A Thesis

Entitled

Effects of Mean Stress and Stress Concentration on Fatigue Behavior of Ductile Iron

By

Nicholas M. Meyer

Submitted to the Graduate Faculty as partial fulfillment of the requirements for the

Master of Science Degree in Mechanical Engineering

__________________________________

Dr. Ali Fatemi, Committee Chair

__________________________________

Dr. Efstratios Nikolaidis, Committee Member

__________________________________

Dr. Lesley Berhan, Committee Member

__________________________________

Dr. Patricia R. Komuniecki, Dean

College of Graduate Studies

The University of Toledo

December 2014
ii
 An Abstract of

Effects of Mean Stress and Stress Concentration on Fatigue Behavior of Ductile Iron

By

Nicholas M. Meyer

Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Master of Science Degree in Mechanical Engineering

The University of Toledo

December 2014

Cast iron tends to be more economical to manufacture than steel, but its

microstructure is riddled with internal defects. With improvements in casting technology,

cast iron can compete with steel in some applications due to the ability to maintain

similar strength. One objective of this study was to analyze data obtained from literature

and make predictive correlations between tensile, microstructural, and fatigue properties.

Another objective of this study was to evaluate tensile and compressive mean stress

effects on smooth and notched fatigue behavior of 120-90-02 ductile cast iron

experimentally, as well as analytically by using predictive models. The relationship

between these two objectives is such that with successful correlations of fatigue limit

with mechanical properties, fewer experiments could be performed to verify analytical

results. This grade of cast iron was selected by Eaton Corporation, who helped fund this

study. The material, 120-90-02 ductile cast iron, was machined and heat-treated as

provided by Eaton Corporation. Neuber’s rule, strain energy density, and finite element

methods were used to obtain stresses and strains at the notch. Modified Goodman, Smith-

iii
Watson-Topper, FKM and the Fatemi-Socie mean stress parameters were used to account

for mean stress effect on fatigue life of both smooth and notched specimens. Mean stress

levels were chosen such that the R-ratios in load-controlled tests were -∞, -7, -3, -1, 0,

1/3, 0.5, and 0.75. By understanding the fatigue behavior of material under varying

degrees of mean stress, design safety can be improved by properly accounting for the

effect of mean stress that a component may be subjected to. The Smith-Watson-Topper

mean stress parameter proved to be the best method for life prediction for both smooth

and notched specimens, based on the nominal stress approach. Using the local approach,

which considers notch root stresses and strains, the Fatemi-Socie fatigue parameter used

with the strain energy density rule provided better results compared to the Smith-Watson-

Topper mean stress parameter.

iv
Acknowledgements

First, I would like to extend my sincere appreciation to Dr. Ali Fatemi, my thesis

and research advisor, for his support, encouragement, and guidance on this thesis study. I

would like to thank Dr. Efstratios Nikolaidis and Dr. Lesley Berhan for serving on my

defense committee. I would also like to thank Eaton Corporation for the supply of

specimens used in testing. I would like to thank my lab-mates for their support and

encouragement. Lastly, I would like to thank my parents for their continued love,

support, and encouragement throughout my entire academic career.

v
Table of Contents

Abstract ............................................................................................................................. iii

Acknowledgements ........................................................................................................... v

List of Tables .................................................................................................................... ix

List of Figures .................................................................................................................. xii

List of Abbreviations ..................................................................................................... xxi

List of Symbols .............................................................................................................. xxii

1 Introduction .................................................................................................................... 1

1.1 Cast Iron and Its Classification ................................................................................1

1.2 Importance of Fatigue Considerations in Design .....................................................3

1.3 Objectives of This Study ..........................................................................................6

2 Literature Review of Fatigue Behavior of Cast Iron and Mean Stress

Effects .......................................................................................................................... 10

2.1 Fatigue Behavior of Cast Iron ................................................................................10

2.2 Correlations for Cast Iron Data Based on Data from Literature ............................22

2.2.1 Correlations Among Mechanical Properties Pertaining to Cast Iron .............. 22

2.2.2 Methods to Predict Fatigue Limit .................................................................... 23

2.3 Mean Stress Effects on Fatigue Behavior ..............................................................25

2.3.1 A Historical Review of Mean Stress ................................................................ 26

2.3.2 Discussion of Mean Stress Equations .............................................................. 28

vi
 2.3.3 Less Commonly Used Mean Stress Equations ................................................ 31

2.3.4 Effects of Mean Stress on Notched Specimens ............................................... 34

2.3.5 Mean Stress and Notch Effects Related to Cast Iron....................................... 38

3 Experimental Program ................................................................................................ 65

3.1 Material Description...............................................................................................65

3.2 Specimen Description ............................................................................................66

3.3 Testing Equipment .................................................................................................67

3.4 Test Methods and Procedures ................................................................................68

3.4.1 Monotonic Tension and Compression Tests ................................................... 68

3.4.2 Incremental Step Test ...................................................................................... 68

3.4.3 Constant Amplitude Smooth and Notched Fatigue Tests ................................ 69

4 Experimental Results and Discussion ........................................................................ 73

4.1 Monotonic Tension and Compression Test Results ...............................................73

4.2 Analysis of the Incremental Step Test Results .......................................................74

4.3 Fully-Reversed Fatigue Test Results of Smooth and Notched Specimens ............76

4.4 Prediction of Fully-Reversed Notched Fatigue Behavior ......................................77

4.4.1 Finite Element Analysis of the Notched Geometry ......................................... 78

4.4.2 Nominal Stress-Based Approach ..................................................................... 79

4.4.3 Local-Based Approaches ................................................................................. 80

4.5 Mean Stress Effects on Smooth and Notched Fatigue Behavior ...........................87

4.6 Correlation of Experimental Data with Mean Stress Parameters...........................90

4.6.1 Predicting Mean Stress Notched Fatigue Behavior from Smooth Fatigue

Behavior ........................................................................................................... 90

vii
 4.6.2 Nominal Stress Approach for Smooth and Notched Specimens ..................... 91

4.6.3 Local Stress Approach for Predicting Notched Behavior ............................... 98

4.6.4 Local Strain Approaches for Predicting Notched Behavior .......................... 100

4.7 Fatigue Notch Factor as a Function of R-ratio and Life ......................................102

4.8 Fatigue Fracture Surfaces .....................................................................................103

5 Summary and Conclusions........................................................................................ 166

References ...................................................................................................................... 171

