technical

Load Distribution Analysis

of Spline Joints
J. Hong, D. Talbot, A. Kahraman
A finite elements-based contact model is developed to predict load distribution along
the spline joint interfaces; effects of spline misalignment are investigated along with
intentional lead crowning of the contacting surfaces. The effects of manufacturing tooth
indexing error on spline load distributions are demonstrated by using the proposed model.

Introduction
Splines are used commonly in power • Quantify the change to baseline load primary contact point r2ij(s2i,t2j) is deter-
transmission systems for coupling two distributions caused by misalignments, mined such that ||r1(s1,t1) – r2(s2,t2)|| is
rotating components such as a shaft and spline tooth (lead and profile) modifi- minimum. The principal curvatures and
its gear. They provide higher load carry- cations and spline helix angle. principal directions of two surfaces at the
ing capacity over keyed shafts, and hence, • Investigate the influence of indexing common contact point p are determined
errors on baseline spline load distribu-
represent better durability performance. in terms of the coefficients of the first
tions.
In addition, they can tolerate a certain and second fundamental form of the sur-
amount of angular misalignment and rel- Computational Model faces. The second phase is to compute the
ative sliding between their internal and A commercial FE-based contact mechan- compliance matrix and set up the contact
external components. The most common ics mo del Helical-3D (Advance d equation to be solved by a modified sim-
failure modes observed in spline joints Numerical Solutions, Inc.) designed spe- plex method. Hertzian theory is used to
include surface wear, fretting corrosion cifically for loaded contact analysis of predict the size of the contact zone and
fatigue, and tooth breakage (Refs. 1, 2). helical gears is modified here to analyze consequently a grid of points in the con-
In spite of such experimental studies, lit- spline joints. The core contact solver of tact zone is laid out on both surfaces.
tle is known about spline failure mecha- this software (CALYX) is based on a for- Then a surface integral method near the
nisms since accurate load distribution mulation by Vijayakar (Ref. 11), which contact zone and a finite element meth-
prediction models are not available, espe- combines the finite element method od away from the contact zone are com-
cially for the cases where the loading is and surface integral method to repre- bined to predict cross compliance terms
three-dimensional as in the case of heli- sent the contact bodies, and calculates between the set of grid points.
cal or cross-axis gear splines (Refs. 3–10). the load distribution and rigid body dis- The displacement u(r ij; r) at a field
These models can predict load distribu- placements by using the linear program- point r caused by a unit normal force at
tion of splines, but they are limited to ming method. Details of the application surface grid point rij is given as (Ref. 11):
simple loading conditions. Other compli- of Helical-3D to analysis can be found in (1)
cating effects such as spline surface mod- Reference 12. A brief description will be u(rij;r) = (u(si)(rij;r) − u(si)(rij;q)) + u(fe)(rijr;q)
ifications and spline tooth manufacturing provided here for completeness purposes. Here q is some location inside the body
errors such as indexing or spacing errors The first phase of contact analysis is to on a matching surface, sufficiently far
are also not considered in these models.) determine the contact zone. CALYX esti- beneath the tooth surface. The first two
This paper aims at developing a finite mates the contact zone by using Hertz’s terms in this equation denote the rel-
elements (FE)-based computational model after locating a set of primary con- ative deflection of r with respect to q,
model of gear-shaft splines. The objec- tact points on the contacting surfaces which is evaluated using the surface inte-
tives of this paper are as follows: and determining relative principal cur- gral formulae. The third term denotes
• Develop a computational model of a vatures and directions. For this, two con- the displacement of q, which is com-
gear-shaft spline interface under com- tacting surfaces, Σ1 and Σ2, are defined in puted using finite element method. The
bined torsion, radial forces and tilting terms of their curvilinear parameters s point q is chosen such that elastic half-
moments. and t as r1(s1,t1) and r2(s2,t2). The primary space assumption will be valid and the
• Establish nominal load distribution contact points are determined and locat- finite element prediction will not be sig-
conditions under pure torsion, spur
ed when r1 and r2 become the closest to nificantly affected by local stresses on
gear loading (torsion and radial force)
and helical gear (torsion, radial force each other (Ref. 11). For this, the surface the surface. The surface integral and
and tilting moment) loading condi- r1(s1,t1) is discretized into a grid of points finite element solutions are combined
tions. r1ij(s1i,t1j). For each of these grid points, a along this matching surface interface, as

