You are on page 1of 8

Materials and Design 54 (2014) 130136

Contents lists available at ScienceDirect

Materials and Design

journal homepage:

Analysis of the torsional strength of hardened splined shafts

I. Barsoum a,, F. Khan a, Z. Barsoum b

Department of Mechanical Engineering, The Petroleum Institute, P.O. Box 2533, Abu Dhabi, United Arab Emirates
Department of Aeronautical & Vehicle Engineering, KTH Royal Institute of Technology, Teknikringen 8, 100 44 Stockholm, Sweden

a r t i c l e

i n f o

Article history:
Received 7 May 2013
Accepted 6 August 2013
Available online 17 August 2013
Induction hardening
Finite element modeling
Torsion strength
Splined shaft

a b s t r a c t
The current study presents a nite element modeling framework to determine the torsion strength of
hardened splined shafts by taking into account the detailed geometry of the involute spline and the material gradation due to the hardness prole. The aim is to select a spline geometry and hardness depth that
optimizes the static torsion strength. Six different spline geometries and seven different hardness proles
including non-hardened and through-hardened shafts have been considered. The results reveal that the
torque causing yielding of induction hardened splined shafts is strongly dependent on the hardness depth
and the geometry of the spline teeth. The results from the model agree well with experimental results
found in the literature and reveal that an optimum hardness depth maximizing the torsional strength
can be achieved if shafts are hardened to half their radius.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
There is an increased demand on power handling capacity of
automotive transmission parts requiring high static and fatigue
strength. Especially in splined shafts transferring high torque commonly used in heavy trucks, the choice of spline geometry and heat
treatment process is crucial in obtaining an adequate strength of
the component. For such connections yielding or ultimate fracture
are potential modes of failure, particularly when overloads are involved. Since hardness of the material is strongly related to the
strength of the material a common method of heat treatment of
splined shafts is induction hardening, which increases the torque
capacity of the shaft. Along with the material aspects associated
with the heat treatment process, the strength of a spline connection is also dependent upon the geometrical design of the spline.
Involute splines are commonly used in automotive industry with
proles similar to those of involute gear teeth. In the past, several
studies [14] have focused on the failure analysis and failure prevention of splined or non-splined shafts; however none have attempted to quantify the simultaneous effect of the spline
geometry and the hardness prole and depth on the torsion
Several studies have focused on determining the fatigue behavior of splined shafts. In [5,6] the uniformity of load distribution on
the spline teeth is studied by means of a nite element model. The
model used accounts for the geometry of the splined shaft with the
sleeve, from which it is found that the fatigue life of the shaft can
be prolonged if a uniform load distribution can be assured. Deter Corresponding author. Tel.: +971 558006641.
E-mail address: (I. Barsoum).
0261-3069/$ - see front matter 2013 Elsevier Ltd. All rights reserved.

mination of the residual stresses during induction hardening process of shafts is carried out in [6] by simulation and X-ray
measurements. Results indicate that hardening depth has an inuence on the residual stress level.
The objective in [7] is to improve the torsional fatigue life of a
steel power transmission shafts. By varying the carburization process parameters such as soaking temperature and time different
carburization depths are achieved. It is found that the carburization depth signicantly inuences the fatigue endurance limit under bending torsional loadings.
In [8] a numerical model was developed to simulate the induction hardening process of a steering pinion shaft by solving the
coupled electromagnetic-thermal problem. They study the effects
of induction parameters i.e. input AC current density, coil velocity
and coil stay time with the model. The computed hardness prole
curves show good agreement with the experimental data.
Similarly in [911] the authors studied experimentally the
effects of induction parameters such as frequency, coil shape and
applied current on nal microstructure, the hardness of the
work-pieces. Their investigation shows that the process parameters inuence the residual stress-state of hardened parts to a great
In [12] fatigue tests of induction hardened shafts are conducted
in which failure occurs mainly due to subsurface crack initiation.
Simulation results showed that the tensile residual stresses in
the core increase with increased hardness depth, which is detrimental for fatigue life and a reason for subsurface crack initiation.
The effect of torque overload on the fatigue life of barreled splined
areoengine mainshaft couplings is investigated experimentally and
numerically in [13,14]. A three dimensional nite element model is
developed, however it does not take into account the material