Appendix A: Tables of Data Compiled from Literature Used in Chapter 2

Correlations .......................................................................................... 178

viii
List of Tables

1.1 Mechanical properties of various grades of cast iron [1]. ............................. 8

2.1 Estimated fatigue limit (σ′w) as predicted by Murakami’s modified
area method compared to the actual fatigue limit (σwo) (where the
fatigue limit was considered as fatigue strength at run-out for ≥ 107
cycles) at R = -1 for JIS FCD400, JIS FCD500, and JIS FCD700
ductile iron [9]. ............................................................................................ 41
2.2 Experimental and predicted fatigue limits as predicted by
Murakami’s area method for cast irons with different
microstructures including vacuum quenched (VQ) and subzero
treatment under rotary bending loads [12]. ................................................. 41
2.3 Non-constant fatigue ratios shown for different matrix
microstructures of an alloyed ductile iron where the fatigue limit is
defined at 107 cycles at R = -1 [14]. ............................................................ 42
2.4 (a) Various models used to estimate the fatigue limit (defined as
fatigue strength at 107 cycles) of nodular cast iron with
microstructural defects at R = -1 and (b) input parameters for models
listed in (a) [17]. .......................................................................................... 43
2.5 Fatigue limit, tensile strength, and fatigue ratio for R = 0.1 of
spheroidal graphite cast iron with different heat treatments at room
temperature, 300 °C, and 500 °C [20]. ........................................................ 44
2.6 Values of aM and bM adapted from the FKM guideline [40]. ...................... 44
2.7 Material properties of the ductile irons containing Ti showing a non-
constant fatigue ratio between material microstructures for R = 0.1
where the fatigue limit is defined as the fatigue strength at 2 x 107
cycles [57]. .................................................................................................. 45

ix
4.1 Summary of monotonic tension and compression properties of 120-
90-02 ductile cast iron. .............................................................................. 106
4.2 Elastic and plastic strain amplitudes obtained from the incremental
cyclic deformation step test in load control. E′ was used in elastic or
plastic strain calculations, E′ = 162 GPa. .................................................. 106
4.3 Summary of fully-reversed (R = -1) constant amplitude fatigue tests
of smooth specimens. ................................................................................ 107
4.4 Summary of fully-reversed (R = -1) constant amplitude fatigue tests
of notched specimens. ............................................................................... 108
4.5 Notch root stresses and strains obtained from FEA. ................................ 109
4.6 Strain-life fatigue and cyclic deformation properties of 120-90-02
ductile iron................................................................................................. 109
4.7 Calculations of Fatemi-Socie (with k = 4) and SWT mean stress
parameters for notched specimens under the fully-reversed (R = -1)
cyclic loading condition with notch root stresses and strains obtained
from Neuber and SED rules for (a) Kf = 2.54 and (b) Kf as a function
of Sa. .......................................................................................................... 110
4.8 Summary of constant amplitude mean stress fatigue tests at different
R-ratios for smooth specimens. ................................................................. 111
4.9 Summary of constant amplitude fatigue tests at different R-ratios for
notched specimens. .................................................................................... 112
4.10 Intercepts and slopes of the best-fit S-N lines including fatigue limits
defined at 106 cycles for each R-ratio. ....................................................... 112
4.11(a) Equivalent fully-reversed (R = -1) local stresses and strains obtained
with Neuber and SED methods with Kf = 2.54 and calculations of
the Fatemi-Socie (with k = 4) and SWT mean stress parameters for
notched mean stress data. .......................................................................... 113
4.11(b) Equivalent fully-reversed (R = -1) local stresses and strains obtained
with Neuber and SED methods with Kf as a function of Sa and
calculations of the Fatemi-Socie (with k = 4) and SWT mean stress
parameters for notched mean stress data. .................................................. 114

x
A.1 Tensile properties from literature used to make correlations in
chapter 2. ................................................................................................... 178
A.2 Fatigue and microstructural data compiled from literature for
correlations in chapter 2. ........................................................................... 181

xi
List of Figures

1-1 Microstructures of: (a) grey iron at 500x, (b) nodular at 200x (or
ductile) iron, (c) white iron at 400x, (d) malleable iron at 150x, and
(e) compacted graphite iron at 100x [1]. ....................................................... 9
2-1 Conventionally determined fatigue lives compared to S-N curves
based on plastic strain amplitude, change in temperature and change
in electrical resistance of the specimen for cast iron EN-GJV-400 at
R = -1 [5]. .................................................................................................... 46
2-2 Fatigue life scatter of pearlitic gray cast iron due to differences in
microstructure from different foundries tested at R = -1 [11]. .................... 46
2-3 Fatigue limit under rotary bending testing conditions of austempered
ductile irons as a function of casting block leg thickness where the
fatigue limit is defined as the fatigue strength at 107 cycles [13]................ 47
2-4 Relationship of the rotating bending fatigue limit under rotary
bending conditions with impact toughness and nodule diameter of
austempered ductile iron where the fatigue limit is defined as the
fatigue strength at 107 cycles [13]. .............................................................. 47
2-5 Fatigue limit as a function of maximum and mean nodule diameters
for spheroidal graphite cast iron under fully-reversed bending where
the fatigue limit is defined as the fatigue strength at 107 cycles [15]. ......... 48
2-6 Fully-reversed bending fatigue limit as a function of tensile strength
for austempered ductile iron compared to steel where the fatigue
limit is defined as the fatigue strength at 107 cycles [15]............................ 48
2-7 Comparison of predicted fatigue limit for nodular cast iron based on
several methods with experimental fatigue limit defined as the
fatigue strength at 106 cycles obtained from 3-point bending tests at
R = 0.3 [19].................................................................................................. 49
xii
2-8 Ultimate tensile strength of cast iron containing spheroidal graphite
with various heat treatments at different temperatures [20]. ....................... 49
2-9 Comparisons of S-N behavior of smooth, notched, and surface rolled
notched specimens for 100-70-03 ductile cast iron under rotating
bending conditions with the fatigue limit defined at 4 x 106 cycles
[21]. ............................................................................................................. 50
2-10 Comparisons of plae bending fatigue test results of as-cast, shot
blast, and fine ground specimens of nodular cast iron at R = 0 where
the fatigue limit was defined at 2 x 106 cycles [22]. ................................... 50
2-11 Effects of Almen intensity on smooth rotating bending fatigue life of
Cu-Ni austempered ductile iron where the fatigue limit was defined
at 107 cycles [23]. ........................................................................................ 51
2-12 Two-step diagram for ductile cast iron under rotating bending
depicting the transition region between surface fracture and internal
fracture investigated for shot-peened conditions [24]. ................................ 51
2-13 Tensile strength vs. Vickers hardness with cast iron data obtained
from literature grouped by average nodule size showing a linear
relationship. ................................................................................................. 52
2-14 Yield strength vs. Vickers hardness with cast iron data obtained
from literature grouped by average nodule size showing a linear
relationship. ................................................................................................. 52
2-15 Fatigue limit defined as fatigue strength at 107 cycles vs. Vickers
hardness grouped by average nodule size showing no correlation. ............ 53
2-16 Fatigue limit defined as fatigue strength at 107 cycles vs. tensile
strength with data from literature grouped by average nodule size
showing no correlation. ............................................................................... 53
2-17 Fatigue limit defined as the fatigue strength between 106 and 107
cycles vs. hardness with cast iron data obtained from literature
grouped by maximum nodule size. ............................................................. 54

