
and it is not immediately obvious (a) whether the tion; this situation implies that the two pressure
maximum equivalent stress occurs within the hub, bumps at the contact extremities do not interact.
and (b) whether the hub hoop stress remains always Consequently, both the stress level along the central
tensile along the shaft–hub contact length. With regard
zone of the assembly and that of the lateral pressure
to problem (a), for all the situation examined in the
bumps become essentially independent of the hub
diagrams reported in section 4 it was numerically
axial length and, therefore, only the hub portion in
found that the maximum equivalent stress occurs
the vicinity of one of its axial extremities may be
within the hub and not within the shaft. With regard
considered in modelling the shaft–hub contact
to problem (b), it was numerically clarified that the
problem. In reference [6] it is noted that, for a nearly
hub hoop stress becomes generally negative in the
square hub cross section, the hub extent in the axial
regions where the hub equivalent stress is maximum.
direction is sufficient to guarantee that ‘increasing
This surprising result is detailed and rationalized in
its length further would not have affected the stres-
section 4.2.
ses in the region of principal interest’, and that ‘[the
With the aid of finite elements, in this paper
stresses] fall close to the common Lamé values at a
design diagrams are presented that express the nor-
distance of about 0.1 [times the shaft diameter]
malized elastic stress concentration factor within
from the edge.’ The results of Fig. 5 of reference [22]
the hub versus the above normalizing parameter
indicate that, for a hub whose axial length is 1.6
for a selection of ratios between the hub inner and
times the shaft radius, a sizeable central zone exists
outer radii. Such diagrams assist the designer in
in which the contact pressure remains uniform in
choosing the most convenient fillet radius for the
the axial direction, and the level of the lateral pres-
rounded edges of the hub, and in assessing the hub
stress level in a keyless shaft–hub press-fit and the sure bumps does not essentially change as the
possible outcome of yielding. hub axial length is increased. In reference [32] the

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
4 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

pressure lateral bumps are interpreted in the light of the hub–shaft contact in the axial direction is often
the Hertzian theory, and it is shown that their extent undefined, since it depends on the assembly proce-
in the axial direction may be estimated by modelling dure. In reference [5] it is noted that, with regard
the hub lateral, rounded portions as two cylinders to a press-fit problem, ‘the radial pressure distribu-
whose radius equals that of the fillet of the hub tion changes little from the frictionless to the fully
bore. According to this modelling, the axial extent of adherent case’. In reference [48] a shrink-fit problem
each lateral bump describing the shaft–hub contact is studied in the assumption of partial slip. Figure 7
pressure is of the order of twice the fillet radius. (In of this study indicates that shaft–hub sticking occurs
reference [11], Fig. 2.7, a shaft–hub press-fit config- along the central part of the contact, whereas the
uration is considered in which the hub axial length mating surfaces slip along the contact lateral zones,
is very small; in this case, the lateral bumps inevita- where the contact pressure bumps take place.
bly interact.) The axial length of the lateral pressure The shear stresses connected to the torque trans-
bumps is reconsidered in section 4.3. mission have been neglected. Such shear stresses
The shaft is supposed to be solid. In [23] p. 8–405, are deemed not to modify the contact pressure
it is noted that, when the Lamé plane solution for referring to a frictionless contact, as they alter the
thick-walled cylinders is employed, the pressure angles and not the radial lengths.
curves valid for a solid shaft are accurate to 5 per The shaft is supposed not to be subjected to
cent even if the shaft is hollow, provided the inside bending, see for example [13]. In addition, the shaft
shaft diameter is not over 25 per cent of the outside. is assumed as static, and the inertial forces due to
A similar error is quoted in reference [3]. In refer- the shaft rotation are neglected, see for example
ence [15] it is found that ‘the value of the inner dia- references [27, 29].
meter of the shaft is of no consequence as long as
the ratio of shaft outer diameter to shaft inner dia-
3 DERIVATION OF THE NORMALIZING
meter is more than 2.5 to 3.’ In references [8] and
PARAMETER
[28], it is observed that in a hollow shaft the stress
concentration is higher than that in a solid shaft.
In this section the normalizing parameter men-
This surprising result is rationalized in [8] by noting
that the outer surface of a hollow shaft deflects in a tioned in section 1 is determined by formulating the
concave fashion, and this bending action transfers frictionless shaft–hub contact problem of Fig. 1(a) in
terms of an integral equation. An example presented
load to the extremities of the contact region.
The materials of the shaft and of the hub are in tabular form evidences the capabilities of the nor-
assumed to behave elastically and to exhibit the malizing parameter.
same Young’s modulus. The approach favoured in
this paper may easily be extended to situations in 3.1 Formulation of the title problem in terms of
which the shaft and hub materials are different. an integral equation
However, the compilation of specific charts addres- It is convenient to address first the simpler situation
sing selected ratios of the two Young’s moduli depicted in Figs 2(a) and (b). The integral equation
would require a considerable number of FE runs (1) expresses the imposition that the contact pressure
and, therefore, this aspect has not been covered for must move the mating surfaces of the shaft and of
the time being. (In reference [25] it is observed that, the hub apart until they do not overlap, under the
although plastic working is not generally admitted restriction that the contact pressure be positive.
by standards, the norm DIN 7190 tolerates plastic The unknown function is the contact pressure p,
regime up to one-third of the hub diameter.) appearing under integration at the left-hand side of
The shaft–hub contact is modelled as frictionless. the integral equation (1); the right-hand side is the
This limit assumption is critically discussed in refer- known term, representing the initial clearance/over-
ence [6] together with the complete adhesion ideali- lapping between the two undeformed mating sur-
zation. The assumption of zero friction is employed faces of the hub and of the shaft
in, for example, references [10, 13, 31, 43], and is
ð +a pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
supported by the following considerations: (a) in I
kðx, y Þpð y Þdy =  r + r 2  x2 (1)
this kind of application the coefficient of friction for a 2
steel components is sufficiently low, rarely exceed-
ing 0.2, see for example references [2, 3, 11, 44, 45] In equation (1), x and y are axial coordinates (see
(in the case of light materials the coefficient of fric- Fig. 2), k represents the Green function describing
tion may reach 0.5, see for example reference [46, the relative radial displacement at the x coordinate
47]); (b) the direction of the shear stresses acting at of the two mating profiles for a collar of a (unit per