This work is the sole property of the Gear and Power Transmission Research Laboratory Department of Mechanical and Aerospace Engineering,
The Ohio State University, 201 W. 19th Avenue, Columbus, OH 43210, USA

44 GEAR TECHNOLOGY | May 2014

[www.geartechnology.com]
described in detail (Ref. 11). The combi- is constrained along its perimeter to onstrate the loads carried by individual
nation of surface integral formulae and represent a purely torsional loading of contact grid cells on each tooth. As this
finite element method described above the spline with no radial force and tilt- type of a representation of the tooth load
provides an accurate and numerically ing moment. The second loading case distributions is not practical, the con-
efficient way of obtaining the compliance (Figure 2b) represents a spline support- tact surface of each tooth of the shaft
matrix for the contacting bodies. ing a gear where T, applied torque, is spline was mapped to a rectangular win-
balanced by (a) the mesh force Fn = 2T/ dow (Figure 3b) with the load distribu-
Finite Element Model of an (dp cosαn) where αn and dp are the normal tion of the tooth surface displayed on
Example Spline and Analysis pressure angle and pitch circle diameter this window. Using this method, load
Results of the gear, respectively, Fn that is acting distributions on all of the teeth can be
Figure 1 shows the contact model for an on the normal plane of the gear along the viewed simultaneously and conveniently
example clearance-fit spline joint that is line of action has a tangent component (Figure 4) for this torsional loading case.
designed according to ANSI Standard Ft = 2T / dp, a radial component Fr = Ft tan Figure 4 shows the load distributions on
B92.1-1996; Table 1 lists its main param- αn / cosβ and an axial component Fa = Ft the spline teeth under pure torsion at
eters. The system model consists of a tanβ, where αn, β and dp are the normal torque levels of T = 2,000 and 4,000 Nm.
shaft, an external spline and an internal pressure angle, helix angle and pitch cir- For instance, at T = 4,000 Nm, maximum
spline. Over the potential contact area, cle diameter of the gear, respectively. This contact stress is predicted to be about
the model shown in Figure 1 uses a con- results in a torsion T about the rotational 123 MPa that occurs at the edge on the
tact grid with M number of elements axis z of the shaft, radial forces Fx = Ft, input side where the torque is applied to
in the face width direction and N num- Fy = -Fr, and tilting moment Mx = T tan β the shaft. Contact stresses reduce signifi-
ber of elements along the profile direc- about the x axis in Figure 2b. In this case, cantly with the axial distance from this
tion. Within each contact element there the moment Mxwas applied in addition to edge. Non-uniform load distributions
are two contact grids in both face width Fx and Fy to the gear tooth and T applied become clearer with increased T while
and profile direction. Width of the con- to the input end of the shaft. the location of maximum stress remains
tact cells is defined such that 2N grids in Influence of loading conditions. In at the input-side edge. It is noted that
the profile direction can capture all the case of pure torsion loading (Figure 2), as the torque increases, the contact area
contact on the tooth. With this, a spline identical load distributions on each spline extends towards the edge of the spline
joint with Z teeth would have a total of tooth are predicted with the load vary- teeth along the profile direction and the
Z × 2M × 2N grid cells defining the con- ing in an exponential manner in the face contact stress increases simultaneous-
tacts along the drive flanks of the teeth. width direction. Figure 3a shows var- ly — with the load distribution pattern
The model allows intentional deviations ious views of the shaft spline to dem- remaining the same.
from the involute spline tooth surfaces
such as profile and lead modifications as Table 1  Example spline design used in this study
in spur and helical gears to prevent any External spline Internal spline
undesirable edge contact conditions. In Number of teeth 25
Spline Module [mm] 3.175
addition, a similar contact grid with the Pressure angle 30°
same resolution can be defined along the Base diameter [mm] 68.732
other Z coast tooth surfaces to capture Major diameter [mm] 82.550 85.725
Form diameter [mm] 76.022 82.728
any back side contacts. Minor diameter [mm] 73.025 76.200
This example spline interface is loaded Circular space width [mm] - 5.055
in two different ways. In the first case, Circular tooth thickness [mm] 4.981 -
Inner rim diameter [mm] 58 95
a moment (torsion) T is applied to the Outer rim diameter [mm] 64 150
end of the shaft (Figure 2a), while the Inner shaft diameter [mm] 45 -
cylindrical disk having the internal spline Outer shaft diameter [mm] 58 -
Profile crown [µm ] 5 0.0