xiii
2-18 Fatigue limit defined as the fatigue strength between 106 and 107
cycles vs. tensile strength with cast iron data obtained from literature
grouped by maximum nodule size. ............................................................. 54
2-19 Correlation of fatigue limit data from literature with the parameter
used in [13] based on impact strength and mean nodule size (also
see Figure 2-4). ............................................................................................ 55
2-20 Fatigue limit of ductile cast iron predicted by Murakami’s method
using maximum nodule size vs. experimental fatigue limit (defined
as fatigue strength at 106 to 107 cycles) with cast iron data from
literature and assuming (a) external defects, (b) internal defects. ............... 56
2-21 Fatigue limit as predicted by de Kazinczy’s method vs. experimental
fatigue limit with cast iron data from literature assuming defects as
pores. ........................................................................................................... 57
2-22 Fatigue limit as predicted by Mitchell’s method with average nodule
size vs. experimental fatigue limit with cast iron data obtained from
literature assuming (a) Kt = 2, and (b) Kt = 3. Reapplication of
Mitchell’s method using maximum nodule size assuming (c) Kt = 2,
and (d) Kt = 3. .............................................................................................. 58
2-23 Comparison of Goodman and Gerber mean stress predictions for
mean stress fatigue data of 316 LN stainless steel in air and mercury
environments [37]. ...................................................................................... 59
2-24 Comparison of experimental data with various mean stress models
for a medium carbon steel [34].................................................................... 59
2-25 Divisions of the Haigh diagram for use with the FKM mean stress
method [31, 40]. .......................................................................................... 60
2-26 Comparison of the FKM mean stress parameter with several other
mean stress parameters for AISI 4340 steel [31]. ....................................... 60
2-27 Stress amplitude vs. combination of mean stress and residual stress
for shot peened and unpeened notched 16MnCr5 steel specimens
where the fatigue limit was defined as the fatigue strength at 107
cycles [46]. .................................................................................................. 61

xiv
2-28 S-N diagrams for high R-ratios of notched (Kt = 1.65) 1045 steel
with Rockwell C hardness of 10, 37 and 50 with the fatigue limit
defined as the fatigue strength at 5 x 106 cycles [47]. ................................. 61
2-29 Correlation of fatigue test results using the Smith-Watson-Topper
mean stress model to account for the development of mean stress
due to the differences between tensile and compressive strengths of
grey cast iron under strain and load-controlled testing [52]........................ 62
2-30 Correlation of fatigue test results of austempered ductile irons with
different heat treatments and R-ratios based on the modified Smith-
Watson-Topper parameter as proposed by Fash and Socie [53]. ................ 62
2-31 Maximum stress vs. number of cycles to failure for different R-ratios
and heat-treatment conditions for austempered ductile irons where
the fatigue limit is defined as the fatigue strength at 106 cycles [54]. ......... 63
2-32 Fatigue limit (defined as the fatigue strength at 2 x 106 cycles)
correlation of ductile cast iron at R = 0 with a geometrical defect
parameter as compared to notch or crack equivalence methods [56].......... 63
2-33 The stress transition level at which failure stems from surface to
subsurface for ductile cast iron at 0.76Sy for (a) iron A1, (b) iron A2,
and (c) iron A3 (see Table 2.7) at R = 0.1 where the fatigue limit is
defined as the fatigue strength at 2 x 107 cycles [57]. ................................. 64
3-1 Geometry of the (a) smooth and, (b) notched axial fatigue
specimens, and (c) monotonic tension specimen. Dimensions are in
mm. Drawings based on Eaton Corporation’s specifications...................... 70
3-2 Finite element mesh of the notch. ............................................................... 71
3-3 (a) MTS load frame with Instron controller (not shown) used for
testing and (b) close up of specimen in gripped position. ........................... 72
3-4 Load block used for testing at R = -∞. ........................................................ 72
4-1 Monotonic tension and compression curves for 120-90-02 ductile
cast iron. .................................................................................................... 115
4-2 Plot of true stress vs. true plastic strain for determining monotonic K
and n values. .............................................................................................. 115

xv
4-3 Strain amplitude verse applied number of cycles during load-
controlled incremental step test showing when cyclic stabilization
occurs. ....................................................................................................... 116
4-4 Composite plot of hysteresis loops from the incremental step test. .......... 116
4-5 Plot of engineering stress amplitude vs. plastic strain amplitude
(calculated from E′) used to determine K′ and n′. ..................................... 117
4-6 Composite plot of cyclic and monotonic true stress-strain curves and
data. ........................................................................................................... 117
4-7 S-N curves of fully-reversed smooth and notched data of 120-90-02
ductile cast iron. ........................................................................................ 118
4-8 Input cyclic stress-strain curve for FEA analysis with comparison of
uniaxial and von Mises output curves showing the effect of a
multiaxial stress state. ............................................................................... 118
4-9 Prediction of notched behavior and comparison with experimental
data using the nominal S-N approach. ....................................................... 119
4-10 Strain-life curves and data including elastic, plastic, and total strain
of smooth specimens. ................................................................................ 119
4-11 Plots of local (a) stress and (b) strain amplitudes vs. nominal stress
amplitude for Neuber’s rule, SED rule, and FEA results. ......................... 120
4-12 Strain-life predictions and scatter plots of fully-reversed (R = -1)
notched fatigue behavior using the local strain approach with
Neuber’s rule, SED rule, and finite element analysis using Kf = 2.54
(a, b), and Kf as a function of Sa (c, d). ...................................................... 122
4-13 Prediction of fully-reversed (R = -1) notched data from the SED rule
using the strain-based version of (a) Fatemi-Socie with k = 4 and Kf
= 2.54, (b) SWT with Kf = 2.54, (c) Fatemi-Socie with k = 4 and Kf
varies with Sa, and (d) SWT with Kf as a function of Sa. .......................... 124
4-14 Prediction of fully-reversed (R = -1) notched data from Neuber’s
rule using the strain-based version of (a) Fatemi-Socie with k = 4
and Kf = 2.54, (b) SWT with Kf = 2.54, (c) Fatemi-Socie with k = 4
and Kf varies with Sa, and (d) SWT with Kf as a function of Sa................ 126

