Professional Documents
Culture Documents
August 8, 2013
Abstract
Multiple interference fits represent high friction-stressed shaft-hub connections.
High surface pressures together with thin-walled intermediate parts lead to high
stresses and showing strong tendencies to plastic deformations. Cylindrical inter-
ference fits and their elasto-plastic behaviour have been analysed in the literature.
However, in the case of multiple interference fits, these methods are not applicable.
In this paper, we derive a new calculation method for elasto-plastically deformed
multiple interference fits with hardening material. The analytical method is vali-
dated by the comparison of the determined stresses and displacements of several
test cases with the solutions of corresponding axisymmetric finite element models.
1 Introduction
A widespread type of shaft-hub connections are interference fits. Due to their versatile
capabilities they are often used to transfer large torsional moments. In many applications
it is sufficient to consider interference fits consisting of two parts. However, in some cases
multiple interference fits consisting of three or more rings have to be used, e.g., for
the connection of machine shafts with a hollow-shaft gearbox. In this case a multiple
interference fit with four parts is applied in order to connect both shafts due to a frictional
connection. In practice, shrink discs are used for such attachments, see RINGSPANN
[2009].
In order to transfer high torsional moments, one has to consider not only elastic de-
formations but also plastic deformations of the material. Unfortunately, the calculation
of an elasto-plastic stressed multiple interference fit is currently not possible without the
numerical solution of nonlinear partial differential equations (PDEs). The aim of this
contribution is to derive a simple analytical method to compute the displacements and
stresses of an elasto-plastic multiple interference fit. Our method requires the solution of
a nonlinear system of equations, see Eq. (12), which is possible using Newtons method.
Note that this computation does involve the computation of definite integrals, see Eq. (9),
1
but avoids the solution of ordinary or partial differential equations. Note that some parts
of this work were published (in German) in an abbreviated form, see Leidich et al. [2010].
2
3 Derivation of the analytical solution
Given the geometry of the interference fit (i.e., the radii and interferences) as well as the
outer and inner pressures, we want to compute the displacements of the rings and the
pressures in the joints. In the case where all rings are in the elastic regime, the solution
is well known and standard. Moreover, analytical solutions for the plastic behavior of
a single interference fit exists, see Section 2. We want to derive an analytical solution
for the plastic behavior of a multiple interference fit. To this end, we give an analytical
relation between the displacements of a single ring and the pressures acting on its inner
and outer surfaces, see Section 3.1. This relation is used in Section 3.2 to derive a
calculation method for a multiple interference fit.
(3)
(2)
(1)
pi r0 r2 po
r1 r3
(j) are labeled by a superscript (j), i.e., the inner and outer radii of the ring (j) are
(j)
ri = rj1 and ro(j) = rj ,
respectively, see Figure 1. Similarly, the inner and outer pressures acting on the ring (j)
(j) (j) (j) (j)
are denoted by pi and po , and the displacements of (j) are denoted by ui and uo .
The rings are coupled by the displacements and the contact pressures in the joints. The
interference j in the joint j is given by
(j+1)
j = ui u(j)
o for all j = 1, . . . , n 1. (1)
3
The case of a clearance fit in the joint j can be modelled by j < 0. The contact pressure
in the joint j is denoted by pj , j = 1, . . . , n 1. Moreover, the multiple interference fit
is subject to an inner pressure pi and an outer pressure po . It is convenient to use the
notations p0 = pi and pn = po .
(iii) elastic behaviour at the outer radius r = ro with Youngs modulus E and Poissons
ratio ,
(vi) we use the von Mises criterion to determine the equivalent stresses and strains,
r = rz = z = 0,
r = rz = z = 0,
(viii) we use the model of Hencky (static) plasticity, i.e., there is a function f : R+ R+
such that eq = f (eq ), see Figure 2,
In the sequel, we derive a relationship between the displacements (ui , uo ) and the pres-
sures (pi , po ) = (r (ri ), r (ro )). To this end, we assume that the displacements (ui , uo )
are known and we derive a formula of the pressures (pi , po ) in terms of the displacements.
Using assumptions (iv) and (v), we have
ur (r) ur (r) 1
const = r (r) + (r) = + = (r ur (r)).
r r r r
Integrating yields
C2
ur (r) = C1 r + . (2)
r
Using
ur (ri ) = ui and ur (ro ) = uo ,
4
eq
f (eq )
0 eq
the constants C1 and C2 are given as the solutions of the linear system
C2 C2
ui = ur (ri ) = C1 ri + , uo = ur (ro ) = C1 ro + .
ri ro
Consequently, we obtain the radial and tangential strains
C2 C2
r (r) = C1 and (r) = C1 + , (3)
r2 r2
respectively. Using z (ro ) = 0 by assumption (ii) and the equations of the elastic regime
at the outer radius r = ro , we obtain
E
0 = z (ro ) = r (ro ) + (ro ) + (1 ) z (ro ) .
(1 + ) (1 2 )
In the model of Hencky plasticity the variable shear modulus is given by, see [Szab,
2000, 16.5c)],
eq (r)
(r) = . (7)
3 eq (r)
5
Consequently, the stress-strain relations imply
where r (ri ) and r (ro ) are determined by (9) and (10), respectively. Note that the
evaluation of pi and po does not require the solution of any differential equation, only
the integral in (9) has to be computed (numerically).