Figure 1 Spline finite element model; (a) spline interface, (b) internal spline, (c) external spline and shaft, (d) potential contact area and contact
elements.

May 2014 | GEAR TECHNOLOGY 45

 technical

Next, consider a helical gear loading metric. With tooth #1 centered below less. This can be explained by the fact
case. Relevant helical gear tooth param- (and closest to) the gear mesh, teeth 1-8 that the spline interface transmitted the
eters are the normal pressure angle of and 21-25 are shown to experience larger same radial load in spur gear loading and
αn = 20°, and the pitch circle diameter of loads while teeth #9-20 bear less load. For helical gear loading conditions. However,
dp = 160 mm. Helix angle β of the gear is instance, the resultant maximum contact a remarkable difference is observed that
varied from 0° (spur gear loading) to 20° stresses at T = 4,000 Nm are 129, 108, 101 axial load distribution on teeth #19-25
at a torque value of 4,000 Nm to predict and 167 MPa for teeth #5, #10, #15 and and #1-2 is biased to the opposite side to
the load distributions shown (Figure 5). #24, respectively, as a direct consequence which torque is applied. This is due to the
For the spur gear loading (Figure 5a), of this unequal loading. In comparison to additional tilting moment Mx transmit-
load sharing along the face width direc- Figure 5a, helical gear loading maintains ted over the spline interface, which causes
tion is still biased towards the side where the same qualitative load sharing char- the load on some teeth to be biased to
torque is applied. Load distributions are acteristics of spline teeth. Figures 5b and the other side to balance it. It is observed
no longer identical for all the spline teeth 5c show that teeth #1-8 and #21-25 carry that as β increases, the load on each tooth
since the loading is no longer axisym- larger load, while teeth #9-20 are loaded gets more concentrated on the side where
the load is biased. For instance, loads
on teeth #3-10 are concentrated to the
side where the torque is applied, while
loads on teeth #20-25 are concentrat-
ed to the opposite side. The maximum
contact stresses are 182 and 334 MPa for
β = 10° and 20°, respectively, with the cor-
responding tilting moments of Mx = 705
and 1,456 Nm. The load concentration
increases significantly because of the
larger resultant tilting moment on the
spline when the β increases.

Figure 2 A spline model with different loading conditions; (a) pure torsion loading
Effect of Design Variations
and (b) gear loading. Misalignments and lead crown modi-
fications. Misalignment of spline cou-
plings has been recognized as harmful to
splines because it causes significant load
concentration on spline teeth, and accel-
erates wear and fretting fatigue of splines
(Refs. 2, 13). Significant load concentra-
tion observed in both misaligned splines
(Figure 6a, a misalignment of φ = 0.12°)
and splines experiencing helical gear
loading (Figure 5b-c) can be potentially
remedied by applying a lead crown modi-
fication along the face width direction.
Figure 6 shows load distributions of a
spline having a misalignment of φ = 0.12°
Figure 3 (a) Load distribution of a spline under pure torsion loading, and (b) mapping of the load
distribution on a tooth to a rectangular window.
along with different lead crown modifi-

Figure 4  Load distributions of the example spline under pure torsion loading at different torque levels.