xvi
4-15 Stress-life predictions and scatter plots of fully-reversed (R = -1)
notched fatigue behavior using the local stress approach with
Neuber’s rule, SED rule, and finite element analysis using Kf = 2.54
(a, b), and Kf as a function of Sa (c, d). ...................................................... 128
4-16 Flow chart to estimate notched specimen behavior. ................................. 129
4-17 Best-fit lines and data at different R-ratios for (a) smooth and (b)
notched specimens. .................................................................................... 130
4-18 Prediction of notched fatigue behavior for different R-ratios from
fully-reversed, smooth specimen fatigue behavior using Modified
Goodman showing predicted S-N lines and experimental notched
data. ........................................................................................................... 131
4-19 Correlations of experimental data at different R-ratios for smooth
specimens using (a) Modified Goodman, (b) SWT, (c) FKM, and (d)
Fatemi-Socie (with k = 4) mean stress models. ......................................... 132
4-20 Correlation of smooth specimen mean stress data with the stress-
based version of the Fatemi-Socie parameter (with k = 4)
corresponding to the equivalent stress approach in Figure 4-19 (d). ........ 133
4-21 Correlations of experimental data at different R-ratios for notched
specimens using (a) Modified Goodman, (b) SWT, (c) FKM, and (d)
Fatemi-Socie (with k = 4) mean stress models. ......................................... 134
4-22 Correlation of notched specimen mean stress data with the stress-
based version of the Fatemi-Socie Parameter (with k = 4)
corresponding to the equivalent stress approach in Figure 4-21(d). ......... 135
4-23 Predicted versus experimental fatigue lives at different R-ratios for
smooth and notched specimens by using the nominal stress-based
approach and (a) Modified Goodman, (b) SWT, (c) FKM, and (d)
Fatemi-Socie (with k = 4) mean stress models. ......................................... 137
4-24 Constant life diagrams and experimental data of smooth specimens
with predictions from (a) Modified Goodman, (b) SWT, (c) FKM,
and (d) Fatemi-Socie mean stress models. ................................................ 138

xvii
4-25 Constant life diagrams and experimental data of smooth specimens
comparing all mean stress models for lives of (a) 103, (b) 104, (c)
105, and (d) ≥ 106 cycles. ........................................................................... 139
4-26 Percentage errors between experimental and predicted stress
amplitude based on Modified Goodman, SWT, FKM, and Fatemi-
Socie (with k = 4) mean stress parameters for smooth specimen
fatigue lives of (a) 104, (b) 105, and (c) ≥106 cycles. ................................ 140
4-27 Constant life diagrams and experimental data of notched specimens
with predictions from (a) Modified Goodman, (b) SWT, (c) FKM,
and (d) Fatemi-Socie (with k = 4) mean stress models. ............................ 141
4-28: Constant life diagrams and experimental data of notched specimens
comparing all mean stress models for lives of (a) 103, (b) 104, (c)
105, and (d) ≥ 106 cycles. ........................................................................... 142
4-29 Percentage errors between experimental and predicted stress
amplitudes based on Modified Goodman, SWT, FKM, and Fatemi-
Socie (with k = 4) mean stress parameters for notched specimen
fatigue lives of (a) 103, (b) 104, (c) 105 and (d) ≥106 cycles. .................... 143
4-30 Correlations of all experimental data for smooth and notched
specimens using the SWT mean stress parameter with equivalent,
fully-reversed notch stress amplitudes obtained from (a) Neuber’s
rule (with Kf = 2.54), (b) the SED rule with (Kf = 2.54), (c) Neuber’s
rule (with Kf as a function of Sa), and (d) SED rule (with Kf as a
function of Sa). .......................................................................................... 145
4-31 Predicted vs. experimental fatigue life of all experimental data using
the SWT mean stress parameter with local stress obtained for
notched specimens from (a) Neuber’s rule (Kf = 2.54), (b) SED rule
(Kf = 2.54), (c) Neuber’s rule (Kf varies with Sa), and (d) SED rule
(Kf varies with Sa). ..................................................................................... 147
4-32 Correlations of all experimental data for smooth and notched
specimens using the Fatemi-Socie mean stress parameter with
equivalent, fully-reversed notch stress amplitudes obtained from (a)

xviii
 Neuber’s rule (Kf = 2.54), (b) the SED rule (Kf = 2.54), (c) Neuber’s
rule (Kf varies with Sa), and (d) the SED rule (Kf varies with Sa). ............ 149
4-33 Predicted vs. experimental fatigue life of all experimental data using
the Fatemi-Socie mean stress parameter with local stress obtained
for notched specimens from (a) Neuber’s rule (Kf = 2.54), (b) SED
rule (Kf = 2.54), (c) Neuber’s rule (Kf varies with Sa), and (d) SED
rule (Kf varies with Sa). .............................................................................. 151
4-34 Smith-Watson-Topper parameter plots of smooth and notched mean
stress data with notch stress and strain values obtained from (a)
Neuber’s rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c) Neuber’s rule
(Kf varies with Sa), and (d) SED rule (Kf varies with Sa)........................... 153
4-35 Predicted verse experimental fatigue lives at different R-ratios for
smooth and notched specimens using the strain-based version of
SWT parameter for notched specimens incorporating (a) Neuber’s
rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c) Neuber’s rule (Kf
varies with Sa), and (d) SED rule (Kf varies with Sa). ............................... 155
4-36 Fatemi-Socie parameter plots with k = 4 for smooth and notched
mean stress data with notch stress and strain values obtained from
(a) Neuber’s rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c) Neuber’s
rule (Kf varies with Sa) and (d) SED rule (Kf varies with Sa). ................... 157
4-37 Predicted verse experimental fatigue lives at different R-ratios for
smooth and notched specimens using the strain-based version of the
Fatemi-Socie parameter for notched specimens incorporating (a)
Neuber rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c) Neuber’s rule
(Kf varies with Sa), and (d) SED rule (Kf varies with Sa)........................... 159
4-38 Fatigue notch factor as a function of (a) fully-reversed (R = -1)
stress amplitude, (b) R-ratio and (c) fatigue life. ....................................... 161
4-39 Fracture surfaces of smooth specimens for each R-ratio and stress
amplitude level tested. Stress amplitude and cycles to failure are
shown below each photo. .......................................................................... 163

