0DVWHU
V'HJUHH7KHVLV
,651%7+$07(;'6(


'HYHORSLQJDQ$GYDQFHG6SOLQH

)DWLJXH3UHGLFWLRQ0HWKRG





 


$EGDOODK=DUDG


'HSDUWPHQWRI0HFKDQLFDO(QJLQHHULQJ
%OHNLQJH,QVWLWXWHRI7HFKQRORJ\
.DUOVNURQD6ZHGHQ



6XSHUYLVRU -RKDQ:DOO%7+
    

 Developing an advanced spline
fatigue prediction method

Abdallah Zarad

A thesis in the field of fracture mechanics

for a Degree of Master of science in mechanical engineering
with emphasis on structural mechanics

Blekinge tekniska högskola

October 2019
2
 ACKNOWLEDGEMENT

I would first like to thank my thesis supervisor Prof. Johan Wall of the Mechanical
engineering department at Blekinge Tekniska Högskola. He was always available
whenever I face a trouble or has a question regarding my research work and reporting.
He steered me in the right direction throughout the research.

I would also like to thank my colleagues of the NVH and Virtual analysis team at
GKN automotive, Köping Eng. Karthik Pingle and Eng. Yi Ma. They were always
there for me by academic assistance and personal support. Without their continuous
participation and advice, the research could not have been successfully conducted.

Finally, I must express my very profound gratitude to my parents for providing me

with unfailing support and continuous encouragement. Thank you to my father,
Mohammed Zarad, for supporting me, believing in me and encouraging me to chase
my dream.

Thank you to my mother, Nagia Ibrahim, for her huge effort and sacrifices
throughout my years of education. I would not have been where I am today and what I
am today if it was not for you.

This one is for you Mom!

3
Abstract

Fatigue failure is one of the most critical issues in industry nowadays as 60 to 90 percent of failures
in metals are due to fatigue. Therefore, different methods and approaches are developed to
estimate the fatigue life of metallic parts. In this research, a case-hardened steel splined shaft is
studied to estimate the fatigue life that the shaft will withstand before failure. The purpose of the
research is to develop an advanced fatigue prediction method for splines.

A static experimental test was performed on the splined shaft for analyzing the load-strain behavior
of the shaft and determining the suitable load cases of the study. A dynamic test of pure torsional
load was carried out to collect experimental results for validating the generated fatigue methods
and investigating the failure behavior of the shaft. Stress analysis was performed on the part for
investigating critical areas and the effect of the different spline teeth designs on the resulting stress.
Two finite element models were analyzed using two software, MSC Marc software with a geometry
of straight spline teeth and Spline LDP with an involute spline teeth model. DIN 5466-1 spline
standard’s analytical solution was used for verification purposes. Stress and strain-based
approaches were used to estimate fatigue life. The most suitable method was evaluated against
experimental test results.

The research findings show that the most critical stress areas on the shaft are the spline root fillet
and relief. When the part fails due to fatigue the crack initiates at the root fillet and propagates to
the relief. It is also shown that involute teeth spline gives higher stress than straight teeth for the
same load due to less contact area.

The conclusion of the research could be summarized in: the stress-based method (Wöhler curve) is
giving good accuracy and proved a reliable method. While among six different approaches used of
strain-based methods, four-point correlation method is giving the best correlation to test results.
Hence, it is recommended to use four-point correlation method for fatigue analysis for its accuracy
and for considering both elastic and plastic behavior of the material.

4
Nomenclature

z Number of teeth
Normal module
Normal pressure angle
d Reference diameter
Base diameter
Major diameter
Form diameter
Minor diameter
Shaft inner diameter
Equivalent diameter
Effective tooth thickness
s Actual tooth thickness
L Spline length
Effective space width
e Actual space width
Geometry factor
Shape factor
Root fillet radius
Torsional shear stress
Maximum torsional shear stress
Torsional moment of resistance
Depth normal to the surface, axial coordinate
Yield strength
Ultimate strength
E Young’s Modulus
Poisson ratio
Stress amplitude
́ Fatigue strength coefficient
True fracture strength
Ultimate tensile strength
Ultimate tensile strength
Yield strength
Material strength at 10 cycles
̀ uncorrected endurance limit
Corrected endurance limit
b Fatigue strength exponent/Wöhler curve slope
k Wöhler curve inverse slope/Wöhler curve factor
Fatigue ductility coefficient
True fracture ductility
Elastic strain amplitude
5
 Plastic strain amplitude
Total strain amplitude
Number of cycles to failure
c Fatigue ductility exponent
HB Brinell hardness
HV Vickers hardness
HRA Hardness at Rockwell scale
T Torque

6
Table of Content
ACKNOWLEDGEMENT ............................................................................................................................ 3
Abstract .................................................................................................................................................. 4
Nomenclature ........................................................................................................................................ 5
Table of Figures ...................................................................................................................................... 9
1 Introduction ...................................................................................................................................... 11
1.1 Background ................................................................................................................................. 11
1.2 Scope .......................................................................................................................................... 14
1.3 Outline ........................................................................................................................................ 15
2 Theory ............................................................................................................................................... 16
2.1 Gear geometry............................................................................................................................ 16
2.2 Fatigue ........................................................................................................................................ 18
2.2.1 Stress-based approach ........................................................................................................ 18
2.2.2 Strain-based approach ........................................................................................................ 20
2.3 Stress calculation ........................................................................................................................ 22
2.4 Mesh ........................................................................................................................................... 23
2.5 Rigid and constraints elements .................................................................................................. 25
3 Method.............................................................................................................................................. 27
3.1 Stress analysis ............................................................................................................................. 28
3.1.1 Finite element analysis........................................................................................................ 29
3.1.2 Contact and root analysis.................................................................................................... 36
3.1.3 Analytical approach ............................................................................................................. 38
3.2 Fatigue ........................................................................................................................................ 40
3.2.1 Stress-based approach ........................................................................................................ 40
3.2.2 Strain-based approach ........................................................................................................ 42
3.3 Experimental tests ...................................................................................................................... 45
3.3.1 Static test............................................................................................................................. 45
3.3.2 Dynamic test ....................................................................................................................... 48
4 Results and discussion ...................................................................................................................... 51

7
 4.1 Stress analysis ............................................................................................................................. 51
4.1.1 MSC Marc ............................................................................................................................ 51
4.1.2 Spline LDP............................................................................................................................ 53
4.1.3 Analytical results ................................................................................................................. 55
4.1.4 Stress comparison ............................................................................................................... 56
4.2 Strain analysis ............................................................................................................................. 57
4.3 Fatigue results ............................................................................................................................ 60
4.3.1 Stress-based fatigue ............................................................................................................ 60
4.3.2 Strain-based fatigue ............................................................................................................ 64
5 Conclusion and future work .............................................................................................................. 71
5.1 Conclusion .................................................................................................................................. 71
5.2 Future work ................................................................................................................................ 72
References............................................................................................................................................ 73