46 GEAR TECHNOLOGY | May 2014

[www.geartechnology.com]
cation magnitudes of δ = 0 and 40 μm at
T = 4,000 Nm under pure torsion load-
ing. It is observed that the crown modi-
fication moves the load from the edge
to the center of the spline, in the process
reducing the maximum contact stresses
significantly. Figure 7 shows load distri-
butions on the splines with different lead
crown modification magnitudes of δ = 0
and 40 µm for helical gear loading with
β = 20° at T = 4,000 Nm. Unlike in the
misaligned spline case, lead crown modi-
fication under a helical gear loading con- Figure 5 Load distributions of a spline under helical gear loading with different helix angles, β, at
dition neither reduces load concentra- T = 4,000 Nm.
tion, nor moves the tooth load from the
edge to the center. This occurs because
the moment acting on the spline remains
a constant for a given torque in helical
gear loading despite the lead crown mod-
ification. The biased load concentration
exists no matter how much lead crown
modification is adopted.
Helical splines: Figure 8 shows the load
distributions of helical splines having dif- Figure 6 Load distributions of a misaligned spline having different lead crown modification
magnitudes, δ, at T = 4,000 Nm and φ = 0.12°.
ferent spline helix angles of γ = -10°, 0°
and 10° under helical gear loading con-
dition with a helical gear (β = 20°) at
T = 4,000 Nm . A negative spline helix
angle represents a left-handed helix and
a positive helix angle represents a right-
handed helix. It is observed that a right-
hand spline helix helps reduce load con-
centration, and a left-hand spline helix
condenses the load concentration. For Figure 7 Load distributions of a spline under helical gear loading having different lead crown
instance, the maximum contact stress- modification magnitudes, δ, at T = 4,000 Nm, and β = 20°.
es are 334 MPa for a spur spline (γ = 0°)
under helical gear loading with a right-
hand helical gear. The maximum con-
tact stress increases dramatically to
754 MPa for left-hand spline helix angles
of γ = -10°. On the other hand, the maxi-
mum contact stresses drop significant-
ly to 211 MPa for right-handed splines
with helix angles γ = 10°. The maximum
contact stress is reduced by more than
35% and more teeth are observed to carry
load (Figure 8c). The dominant factor is
that the spline helix introduces an axial
load that can either increase the tilting
Figure 8 Load distributions of helical splines having different spline helix angles, γ, under helical
moment or counterbalance the tilting gear loading with a gear helix angle, β = 20°, at T = 4,000 Nm.
moment, depending on its direction. For
a spline loaded by a right-handed gear,
a right-hand spline helix would induce
an axial load to counterbalance the tilt-
ing moment thus reducing the load con-
centration. On the other hand, a left-
hand spline helix would induce an axial
load, which would increase the tilting