xix
4-40 Fracture surfaces of notched specimens for each R-ratio and stress
amplitude level tested. Stress amplitude and cycles to failure are
shown below each photo. .......................................................................... 165

xx
List of Abbreviations

ADI ........................................Austempered Ductile Iron

ASTM ....................................American Society for Testing and Materials

BHN .......................................Brinell Hardness Number

CB ..........................................Cantilever Bending
CGI.........................................Compact Graphite Iron

FEA ........................................Finite Element Analysis

FS ...........................................Fatemi-Socie Model
Fq ...........................................Frequency

HRC .......................................Rockwell Hardness C Scale

Hv...........................................Vickers Hardness

NG ..........................................Nodular Graphite

RTB ........................................Rotating Bending

SG ..........................................Spheroidal Graphite
SWT .......................................Smith-Watson-Topper

TCD........................................Theory of Critical Distance

VC ..........................................Vanadium Carbide

YS ..........................................Yield Strength

xxi
List of Symbols

A.............................................Coefficient of the S-N line

a ..............................................Peterson’s Material Characteristic Length
B .............................................Slope of the S-N line
b..............................................Fatigue Strength Exponent
c ..............................................Fatigue Ductility Exponent
Do ...........................................Original Diameter of Specimen
E .............................................Modulus of Elasticity
E′ ............................................Cyclic Modulus of Elasticity
e ..............................................Engineering Strain
%EL .......................................Percent Elongation
K.............................................Strength Coefficient
k..............................................Material Constant for Fatemi-Socie Equation
K′ ............................................Cyclic Strength Coefficient
Kf ............................................Fatigue Notch Factor
Kt ............................................Stress Concentration Factor
Lo ............................................Original Gage Length
N.............................................Applied Number of Cycles
Nd ...........................................Nodule Diameter
n..............................................Strain Hardening Exponent
n′.............................................Cyclic Strain Hardening Exponent
Nf ............................................Cycles to Failure
2Nf ..........................................Reversals to Failure
R .............................................R-ratio (ratio of minimum to maximum stress)
r ..............................................Notch Root Radius
q..............................................Notch Sensitivity Factor
S .............................................Engineering Stress
Sa ............................................Stress Amplitude
Sar ...........................................Equivalent Fully-Reversed Alternating Stress
Sf ............................................Fatigue Strength
Sm ...........................................Mean Stress
Smax .........................................Maximum Stress
Smin .........................................Minimum Stress
Sy ............................................Yield Strength
Sy′ ...........................................Cyclic Yield Strength
Su ............................................Ultimate Tensile Strength

xxii
γ ..............................................Empirical Walker Equation Fitting Constant
γa.............................................Shear Strain Amplitude
δa ............................................Displacement Amplitude of Loading Actuator
ε ..............................................True/Local Strain
Δε/2 ........................................True/Local Strain Amplitude
Δεe/2 .......................................True/Local Elastic Strain Amplitude
εf .............................................True Fracture Strain
εf′ ............................................Fatigue Ductility Coefficient
Δεp/2 .......................................True/Local Plastic Strain Amplitude
νe, νp, νeff .................................Elastic, Plastic, Effective Poisson Ratios
ρ..............................................Neuber’s Material Characteristic Length
σ .............................................True/Local Stress
Δσ/2 ........................................True/Local Stress Amplitude
σ0 ............................................Defect Free Fatigue Limit
σf ............................................True Fracture Stress
σf′............................................Fatigue Strength Coefficient
σn,max .......................................Maximum Normal Stress for Fatemi-Socie Equation

xxiii
Chapter 1

Introduction

1.1 Cast Iron and Its Classification

Cast iron, a ferrous metal, is divided into five classifications: grey, white,

malleable, ductile (or nodular), and compacted graphite iron. Cast irons generally contain

3.0 to 4.5 weight percent of carbon as well as other alloying elements. Grey iron contains

an α-ferrite or pearlite matrix surrounding graphite flakes; it is also weak and brittle in

tension due to high stress concentration points caused by the graphite flakes [1].

Typically, grey iron is not welded due to its tendency to crack if not carefully preheated.

The tensile strength of grey iron ranges from 100 MPa to 400 MPa. The compressive

strength can be 3 to 4 times greater than the tensile strength, because in compression, the

effect of the stress concentration from the graphite flakes is negligible [2].

White iron contains mostly cementite and very little graphite, resulting in a white

appearance on the fracture surface. Due to the high hardness of white iron, the material is

extremely difficult to machine, resulting in limited applications, which require high wear

resistance. White iron, though not typically used in design applications, is made as a

precursor for malleable iron [1].

1
 Malleable iron is formed when white iron is heated between 800 ºC to 900 ºC

(1470 ºF to 1650 ºF). During the time in heat treatment, the cementite breaks down into

graphite resulting in a microstructure similar to ductile iron, which attributes to the

similarities of strength and ductility to that of ductile iron. Applications of malleable iron

can be found in the automotive industry, pipes and pipe-fittings, as well as other heavy-

duty areas [1].

With the addition of magnesium and/or cerium to grey iron before the casting

process, ductile (or nodular) iron is formed, resulting in a noticeably different matrix.