I. Barsoum et al. / Materials and Design 54 (2014) 130136

gradation resulting from the heat treatment process of the splined

shaft. As reported in [15,16], engineering failure analysis and prevention of splined shaft failures is of great interest for the automotive industry when designing these components. However, none of
the studies has so far addressed the combined effect of the geometry of the splined shaft and the material gradation in strength
through the cross section of the shaft as a result of the induction
hardening process.
Recently Kang et al. [17] performed nite element simulation of
the induction hardening process of axle shafts with the objective to
study the effect of surface hardening depth and distribution of
residual stresses on the torsional strength of the shaft. The model
they utilized is highly sophisticated taking into account the thermo-mechanical and metallurgical phase transformation behavior
of the material during the induction hardening process. The shaft
geometry they considered is circular in cross section, disregarding
the effect of the spline design. Their results indicate that there is a
strong effect of the induction hardening depth on the torsional
strength. The importance of the hardening depth on the static
and fatigue strength of induction hardened shafts has also been
pointed out by Fett [18], where the relationship between induction
hardening depth and torsional strength and fatigue life is determined though an extensive experimental program on induction
hardened shafts of various material grades. Tortorella et al. [19]
also performed static torsion experiments on induction hardened
shafts and arrive upon the same conclusion as in [18] that the
induction hardening depth has a paramount effect on the static
Hence, the objective of this study is to develop a modeling
framework which addresses some of the shortcomings of the models used in the literature to model the strength of hardened splined
shafts. The analysis presented in this study aims to quantify the
torsional strength of splined shafts by taking into account the detailed effect of the geometry of the spline and the gradation in
material property, such as the hardness prole and hardness depth
resulting from the hardening process.
2. Geometry of the involute spline
There are various international standards in which a thorough
denition of the spline geometry can be found [2022]. Here the
geometry of the involute spline and the geometric quantities that
dene the spline will be briey described. The terminology of an
involute splined shaft is shown in Fig. 1. The pitch circle with radius rp is the theoretical circle upon which several geometric
parameters are dened such as the pressure angle a. The number
of spline teeth is denoted Z and together with rp denes the module
m of the spline as

Table 1
Chemical composition of SAE 4140 (42CrMoS4).
C (%)

Si (%)

Mn (%)

Cr (%)

Mo (%)






m 2rp =Z

which is an indication on the size of the spline tooth. The top and
bottom land of the tooth are dened by the addendum radius ra
and the dedendum radius rd, respectively.
In the present study six different spline geometries are considered comprising of two different pressure angles a = 30 and 45
and three different tooth numbers Z = 20, 30 and 40. Based on
the polar moment of inertia of a circular shaft the static torsional
yield strength scales with the cube of the spline dedendum radius
(rd3). Therefore the addendum radius ra of the spline has been kept
xed for all cases while the dedendum radius rd is varied by changing the module m. Consequently the pitch radius rp is given by Eq.
(1) once the module m is set. The values of the module considered
here are m = 1.5, 1.0 and 0.75 mm. These values for the module
along with number of teeth given above gives a constant value of
the pitch radius rp = 15 mm.
3. Mechanical properties of SAE 4140 steel
The material considered in the shaft is a medium-carbon steel
denoted SAE 4140 (42CrMoS4), which is commonly used in induction hardened transmission components and drive shafts in the
automotive industry [23,24]. The chemical composition of SAE
4140 is given in Table 1.
By varying the process parameters in the induction hardening
process various hardness depths and proles can be achieved.
Mechanical properties such as yield and ultimate strength are closely related to the hardness where the yield strength is approximately one third of the hardness in HV (Vickers hardness). In
order to establish a relation between stressstrain behavior and
the hardness the materials simulation software JMATpro [25] is
utilized. The software is based on thermodynamics modeling combined with theoretical materials models and properties databases
allowing for a quantitative calculation of wide range of thermophysical and mechanical quantities [26]. Given the chemical composition of a specic steel alloy, the stressstrain behavior can be
determined with JMATpro. In Fig. 2(a) the stressstrain curves
show that the work-hardening behavior as well as the yield and
ultimate strength strongly depend on the hardness level where
each stressstrain curve correspond to a hardness value. Also the
yield and ultimate strength for each hardness value can be



Fig. 1. Parameters dening the spline geometry.