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
Normalization of the stress concentrations 5

 
ð þa k x
, y    
ri ri ri y y I x2 (4)
p ri d = 
a
ri
E ri ri 2 2r

The following passages aim at normalizing the two


members of the integral equation (4). The contact
pressure p may be normalized with respect to the
reference contact pressure p0, computed by model-
ling the press-fit problem as plane and by employing
the Lamé equations for thick-walled cylinders. The
reference contact pressure p0 describes the contact
pressure along the central portion of the actual
shaft–hub press-fit, and it may be expressed in terms
of the diametral interference I and of the hub radial
Fig. 2 The shaft-hub contact problem
aspect ratio as follows (see for example [27])
unit circumferential length) linear concentrated  
EI 1 EI ri2
force applied at the y coordinate (in the integral p0 = = 1 2
2ri 1 + ri2 + ro2 4ri ro
equation (1), the y coordinate is superposed to the x ro2 ri2
     (5)
coordinate, and it is a dummy (internal) variable), a EI 1 r2 EI ri
is the contact semiwidth (see Fig. 2(b)), I indicates = 1  i2 = f
ri 4 ro ri ro
the diametral interference, ri is the hub inner radius
(it nominally coincides with the shaft radius), ro is
where f denotes a function of the (dimensionless)
the hub outer radius, and r is the radius of the
hub radial aspect ratio ri/ro. By normalizing p over
rounded profile. Since p represents the contact pres-
p0, the integral equation (4) may be recast as
sure, p(y)dy may be interpreted as a force per unit
circumferential length, distributed along a collar of      
ðþa k x
, y
p ryi   r I  x2
infinitesimal axial length dy. ri ri ri y i 2 2r
ri d =   (6)
It is convenient to simplify the right-hand side of ra E p0 ri EI f rroi
i
equation (1), by employing the classical parabolic
approximation adopted in Hertzian contacts (see for
example [22, 31, 33]) (Consistent with the above normalization, in the
definition of the stress concentration factor Kt, the
ð +a contact pressure p0 has been assumed as the nor-
I x2
k ðx, y Þpð y Þdy =  (2) malizing parameter.) Consequently
a 2 2r
 
Since the title problem is essentially cylindrical  p y   I x 2

ðþ a y ri

(the shaft is cylindrical, and the hub is cylindrical, ri k x


, d ryi = I 2f 2rri
ri ri

p0 ðr Þ
apart from the bore radiused edges), for a solid shaft  2 h 2 i o (7)
ra 1 x ri
i = 2f ð
ri 1
and for a hub of prescribed radial aspect ratio ri/ro ro Þ ri rI

the Green function k is expected to be independent


of the shaft radius (recalling that reference is made An attentive examination of the normalized inte-
to a unit force per unit circumferential length), pro- gral equation (7) reveals that, if two press-fit prob-
vided that coordinates normalized with respect to lems possess the same hub radial aspect ratio ri/ro
the shaft radius ri are employed, namely x/ri and and the same normalized radius of the rounded pro-
y/ri. Moreover, the Green function is inversely pro- file r/ri, but they differ in the values of the shaft
portional to the Young’s modulus E common to the radius, ri, of the radius of the rounded profile, r, and
two materials. Therefore of the diametral interference, I, and if the ratio
2

  ri =ðrI Þ is the same for the two cases, then the nor-
k x y malized contact pressure p/p0 and, more generally,
k ðx, y Þ = , (3)
E ri ri the normalized stress state is the same for the two
cases, and the normalized contact width, defined by
2
where k is a normalized Green function referring a/ri, is the same. Consequently, F = rI=ri is a nor-
to a unit Young’s modulus. For a prescribed hub malizing parameter, summarizing the simultaneous
radial aspect ratio ri/ro, the integral equation (2) effects of the radius r, the diametral interference
thus becomes I, and the shaft radius ri. (The reciprocal of the

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
6 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