8
Table of Figures
Figure 1: GKN Twinster unit ................................................................................................................. 13
Figure 2: eTwinster spline shaft ........................................................................................................... 14
Figure 3: Shaft and hub ........................................................................................................................ 14
Figure 4: Gear geometry and parameters [2] ...................................................................................... 17
Figure 5: Illustration of Wöhler curve .................................................................................................. 20
Figure 6: Nomenclature for hysteresis loop [1] ................................................................................... 21
Figure 7: Illustration of strain-life curve .............................................................................................. 22
Figure 8: Illustration of aspect ratio calculation method .................................................................... 23
Figure 9: Illustration of skew calculation method ............................................................................... 23
Figure 10: Tria and quad elements ...................................................................................................... 24
Figure 11: 3D element types [6]........................................................................................................... 24
Figure 12: First and second order tetrahedral elements [10] ............................................................. 25
Figure 13: Force distribution using RBE2 [12]...................................................................................... 26
Figure 14: Force distribution using RBE3 [12] ...................................................................................... 26
Figure 15: Cad model of the ODC shaft ............................................................................................... 30
Figure 16: Cad model of the companion flange................................................................................... 30
Figure 17: Shaft with mesh shell of linear triangular elements ........................................................... 31
Figure 18: Linear triangular element ................................................................................................... 31
Figure 19: Shaft with tetra mesh of second order (tet10) ................................................................... 32
Figure 20: Second order tetrahedral element (Tet10 element) .......................................................... 32
Figure 21: ODC shaft and companion flange fitting............................................................................. 33
Figure 22: RBE2 and RBE3 applied on the companion flange and the shaft, respectively. ................. 35
Figure 23: RBE2 and RBE3 elements .................................................................................................... 35
Figure 24: Input data for external and internal splines ....................................................................... 37
Figure 25: Five teeth of the internal and external splines mesh ......................................................... 37
Figure 26: Illustration of the torsional load applied on the ODC shaft................................................ 38
Figure 27: Static test setup with the strain gauges attached to the spline ......................................... 46
Figure 28: Strain gauges position on the shaft [17] ............................................................................. 46
Figure 29: T-ε curve at the bearing seat .............................................................................................. 47
Figure 30: T-ε curve at spline relief ...................................................................................................... 47
Figure 31: Torsional dynamic test setup [17] ...................................................................................... 48
Figure 32: Torsional dynamic test results ............................................................................................ 49
Figure 33: Dynamic test results with B10 and B50 fitting.................................................................... 50
Figure 34: Initial contact analysis between the teeth ......................................................................... 51
Figure 35: Contact of the drive side on the spline shaft ...................................................................... 51
Figure 36: Stress distribution on the shaft, T = 3800 Nm .................................................................... 52

9
Figure 37: Stress distribution at critical areas, T = 3800 Nm ............................................................... 52
Figure 38: Stress results using MSC Marc ............................................................................................ 53
Figure 39: Contact stress distribution along roll angle and spline face width for one tooth, T = 3800
Nm ........................................................................................................................................................ 54
Figure 40: Contact stress and load distribution along each tooth, T = 3800 Nm ................................ 54
Figure 41: Root stress distribution, T = 3800 Nm ................................................................................ 55
Figure 42: Root fillet maximum stresses calculated by different tools ............................................... 57
Figure 43: Strain distribution on the shaft, T = 2800 Nm .................................................................... 58
Figure 44: Strain distribution at critical areas, T = 2800 Nm ............................................................... 58
Figure 45: Maximum value of calculated and measured strain at the spline relief ............................ 59
Figure 46: Strain values at spine relief ................................................................................................. 60
Figure 47: Generated S-N Wöhler curve .............................................................................................. 61
Figure 48: Correlation between experimental life and predicted life with Wöhler curve .................. 62
Figure 49: Correlation between B10 experimental life and predicted life with Wöhler curve ........... 63
Figure 50: Correlation between B50 experimental life and predicted life with Wöhler curve ........... 63
Figure 51: Maximum strain at spline relief and root fillet ................................................................... 64
Figure 52: Elastic, plastic and total strain fatigue curves - Original universal slopes method ............ 65
Figure 53: Total strain fatigue curve - Original universal slopes method ............................................ 65
Figure 54: Total strain fatigue curves for different approaches .......................................................... 66
Figure 55: Experimental and predicted life correlation at spline relief - Strain-based approaches ... 67
Figure 56: Experimental and predicted life correlation at root fillet - Strain-based approaches ....... 68
Figure 57: B10 experimental and predicted life correlation at root fillet – Strain-based approaches 69
Figure 58: B50 experimental and predicted life correlation at root fillet – Strain-based approaches 69

10
1 Introduction

This thesis work was carried out at the product development department of GKN Driveline Köping
AB, Köping, Sweden. GKN is a global company active in automotive, aerospace, powder technology
and land systems. GKN Driveline is the world’s leading manufacturer of automotive driveline
components and the factory in Köping is focusing on manufacturing and developing gears and
components for AWD (All-Wheel-Drive) vehicles.

1.1 Background

Fracture is a problem that society has faced since there were simple structures and parts, even
before engineers start using machines in building more complicated ones. And despite the wide
improvement in science and the availability of more tools, the problem is worse today than in
previous centuries due to the complexity of the current used technology and structures.

Components subjected to repeated or fluctuating load may break. This type of failure is called
fatigue. Despite that this phenomenon has been known for more than 150 years, the fatigue failure
is still one of the most common threats for metals as 60 to 90 percent of mechanical failures are
due to fatigue. [24]

There are different types of fatigue, the most common one is mechanical fatigue. This type of
fatigue is related to the mechanical strength. It is usually caused by inappropriate material choice,
manufacturing faults, weak welds, or design inadequacy.

Most authors and researchers within the fatigue area have agreed upon defining three stages of
fatigue fracture –regardless the type of fatigue-: initiation, propagation and final rupture, [15].
Indeed, this is to simplify the study of fatigue as it can become complex.

1. Stage 1- Initiation. It is the most complex and most studied stage of fatigue fracture. There
are irreversible changes that happen in this stage due to cyclic shear stresses. The
cumulative damage that happens over a large cycle of load is the most important subject of
study in this stage.

11
 2. Stage 2- Propagation. In this stage, the microcrack changes direction and grow perpendicular
to the tensile stress. It is easier to identify the crack and the fatigue failure in this stage.
3. Stage 3- Final rupture. The component is weakened significantly that complete fracture can
occur with few more load cycles in this stage. As the fatigue crack gradually reduce the
cross-sectional area of the part or component.

There are different methods and approaches one can follow to predict the fatigue life. Depending
on the material, the geometry, the type of load applied and different parameters the number of
cycles to failure can be calculated.

Some engineers dedicated their whole career to specialize in fatigue fracture. Since it is one of the
most important concerns in all the manufactured parts subjected to frequent cyclic stresses in
different industries. Examples of industries include:

ƒ Aerospace

The components used in the aerospace industry usually subjected to intense stress and extreme
temperature fluctuations. This makes material choice and prototyping of high importance.

ƒ Automotive

Much like aerospace, parts and components used in manufacturing of automobiles experience a
wide range of environmental and operational risks which subject it to different types of fatigue.

During the past 20 years there has been a high interest in improving quality and reliability of
manufactured products in the automotive industry due to the strong competition within the market
and the high customer demands and requirements.

GKN developed Twinster technology which offers enhanced traction, driving dynamics and stability
for all-wheel drive. Twinster makes the drive safe in all weather and road conditions by constantly
managing the distribution of torque between the front and rear axles and between the rear wheels,
see Figure 1 [19]. In this research, the outer disc carrier (ODC) spline of a Twinster unit is studied.

12
Figure 1: GKN Twinster unit

13
1.2 Scope

The purpose of this thesis project is to develop a fatigue life prediction method for a spline shaft,
Figure 2. The shaft is fitting in a companion flange which rotates with the shaft during operation,
see Figure 3. The shaft is the outer disc carrier of a Twinster, see Figure 1. Considering the
geometry, material properties and torsional load applied it should be possible to calculate the
number of cycles that the part could withstand.