I. Barsoum et al. / Materials and Design 54 (2014) 130136


R p 02



Fig. 2. Mechanical properties of 42rMoS4.

obtained as shown in Fig. 2(a). The stressstrain curves can be tted to the plasticity model given by

r rN

where E = 210 GPa is the elastic modulus, K the plastic modulus and
N the hardening exponent respectively. Hence, the yield strength
Rp02, the ultimate strength Rm and the plasticity model parameters
K and N can be obtained as function of Vickers hardness denoted
H here, as shown in Fig. 2(b) and (c). The plasticity model parameters K and N as function of hardness H are given by Eqs. (3) and (4).

N n0 en1 H

K k0 H k1

The constants in Eqs. (3) and (4) are obtained by a least-square

t to the data in Fig. 2(b)-(c) given by: n0 = 0.4911, n1 = 0.00232 1/
HV, k0 = 0.0034 MPa/HV and k1 = 0.3 MPa. Now, for a specic hardness value given in HV the stressstrain behavior can be determined from Eqs. (2)(4), which in combination with the
hardness prole of the induction hardened shaft gives the gradient
in mechanical behavior of the shaft, e.g. the stressstrain behavior.
This is in accordance with well established ndings in the litera-


ture [30]. The tensile yield strength Rp02 and tensile ultimate
strength Rm can also be tted to simple expression as function of
hardness H as

Rp02 J Rp Hn2

and Rm J m H

where n2 = 1.25, JRp = 0.56 and Jm = 3.15 based on the best t to data
in Fig. 2(b).
4. Hardness prole and depth
Fig. 3(a) shows a non-hardened splined shaft cut in half where
the surface and core are identied. In order to measure the hardness prole, transverse cuts of the hardened shafts where made,
which where molded and polished as shown in Fig. 3(b) before
the indentation testing was conducted.
Fig. 4(a) shows measured hardness H in Vickers hardness value
(HV) against depth below surface (c.f. Fig. 1) of an induction hardened shaft. It is notable that the hardness is highest at the outer
surface with a value of 630 HV and decreases towards the center
of the shaft reaching the value of the core hardness 230 HV. To
quantify the depth of the hardness most often an effective hardness depth is dened. Here the effective hardness depth DH is dened as the depth below the surface at which the hardness level
reaches a value of 400 HV. This is common practice used in the
automotive industry [18], where 400 HV corresponds to about 40

Fig. 3. A radial cut of the splined shaft.


I. Barsoum et al. / Materials and Design 54 (2014) 130136



Fig. 4. (a) Measured hardness prole of a splined shaft; (b) hardness proles given by Eq. (5).

HRC (Rockwell C scale hardness value). For the case shown in

Fig. 4(a) DH = 5 mm, which is about 34% of the shaft radius. The
shaft radius refers to the pitch radius of the spline (rp = 15 mm).
The method used here to represent the hardness prole as a function of the depth below the surface is similar to the one used in
[27,28], which is given by the expression

Hr Hcase gr Hcore 1  gr

where Hcase and Hcore are the case and core hardness, respectively,
which are here given in Fig. 4(a) as Hcase = 630 HV, Hcore = 230 HV.
The function g(r) is given by



rr CS
gr 1 c1 rr


r < r CR
rCR 6 r 6 r CS

r > r CS

The choice of g(r) satises the following constraints by default:

g(rCS) = og(rCS)/@r = 0. However, additional conditions are imposed
on the function g(r) and H(r), which are

gr rCR

@gr r CR


Hr r p  DH 400

Using the rst condition in Eq. (8) yields an expression of g(r) in

terms of rCS and rCR and Eq. (7) can be rewritten as

gr 1

rCS r  rCS 2r  3 rCS r CS

r CS 2 r CS r CS  r CS 2

for r CR 6 r 6 r CS

In Eqs. (7)(9), rCS and rCR are the size of the case and core respectively as indicated in Fig. 4(a). By using Eq. (9) and the second condition of Eq. (8) in Eq. (6) the entire hardness prole will be
controlled by the choice of the hardness depth DH and core size
rCR. Now, by varying the hardness depth DH different hardness proles can be achieved with Eqs. (6) and (7). As shown in Fig. 4(b) ve
different effective hardness depths are constructed and considered
with DH = 2.5, 5.0, 6.0, 7.0 and 8.0 mm. Two additional hardness
depths values were considered, DH = 0 and DH = 1, corresponding
to a non-hardened shaft with a uniform hardness of 230 HV and
through-hardened shaft with a uniform hardness of 630 HV. These
representations of non-hardened and through hardened shafts are
rather ideal. However, they will here serve as the lower and upper
limits respectively on the effective hardness depth DH.