parameter previously reported has been used in the


definition of F, since it is more convenient that the
radius r of Fig. 2(a), which describes the fillet radius
of Fig. 2(c) and may become very small for nearly
sharp edges of the hub bore, appears at the numera-
tor.) In fact, for a prescribed hub radial aspect ratio
ri/ro, and a normalized radius of the rounded profile
r/ri, it is possible to prepare design diagrams report-
ing the normalized von Mises peak stress within the
hub and the normalized contact semiwidth a/ri, ver-
sus the normalizing parameter F. The Young’s mod-
ulus is not included into the parameter F, but it
affects the reference pressure p0, and it manifests
itself when p is computed as (p/p0) 3 p0.
The approach employed to analyse the simplified
configuration of Fig. 2(a) may immediately be
extended to Figs 2(c) and (d), that model the prob-
lem of Fig. 1(a), where r is the fillet radius. In fact,
its modelling in terms of two integral equations is
ð +a Fig. 3 Kt versus F for the geometry of Fig. 1 (a) for
I ri/ro= 0.5, and for a selection of r/r ratios
kðx, y Þpð y Þdy = <0
2
ða
+a pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (8)
I
kðx, y Þpð y Þdy =  r + r 2  x2 0<x< þ a
a 2 insensitive to r/ri, provided that the considered
interval of the normalized fillet radii is not too
to which the above procedure may be applied, thus wide. In fact, Fig. 3 reports the FE predictions for
confirming that the normalizing parameter F is the stress concentration factor Kt, defined as the
suitable for the problem of Fig. 1(a) too. von Mises peak stress within the hub, normalized
In addition, the integral approach may be with respect to the reference contact pressure p0,
employed to define the normalizing parameter F for versus the normalizing parameter F, for the com-
the configuration of Fig. 1(b), where this time r monplace hub radial aspect ratio ri/ro = 0.5, and
represents the radius of the filleted extremity of the for a selection of values of the normalized fillet
solid shaft. Space restrictions do not allow the pre- radius r/ri, namely 0.0025, 0.005, 0.01875, 0.025,
sentation of the mathematical details for this case. 0.05, 0.1, 0.15, 0.2.
Returning to the problem of Fig. 1(a), it has The selection of p0 as the normalizing parameter
already been underlined that the diagrams reporting in Kt has been discussed with reference to equation
the normalized von Mises peak stress within the (6), and it is consistent with the requirement that
hub versus the normalizing parameter F are strictly the normalizing parameter be easily computable.
valid for prescribed values of the hub radial aspect The present authors are however aware of the fact
ratio ri/ro and of the normalized fillet radius r/ri. that other authors use a different normalizing para-
The following observations clarify the role of the meter, see for example [28].
parameter r/ri. Since the hub bore edge is rounded, Figure 3 confirms that, for a prescribed value of
the axial length of the shaft–hub contact is lower F, Kt increases with r/ri, as a result of the stiffening
than the hub length. The hub lateral axial portions effect of the hub projections beyond the contact
that are not subjected to the contact pressure may zone. However, if r/ri is comprised within the tech-
be interpreted as hub projections beyond the con- nically sufficiently wide interval 0.025–0.15 (see sec-
tact zone, and their presence increases the hub tion 4.1), the maximum error between the two Kt
radial stiffness. Since the filleted edge of the hub curves is about 30 per cent, the minimum error is
bore is assumed to be described by a quarter of 12 per cent, and in most practical cases the error is
circumference, this stiffening effect is expected to within 20 per cent.
increase with the normalized fillet radius for pre- Since the compilation of design diagrams
scribed values of F and ri/ro. However, a FE investi- depending upon the two independent parameters
gation shows that, while the normalized peak stress ri/ro and r/ri would require a considerable number
strongly depends on F and on ri/ro, it is relatively of FE runs, it was decided to accept the error

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
Normalization of the stress concentrations 7

Table 1 Comparison between test cases with different pressure p0, that incorporates the Young’s modulus
geometries characterized by the same value effect.
of the coefficient F. Table 1 confirms the analytical expectations
according to which, for different radial dimensions
Case 1 Case 2
of the shaft and the hub, different fillet radii and ini-
r 1 mm 0.375 mm tial clearances, different Young’s moduli, and for
ri 20 mm 15 mm
ro 40 mm 30 mm
different normalized fillet radii r/ri, albeit falling
r/ri 0.05 0.025 within a sufficiently restricted interval, but for a
ri/ro 0.5 0.5 fixed hub radial aspect ratio ri/ro, the normalized
I 0.04 mm 0.06 mm
F = rI=ri
2
0.0001 0.0001 maximum contact pressure, pmax/p0, the hub nor-
E 210 000 MPa 100 000 MPa malized hoop stress, sc/p0, the normalized maxi-
po = EI*(1- ri2/ro2) /(4 ri) 78.75 MPa 75.00 MPa
pmax 981.40 MPa 825.20 MPa
mum equivalent stress within the hub, Kt = seq,max /
seq,max,hub 612.30 MPa 530.80 MPa p0, and the normalized contact width, a/ri, are rea-
sc,hub| where seq, is maximum –155.4 MPa –120.0 MPa sonably similar for Cases 1 and 2, provided that the
pmax/po 12.46 11.00
sc,hub/po –1.97 –1.60 normalizing factor F is the same.
Kt = seq,max /po 7.78 7.08 The results reported in Table 1 also evidence that
a 0.022 mm 0.014 mm the contact width a is poorly approximated by a
a/ri 0.0011 0.0009
Winkler-type formula, see for example [49],
obtained by putting x = a in the right-hand side of
incurred in neglecting the influence of r/ri, and to equation (2), and by imposing that this termpbe ffiffiffiffi null.
prepare design diagrams reporting the stress con- This approach provides the expression
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi a = Ir . For
centration factor Kt versus the normalizing para- Case 1 of Table 1, a = 0:0431 = 0:2mm, which con-
meter F, computed for selected values of the radial siderably differs from the FE forecast a = 0.022 mm.
aspect ratio ri/ro, and for the specific value of the The differences of about 12 per cent, 5 per cent,
normalized fillet radius r/ri = 0.05, that in Fig. 3 is and 9 per cent between the two values of pmax/p0, of
straddled by the two curves referring to r/ri = 0.025 sc/p0, and of Kt, respectively, for the two cases of
and to 0.15. Consequently, if r/ri is comprised with- Table 1 are imputable to five causes, namely (a) the
in the technically sufficiently wide interval 0.025– errors incurred in the FE numerical solution; (b) the
0.15, the maximum error is halved, and it becomes fact that in contact problems the large deformation
15 per cent. The numerical examples of section 5 theory is adopted in the FE solution, whereas the
confirm this forecast. normalization upon which F relies, is valid in linear
elasticity, see for example reference [50]; (c) the
parabolic approximation of the circular fillet, adopt-
3.2 Numerical validation of the normalizing
ed in equation (2); (d) the idealization of the hub
parameter F
cross-section in terms of a rectangle. In fact, the
With reference to the frictionless contact problem of normalization developed in this paper requires that
Fig. 1(a), a detailed numerical example is presented the Green function refer to a rectangular hub cross-
in tabular form to evidence the normalizing poten- section, i.e. it be exactly the same independent of
tials of the parameter F. A mechanical analysis of the value of the fillet radius, whereas in the FE solu-
two different shaft–hub assemblies, named Case 1 tion the equivalent of the Green function refers to
and Case 2, has been carried out with finite ele- the actual geometry of the hub cross-section, and,
ments and the numerical forecasts are collected in therefore, it (moderately) changes with the fillet
Table 1. The shaft radius ri (nominally coinciding radius; (e) the influence of the different values of
with the hub bore radius), the hub outer radius ro, the normalized fillet radius r/ri for the two cases
the fillet radius r, the diametral interference I, and analysed in section 3.1 with regard to Fig. 3. The
the Young’s modulus E, are different for the two cas- above errors are consistent with the comments on
es. In addition, the normalized fillet radius r/ri is Fig. 3 in section 3.1.
different for the two cases. However, the hub radial The discussion of the five sources of error refers
aspect ratio ri/ro and the normalizing coefficient F to the situation illustrated in Fig. 1(a). Moving to
are the same for the two cases. Fig. 1(b), the source of error (e) is negligible, since
In Table 1 the maximum contact pressure, pmax, the hub is assumed to be infinitely long, and the sol-
the hub maximum equivalent stress, seq,max, and the id shaft is much stiffer than the hub, see for example
hoop stress within the hub, sc, computed where the reference [16], p. 698. Consequently, the predictions
hub equivalent stress is maximum, have been nor- referring to Fig. 1(b) are generally more accurate
malized with respect to the reference contact than those dealing with Fig. 1(a), as confirmed by