Figure 2: eTwinster spline shaft

Figure 3: Shaft and hub

14
To achieve best results and understand the phenomenon, static and dynamic physical tests were
performed on the studied part, see section 3.3.

The outer disc carrier spline shaft is carrying high torsional load during operation, which increase
the chance of failure. Hence, an accurate study and analysis of the part needs to be performed in
order to check if the designed part meets the customer requirements and achieve the required
number of cycles.

The analysis of the shaft together with the physical tests are used to develop a fatigue prediction
method which could be generalized and used for different splines. Different designs of the spline
teeth are also studied in order to investigate the effect of involute teeth geometry on the root
bending stress.

The following research questions are used as guidelines to achieve the goal of the project:

x How does involute tooth design affect the stress on splines?

x What approaches could be followed to predict fatigue lives for the ODC splined shaft, and
how accurate are those approaches?

1.3 Outline

x Literature research and understanding of the spline strength calculation standard (DIN 5466-
1) which considers operating conditions, calculations of the operating states and stresses,
durability and wear estimation of splines.
x Comparison and selection of appropriate calculation methods from literature, previous
researches and developed approaches. Then scripting the methodology using MATLAB.
x Stress analysis of the studied shaft using FEM and theoretical calculations.
x Combining with the existing test results and developing fatigue prediction method using
scripting.
x Comparison of developed methodology vs classic FEM.
x Validating the developed method against test results.

15
 2 Theory

In this chapter, the scientific principals behind the used methods are explained. Basic definition of
gear geometry is presented. Then the theory behind the used fatigue prediction methods is
explained in a detailed way. Finally, the used finite element analysis mesh and elements are
discussed. The purpose of this chapter is to shed light on the data in a scientific manner.

2.1 Gear geometry

As the spline shaft and hub have straight teeth, they can be dealt with as spur gears.

ƒ Module , the reference diameter divided by the number of teeth, equation (1).
(1)
= =

ƒ Pitch
-Pitch circle p, the distance between beginning of two close teeth on the reference circle,
see equation (2) and Figure 4.
∙ (2)
=

-Base circle , the distance between beginning of two close teeth on the base circle, see
equation (3) and Figure 4.
∙ (3)
=

16
 Figure 4: Gear geometry and parameters [2]

From Figure 4 we can see the different parameters that describe the geometry of a gear.

17
2.2 Fatigue

Fatigue life prediction methods can be divided into two main groups according to the used
approach. The first group consists of models based on the prediction of the crack initiation and
propagation using a criterion based on the stress/strain of components. The remarkable point about
this approach is the lack of dependence from loading and specimen geometry. Which could be
useful in this research as the generated method could be generalized and used with different
splines’ geometries subjected to different loads.

The approach of the second group is based on the continuum damage mechanics, in which fatigue
life is calculated by determining the damage parameter cycle by cycle. [3] The first group approach
will be used in this study.

In this research, a uniaxial fatigue analysis is assumed. A uniaxial fatigue analysis is used to estimate
the fatigue life of a specimen under cyclic loading, when the crack is due to a uniaxial state of stress.
[1] Uniaxial stress and strain-based methods are used in this study. Where the stress-based method
is a generated relation (curve) between the number of cycles and the applied stress amplitude. And
the strain-based approach is a relation between the number of reversals and strain amplitude.

2.2.1 Stress-based approach

The very first systematic study of metal fatigue is the S-N curve (also known as Wöhler curve), which
is a relation between stress amplitude and number of cycles that the specimen could withstand
before failure. The S-N curve is dependent on the material without considering the geometry or the
way of loading. The stress and life relation can be generated by performing a physical test on a
uniform material specimen at a constant amplitude load. But unfortunately, this data is not
available for all existing materials. So, approximated values that match the studied material the
most are used. For example, if it is needed to plot the S-N curve for low-alloy steel part that consists
of chromium, Nickel, Magnesium and carbon. A S-N curve that describes a similar low alloy material
could be found which can be applied to the case of study. Basically, those existing methods are
somehow generalized to be applied on different materials with different specifications but only with
a small variation in the material property. And the fatigue analysis is always about finding the best
approximation method which will lead to the most accurate results.

The equation that describes Wöhler curve is as follow

(4)
= ́

18
The curve is usually divided into three regions:

x Low cycle fatigue: Between 1 and 10 cycles.

x High cycle fatigue: From 10 to 10 cycles.
x Infinite life region: For cycles more than 10 , see Figure 5.

The fatigue strength obtained from standard fatigue-test specimens or from estimates based on
static test needs to be modified. There are differences that need to be considered between the test
specimen and the actual part. Those differences are represented by strength-reduction factors that
are multiplied by the theoretical estimate, equation (5), to obtain corrected fatigue strength for the
case of study. [14]
= (5)
̀

Where
x : a reduction factor depends on the type of load (bending, axial or torsional).
x : The rotating beam and static test specimens are small, if the part is too large, a
strength reduction size factor needs to be applied to consider the fact that larger parts have
high probability to fail at lower stress due to the large stressed volume.
x : Rougher surface finishes will lower the fatigue strength because of the stress
concentrations and the alternation of the physical propertied of the surface layer.
x : The fracture toughness decreases at low temperature and increases at high
temperatures.
x : The 50% reliability has a factor of 1 and the factor reduces as higher reliability is
chosen.

For calculations of the previous factors, see section 3.2.1.

19
 Figure 5: Illustration of Wöhler curve

2.2.2 Strain-based approach

In section 2.2.1 Stress-based approach, the stress-based fatigue analysis is presented. This approach
works well for situations in which only elastic stresses and strains take place. While in some cases
local cyclic plastic deformation is presented. So, another approach that uses the local strain as the
governing fatigue parameter was developed in the late 1950s and has shown to be more effective in
predicting the fatigue life of a component. [15]

This approach usually starts by studying the hysteresis loop, which describes the elastic and plastic
behavior of the material under loading and unloading. The material deformation during a fatigue
test is measure with the hysteresis loop. Each strain range tested has a corresponding measured
stress range. The hysteresis loop is a plot of all this data, see Figure 6. [20]

20
 Figure 6: Nomenclature for hysteresis loop [1]

Based on Morrow’s proposal (1965), the relation between the total strain amplitude and fatigue life
can be expressed as in equation (6) below

́ (6)
= + = 2 + 2

́ = The fatigue strength coefficient

= The fatigue ductility coefficient

b = The fatigue strength exponent

c = The fatigue ductility exponent

The previous equation is called the strain-life equation for zero mean stress case. This equation is a
summation of two separate curves, the elastic strain amplitude versus life ( -2 ) –developed by
Basquin, 1910- and the plastic strain amplitude versus life ( -2 ) –developed by Manson, 1953
and Coffin, 1954- [1], see Figure 7.

21
 Figure 7: Illustration of strain-life curve

2.3 Stress calculation

There are different analytical and experimental ways for calculating the stress. One of the most
critical stress areas for a gear or spline is the root fillet, which is subjected to a root bending stress.
The root bending stress depends on the type of loading, the geometry of the teeth (root and tip
diameters-root fillet radius-module) and the number of teeth. [4] and [5]

Finite element analysis is proved an accurate and reliable method for structural analysis. The
general concept of FEM is to divide the whole geometry of interest which has infinite degrees of
freedom into finite number of elements which are analyzed individually then combine the resulting
values to study the component’s behavior.