Ductile iron contains small graphite particles that are spherical in nature, which develop

during the casting process. The graphite is surrounded by either a pearlite or a ferrite

microstructure depending on the heat treatment. In order to form a pearlite matrix, a

controlled shakeout from the casting mold is required. The softening of ductile iron

through an annealing process forms a ferrite matrix by removing the carbon present in the

material. Normalizing ductile iron produces more uniform properties through means of

austenitizing, which will form a fully pearlite matrix [3]. The mechanical properties of

ductile iron with a ferrite microstructure are near those of steel [1]. The modulus of

elasticity for ductile iron is near 172 GPa compared with steel, whose modulus is about

200 GPa. The ductility of nodular iron is similar to that of malleable iron, but if annealed

and followed by a slow cool, the ductility can become even greater. The compressive

strength of nodular iron is higher than its tensile strength, but the difference is small

compared to grey iron [2]. Several applications for the use of ductile iron have been

realized because it is more economical to manufacture while maintaining similar

mechanical properties as those of steel. Typical applications for ductile iron are in high

2
wear environments such as valves, pump bodies, automotive parts, machine components,

gears, shafts [1].

Compacted graphite iron (CGI) is a newer addition to the category of cast iron,

with a microstructure in between grey iron and nodular iron; the graphite present in CGI

is worm-like. The matrix for CGI will contain pearlite, ferrite or a combination thereof

[1]. It has been observed that increasing the nodularity increases CGI’s ductility and

strength; however, ferritic matrices result in lower mechanical properties compared to

pearlitic matrices. Some applications of CGI are found in the automotive industry with

parts such as engine blocks, high-speed brake discs and flywheels. Figure 1-1 shows the

microstructures of the cast iron classifications discussed, and Table 1.1 shows some

mechanical properties of various cast irons [1].

1.2 Importance of Fatigue Considerations in Design

The American Society for Testing and Materials (ASTM) defines fatigue as a

“progressive localized” process that results from fluctuating stresses and strains. This

implies that the fatigue process happens over time in a localized area of a component.

Fatigue failure is often sudden without any initial indications that the component is going

to fail. The mechanisms that cause fatigue failure may have been initiated since the

component was put into service [4].

The properties obtained from monotonic tension tests provide a great deal of

insight to a material. Important properties such as Young’s modulus, tensile and yield

strengths, and ductility are some of the properties obtained from a tension test. However,

these properties are considered for components or structures under static loads, and most

engineering designs must withstand cyclic loads applied for the majority of their useful

3
lives. Analogous to monotonic properties, cyclic properties must be considered when

designing for cyclic loading conditions, as these properties may be drastically different

from the monotonic values. Components such as gears, wings of an aircraft, and

suspension components are examples of designs that are cyclically loaded.

There are stress-based, strain-based, and fracture mechanics-based approaches for

life prediction of specimens or components. A very common method of characterizing

fatigue behavior is by performing crack initiation tests. These tests are the basis of stress-

or strain-life approaches where the majority of life is spent initiating a crack compared to

the further number of cycles which cause fracture. The fracture mechanics approach,

which is mainly concerned with crack growth, assumes a crack is initially present in the

material, and once that crack reaches a critical length, failure occurs.

The stress-based approach was first investigated by Wöhler in the 1860s from

failure studies of railroad axles. In this approach, specimens (smooth or notched) are

subjected to rotating bending, reversed bending, or uniaxial loading. The stress-based,

also known as the S-N approach relates cycles to failure, Nf, to the nominal stress

amplitude, Sa. The relationship is as follows:

𝐵 (1.1)
𝑆𝑎 = 𝐴(𝑁𝑓 )

In equation 1.1, A is the coefficient of the S-N line, which is the fatigue strength at one

cycle. Variable B denotes the slope of the S-N line.

The strain-based approach was developed by Coffin and Manson in the late

1950s. Unlike the S-N approach which is based on nominal stress, this method relates

reversals to failure to the total strain-amplitude at the critical fatigue location. The

relationship is as follows:

4
∆𝜀 ∆𝜀𝑒 ∆𝜀𝑝 𝜎𝑓′ 𝑏 𝑐
= + = (2𝑁𝑓 ) + 𝜀𝑓′ (2𝑁𝑓 ) (1.2)
2 2 2 𝐸

This method accounts for both elastic and plastic deformation. Four properties are needed

in this equation: fatigue strength coefficient, σf′, fatigue strength exponent, b, fatigue

ductility coefficient, εf′, and fatigue ductility exponent, c. The strain-based approach is

considered more comprehensive since it considers any plastic deformation that occurs in

localized regions.

Damage tolerant designs require the use of fracture mechanics. The stress and

strain-based approaches discussed previously are mainly concerned with crack initiation,

whereas the fracture mechanics approach can be viewed as a crack growth approach.

Griffith is considered the “early father” of fracture mechanics by showing the product of

the far field stress is related to the energy release rate G. Irwin, the “modern father” of

fracture mechanics used the stress intensity factor K, to quantify the driving force of the

crack tip. Irwin used the stress intensity factor to determine the energy release rate for

plane stress and plane strain conditions. Application of fracture mechanics in design

relates the growth of the crack length over time to the stress intensity factor range. The

crack growth rate, da/dN, is described by the following relationship:

d𝑎
= 𝐴(∆𝐾)𝑛
d𝑁 (1.3)

In equation 1.3, ΔK is the stress intensity factor range, which can be determined by

knowing the stress range and crack length. A is the coefficient of the crack growth line

and n is the slope of the line. The main approach used in this study is the stress-based

approach.

5
1.3 Objectives of This Study

This study had several objectives. The first objective was to provide a literature

review of topics covering fatigue behavior, effect of mean stress, and notch effects on

cast iron. Using data obtained from literature, tensile properties were correlated with

hardness and the fatigue strength was attempted to be correlated with hardness or tensile

strength. Data were also separated based on casting characteristics to observe any trends

related to the microstructure. The second objective of this study was to analyze data from

experimental work performed on 120-90-02 ductile cast iron in this study. The purpose of

the experimental work was to study the fatigue behavior on smooth and notched

specimens under the influence of mean stress. Neuber’s rules, strain energy density, and

finite element methods were used to obtain stresses and strains at the notch. Four mean

stress parameters were investigated to determine which method best predicts the fatigue

behavior under the presence of mean stress. For notched behavior predictions, both

nominal and local approaches were considered. The four mean stress parameters

investigated were Modified Goodman, Smith-Watson-Topper (SWT), FKM, and Fatemi-

Socie (FS) mean stress models. In addition to evaluation of the mean stress models,

stress- and strain-based methods were used to predict notch behavior.