(Fig. 1) and hardness depth (Fig. 4(b)) a 3D nite element model

was built using the commercial package ABAQUS [29]. Since the
geometry and torsional loading exhibits cyclic symmetry, only
one tooth of the splined shaft along with the corresponding cyclic
symmetric segment of the hub was modeled as shown in Fig. 5. The
back faces of the hub as well as that of the spline are kinematically
coupled to reference points. The reference point of the hub is given
a xed displacement boundary condition were all six degrees of
freedom (3 displacements and 3 rotations) are set to zero. The reference point of the shaft is given a xed rotation about the x-axis
which will give rise to a reaction torque denoted T. The remaining
degrees of freedom at the reference point of the shaft are set to
zero. A contact denition is established between the hub and
spline surfaces which are in contact with a friction coefcient of
0.1 to mimic dry steel to steel contact. The nite element mesh
used consisted of about 26,000 eight-node continuum elements
(C3D8) and the analysis was ran using the ABAQUS/Standard solver. The hub is modeled as an elastic material (E = 210 GPa and
v = 0.3) whereas the spline is modeled as elasticplastic material
with isotropic hardening. The splined shaft is sliced into 13 different layers radially ranging from the non-hardened core layer at the
center and the case with highest hardness at the outer layer, which
includes the spline tooth. Each sub-surface layer had a thickness of
1 mm whereas the surface layer (case) including the tooth of the
spline had a thickness of 2.5 mm. Depending on the hardness
depth DH considered, each layer in Fig. 5 will have different hardness values due to the varying hardness, which depends on the
hardness depth DH (cf. Fig. 4(b)). Consequently, since the plastic



Layer 1
Axis of Cyclic

5. Finite element model for a splined shaft

In order to determine the static strength in torsion of an induction hardened shaft accounting for the effect of spline geometry


Layer 11

Fig. 5. The cyclic symmetric FEM model.


I. Barsoum et al. / Materials and Design 54 (2014) 130136

Fig. 6. Torque T versus shear strain c for pressure angle a = 30, various spline teeth Z and hardness depth DH.

behavior is linked to the hardness by Eqs. (2)(4), each layer in the

model will be assigned an elasticplastic true stressstrain curve
corresponding to its hardness value following the hardness prole
set by DH.
6. Results and discussion
The results from the nite element models are generated for six
different geometries of the spline on the shaft comprising of three
different number of teeth (Z = 20, 30, 40) and two different pressure angles (a = 30, 45) with seven different hardness depths values DH = 0, 2.5, 5.0, 6.0, 7.0, 8.0 and 1, summing up to a total of 42
different models. For each run the torque T and the end twist of the
shaft is monitored, from which the corresponding shear strain c
can be calculated. In Fig. 6 the torque T versus the shear strain c
is plotted, which pertains to results with a = 30, all Z and DH values. As the gure reveals the hardness depth has a strong effect on
the mechanical response T vs. c of the splined shaft. Also the number of spline teeth Z affects the level of static torque the shaft can
carry. It was found that the pressure angle a has a very minor inuence on the torsion strength and hence the results for a = 45 are
not reported here.
The torque value dening the torsional strength is here determined with the 0.2% offset method applied to Fig. 6 and is denoted
Tp02. In Fig. 7 the torsional strength Tp02 as function of the normalized hardening depth DH/rp and the number of spline teeth Z is
shown. Experimental results of torsion strength obtained from
[18] for two different steels, SAE 4140 and AISI 1040, are also
shown in Fig. 7. Although the static torsion tests in [18] are performed on round non-splined shafts with various hardness depths,
the agreement between the experiments and the FEM results are
reasonably well.
As evidenced from Fig. 7, the hardness depth has a large impact
on the increase in torsional strength. Especially in the range of
hardness depths 0 6 DH/rp 6 0.5 where the increase is most noted.
For depths larger than 50% of the pitch radius rp the torsional
strength increases marginally. In view of these results it seems that
a ratio of hardness depth over pitch radius (DH/rp) of about 50%
would generally give an optimum increase in torsional strength.
This optimum is closely associated with the transition in failure
mode of hardened shafts subjected to torsion load. The two failure
modes that can occur are either sub-surface failure or failure initiated at the surface both of which take place at the location where

Fig. 7. Torque T versus hardness depth DH and number of spline teeth Z = 20, 30 and

the applied shear stress exceeds the shear strength. The torsional
yield strength can be obtained from Eq. (5) by the approximation
0.577 Rp02 whereas the shear stress is given by 2Tr=pr4p . These
two expressions are plotted in Fig. 8 for DH/rp = 0.26 and 0.63.
For DH/rp = 0.26 the stress (dashed line) exceeds the strength (solid

Fig. 8. Strength versus stress and normalized hardness depth DH/rp.