22
2.4 Mesh

Meshing the geometry is the main step in preparing the model before performing a finite element
analysis. Good quality mesh increases the similarity between the original component and the
created simulation which gives more accurate results and increases the reliability. The mesh is
described by the element shape. There are two 2D elements types – tria (triangular) and quad
(quadrilateral), which has 3 and 4 nodes, respectively, see Figure 10. Since 2D elements can only
support translational DOF and in-plane loading, the thickness of the analyzed model must be small
relative to its length and width. [7] Each linear triangular element has six degrees of freedom where
each node contributes two degrees of freedom. The mesh quality of 2D triangular elements is
measured by different factors:

x Aspect ratio: the ratio of the longest edge of an element to either its shortest edge or the
shortest distance from a corner node to the opposing edge (‘’Highest to closest node). See
Figure 8, [9]

Figure 8: Illustration of aspect ratio calculation method

x Skew: skew of triangular elements is calculated by finding the minimum angle between the
vector from each node to the opposing mid-side, and the vector the two adjacent mid-sides
at each node of the element. See Figure 9, [9]

a b

Figure 9: Illustration of skew calculation method

–
x Equilateral-volume-based Skewness = [8]
x Minimum and maximum angle of trias elements.

23
 Figure 10: Tria and quad elements

There are four different 3D element types – tets, bricks, prisms and pyramids, see Figure 11. Those
four elements can always be used, in many combinations, to mesh any 3D model. Tetrahedral
elements are the most popular type of elements for its ability to mesh any 3D volume, regardless of
its shape or topology. They are also the only kind of elements that can be used with adaptive mesh
refinement. On the other hand, the three elements types (bricks, prisms and pyramids) require
some effort from the user to create such a mesh. So, they should not be used until it is motivated to
do so. One of the main motivations for using bricks and prisms is reducing the number of elements
in the mesh. [6] Tetrahedral elements can be divided according to the shape function into two
different types:

x First order elements, a 4-node element with three degree of freedom at each node resulting
in a total of 12 DOF, see Figure 12.a.
x Second order elements, a 10-node element with 30 DOF, see Figure 12.b.

Figure 11: 3D element types [6]

24
 Figure 12: First and second order tetrahedral elements [10]

One of the quality checks for tetra elements is the tet collapse value, which is calculated by dividing
the distance from each of the four nodes to the opposite side of the element by the square root of
this area, the minimum value found is normalized by 1.24 and then reported as tet collapse value.

2.5 Rigid and constraints elements

Rigid element (node to node connection) transfers all forces and moments from one node to
another as it is. There are different types of rigid elements as Rbar, Rigid, Rigid link. While on the
other hand there are different multi point constraints (single node connected to many nodes) as
RBE2 and RBE3, which are preferred over rigid elements. [12]

Rigid and RBE2 add stiffness to the original structure, while RBE3 does not. One important
difference between RBE2 and RBE3 is that RBE2 connects one independent node to one or more
dependent nodes, while RBE3 connects one dependent node to one or more independent nodes.
So, if we assumed 100 N force is applied at RBE2 element as shown in Figure 13, it will get equally
distributed between the 2 nodes. While in RBE3 as shown in Figure 14, it will be distributed as 75 N
at point B and 25 N at point A. [12] Hence, RBE2 is usually used for fixing elements while RBE3 are
more common in applying a load through elements.

25
Figure 13: Force distribution using RBE2 [12]

Figure 14: Force distribution using RBE3 [12]

26
3 Method

In this chapter, the different parts of the research are tied together. As explained in chapter 2
Theory, the stress or the strain is the input and the number of cycles is the output in followed
fatigue approaches. The different methods, conditions and assumptions for calculating stress, strain
and fatigue lives are shown and explained in this chapter.

The work stages through the research could be divided as follows:

Work stages

Strain-based
Stress analysis Stress-based fatigue
fatigue

Basquin-Coffin-
MSC Marc Manson equation
Wöhler curve
Original
universal
Spline LDP slopes

Uniform
material law
DIN 5466-1

Four-point
correlation

Modified four-
point
correlation

Modified
universal
slopes

Hardness

27
 3.1 Stress analysis

For studying the fatigue failure of the twinster ODC (Outer Disk Carrier) shaft, the stress acting on
the part needs to be investigated. As stress is the translation of the applied load and the input
parameter in fatigue prediction methods. The stress analysis was carried out using different
methods and tools for verification and accuracy comparison of those methods. In this study,
mathematical and experimental methods will be used to calculate the stress using analytical
calculations -Spline standard- and CAE tools -MSC Marc and Spline LDP-. The choice of those tools
depends on the working conditions, the geometry of the studied part and the availability of the
tools. MSC Marc is a widely used finite element solver which were tested throughout previous
analysis and experiments. Spline LDP software is a user-friendly tool that considers the micro-
geometry of splines which gives more accurate stress results. Spline standard is used for checking
how reliable is the analytical solution compared to numerical and experimental ones.

The following material data are given:

Table 1: Material data

Material Youngs Modulus Poission ratio Yield strength Ultimate tensile

[GPa] [MPa] strength [MPa]
18CrNiMo7-6H 210 0.3 2130* 2464*
Case hardened
*Material parameters based on surface hardness of material, which is 746 HV ~ 80-83 HRA

28
 3.1.1 Finite element analysis

3.1.1.1 Hypermesh

Hypermesh software is used in this part due to its wide variety of choices and flexibility. It can also
deal with complex geometries. Some steps were followed to generate a tetra mesh: [11]

1) Importing the cad models –.stp file format- of the ODC shaft and the companion flange to
Hypermesh’s graphical interface as in Figure 15 and Figure 16.
2) Creating a new component in which a 2D triangular element mesh will be saved.
3) Create a mesh shell of triangular elements. See Figure 17 and Figure 18
4) Check the mesh for free edges and T-connections and fix it.
5) Set aspect ratio to < 5, skew value to < 60, Equilateral Skew to < 0.6, minimum and
maximum angles limits of the trias elements to 20 and 120 degree, respectively.
6) Convert the tria mesh to tetra mesh.
7) Perform quality checks for the tetra mesh (tet collapse > 0.1) and improving the quality of
the mesh if required.
8) Convert generated first order tetra mesh to second order, see Figure 19 and Figure 20. This
finally resulted in 2701629 nodes and 1787981 elements for the shaft and the companion
flange. The shaft has more fine mesh -1191643 elements- as it is the part of interest.

29
 Figure 15: Cad model of the ODC shaft

Figure 16: Cad model of the companion flange

30
Figure 17: Shaft with mesh shell of linear triangular elements

Figure 18: Linear triangular element

31
 Figure 19: Shaft with tetra mesh of second order (tet10)

Figure 20: Second order tetrahedral element (Tet10 element)

32
The ODC shaft and the companion flange are fitted together after the parts being meshed. The
tolerance between the teeth is adjusted by fixing the companion flange and changing the
coordinates and angle of the shaft until a good mesh is observed. At this level, the most important
consideration is avoiding any interaction or overlapping between the teeth, even if it is not the
perfect mesh. See Figure 21

Figure 21: ODC shaft and companion flange fitting

33
 3.1.1.2 MSC Marc

MSC Marc Mentat is an advanced simulation software which will be used in this part as the finite
element analysis tool. As most of FEA software there are some steps that need to be followed for
preparing the model and carrying out the analysis. Those steps vary slightly from one software to
another. The following steps are followed in Marc Mentat:

1) Importing the geometry from Hypermesh as .bdf file format.