(j) (j)
for the unknowns ui , uo , j = 1, . . . , n and pj , j = 1, . . . , n 1. Note that the inner
pressure p0 = pi and the outer pressure pn = po are given. Hence, we arrived at a
nonlinear system with 3 n 1 equations and 3 n 1 unknowns. This system of nonlinear
equations can be solved by Newtons method. The solution of the system (12) yields the
6
(j) (j)
displacements ui , uo , j = 1, . . . , n. By using (8) and (9) the stresses r and can
be evaluated. Using arguments similar to those leading to (9), a formula for the axial
stresses z can be derived. Let us briefly highlight the changes which are necessary in
case of a solid shaft, i.e. r0 = 0. In this case the inner pressure p0 = pi is unknown, but
(i)
the inner displacement of the solid shaft ui is zero due to continuity considerations, i.e.,
(1) (1)
ui = 0. Hence, the known inner displacement ui has to be replaced by the unknown
inner pressure pi = p0 . The number of unknowns remain 3 n 1.
(ii) finite element model of a standard interference fit with elasto-plastic behaviour,
(iii) finite element model of a multiple interference fit with elastic behaviour,
(iv) finite element model of a multiple interference fit with elasto-plastic behaviour.
Each step include several test cases. These test cases differ in the size of the radii, inter-
ferences and in the materials of the rings. Due to the rotational symmetry of a multiple
interference fit, it is sufficient to consider a axisymmetric two-dimensional model. The in-
terferences (resp. clearance fit) in the joints are modelled using the function interference
fit of Abaqus.
In practice, the clearance fit (modelled as a negative interference) is necessary for the
assembly of two shafts (e.g. one solid and one hollow or two hollow). To realize the
positive interference, one uses a shrink disc which includes a taper interference fit. In the
practical assembly as well as in the finite element model, the interference between the
two outer shafts is gradually increased by the shrink disc. This will cause the clearance
fit (between the inner rings) to be closed once the interference (between the outer rings)
is large enough. In contrast, the assembly in the analytical model is considered to happen
simultaneously, since we are using the material law of Hencky (static) plasticity.
Figure 3 shows the finite element model of the multiple interference fit with boundary
conditions.
7
Figure 3: Finite element model of the multiple interference fit.
stress-strain curves are used. In the test Cases 1 and 3 we consider a tempering steel
material 1, whereas the construction steel material 2 is used in Case 2. The test cases
k.1, k = 1, 2, 3, are similar to the test cases k. They do not show a qualitatively different
behaviour.
The material parameters are
( (
949 MPa, 1.95 105 MPa for material 1,
= 0.3, yield stress = and E =
432 MPa, 2.06 105 MPa for material 2.
The function f , which describes the stress-strain curve, see assumption (viii) and
Figure 2, is obtained by a static tensile test. These stress-strain curves are shown in
Figure 4.
Results. The distributions of the stresses r , and eq are shown in Figure 5 and
Figure 6 for Case 2 and Case 3, respectively. In test Case 2 plastic deformations occur
in the thin-walled intermediate part. In Case 3, plastic deformations occur in the inner
and outer rings of the multiple interference fit, whereas the middle ring is elastically
deformed. The analytically and numerically obtained stresses are almost identical. In
particular the radial stresses at the joints (i.e. the joint pressure) match very well. The
relative errors in the radius of the zone of plasticity stay below 2% and the relative errors
in the stresses stay below 3%. However, large relative errors occur in the axial stresses
in the zone of plasticity. This is shown in Figure 7. In this figure we have detailed the
zone of plasticity in the outer ring. The radial displacements for Case 3 are shown in
8
Figure 4: Stress-strain curves of the used materials.
9
Figure 6: Analytically (, + and 3) and numerically (solid lines) determined distributions
of the stresses for Case 3 of Table 1.
10
Radii of the rings Interferences
r0 r1 r2 r3 1 2
Name [mm] [mm] [mm] [mm] [m] [m]
Case 1 0 18 20 40 -25 420
Case 1.1 0 18 22.5 45 -25 473
Case 2 0 36 40 80 -122 300
Case 2.1 0 36 45 112.5 -130 320
Case 3 18.0 36 40 80 -60 740
Case 3.1 25.25 36 40 80 -50 620
Table 1: Parameters for six elasto-plastic stress three-part multiple interference fits, cf.
Figure 1.
Figure 8. Similar to the stresses, only small differences between the numerical and the
analytical solution occur. The relative errors for the displacements stay below 1%.
6 Conclusions
Using the analytical solution for multiple interference fits given in Section 3, it is possible
to compute an elasto-plastic deformed multiple interference fit without solving (nonlin-
ear) PDEs. The analytical method requires the solution of a nonlinear system of equa-
tions, which can be done using Newtons method. The analytical method is validated by
a comparison of the determined stresses and displacements of several test cases with the
solutions of corresponding axisymmetric FE model. Up to the axial stresses, the relative
errors in the stresses and displacements stay below 3%.
The analytical solution presented in this paper is one ingredient to calculate complex
shaft-hub connections or shrink discs, see RINGSPANN [2009], if plastic deformations
are present. Up to now, one has to solve the associated PDEs numerically in order to
compute a multiple interference fit under plastic deformations. Using our method, one
could avoid the computationally expensive solution of PDEs, see, e.g., the comparison in
Leidich and Ltzer [2011].
Acknowledgement
The authors gratefully acknowledge the financial support by AiF (German Federation of
Industrial Research Associations) and by FVA (Research Association for Drive Technol-
ogy).
References
N. Baldanzini. A General Formulation for Designing Interference-Fit Joints With Elastic-
Plastic Components. Journal of Mechanical Design, 126(4):737743, 2004.
11
Figure 8: Analytically () and numerically (solid line) determined radial displacement
for Case 3 of Table 1 in dependence of the radius r.
12
G. Lundberg. Die Festigkeit von Presssitzen. Das Kugellager, 19(1/2):111, 1944.
13