I. Barsoum et al. / Materials and Design 54 (2014) 130136


Fig. 9. Contour plot of PEEQ for hardness depth DH = 5, 6, 7 and 8 mm.

line) at the sub-surface whereas for DH/rp = 0.63 the stress exceeds
the strength at the surface. The transition between the two failure
modes occurs at the optimum hardness depths at about DH/rp = 0.5.
Increasing the hardness beyond this depth would generally not
give higher torsional strength since it is mainly controlled by the
hardness level of the surface. For DH/rp values lower than the optimum the torsional strength is a strong function of the hardness
depth but also the hardness level of the core.
It should however be pointed out that this analysis and its results are for a splined shaft loaded with a static torque. In the case
the shaft is exposed to dynamic loading the failure is controlled by
fatigue or wear of the splines for which shear stress and strength
are not suitable quantities. A similar analysis as in Fig. 8 can be
undertaken, however with stress measures in accordance with
e.g. the Findley criterion [31] or an equivalent multiaxial fatigue
stress measure [32].
As shown in Fig. 7 the number of spline teeth Z has also an inuence on the overall increase in torsion strength of the shaft. However, this is not as marked as in the inuence of hardness depth on
strength. An increased number of involute spline teeth Z, i.e. decrease gear module m, give an increase in torsion strength. The increase in strength with number of teeth is attributed to that the
torque is shared by a larger number of teeth also resulting in a
more uniform distribution of load. Nevertheless, the results from
this FEM model for an induction hardened shaft where the spline
geometry is accounted for suggests that the largest contribution
to the increase in strength is due to the material hardness depth.
In Fig. 9 contours of the equivalent plastic strain (PEEQ) through
the cross section shown for one spline and hardness depths of
DH = 5, 6, 7 and 8 mm. The contours represent the instant of initial
yielding corresponding to the torque at a 0.2% offset shear strain in
the torque versus shear strain diagram in Fig. 6. One can note that
the larger the hardness depth the lesser the extent of the plastic
zone and an increased resistance to yielding. With further increase
in the hardness depth, the extent of the plastic zone hardly changes
which is manifested in the minor increases in torsion strength for
hardness depth values larger than DH/rp = 0.5 as noticable in Fig. 7.
7. Conclusions
The current study presents a nite element modeling framework to obtain the torsion strength of a hardened splined shafts

by taking into account both the detailed geometry of the spline

and the material gradation in the model. It can be concluded that
increasing the number of teeth Z, which corresponds to increasing
the dedendum radius rd and keeping the addendum radius ra constant, increases the torsion strength of the shaft. This increase is
also attributed to that the torque is distributed over a larger number of spline teeth. It is also found that the pressure angle a has no
marketable effect on the strength. The largest contribution to the
increase in torsion strength is however due to the increase in hardness depth DH. The numerical results of this study and experimental results in [18] agree well and suggest that there is an optimum
hardness depth for hardened shafts, which is approximately half
the shaft radius. A hardness depth larger than this would not have
any higher inuence on the torsional strength.
The authors would like to thank Mr. Gregory A. Fett at Dana
Corporation in Maumee, Ohio, for sharing his thoughts on the
experimental results.
[1] Yang DCH, Tong SH. On the prole design of transmission splines and keys.
Mech Mach Theory 2007;42(1):827.
[2] Yan HC, Lin LD, Fang W, Tao CH. Fracture analysis of spline shaft. J Mater Eng
[3] Limmer L, Nowell D, Hills DA. A combined testing and modeling approach to
the prediction of the fretting fatigue performance of splined shafts. J Aerospace
Eng 2001;215(2):10512.
[4] Clarke CK, Halimunanda D. Failure analysis of induction hardened automotive
axles. J Fail Anal Prev 2008;8(4):38696.
[5] Tjernberg A. Load distribution and pitch errors in a spline coupling. Mater Des
[6] Tjernberg A. Load distribution in the axial direction in a spline coupling. Eng
Fail Anal 2001;8:55770.
[7] Loganathan TM, Purbolaksono J, Inayat-Hussain JI, Wahab N. Effects of
carburization on expected fatigue life of alloys steel shafts. Mater Des
[8] Zabett A, Azghandi SHM. Simulation of induction tempering process of carbon
steel using nite element method. Mater Des 2012;36:41520.
[9] Kristoffersen H, Vomacka P. Inuence of process parameters for induction
hardening on residual stresses. Mater Des 2001;22:63744.
[10] Kayacan MC, Colak O. A fuzzy approach for induction hardening parameters.
Mater Des 2004;25:15561.
[11] Tjernberg A. Comparison of methods for determining residual stresses in
induction-hardened transmission shafts. Scand J Metall 1999;28:1628.