2) Define the geometrical properties of the parts as structural 3D solid.
3) Define the material as elastic-plastic isotropic case-hardened steel with a Young’s modulus
of 210 GPa and Poisson ratio of 0.3.
4) Define the contact bodies (ODC shaft and companion flange) as deformable bodies.
5) Create two contact interactions, one for initial contact and the other for stress analysis. Set
up the contact interaction between the two parts:
a. Contact type: touching
b. Contact tolerance = 0.1
c. Bias factor = 0.85
d. Stress-free projection onto contact surface at initial contact.
e. Friction coefficient = 0.12
6) For applying boundary conditions: Create RBE2 for fixed displacement and RBE3 for torsional
load, see section 2.5 Rigid and constraints elements,Figure 22 and Figure 23.
7) Lock the companion flange in all displacements and rotations.
8) Create three different structural point loads of 3800, 3300, 2800 Nm, see section 3.3.1.
9) Create structural static load cases.
10) Create one job for initial contact analysis and three jobs for the three different load cases.
11) Run the jobs one by one and save the results to new files. Casi-iterative solver type was used
as it is generally better for dense models (Normally dominated by 3 Dimensional elements)
[23].
12) Check the initial contact analysis and if needed readjust the teeth mesh.

34
Figure 22: RBE2 and RBE3 applied on the companion flange and the shaft, respectively.

Figure 23: RBE2 and RBE3 elements

35
 3.1.2 Contact and root analysis

Spline LDP software is used at this point. Spline LDP is a quasi-static load distribution program for
splines. It can analyze side fit splines with and without back-side contact or major/minor fit. The
influence of micro-geometry (profile and lead modification) can also be included in the analysis. [13]
The remarkable strength point in Spline LDP for this analysis is that it considers the involute of the
teeth, which was not considered in the cad model used in the FEA in Marc mentat. So, Spline LDP
will be used to study the contact stress and the root bending stress which will be compared with the
results from Marc.

The following steps are followed for setting up the problem:

1) Choose system units and the type of fit.

2) Entre external and internal spline data. See Figure 24
3) Run an initial calculation and check the mesh. See Figure 25
4) Define the operating conditions as pure torsional load with the three different load cases of
interest (3800, 3300, 2800) Nm. See Figure 26
5) Setup the program control as:
a. Beginning of rotation position = 0.00 degree
b. End of rotation position = 360 degrees
c. Number of rotation increments = 2
d. Number of grids along facewidth = 40
e. Number of grids along profile = 20
6) Include FEA root stress.

36
 Figure 24: Input data for external and internal splines

Figure 25: Five teeth of the internal and external splines mesh

37
 Figure 26: Illustration of the torsional load applied on the ODC shaft

3.1.3 Analytical approach

For calculating the maximum torsional stress acting on the spline, spline standard ‘’DIN-5466-1’’ [4]
is used. For calculating the maximum stress of a twisted toothed shaft, the outer diameter of a
smooth unnotched shaft with the same torsional moment is determined. It is called
equivalent/replacement diameter and depends on the root diameter, tip diameter and a geometry
factor, see equation (7). [4]

(7)
= + ∙ ∙( − )

The geometry factor for calculating the equivalent diameter depends on the tooth geometry
of the shaft, see equation (8). [5]

( . ) (8)
= 1.3 ∙ + 0.06 ∙ + 0.23

From the torsional moment of resistance of a smooth shaft with equal torsional moment of
inertia, the torsional stress is calculated, see equation (9).

16 ∙ (9)
= = ∙
∙( − )

38
The maximum torsional stress in the roof fillet of the notched shaft is calculated from the
torsional stress and a geometry-dependent form factor , see equation (10) and (11).

= ∙ (10)
Where, [5]
. (11)
= 1.39 ∙ + 0.08 − 0.03 ∙

For calculating the maximum torsional stress using previous equations, the inner diameter of the
shaft should not exceed the maximum allowable diameter, see equation (12) and (13).

0≤ ≤ (12)
Where, [5]
. (13)
= ∙ (0.7 ∙ − 2.7 + 2 ∙ )

While the geometry factor, the equivalent diameter and the form factor are independent of the
inner diameter.

39
 3.2 Fatigue

Analytical solution is used for fatigue analysis in this research. As one of the purposes of the
research is to evaluate the experimental methods using CAE and all the software methodologies are
basically depending on analytical solutions and the physics behind it. Two different approaches are
followed, stress-based fatigue prediction approach and strain-based fatigue prediction approach.

3.2.1 Stress-based approach

For simplification purposes, the ODC shaft is assumed to be subjected to pure torsional load. There
is still a bending load from fixing the specimen, but this can be neglected since it only comes from
the weight. Therefore, uniaxial fatigue analysis methods are followed.

As discussed in section 2.2.1 Stress-based approach, the S-N diagram (Wöhler curve) will be used.
For creating an estimation of S-N diagram, and need to be calculated.

For ≥ 1400 , ̀ = 700 in case of steel. [14]

x Loading effect

The strength reduction load factor = 1, for pure torsion cases. [14]

x Size effect
According to the metal fatigue analysis handbook [1], for all steels:

= 1.0 for ≤ ,

For case hardened steel the effective diameter is calculated as equation (14). [1]
4 (14)
=

, = 16

Where and are the volume and surface of the section of the component of interest, see
equations (15) and (16). For the ODC shaft:
(15)
= ∙( − )∙
4

= ∙ ∙ (16)

40
 = 15.52
Since ≤ ,
Then = 1.0

x Surface effects

The splines of the shaft have not been subjected to surface finish. So, the surface factor will
be taken as one.

x Temperature effect

For ≤ 450 ℃, = 1 [14]

x Reliability effect

50 % Reliability is chosen as the calculations are compared to experimental test results. For 50 %
reliability, = 1 [14]

Multiplying all the previous factors by ̀, the corrected endurance limit is obtained.

= ̀ = 700

For estimating the strength at 10 cycles, according to [14]:

= 0.90 = 2215.8

Using this value of material strength and the corrected endurance limit , the S-N diagram can be
plotted in the high cycle fatigue region between 10 and 10 cycles. Then the curve is continued
horizontally beyond = 10 in the finite life region. See Figure 5

As mentioned above (4), the equation that describes Wöhler curve is as follow

= ́

For determining the unknown parameters in the latter equation, see equations (17), (18) and (19).
[14]

1 (17)
= log( )
Where

= log( ) − log( ) = −3 (18)

41
 = −0.1668

log ́ = log( ) − log( ) = log( )−3 (19)

log ́ = 3.84

́ = 7014

3.2.2 Strain-based approach

Different methods were presented and developed within this approach. The general concept behind
those methods is the same while the differences are in the best way to use those equations
according to the operating conditions, material properties and analyzed part manufacturing and
state. As explained in section 3.2.2, Basquin, Coffin and Manson equation is used in the strain-based
fatigue analysis for zero mean stress case. Six different methods are used for estimating fatigue life,
see Table 2.

Table 2 Overview of main estimation methods and their key parameters [16]

Estimation method ́
Original universal slopes method −0.12 −0.6
Modified universal slopes method , −0.09 , , −0.56
Four-point correlation method , , , , , , ,
Modified four-point correlation method , , , , ,
Uniform material law −0.087 , −0.58
Hardness method ( ) −0.09 , −0.56

The strain-based approach usually considers the number of reversals 2 as the independent
variable instead of the number of cycles (1 Reversal = cycle).
The true fracture strength and true fracture ductility are two important quantities that the strain-
based methods are depending on. The true fracture strength is the true stress at final fracture. This
can either be estimated or determined from the load-strain curve at fracture point. From a static
test that was performed on the ODC shaft the true fracture strength was determined, =
2547.2 .
While true fracture ductility is the true strain at final fracture, which was similarly determined from
a load-strain curve at fracture point, = 0.01106 .