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
8 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

equation approach is too complex to be employed


in the evaluation of the hub stress field; finite ele-
ments are a more convenient means for forecasting
such stresses. The design charts are prepared by
varying the normalizing parameter within an inter-
val covering technically significant shaft–hub press-
fit configurations. From a practical viewpoint, it is
sufficient to vary only one of the three variables
constituting the parameter F; it is particularly con-
venient to keep fixed the nominal geometries of the
hub and shaft, and to modify the interference alone.
This straightforwardness in preparing the diagrams
stems from the summarizing capabilities of F.

4.1 Design recommendations for the shaft–hub


press-fit
To correctly define the range of the normalizing
parameter F covering practically relevant situations
Fig. 4 Contact pressure for the two cases of Table 1 of the title problem, in this subsection the available
recommendations are collected that address the
optimal radial aspect ratio for the hub, the interval
of the fillet radii adopted for the hub bore edges,
the numerical examples presented in section 5. In and the selection of the shaft–hub interference.
fact, the numerical forecasts reported in Fig. 3, deal- The hub radial aspect ratios ri/ro considered in [4,
ing with the geometry of Fig. 1(a) and with various 6, 21, 24, 43] range between 0.33 and 0.75 in stan-
normalized fillet radii, show that, as the normalized dard mechanical applications. In [19], it is demon-
fillet radius is increased, the normalized maximum strated that the optimal aspect ratio between the
equivalent stress within the hub approaches the hub inner and outer radii, that maximizes the load
forecasts dealing with Fig. 1(b). This result may transmitted by the joint, falls within the range 0.5–
be rationalized by observing that the problem of 0.7. In the present paper the interval of aspect ratios
Fig. 1(b) may be interpreted at that of Fig. 1(a), in comprised between 0.3 and 0.7 has been investi-
which the hub projections beyond the contact zone gated. (Hubs of radial thickness smaller that those
(see section 3.1) become infinite. covered in the diagrams here presented are met in
It is difficult to examine separately the above five dental implants, see for example [51].)
sources of error. The numerical examples of Table 1 The radius of the rounded edge of the hub bore is
and those presented in section 5 indicate that in addressed in the following. By considering [22, 24],
most cases the global error between the FE forecast reference [35] referring to a plane indentation, and
and the design curves of Figs 7 and 8 is within, say, the ball bearing realm, the inference may be derived
10 per cent, see also the maximum error estimates that the fillet radii adopted for the hub edges gener-
presented in section 3.1. An applicability limit valid ally range between 0.025 and 0.15 times the shaft
for the analogous Hertzian theory, expressed in radius, and that the widest technical interval is
terms of the ratio between the contact width and 0.0025<r/ri<0.2.
the fillet radius, is reported in reference [14], p. 57. The selection of the shaft–hub interference is
Figure 4 details the contact pressure bumps for addressed in reference [23], p. 8–405, where the
the two cases of Table 1. Figure 5 shows the FE out- ANSI B4.2-1978 standards are reported. Following
put in terms of von Mises equivalent stress for Case an order of increasingly demanding applications,
1 of Table 1. the interference fits are classified in three groups,
namely locational interference fits, medium drive
fits, and force fits. For the first group (employed for
4 DESIGN CHARTS locating ball and roller bearings), and for steel shaft
diameters in the region of 20 mm, the average dia-
The formulation of the title problem in terms of metral interference I is about 1/1000 the shaft
an integral equation has allowed the normalizing diameter; for the second group the coefficient
parameter to be determined; however, the integral becomes 1.5/1000, and for the third group the

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
Normalization of the stress concentrations 9