This thesis starts with a literature review of cast iron. The literature review

introduces fatigue behavior on smooth and notched cast iron specimens, including

methods that estimate the fatigue strength. The next section presents attempted

correlations of tensile and fatigue properties and an evaluation of methods found in the

literature that are suggested for predicting the fatigue limit. A review of mean stress is

discussed next. A historical review, discussion of mean stress equations, and mean stress

6
related to fatigue research performed on materials including cast iron follow. The

experimental program is discussed next; topics include material used, specimen design,

testing equipment, methods and procedures to obtain tensile and fatigue data.

Experimental results are presented and discussed next starting with monotonic tension

and compression tests, followed by fully-reversed fatigue testing, notched behavior

prediction, and mean stress effect evaluation. The study ends with a summary of the

entire thesis, as well as conclusions made. The appendix contains all the data used from

literature for correlations.

7
Table 1.1: Mechanical properties of various grades of cast iron [1].

Tensile Strength Yield Strength Ductility

Grade Matrix Structure
[MPa (ksi)] [MPa (ksi)] [%EL in 50mm (2 in.)]
Gray Iron
SAE G1800 Ferrite + Pearlite 124 (18) - -
SAE G2500 Ferrite + Pearlite 173 (25) - -
SAE G4000 Pearlite 276 (40) - -

Nodular (or Ductile) Iron

ASTM A536
60-40-18 Ferrite 414 (60) 276 (40) 18
100-70-03 Pearlite 689 (100) 483 (70) 3
120-90-02 Tempered Martensite 827 (120) 621 (90) 2

Malleable Iron
32510 Ferrite 345 (50) 224 (32) 10
45006 Ferrite + Pearlite 448 (65) 310 (45) 6

Compacted Graphite Iron

ASTM A842
Grade 250 Ferrite 250 (36) 175 (25) 3
Grade 450 Pearlite 450 (65) 315 (46) 1

8
 (a) (b) (c)

(d) (e)

Figure 1-1: Microstructures of: (a) grey iron at 500x, (b) nodular at 200x (or ductile)
iron, (c) white iron at 400x, (d) malleable iron at 150x, and (e) compacted
graphite iron at 100x [1].

9
Chapter 2

Literature Review of Fatigue Behavior of Cast Iron and

Mean Stress Effects

2.1 Fatigue Behavior of Cast Iron

Germann et al. [5] studied the fatigue behavior and lifetime calculations of EN-

GJL-250 (lamellar graphite), EN-GJS-600 (nodular graphite), and EN-GJV-400

(compacted graphite) cast irons. Experiments consisted of stress-controlled load increase

tests and constant amplitude tests. The hysteresis, temperature, and electrical resistance

were recorded in order to characterize the fatigue behavior. They used the PHYBAL

method, which is a physically based life prediction method that calculates S-N curves by

generalizing Morrow and Basquin equations. Two constant amplitude fatigue tests and

one load increase test are required to predict fatigue behavior with this method. The

PHYBAL method was modified such that it allowed calculation of stress-life curves for

different failure probabilities. The endurance limit was estimated with one specimen with

load increase tests. Figure 2-1 shows S-N curves for EN-GJV-400 which were calculated

using the PHYBAL method incorporating plastic strain amplitude, change in temperature

and electrical resistance of the specimen.

10
 Salman et al. [6] studied the effects of various austempering temperatures on the

fatigue properties of ductile iron. Fully-reversed, strain-control testing was performed on

ADIs of various austempering temperatures and the endurance limit was considered to

range between 300 MPa and 350 MPa. It was found that the hardness of the ADI

decreases with the austempering temperature; and so did the cycles to failure.

Petrenec et al. [7] studied fatigue crack initiation on austempered ductile iron

under fully-reversed, strain-controlled testing at temperatures of 23 °C and -45 °C. It was

observed that the material cyclically hardened due to martensitic transformation of

unstable retained austenite caused by cyclic strain near graphite nodules at low

temperatures. Softening effects were observed at both temperatures due to crack initiation

near graphite nodules and plastic strain localization into slip bands. The main damage

mechanism was caused by the coalescence of short cracks, which initiated early from

graphite nodules around 5% of life into long cracks. The stress concentration factor of the

graphite nodules also increased due to the effects of low temperature. Numerical

simulations performed through FEA show that nodules located near the surface had a

higher damaging effect compared to an internal particle.

Čanžar et al. [8] used four types of nodular cast irons to study the effect of the

microstructure on the fatigue behavior. Tests were performed under axial strain-control.

Under fully-reversed fatigue testing, it was found that a mean strain had no significant

effect on the cyclic hardening behavior of the material. A circularity factor, proposed by

Russ, is shown in equation 2.1 and it takes into account geometry of the defect. Equation

2.1 was used to quantify those nodules from which cracks initiated. If the circularity

factor (CSF) is close to zero, the geometry is considered highly irregular. If the CSF is

11
equal to one then the geometry would be considered perfectly spherical. In the case of

ductile iron, the nodule is the considered geometry. The equation, where A is the area and

p is the perimeter of the nodule is as follows:

4𝜋𝐴
𝐶𝑆𝐹 =
𝑝2 (2.1)

It was found that larger, more irregularly shaped nodules promoted higher crack growth

rates and smaller values of the critical stress intensity factor compared to irons containing

a higher frequency of smaller, more regularly shaped nodules.

Yamabe and Kobayashi [9] studied the influence the casting surface had on the

fatigue strength of three cast irons which were cast in sand molds and then shot blasted.

Using statistics of extremes to determine the maximum defect size, the lower bound

fatigue limit could be estimated. Murakami’s method [10] was modified by considering

not only the defect size, but also the surface roughness. The authors proposed that the

fatigue limit could be determined without the use of fatigue tests, but rather through non-

destructive inspection which incorporates the surface roughness and nodule diameter.

The modified parameter, equation 2.2, is as follows:

∗ (2.2)
√𝑎𝑟𝑒𝑎𝑒𝑓𝑓 = √𝑎𝑟𝑒𝑎𝑅,𝑚𝑎𝑥 + √𝑎𝑟𝑒𝑎𝐷,𝑚𝑎𝑥

∗
The√𝑎𝑟𝑒𝑎𝑒𝑓𝑓 term incorporates surface conditions of the specimen and is the effective

defect size. The surface condition as it changes with shot peening is captured

by√𝑎𝑟𝑒𝑎𝑅,𝑚𝑎𝑥 and the defect size is captured by√𝑎𝑟𝑒𝑎𝐷,𝑚𝑎𝑥 . Table 2.1 contains the

estimates of actual experimental fatigue data. It can be seen that by increasing the number

of specimens the ratio of the estimated fatigue limit to the experimental fatigue limit

decreases.