42
The governing equations for the different methods in Table 2 can be solved using MATLAB due to its
complexity and for generating the strain-life curve.

x Original universal slopes method [16]

This method works for all materials with ultimate strength between 110 and 2830 MPa. The
corresponding expression is shown in equation (20).
. . (20)
.
= 1.9018 2 + 0.7579 2

x Modified universal slopes method [16]

It is also a method that works for all materials but without an ultimate strength limits, see equation
(21).

. . (21)
. . .
= 0.623 2 + 0.0196 2

x Four-point correlation method [16]

A method for all material, 110 < < 2830 , see equations (22) and (23).
. (22)
.
.
1 ×
= 10 2
2
. ∗
1 .
( )
+ 10 2
2
Where,
. (23)
∗ ×
= 10

x Modified four-point correlation method [16]

A method for all materials without a defined ultimate strength region, see equations (24) and (25).
∆ ∗ (24)
.
.
. .
= 2 + 2

Where,
. (25)
∆ ∗ .
= 10
2

x Uniform material law method [16]

43
There are two different equations for this method, the first one deals with Al and Ti alloys and the
other deals with unalloyed and low alloy steel. The second equation is best suitable for this study,
see equations (26) and (27).

. . (26)
= 1.5 2 + 0.59 2

(27)
1 ≤ 0.003
=
1.375 − 125 ≥ 0.003

Since = 0.011 ≥ 0.003, then =0.09

x Hardness method [16]

The hardness method works for steel of hardness between 100 and 700 HB, see equation (28).

4.25 + 225 . 0.32 − 487 + 191000 . (28)

= 2 + 2

44
 3.3 Experimental tests

For determining the load cases values and examine the behavior of the shaft a static test was
performed. The shaft is loaded gradually until fracture takes place, data are collected during the test
and analyzed after failure.

A dynamic test is carried out for determining the fatigue lives experimentally. The dynamic test
results are the verification tool for the estimated lives.

3.3.1 Static test

A static test was performed in the lab on the ODC shaft. Strain gauges were installed on different
positions of the shaft to record the strain values with the torque, See Figure 27 and Figure 28. The
load is increased gradually until failure occurred. The Torque-strain ( - ) curve was plotted for
analysis and data collection. See Figure 29 and Figure 30
Failure in spline shaft is unlikely to happen at the shaft’s bearing seat. Since the stress and strain are
higher at the spline relief and root fillet. See Figure 36 and Figure 37 for stress difference - Figure 29
and Figure 30 for strain difference. Therefore, the spline relief Torque-strain ( - ) curve will be
studied. As shown in Figure 30, the curve behave linearly for = [0 3500] , while there is a
non-linear behavior between [3500 5000] and the lead to failure region at the end until
6800 .
The frequency was variated between 0.5 and 5 to ensure right test conditions, then a frequency
of 3 was chosen as a suitable frequency for the dynamic test. The load cases were decided
depending on the static test results within a suitable range of [2800 3800] .

45
Figure 27: Static test setup with the strain gauges attached to the spline

Figure 28: Strain gauges position on the shaft [17]

46
Figure 29: T-ε curve at the bearing seat

Figure 30: T-ε curve at spline relief

47
 3.3.2 Dynamic test

A torsional fatigue dynamic test was performed on the ODC shaft for verifying the analytical results.
See test setup in Figure 31. Ten different test specimens were used in the test, three specimens for
each load case and one spare part to be used if needed. The number of cycles to crack initiation and
failure were recorded. The test was performed on a frequency of 3 as decided in section 3.3.1;
see test results in Figure 32.

Figure 31: Torsional dynamic test setup [17]

48
 Figure 32: Torsional dynamic test results

Using Weibull probability plot for cycles, the test results data are fitted to B10 and B50 fitting as
shown in Figure 33. Where B10 and B50 means a reliability of 90 and 50 percent, respectively. The
analytical calculated life is more likely to be compared against B50 fitting values as the reliability
cannot be so high with only nine test specimens.

49
Figure 33: Dynamic test results with B10 and B50 fitting

50
4 Results and discussion
4.1 Stress analysis
4.1.1 MSC Marc

4.1.1.1 Initial contact

As explained in section 3.1.1.2, the analysis in Marc starts by ensuring uniform contact between the
teeth; see Figure 34 and Figure 35.

Figure 34: Initial contact analysis between the teeth

Figure 35: Contact of the drive side on the spline shaft

4.1.1.2 Loads applied

Running the jobs for the three different load cases, the following stress distribution is obtained:

51
 Figure 36: Stress distribution on the shaft, T = 3800 Nm

Spline relief
Root fillet

Figure 37: Stress distribution at critical areas, T = 3800 Nm

The stress distribution shows that the maximum stress is at the root fillet of the splines’ drive sides.
There is also a high stress distributed uniformly around the spline relief. Hence, those two spots are
the critical areas for the splined shaft, see Figure 36 and Figure 37.

52
Comparing the stresses for all load cases, the stresses are found to have a linear relation with the
applied load that could be expressed as a constant stress to load ratio of 0.37 for root fillet and
0.26 for spline relief. See Figure 38

Figure 38: Stress results using MSC Marc

During gear load transmission, the teeth are subjected to tensile stresses at the drive side and
compressive stresses at the coast side. Tooth fracture occurs at the fillet of the tensile-stress side.
So, the root fillet stress at the drive side is expected to be higher than the coast side, which is shown
clearly in Figure 38.

4.1.2 Spline LDP

4.1.2.1 Contact stress

The contact stress between the teeth was analyzed and the distributions along roll angle, face width
and tooth drive side are shown in Figure 39 and Figure 40.

The involute roll angle is the angle whose arc on the base circle of radius unity equals the tangent of
the pressure angle at a selected point on the involute. While the face width is the length of the

53
splines in the axial plane. The roll angle and face width values are used to locate a point on the
spline teeth.

Figure 39: Contact stress distribution along roll angle and spline face width for one tooth, T = 3800 Nm

Figure 40: Contact stress and load distribution along each tooth, T = 3800 Nm

The contact stress distribution shows that a maximum stress of 2715 is at the beginning of the
face width at roll angle of 31.5° for the first torque level 3800 . The stress distribution is similar

54
for all applied loads with a maximum value of 2358 and 2000.8 for torques of 3300 and
2800 , respectively.

All the teeth have the same stress behavior. The detailed results shown in Figure 39 is presented in
Figure 40 for all the 25 teeth. The stress is shown in contour plot and the maximum contact stress
and force are the same for all the teeth.

This very high edge loading results from manufacturing deviations and misalignment and needs to
be avoided by micro-geometry modifications. And that is the case in most of the manufactured
gears and splines. The important observation from the contact analysis is that the distribution is
similar for all the teeth which will not result in a different probability of failure.

4.1.2.2 Root stress

The root stress distribution in Figure 41 shows the stress along the face width of the spline. The
maximum value is at the beginning of the face width, which verifies the finite element analysis done
in section 4.1.1.2 Loads applied.

Figure 41: Root stress distribution, T = 3800 Nm

The maximum values of root stress for torque levels 3800, 3300, 2800 are 1524, 1323 and
1123 , respectively.