Fig. 5 FE output in terms of von Mises equivalent stress for Case 1 of Table 1

coefficient is 2/1000. Consistently, in [18] the rule of 4.2 The sign of the hub hoop stress in the
thumb is quoted, according to which the diametral vicinity of the bore edges
interference I must fall within the interval 1/1000–2/ It has already been noted in section 1 that, consis-
1000 the shaft diameter. Analogously, in reference tent with the Lamé equations for thick-walled cylin-
[4] the interval suggested is 1/1000<I/(2ri)<3.75/ ders, the hub radial stresses are compressive
1000. whereas the hoop stresses are tensile along the cen-
The ratio I/r does not appear to have been direct- tral part of the hub bore surface. Moving to the hub
ly addressed in the quoted references. From the pre- lateral zones, the presence of infinite pressure peaks
viously reported intervals 0.025<r/ri<0.15 and 2/ when the bore edges are sharp is responsible for the
1000<I/ri<4/1000, it may be deduced that 0.013<I/ hub hoop stresses becoming locally negative. This
r<0.16, but it is difficult to suggest an optimized interesting result may be rationalized with the aid of
value of this parameter. The applicability limit a/ Hooke’s law, according to which ‘the circumferen-
r<0.1 is reported in reference [14], p. 57 with regard tial strain is proportional to the difference between
to the Hertzian theory, where a may be interpreted hoop stress and the Poisson’s ratio times the radial
as the contact semiwidth. The following rough esti- stress, the axial component being negligible. When
mates may be made. Accordingpto ffiffiffiffi a Winkler-type the radial stress tends to minus infinity, this rela-
model, e.g.
pffiffiffiffiffiffiffi reference [49], a = Ir . Consequently, tionship requires the circumferential component to
a=r = I=r . By employing the previous numerical approach minus infinity as well, in order to keep the
bounds, the inequality a/r<0.4 should hold, which hoop strain finite’, see reference [52].
does not respect the applicability limit a/r<0.1. A more physical explanation of the outcome of
However, from the discussion of the results of compressive hoop stresses at the hub bore edges is
Table 1, it emerges that the Winkler model notice- presented in the following. The hub’s most stressed
ably overrates the contact width, up to ten times, so zones fall in the vicinity of the bore edges, whereas
that the applicability limit favoured in reference [14] the remaining hub regions are relatively under-
should be verified for most cases. stressed, see Fig. 6(a). This situation may be idea-
From the results collected above, it may be con- lized by describing the understressed regions as
cluded that the interval of the normalizing parameter rigid, since their strains are small; conversely, the
F = comprised between 2 3 10-6 and 10-2 generously most stressed hub zones are idealized as deformable
covers the commonest practical applications. rings, whose inner radius is loaded by the local

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
10 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

Fig. 6 Stress concentrations at the hub bore edges: (a)


real geometry, (b) idealization

contact pressure p, whereas their outer radius is


radially restrained by the hub understressed zones,
see Fig. 6(b). The axial stresses within the ring may
be assumed to be null, as a result of the proximity
of the most stressed zones to the lateral free surface.
By computing the stresses within the deformable
ring with the Lamé equations, it may be shown
that, for small radial thicknesses of the ring, the
ring hoop stress is compressive, see reference [16],
p. 705. Fig. 7 Kt versus F for the geometry of Fig. 1 (a)
The outcome of compressive hoop stresses at the
hub bore edges is photoelastically confirmed in [53], with F for a prescribed value of ri/ro, whereas it
addressing the insertion of a conical stem into the increases with ri/ro for a prescribed value of F. In
conical cavity of a head in a hip replacement; it is addition, for small (large) values of F, the hub cir-
noted that ‘at the ends of the taper contact there cumferential stress becomes negative (positive).
were pressure concentrations and compressive Consequently, for decreasing values of F, two con-
hoop stresses’. flicting trends coexist: (a) an increase of Kt takes
When the hub bore edges are no longer sharp but place, and this fact is negative from the viewpoint of
rounded, the contact pressure peaks are mitigated the mechanical strength of the coupling; (b) a nega-
together with the trend of the hub hoop stresses tive hoop stress arises within the hub, and this
towards negative values. aspect is positive, since most materials exhibit high-
The published FE forecasts reflect only qualita- er strength in compression. Therefore, each config-
tively the local trend of the hub hoop stresses uration constitutes a compromise between these
towards negative values, as a result of the weak two conflicting aspects. It is however difficult to
character of this numerical method and of the devise an optimal value of the fillet radius, one of
coarse meshes unavoidably employed in the past the reasons being the fact that most configurations
years. In fact, in references [17] and [52] it is are imposed by technological requirements and by
numerically found that the hub hoop stress sharply the designer personal background.
decreases in the vicinity of the contact extremity,
albeit without attaining negative values. Similarly,
4.3 Compilation of the design diagrams
Fig. 8 of reference [20] shows that the numerical
hoop stresses along the internal surface of the hub, The commercial FE program MSC Marc 2007r1 has
positive at the hub central part, drastically diminish been employed in this study. The problem has been
at the contact extremities, and for some configura- modelled as axisymmetric. The mesh adopted was
tions they become negative, albeit moderately so. very fine, of about 80 000 nodes, to ensure that the
Since, for an imposed hub radial aspect ratio ri/ro, number of nodes defining the progressive contact
the elastic stress profile in the hub depends nearly zone under the imposed interference be constantly
exclusively on the normalizing parameter F (see the higher than 10, see reference [54]. The mean dimen-
comments with regard to Fig. 3), it is now possible to sion of the elements modelling the hub/shaft filleted
determine which range of F produces negative hoop zones is 2.5 3 1025 3 ri. In some test cases, an even
stresses in the vicinity of the hub bore edges. The finer mesh was also employed to assess the numeri-
corresponding results are presented in Figs 7 and 8. cal convergence. The interference at the frictionless
From the diagrams of Figs 7 and 8, it emerges unilateral contact interface has been mimicked by
that the stress concentration factor Kt decreases imposing a thermal expansion to the shaft. In order

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
Normalization of the stress concentrations 11