May 2014 | GEAR TECHNOLOGY 47

 technical
Performance of Splined Shafts,” Proc. Instn.
Mech. Engrs., Part G, 215 (2001) 105-112.
6. Kahn-Jetter, Z. and S. Wright. “Finite Element
Analysis of an Involute Spline,” J. Mech. Des.,
122 (2000) 239-244.
7. Tjernberg, A. “Load Distribution in the Axial
Direction in a Spline Coupling,” Engineering
Failure Analysis, 8 (2000) 557-570.
8. Leen, S.B., I.J. Richardson, I.R. McColl,
E.J. Williams and T.R. Hyde. “Macroscopic
Fretting Variables in a Splined Coupling under
Combined Torque and Axial Load,” J. Strain
Analysis 36 (2001) 481-497.
9. Adey, R.A., J. Baynham and J.W. Taylor.
“Development of Analysis Tools for Spline
Couplings,” Proc. Instn. Mech. Engrs., Part G,
214 (2000) 347-357.
Figure 9 Load distributions of a spline having random tooth indexing errors at different torque levels.
10. Medina, S. and A.V. Olver. “Regimes of Contact
moment thus resulting in larger load con- under combined torsional load, radial in Spline Coupling,” J. of Trib.124 (2002) 351-
centration on the spline teeth. The result load and tilting moment was proposed. 357.
11. Vijayakar, S. “A Combined Surface Integral
implies that in helical gear loading con- Load distributions of the baseline system and Finite Element Solution for a Three-
ditions, selecting a helical spline with the of the spline coupling under pure tor- Dimensional Contact Problem,” Int. J. for
same helix direction as that of the helical sion, spur gear loading and helical gear Numerical Methods in Engineering 31 (1991)
gear could reduce the load concentration, loading were characterized. Pure torsion 525-545.
12. Hong, J., D. Talbot and A. Kahraman. “Load
and improve load distribution. loading results showed identical load dis- Distribution Analysis of Clearance-fit Spine
tributions on all spline teeth, with each Joints using Finite Elements,” Mechanisms and
Effect of Spline Tooth Indexing tooth exhibiting exponentially decreas- Machine Theory, 74 (2014) 42-57.
Errors ing load in axial direction while helical 13. Medina, S. and A.V. Olver. “An Analysis of
Misaligned Spline Coupling,”Proc. Instn. Mech.
Investigations of spline load distributions gear loading led to cyclic load concentra- Engrs., Part J, 216 (2002) 269-279.
discussed above all assumed perfect tooth tion oscillating across the face width of 14. Chase, K.W., C.D. Sorensen and B.J.K.
geometry, ignoring the effect of tooth the spline teeth. The effectiveness of lead DeCaires.“Variation Analysis of Tooth
indexing errors. However, it occurs fre- crown in improving load distributions Engagement and Loads in Involute Splines,”
IEEE Trans. Autom. Sci. Eng. 7 (2010) 54-62.
quently that some spline teeth exhibit of misaligned splines was demonstrated.
heavier damage than others, which might Selecting a helical spline with the same
be caused by spline tooth indexing errors helix direction as that of the helical gear
For Related Articles Search
(Refs. 12-14). In order to demonstrate loading the spline was shown to reduce
the effect of indexing errors on spline the load concentration and improve load splines
load distributions, a spline joint with a distribution of splines undergoing helical at www.geartechnology.com
random tooth indexing error distribu- gear loading. Finally, effects of indexing
tion is considered next. Figure 9 shows a errors of spline teeth were investigated Jiazheng Hong is a Ph.D.
random tooth indexing error sequence for the pure torsion loading case to show student at Mechanical and
considered together with the correspond- significant unequal load sharing at spline Aerospace Engineering at
the Ohio State University.
ing spline load distribution at T = 2,000 teeth due to indexing errors. He conducts research
and 4,000 Nm under pure torsion condi- Acknowledgement. The authors thank projects in gear and spline
tion. Load concentrations are observed Dr. Sandeep Vijayakar of Advanced contact mechanics for Gear
and Power Transmission
on spline teeth having larger indexing Numerical Solutions, Inc. for making Research Laboratory.
error, such as spline tooth #6, #9, #12 Helical-3D available.
David Talbot is a
and #19. Teeth with larger indexing error Research Scientist
have smaller clearance, and they will References at Mechanical and
engage first. Teeth with smaller indexing 1. Ku, P.M. and M.L. Valtierra. “Splice Wear Aerospace Engineering at
Effects of Design and Lubrication,” J. Eng. for the Ohio State University.
error have larger clearance and they will Ind., 97 (1975) 1257-1263. He is specialized in
gradually come into contact when the 2. Brown, H.W. “A Reliable Spline Coupling,” J. gear system efficiency,
torque increases to a certain level. Figure Eng. for Ind., 101 (1979) 421-426. gear manufacturing
and geometry and load
9 emphasizes that the load concentration 3. Volfson, B.P. “Stress Sources and Critical Stress
distribution simulation.
Combinations for Splined Shaft,” J. Mech. Des.,
resulting from spline tooth indexing error 104 (1982) 551-556. Ahmet Kahraman is a
is significant and must be accounted for 4. Barrot, A., M. Paredes and M. Sartor. Professor of Mechanical
in the design of a spline at a certain qual- “Determining Both Radial Pressure and Aerospace Engineering
Distribution and Torsional Stiffness of Involute at the Ohio State University.
ity level. He is the Director of
Spline Couplings,” Proc. IMechE, Part C: 220
(2006) 1727-1738. Gleason Gear and Power
Summary and Conclusions 5. Limmer, L., D. Nowell and D.A. Hills. “A
Transmission Research
Laboratory. He also directs
A finite element-based computational Combined Testing and Modeling Approach Pratt & Whitney Center of
model of a gear spline-shaft interface to the Prediction of the Fretting Fatigue Excellence in Gearbox Technology.

48 GEAR TECHNOLOGY | May 2014

[www.geartechnology.com]