4.1.3 Analytical results

Using the analytical equations as explained in section 3.1.3, the following spline parameters are
calculated: -See appendix-1 for spline and companion flange data-

55
Geometry factor:

= 0.273

Equivalent diameter:

= 37.3

Torsional stress for the three loads:

= 564.85

= 490.53

= 416.2

Geometry factor:

= 1.94

Maximum torsional stress

, = 1093

, = 949.4

, = 805.6

4.1.4 Stress comparison

The root fillet stress obtained using previous tools is shown in Figure 42. Results from spline LDP
and Marc are close to each other with 10% difference for spline LDP stress values. This difference is
a result of the involute shaped teeth model used in spline LDP. Since the contact area between the
teeth is less with an involute more than it is without, and with the same load applied in both cases,
the resulting stress is expected to be higher according to the relation: = , which is illustrated in
Figure 42 below.

The analytical results have low values compared to numerical results. This difference is a result of
the many approximations made in the analytical solution which led to inaccuracy in the method and
the calculations, while the numerical method has dealt with a simulation of the real geometry.
Hence, the experimental analysis is more accurate than the analytical one in this study and this
difference in the results was expected.

56
 Figure 42: Root fillet maximum stresses calculated by different tools

4.2 Strain analysis

Since the strain is the dependent parameter in strain-based fatigue approaches, it needs to be
measured and analyzed. An experimental test was performed on the shaft to measure strain as
explained in section Static test for verification purposes the strain was calculated with Marc to be
compared to the measured strain. See Figure 43.

57
 Figure 43: Strain distribution on the shaft, T = 2800 Nm

Figure 44: Strain distribution at critical areas, T = 2800 Nm

A difference between the calculated and measured values is observed, see Figure 45. Investigating
the reasons of this difference, it was found that the maximum strain is at the beginning of the relief;
see Figure 44, while the strain gauges were placed in the middle of the relief. See Figure 28.

58
 Figure 45: Maximum value of calculated and measured strain at the spline relief

So, the strain was calculated with Marc at the same locations on the shaft where the gauges were
placed; see Figure 46. The calculated strain with Marc correlates very well with the experimental
values, which verifies the FE model results.

59
 Figure 46: Strain values at spine relief

4.3 Fatigue results

4.3.1 Stress-based fatigue

Applying the method explained in section 3.2.1, the S-N curve is plotted for the finite life in high
cycle region and infinite life as shown in Figure 47.

60
 Figure 47: Generated S-N Wöhler curve

The material and conditions of the ODC shaft gave a Wöhler curve factor of = 6. While the B10
and B50 fit curves has a of 6 and 7.12, respectively. See Figure 33

Using stress values calculated in section 4.1 Stress analysis as the dependent values in Wöhler curve
Figure 47, fatigue life can be estimated as follows:

Table 3: Fatigue life using Wöhler S-N curve

Stress
Torque Marc-Root fillet Marc-Spline relief Spline LDP DIN 5466-1
3800 13180 12590 9412 69190
3300 34170 29110 22000 160900
2800 99600 77740 58780 430800

61
For visualizing the correlation with the dynamic test results a plot of correlation was made, which
shows the experimental life on x-axis and the estimated life on y-axis with a line of equality of slope
1. So, the ideal correlation happens when a point is plotted exactly on the line while a point below
the line is acceptable as it indicates an under estimation of fatigue life. On the contrary a point
above the line indicates an over estimation of fatigue life.

Figure 48: Correlation between experimental life and predicted life with Wöhler curve

Based on the above clarifications, the Stress calculated with Marc at the root fillet is giving the best
correlation to test results. Spline LDP stress has more conservative –and acceptable- fatigue life
which is expected due to the higher value of stress. While the stresses from DIN standard and Marc
at the relief overestimate fatigue life, see Figure 48.

The fatigue lives are also checked against B10 and B50 lives as shown in Figure 49 and Figure 50
below. The B50 results are more conservative than B10 as the reliability is less and the number of
experimental cycles is higher as shown in Figure 33. The stress at the fillet by Spline LDP has the
best correlation in B10 results, while in B50 results both stresses from Spline LDP and Marc are
giving good correlation with a preference for Marc.

62
Figure 49: Correlation between B10 experimental life and predicted life with Wöhler curve

Figure 50: Correlation between B50 experimental life and predicted life with Wöhler curve

63
4.3.2 Strain-based fatigue

The calculated maximum strain at the root fillet and relief is taken as the dependent variable to
predict the fatigue life; see Figure 51.

Figure 51: Maximum strain at spline relief and root fillet

As explained in section 3.2.2, different approaches are followed to calculate fatigue life. Solving all
the equations from section 3.2.2 using MATLAB, the following curves are obtained.

All the curves have similar behavior. The plastic line is higher than the elastic line at low fatigue life
(High strain value). While at low strain, the elastic line is higher than plastic line, which is expected
as a material behavior. The point where the elastic and plastic lines intersect is known as the
transition point at which the material’s plastic deformation starts. For strong materials, as in this
case, the transition point is at low fatigue life, see Figure 52. The sum of the elastic and plastic lines
gives the total strain curve, which is used to estimate fatigue life, see Figure 53.

64
Figure 52: Elastic, plastic and total strain fatigue curves - Original universal slopes method

Figure 53: Total strain fatigue curve - Original universal slopes method

65
All the six used approaches have similar behavior, the total strain lines are plotted in Figure 54. The
figure shows that the curves are within the same region with obvious deviations at the low cycle
fatigue region before the transition point.

Figure 54: Total strain fatigue curves for different approaches

Using the maximum values of strain at the spline relief -Figure 51- in all the previous curves, fatigue
life is estimated. For checking the correlation to test results the predicted lives are plotted against
experimental lives on a correlation curve, as shown in Figure 55.

66
 Figure 55: Experimental and predicted life correlation at spline relief - Strain-based approaches

All predicted fatigue lives at the spline relief are overestimated, which highlights the observation
that the part will fail before achieving those numbers of cycles due to a higher value of strain. It is
also seen that the results variation range has an inverse relation with the load.

Estimating fatigue life in the same way using the strain values at the root fillet and plotting the
results on a correlation curve, Figure 56 is obtained. The results are conservative, and the
correlation is much better for the fillet strain than the relief strain which was expected as the stress
and strain are much higher at the fillet and the crack propagation is expected to initiate at the fillet.

67
 Figure 56: Experimental and predicted life correlation at root fillet - Strain-based approaches

Four-point correlation method has the best correlation to test results. While the hardness method
has the worst correlation despite that it is still giving underestimated fatigue lives.

For further clarification of the results, the correlation curve is plotted with B10 and B50
experimental life as shown in Figure 57 and Figure 58 below.

68
Figure 57: B10 experimental and predicted life correlation at root fillet – Strain-based approaches

Figure 58: B50 experimental and predicted life correlation at root fillet – Strain-based approaches

69
The B10 and B50 results correlation behaves in a similar way to Wöhler curve results correlation;
see Figure 49 and Figure 50. B50 results are more conservative than B10, while strain-based
approach in general gives more conservative results and better correlation than stress-based
approach.

In the B10 correlation figure, original universal slopes and four-point correlation methods have the
best correlation. While in B50 correlation figure, the original and modified four-point correlation
methods have the best correlation to test results with a clear preference to the modified four-point
correlation method.

It is also clear from Figure 56, Figure 57 and Figure 58 that the modified universal slopes results are
too close to original universal slopes results with a slight increase in the quality of correlation.
Despite that both the methods are giving conservative results, they are not as good as other
methods in terms of correlation.