to preclude any undesired discretization disalign- Similarly, Fig. 8 refers to the geometry depicted in
ment at the mating surfaces as a result of the shaft Fig. 1(b). To get forecasts independent of the length
axial expansion, a fictitiously orthotropic thermal by which the hub protrudes from the shaft
expansion coefficient has been adopted. extremity, a length higher than the shaft diameter
The normalizing approach favoured in this study has been adopted in the numerical simulation.
requires that the hub radiused zone not be Formula (10) for Kt interpolates the forecasts of
described by the exact expression employed in Fig. 8 with a maximum error of about 5 per cent,
equation (1), but by its parabolic approximation of which remains lower than 3 per cent for most
equation (2), which therefore has been adopted in applications
numerically modelling the two mating profiles.
h i
Diagrams have been prepared for the two geome-
F ðri =ro , FÞ = 11:62  0:878 logF + 0:142ðlogFÞ2
tries of Figs 1(a) and (b) and for the following selec-
tion of radial aspect ratios ri/ro: 0.3, 0.4, 0.5, 0.6, 0.7. 3(ri =ro 0:5)3(ri =ro 0:3)
h i
Figure 7 refers to the geometry of Fig. 1(a). For each + 6:67 + 3:530 logF0:198 ðlogFÞ2
radial aspect ratio, the dimensions of the two contact-
ing bodies have been kept fixed and equal to those of 3(ri =ro 0:7)3(ri =ro 0:3)
h i
Case 1 of Table 1, whereas only the interference has + 0:58  1:906 log F + 0:095 ðlogFÞ2
been varied in such a way that the parameter F,
reported along the x-axis, covers the required interval 3(ri =ro 0:7)3(ri =ro 0:5)
between 2 3 10-6 and 10-2. To get forecasts indepen- Kt = e F ðri =ro , FÞ (10)
dent of the shaft length, in the numerical simulation
the shaft protrudes from the hub lateral walls by a
In formulae (9) and (10), log is the natural logarithm.
distance higher than the shaft diameter.
Finally, Figure 5, already referred to in section
It is noted that the fillet radius of Case 1 in
3.2, shows the FE output in terms of von Mises
Table 1, normalized versus the shaft radius, is a
equivalent stress for Case 1 of Table 1.
mid-value with respect to the interval of practically
Diagrams detailing the shaft–hub contact width
employed normalized fillet radii discussed in sec-
and the axial extent of the contact pressure bumps
tion 4.1. The adoption of this mid value allows the
are omitted for brevity. As an alternative, the follow-
maximum error due to the cause (e) discussed in
ing concise indications are supplied.
section 3.2 to be minimized.
For the geometry of Fig. 1(a), the axial extent of
The hub elastic maximum equivalent stress nor-
the pressure bumps may be conventionally defined
malized over the reference pressure p0 is reported
as follows. Moving from one of the shaft–hub con-
along the y-axis. Each curve is defined by 21 points.
tact extremities towards the contact centre, the con-
A bilogarithmic scale has been adopted. In Fig. 7 a
tact pressure is null at the contact extremity, it then
transition curve has been introduced to evidence
displays a sharp rise, then it declines and it gradu-
the zones in which the sign of the hoop stress, com-
ally approaches its central value p0, see Fig. 4. The
puted where the maximum elastic equivalent stress
axial extent of the pressure bumps is here defined as
takes place, is positive or negative.
the distance between the contact extremity identi-
Formula (9) for the elastic stress concentration
fied by a in Fig. 2(d) and the axial coordinate at
factor Kt interpolates the forecasts of Fig. 7 with a
which the contact pressure equals 1.05 p0. For all
maximum error of about 5 per cent, which remains
the examined geometries of Fig. 1(a) considered in
lower than 3 per cent for most geometries
Fig. 7, the axial extent of each contact pressure
bump rarely exceeds 3 times the fillet radius r of the
h i hub bore edge.
F ðri =ro , FÞ = 18:40 + 0:887 log F + 0:208 ðlog FÞ2 Moving to the geometry of Fig. 1(b), the axial
3(ri =ro 0:5)3(ri =ro 0:3) extent of the contact pressure bump is generally
h i much higher, and it may reach 20 times the fillet
+ 19:27 + 0:474 log F0:328 ðlog FÞ2 radius r of the shaft extremity. Consequently, a hub
3(ri =ro 0:7)3(ri =ro 0:3) of finite axial length cannot be modelled, numeri-
h i cally or analytically, by an infinitely long hub.
+ 5:30  0:823 log F + 0:140ðlog FÞ2 Moving to the stress concentrations, for small val-
3(ri =ro 0:7)3(ri =ro 0:5) ues of the normalizing coefficient F, Kt of Fig. 8 is
about twice that of Fig. 7, whereas for high values of
Kt = e F ðri =ro , FÞ (9) F the two stress concentrations become similar.

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
12 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

the maximum von Mises equivalent stress within the


hub according to Fig. 8 is 5.95 3 147.52 = 877.74 MPa.
The maximum von Mises equivalent stress forecast
by FE is 908 MPa, with an error of about 3 per cent
between the predictions based on the normalizing
coefficient F and the FE forecasts.