I. Barsoum et al. / Materials and Design 54 (2014) 130136

[12] Tjernberg A. Fatigue lives for induction hardened shafts with subsurface crack
initiation. Eng Fail Anal 2002;9:4561.
[13] Leen SB, McColl IR, Ratsimba CHH, Williams EJ. Fatigue life prediction for
barrelled spline coupling under torque overload. Proc Instit Mech Eng Part G: J
Aerospace Eng 2003;217(3):12342.
[14] Sum WS, Leen SB, Williams EJ, Sabesan R, McColl IR. Efcient nite element
modelling for complex shaft couplings under non-symmetric loading. J Strain
Anal Eng Des 2005;40(7):65573.
[15] Nanaware GK, Pable MJ. Failures of rear axle shafts of 575 DI tractors. Eng Fail
Anal 2003;10(6):71924.
[16] Lin CY, Hung JP, Hsu TC. Failure analysis of reverse shaft in the transmission
system of all-terrain vehicles. J Fail Anal Prev 2008;8(1):7580.
[17] Kang DH, Lee BJ, Yun CB, Kim KW. Study on torsional strength of inductionhardened axle shaft. Trans Korean Soc Mech Eng A 2010;34(5):6459.
[18] Fett GA. Importance of induction hardened case depth in torsional
applications. Heat Treating Prog 2009;9(6):159.
[19] G. Tortorella, W. Passos, G. Flesch, Torsional strength for induction hardened
shafts applied in homocinetic half shafts, SAE technical paper series 2004013231; 2004. p.16.
[20] Straight cylindrical involute splines Metric module, side t Part 1:
Generalities. ISO 41561:2005. 200510.
[21] Involute splines based on reference diameters Part 1: General. DIN 54801,

[22] Involute splines based on reference diameters Part 2: Nominal and

inspection dimensions. DIN 54802, 200603.
[23] H. Kristoffersen, S. Haglund, Inuence of straightening on bending fatigue
properties of induction hardened shafts in steel 42CrMo4 (SAE 4140). In:
Proceedings of the 6th International Quenching and Control of Distortion
Conference, Chicago, IL; 2012.
[24] Starke P, Walther F, Eier D. New fatigue life calculation method for quenched
and tempered steel SAE 4140. Mater Sci Eng A 2009;523(12):24652.
[25] JMATpro 4.1. The Materials Property Simulation Package, Sente Softare Ltd.
[26] Saunders N, Guo Z, Li X, Miodownik AP, Schill JPh. Using JMatPro to model
materials properties and behavior. JOM 2003;55(12):605.
[27] MackAldener M, Olsson M. Tooth interior fatigue fracture computational and
material aspects. Int J Fatigue 2000;23:28392.
[28] Fajdiga G, Flaker J, Glodez S, Hellen TK. Numerical modelling of micro-pitting
of gear teeth anks. Fatigue Fract Eng Mater Struct 2003;26(12):113543.
[29] ABAQUS 2009. Standard users manual. version 6.9. Dassault Systmes
[30] ASM metals handbook: properties and selection, 8th ed. 1, OH: ASM; 1961 p.
[31] Findley WN. A theory for the effect of mean stress on fatigue of metals under
combined torsion and axial load or bending. J Eng Ind 1959:3016.
[32] Socie DF, Marquis GB. Multiaxial Fatigue. SAE International; 2000.


Analysis of the torsional strength of hardened splined shafts