70
5 Conclusion and future work
5.1 Conclusion

The thesis aimed to develop an accurate and advanced fatigue prediction method for an outer disc
carrier spline shaft. Stress and strain analyses were performed on the shaft. Stress and strain-based
fatigue prediction methods were developed and evaluated against experimental test results.

Finite element analysis is the best approach for structural analysis since it is giving accurate and
reliable results. The stress results from Marc and Spline LDP were verified against each other and
showed good correlation with 10% difference for Spline LDP, which is explained by the involute
teeth model used in Spline LDP. As the involute teeth design has less contact area between the
teeth which results in higher root bending stress. The FEA was also verified through comparing the
calculated and measured strain values, which match perfectly.

Among the different methods and approaches used in the analysis, Wöhler curve and the four-point
correlation method are giving the best correlation to test results. The loads applied on the ODC
shaft in this study were within the elastic region, so the stress-based approach still gives good
correlation. But it is recommended to use the strain-based approach for its ability to consider both
elastic and plastic fatigue lives. So, the four-point correlation method has the preference over
Wöhler curve. It also has the advantage that it is applicable for all materials with a wide range of
ultimate tensile strength.

One of the findings of this study is that during fatigue failure of splines the crack initiates at the end
of the root fillet where the maximum stress is and propagates towards the relief. This finding was
verified through the good correlation of fatigue life calculated with stress values at the root fillet,
while the stress at the relief is giving bad correlation with an overestimated life.

71
5.2 Future work

The stress and strain at the root fillet are really high that it affects the fatigue life significantly. This
could be seen from the difference between estimated lives using strain at root fillet and spline
relief; Figure 55 and Figure 56. The fatigue life could reach over 10 million cycles at the relief strain.
Hence, reducing the stress, and subsequently the strain, at the root fillet is highly recommended.
One of the suggestions to do so is performing a topology optimization on the spline shaft with
maintaining the minimum requirement. Increasing the number of teeth is also another option for
reducing the bending stress.

For comparing the obtained results against B10 fitting, more experimental test data is required for
such a high reliability.

Performing a finite element fatigue analysis using FEMFAT software for instance would be a good
tool either for verifying the developed methods or checking the reliability of the software.

The fatigue prediction methods that give best fatigue life were developed, checked and analyzed for
this specific spline shaft using its material properties and geometry-dependent stresses, which
raises the question: Could the four-point correlation method be generalized and used for different
parts? Hence, applying the developed methods on different physical parts with different material
properties for deciding if those methods could be generalized or not is recommended.

72
 References

[1] Y.-L. Lee, M. E. Barkey, and H.-T. Kang, Metal Fatigue Analysis Handbook:
Practical Problem-solving Techniques for Computer-aided Engineering.
Elsevier, 2011.
[2] Ario Hardjosuwito, “Gear and transmission technology,” RWTH Aachen
University.
[3] “A Review on Fatigue Life Prediction Methods for Metals.” [Online].
Available: https://www.hindawi.com/journals/amse/2016/9573524/.
[Accessed: 24-Jun-2019].
[4] “DIN 5466-1:2000-10, Tragfähigkeitsberechnung von Zahn- und Keilwellen-
Verbindungen_- Teil_1: Grundlagen,” Beuth Verlag GmbH.
[5] “Beitrag zur Ermittlung der Kerbwirkung an Zahnwellen mit freiem und
gebundenem Auslauf,” ResearchGate. [Online].
Available: https://www.researchgate.net/publication/44229236_Beitrag_zur_
Ermittlung_der_Kerbwirkung_an_Zahnwellen_mit_freiem_und_gebundenem_
Auslauf. [Accessed: 17-Jun-2019].
[6] “Meshing Your Geometry: When to Use the Various Element Types,” COMSOL
Multiphysics. [Online]. Available: https://www.comsol.com/blogs/meshing-
your-geometry-various-element-types/. [Accessed: 26-Jun-2019].
[7] “Element Types.” [Online]. Available: http://fea-cae-engineering.com/fea-
cae-engineering/element_types.htm. [Accessed: 26-Jun-2019].
[8] “Skewness.” [Online].
Available: https://www.sharcnet.ca/Software/Ansys/17.0/en-
us/help/wb_msh/msh_skewness.html. [Accessed: 26-Jun-2019].
[9] “Shell meshing-Checking and editing mesh” [Online].
Available: https://altairuniversity.com/wp-
content/uploads/2012/04/HM_Quality_extract.pdf. [Accessed: 26-Jun-2019].
[10] O. C. Zienkiewicz, R. L. Taylor, and D. Fox, “Appendix A - Isoparametric
Finite Element Approximations,” in The Finite Element Method for Solid and
Structural Mechanics (Seventh Edition), O. C. Zienkiewicz, R. L. Taylor, and
D. Fox, Eds. Oxford: Butterworth-Heinemann, 2014, pp. 597–603.
[11] “3D meshing” [Online]. Available: https://altairuniversity.com/wp-
content/uploads/2014/02/3DMeshing.pdf. [Accessed: 26-Jun-2019].
[12] N. S. Gokhale, Practical Finite Element Analysis. Finite To Infinite, 2008.
[13] “GearLab Software for Consortium Members,” Gear and Power Transmission
Research Laboratory. [Online].
Available: https://gearlab.osu.edu/research/gearlab-software-consortium-
members. [Accessed: 28-Jun-2019].
73
[14] R. L. Norton, Machine Design: An Integrated Approach. Prentice Hall, 2011.
[15] Y.-L. Lee, J. Pan, R. Hathaway, and M. Barkey, Fatigue Testing and Analysis:
Theory and Practice. Elsevier, 2011.
[16] R. Basan, M. Franulović, I. Prebil, and N. Črnjarić-Žic, “Analysis of strain-life
fatigue parameters and behaviour of different groups of metallic
materials,” International Journal of Fatigue, vol. 33, no. 3, pp. 484–491, Mar.
2011.
[17] GKN company report.
[18] S. Wang, G.R. Liu, G. Y. Zhang, and L. Chen, “Design of Asymmetric Gear
and Accurate Bending Stress Analysis Using the Es-Pim with Triangular
Mesh,” International Journal of Computational Methods, Nov. 2011.
[19] https://www.gknautomotive.com/en/company/media-centre/news-
releases/2017/gkns-twinster-provides-intelligent-all-wheel-drive-torque-
vectoring-for-new-opel-insignia/
[20] https://www.efatigue.com/glossary/?word=cyclic_stress_strain_curve
[21] “MSC SimCompanion - Marc 2014.2 User’s Guide.” [Online].
Available: https://simcompanion.mscsoftware.com/infocenter/index?page=co
ntent&id=DOC10799&cat=MARC_DOCUMENTATION_2014.2&actp=LIST.
[Accessed: 21-Aug-2019].
[22] Hong, Jiazheng. "A Semi-Analytical Load Distribution Model of Spline Joints."
Electronic Thesis or Dissertation. Ohio State University, 2015.
[23] “MSC SimCompanion -.” [Online].
Available: https://simcompanion.mscsoftware.com/infocenter/index?page=co
ntent&id=S:KB8022085&actp=LIST. [Accessed: 25-Sep-2019].
[24] T. Dahlberg and A. Ekberg, Failure Fracture Fatigue: An Introduction. Lund:
Studentlitteratur AB, 2002.

74









'HSDUWPHQWRI0HFKDQLFDO(QJLQHHULQJ 7HOHSKRQH 
%OHNLQJH,QVWLWXWHRI7HFKQRORJ\ (PDLO LQIR#EWKVH
6(.DUOVNURQD6:('(1