6 CONCLUSIONS

The elastic stress concentrations developed from


the keyless, frictionless, static press-fit of a solid
shaft into a hub have been addressed. Two config-
urations have been examined, namely (a) an infinite
shaft press-fitted into a hollow hub with bore
rounded edges, and (b) a shaft with filleted extre-
mity, partially inserted into a hub. Derived from an
analytical approach, a normalizing parameter has
been proposed that accounts for the combined
effects on the stress concentrations of the fillet
Fig. 8 Kt versus F for the geometry of Fig. 1 (b) radius, the shaft radius, the interference, and the
Young’s modulus. With the aid of finite elements,
various design charts have been compiled that
5 NUMERICAL EXAMPLES report the elastic stress concentrations within the
hub versus the proposed normalizing parameter.
Selected numerical examples are presented in this Each curve is valid for a fixed ratio of inner to outer
section, aimed at allowing the nonspecialist to cor- hub radii. Such charts assist the designer of a press-
rectly employ the design charts without excessively fitted assembly in selecting the most convenient fil-
dwelling on the underlying theory. let radius, in evaluating the maximum equivalent
The first example deals with locational interfer- stress, and in assessing its mechanical strength and
ence fits and geometries encountered in the ball the possible outcome of yielding.
bearing realm. The shaft–hub complete press fit illu-
Ó Authors 2011
strated in Fig. 1(a) is examined. The following values
are considered: ri = 5 mm, ro = 13 mm, r = 0.5 mm,
I = 0.0025 mm, E = 205 000 MPa. Consequently, ri/ro = REFERENCES
0.385 and F = 0.000 050. The elastic stress concen-
tration factor Kt predicted by the interpolating for- 1 Horger, O. J. and Nelson, C. W. Design of press-
and shrink-fitted assemblies. J. Appl. Mech., 1937,
mula (9) is 8.97. By employing expression (5), the
4, A183–A187.
reference contact pressure p0 is (205 000 3 0.0025)/ 2 Jordan, A. Designing interference fits. Machine
(4 3 5) 3 [1-(5/13)2] = 21.83 MPa. Consequently, the Design, 1974, 46, 68–72.
maximum von Mises equivalent stress within the 3 Baugher, J. W. Transmission of torque by means of
hub according to Fig. 7 is 8.97 3 21.83 = 195.85 MPa. press and shrink fits. ASME Trans., 1931, 53, 85–92.
The maximum von Mises equivalent stress forecast 4 Trock, B. Shrink fits: holding power can be
by FE is 212 MPa, with an error of about 7.6 per cent increased. Iron Age, 1953, 179, 179–175.
5 Conway, H. D. and Farnham, K. A. Contac stresses
between the predictions based on the normalizing
between cylindrical shafts and sleeves. Int. J. Engng
coefficient F and the FE forecasts. Sci., 1967, 5, 541–554.
The second example addresses an incomplete, 6 White, D. J. and Humpherson, J. Finite-element
force press-fit illustrated in Fig. 1(b) and defined by analysis of stresses in shafts due to interference-fit
the following values: ri = 12 mm, ro = 22 mm, r = 2 mm, hubs. J. Strain Analysis., 1969, 4, 105–114.
I = 0.048 mm, E = 210 000 MPa. Consequently, ri/ro = 7 Bhonsle, S. R. and Work, C. E. Evaluation of three-
0.546 and F = 0.000 667. The elastic stress con- dimensional stresses in shrink-fit problems by scat-
tered-light photoelasticity. Exp. Mech., 1970, 10,
centration factor Kt predicted by the interpolating 19N–28N.
formula (10) is 5.95. By employing expression (5), the 8 Mather, J. and Baines, B. H. Distribution of stress
reference contact pressure p0 is (210 000 3 0.048)/ in axially symmetrical shrink-fit assemblies. Wear,
(4 3 12) 3 [1-(12/22)2] = 147.52 MPa. Consequently, 1972, 21, 339–360.

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
Normalization of the stress concentrations 13

9 Bernasconi, G. Axial-symmetric elastic strains and 28 Croccolo, D. and Vincenzi, N. Stress concentration
stresses in tight cylindrical hub–shaft couplings. factors in compression-fit couplings. Proc. IMechE,
Meccanica, 1972, 7, 54–55. Part C: J. Mechanical Engineering Science, 2009,
10 Prasad, S. and Dasgupta, S. Axisymmetric shrink 224, 1143–1152.
fit problems of the elastic cylinder of finite length. 29 Luo, H, Liu, C., and Li, K. Contacting model con-
J. Elast., 1977, 7, 225–242. siderations of interference fit for shaft–hub system.
11 Selvage, C. C. Assembly of interference fits by Adv. Mater. Res., 2010, 118-120, 294–298.
impact and constant force methods. MSc Thesis, 30 Galin, L. A. Contact problems in the theory of elasti-
MIT, 1979. city. 1953 (Gostekhizdat, Moscow). English transla-
12 Grimm, T. R. and Chiu, A. C. Design of hubs to tion by the Department of Mathematics, North
minimize interference stresses – a finite element Carolina State College, Raleigh.
study. Comput. Engng, 1988, 3, 85–91. 31 Goodier, J. N. and Loutzenheiser, C. B. Pressure
13 Garnett, M. D. and Grimm, T. R. Finite Element peaks at the ends of plane strain rigid die contacts
analysis of interference-fitted shafts subjected to (elastic). J. Appl. Mech. 1965, 32, 462–463.
bending. Comput. Engng, ASME, 1989, 2, 273–280. 32 Strozzi, A. Static stresses in an unpressurized,
14 Schmelz, F., Seherr-Thoss, H. C., and Aucktor, E. rounded, rectangular, elastomeric seal. ASLE
Universal joints and driveshafts, 1992 (Springer, Trans., 1986, 29, 558–564.
Berlin). 33 Ciavarella, M., Hills, D. A., and Monno, G. The
15 Prasad, N. S., Sashikanth, P., and Ramamurti, V. influence of rounded edges on indentation by a flat
Stress distribution in interference joints. Comput. punch. Proc. IMechE, Part C: J. Mechanical Engi-
Struct.s, 1994, 51, 535–540. neering Science, 1998, 212, 212–319.
16 Strozzi, A. Costruzione di Macchine, 1998 (Pita- 34 Jäger, J. New analytical solutions for a flat rounded
gora, Bologna). punch compared with FEM. Computational Meth-
17 Zhang, Y., McClain, B., and Fang, X. D. Design of ods in Contact Mechanics V (WIT Press, South-
interference fits via finite element method. Int. J. ampton), 2001, pp. 307–316.
Mech. Sci., 2000, 42, 1835–1850. 35 Bijak-Zochowski, M., Marek, P., and Tracz, M. On
18 Norton, R. L. Machine design: an integrated methods of reduction and elimination of stress sin-
approach, 2000 (Prentice-Hall, New York). gularities in some elastic contact problems. Int. J.
19 Castagnetti, D. and Dragoni, E. Optimal aspect Mech. Sci., 1994, 36, 279–296.
ratio of interference fits for maximum load transfer 36 Oda, J., Sakamoto, J., and San, K. A method for
capacity. J. Strain Anal., 2005, 40, 177–184. producing a uniform contact stress distribution in
20 Özel, A., Temiz, S., Aydin, M. D., and Sen, S. Stress composite bodies with interference. Struct. Optim.,
analysis of shrink-fitted joints for various fit forms 1991, 3, 23–28.
via finite element method. Mater. Design, 2005, 26, 37 Benuzzi, D. and Donzella, G. Prediction of the
281–289. press-fit curve in the assembly of a railway axle
21 Kanber, B. Boundary element analysis of interfer- and wheel. Proc. IMechE, Part F: J. Rail and Rapid
ence fits. Turkish J. Engng Env. Sci., 2006, 30, Transit, 2004, 218, 218–51.
323–330. 38 Buciumeanu, M., Miranda, A. S., Pinho, A. C. M.,
22 Kolonits, F. The influence of rounded edges on the and Silva, F. S. Design improvement of an automo-
shaft–hub contact. Periodica Polytechnica Ser. tive-formed suspension component subjected to fret-
Transp. Engng, 2007, 51, 35–44. ting fatigue. Engng Failure Anal., 2007, 14, 810–821.
23 Avallone, E. A., Baumeister, T., and Sadegh, A. M. 39 Ciavarella, M., Baldini, A., Barber, J. R., and
Mark’s standard handbook for mechanical engi- Strozzi, A. Reduced dependence on loading para-
neers, 2007 (McGraw-Hill, New York). meters in almost conforming contacts. Int. J. Mech.
24 Francis, M., Henri, C., and Hoa Ngan, N. Stress Sci., 2006, 48, 917–925.
concentration of shrink fit assembly using finite 40 Ciavarella, M., Strozzi, A., Baldini, A., and
element method. In IASTED Asian Conference on Giacopini, M. Normalization of load and clearance
Modelling and Simulation, Beijing, China, 8–10 effects in ball in socket-like replacements. Proc.
October 2007, pp. 67–70. IMechE, Part H: J. Engineering in Medicine, 2007,
25 Ponetchi, P. P. and Vela, I. Stress and strain of 221, 221–601.
solid shafts in interference fit couplings. Analele 41 Pioli, A., Strozzi, A., Baldini, A., Giacopini, M.,
Universitatii, ‘Eftimie Murgu’ Resita, 2008, 15, and Rosi, R. Influence of the initial clearance on
331–338. the peak stress in connecting-rod small ends. Proc.
26 Lanoue, F., Vadean, A., and Sanschagrin, B. Finite IMechE, Part D: J. Automobile Engineering, 2009,
element analysis and contact modelling considera- 223, 223–769.
tions of interference fits for fretting fatigue strength 42 Strozzi, A., Baldini, A., Giacopini, M., Rosi, R., and
calculations. Simul. Modell. Pract. Theory, 2009, 17, Bertocchi, E. Contact stresses within a split ring
1587–1602. inserted into a circular housing. J. Strain Analysis.,
27 Croccolo, D. and Vincenzi, N. A generalized theory 2009, 44, 1–18.
for shaft–hub couplings. Proc. IMechE, Part C: J. 43 Okubo, H. The stress distribution in a shaft press-
Mechanical Engineering Science, 2009, 223, 2231– fitted with a collar. Z. Angew. Math. Mech., 1952,
2239. 32, 178–186.

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016
14 A Strozzi, A Baldini, M Giacopini, E Bertocchi, and L Bertocchi

44 Booker, J. D. and Truman, C. E. A statistical study Proc. IMechE, Part H: J. Engineering in Medicine,
of the coefficient of friction under different loading 1989, 203,
regimes. J. Physics D: Appl. Phys., 2008, 41, 1–12. 15–34.
45 Truman, C. E. and Booker, J. D. Analysis of a 54 Olukoko, O. A., Becker, A. A., and Fenner, R. T.
shrink-fit failure on a gear hub/shaft assembly. Three benchmark examples for frictional contact
Engng Failure Anal., 2007, 14, 557–572. modelling using finite element and boundary
46 Croccolo, D., De Agostinis, M., and Vincenzi, N. element methods. J. Strain Analysis, 1993, 28,
Recent improvements and design formulae applied 293–301.
to front motorbike suspension. Engng Failure
Anal., 2010, 17, 1173–1187.
47 Croccolo, D., De Agostinis, M., and Vincenzi, N.
APPENDIX
Static and dynamic strength evaluation of interfer-
ence fit and adhesively bonded cylindrical joints.
Int. J. Adhesion Adhesives, 2010, 30, 359–366. Notation
48 Ast, M., Rösle, H., and Schenk, R. FEM-Analyse
a contact extent
reibschlüssiger Welle-Nabe-Verbindungen. VDI-
Berichte 1384, ‘Welle-Nabe-Verbingunen - System- E Young’s modulus
komponenten im Wandel’, Fulda, Germany, ISBN F interpolating function
3-18-091384-3, 1998, pp. 227–244. I diametral interference
49 George, A., Strozzi, A., and Rich, J. I. Stress fields k Green function
in a compressed unconstrained elastomeric O-ring k normalized Green function
seal and a comparison of computer predictions
Kt stress concentration factor
and experimental results. Tribol. Int., 1987, 20,
237–247. p contact pressure
50 Strozzi, A. and Unsworth, A. Observations on the p0 reference contact pressure
Paper by O#Carrol et al.. Proc. IMechE, Part H: J. r fillet radius
Engineering in Medicine, 1995, 209, 203–205. ri shaft radius, hub inner radius
51 Bozkaya, D. and Muftu, S. Mechanics of tapered
ro hub outer radius
interference fit in dental implants. J. Biomech.,
2003, 36, 1649–1658. x axial coordinate
52 Andrisano, A. O., Dragoni, E., and Strozzi, A. Axi- y axial coordinate
symmetric mechanical analysis of ceramic heads
sc circumferential stress
for total hip replacement. Proc. IMechE, Part H: J.
Engineering in Medicine, 1990, 204, 157–167. seq von Mises equivalent stress
53 Fessler, H. and Fricker, D. C. A study of stresses in F normalizing parameter
alumina universal heads of femoral prostheses.

J. Strain Analysis Vol. 46


Downloaded from sdj.sagepub.com at PENNSYLVANIA STATE UNIV on May 9, 2016

You might also like