12
 Tayanç et al. [11] studied the effects of different matrix structures using rotating

bending fatigue tests on austempered ductile iron. An increase in fatigue strength with

heat treatment temperatures between 230 °C and 330 °C was reported; however there was

a decrease in fatigue strength when the temperature range of heat treatment was between

330 °C and 430 °C. Fractography revealed that the specimens with the highest fatigue

strength exhibited ductile fracture, while specimens with the lowest fatigue strength

revealed mostly brittle cleavage cracking. Two distinct observations were made about the

fatigue crack. The first observation was that the fatigue crack grows through the matrix

by connecting nodules along the austenite/ferrite interfaces. The second observation was

that the crack would cut through the lath of the austenite and ferrite layers. The

specimens containing a martensitic matrix had the lowest fatigue strength.

Collini et al. [12] studied how the microstructure affected the mechanical and

fatigue properties of identical pearlitic grey cast irons from three different foundries. A

Weibull-type distribution was applied to compare mean values and scatter of the data

points. Experiments showed a high degree of scatter of S-N data among the three

different foundries. The shape and dimensions of the graphite lamellas influence the

static and dynamic strength of the specimens since they can be considered as notches or

cracks. The ultimate tensile strength was highest in microstructures with large primary

dendrites and lowest when the dendrites were short and globular. It was also noticed that

the ultimate tensile strength and fracture toughness decreased with the size of the eutectic

cell size, but the fatigue strength increased. As the structure of the eutectic cells becomes

finer and as phosphorus decreases, the fatigue strength increases. The graphite, as well as

13
the casting process and expertise of the foundry influence the properties of the iron.

Figure 2-2 shows these differences on an S-N diagram.

The fatigue behavior of spheroidal graphite cast iron containing vanadium

carbides in a martensitic microstructure was investigated by Tokaji et al. [13]. Rotating

bending specimens were prepared with one group containing a fully martensitic matrix

and the other containing a martensitic/bainitic microstructure. From the experimental

tests, it was determined that the fully martensitic structures had a higher fatigue strength.

The fatigue limit was noticed to increase with Vickers hardness, but would decrease

above a hardness of 650. Most cracks initiated from casting defects. It was stated that

estimation of the fatigue strength by Murakami’s method is suitable for as-cast materials.

For heat-treated materials, as the hardness increased, the predicted fatigue limit was

higher than the experimental limit and this difference became greater as the hardness

increased, as shown in Table 2.2.

The fatigue limit based on casting size and various austempering temperatures in

the high cycle fatigue regime was investigated by Lin and Wei [14]. The austempering

treatments increased the ultimate strength, yield strength and hardness, compared to the

as-cast material. The ultimate strength is strongly dependent on the morphology of the

ausferritic matrix and the amount of micro-shrinkages, which is correlated to the casting

size. As the size increases, the nodule count and nodularity decrease, but the nodule size

increases. This further reduces the fatigue limit of the specimen, as shown in Figure 2-3.

The fatigue limit was found to decrease linearly as the mean nodule diameter increased.

However, a unique relationship did not exist between fatigue limit and nodule diameter

because the fatigue strengths were significantly different at both austempering

14
temperatures. It has been observed that the fatigue limit, Sf, is related to impact toughness

and mean nodule diameter by the following equation with R2 = 0.91:

𝑆𝑓 = 371.51𝐼 0.31 𝑑−0.40 (2.3)

where I is the impact toughness and d is the mean nodule diameter. Equation 2.3 states

that the fatigue limit increases as impact toughness increases and nodule size decreases.

Results of equation 2.3 can be seen in Figure 2-4

The influence of the matrix on an alloyed ductile iron was investigated under

rotating bending fatigue tests. Similar to [14], Toktaş et al. [15] reported that the fatigue

strength is not proportional to tensile strength and hardness, but rather impact toughness

and nodule diameter using equation 2.3. A correlation between the fatigue ratio and

matrix structure was not found, as the fatigue ratio varied from 0.222 to 0.354, see Table

2.3. The fatigue ratio was defined as the fatigue strength normalized by the tensile

strength. The tensile properties, hardness, and fatigue strength were found to increase

with higher contents of pearlite. Low austempering temperatures were found to give

higher strength, but lower ductility. Austempering at 365 °C resulted in the formation of

carbide precipitates, causing a reduction in the mechanical properties. Fatigue cracks

were found to initiate at the nodule/matrix interface due to a weak interface bond and the

fact that the nodules have a very low Young’s modulus.

Tanaka et al. [16] investigated the fatigue limit of spheroidal graphite cast iron

under plane bending using plate specimens in fully-reversed bending, the relation

between fatigue strength and microstructural factors, and developed prediction equations

for the fatigue limit. It was found that the fatigue strength linearly decreased with an

increase of the maximum and mean nodule diameters for irons with the same matrix

15
structures. However, if the iron was annealed, the fatigue limit became less sensitive to

nodule size. Fatigue limit of matrices which were harder or were quenched and tempered,

had a greater dependence on the nodule size, as can be seen from Figure 2-5. Figure 2-6

shows that the relationship between fatigue strength and tensile strength is not simple, but

that it also depends on nodule size. During cyclic loading, it was also found that the

hardness of the matrix increases.

Lin and Lee [17] examined the effects of highly stressed volume on the fatigue

strength of austempered ductile iron (ADI) using both notched and smooth specimens.

Results showed that the fatigue strength in axial loading was significantly lower than in

rotating bending tests. It was also shown that the fatigue strength decreased due to notch

effects and the fact that ADI has a harder, more brittle matrix microstructure. A universal

relationship between the fatigue limit and the highly stressed volume for various irons,

loading modes, and specimen geometries was developed and given by:

𝐾𝑡 𝑆𝑓
= 11.1𝑉 −1
𝐼 0.747 (2.4)

where I is the impact toughness of the material, Sf is the fatigue limit, V is the highly

stressed volume, and Kt is the stress concentration factor.

Nadot et al. [18] studied the effects of casting defects of nodular iron on high

cycle fatigue properties. Surface and internal shrinkages were analyzed to study the

effects on the fatigue limit. Experimental fatigue data were also compared with six

prediction models. Table 2.4(a) shows the analytical models used and Table 2.4(b) shows

the input parameters required by each model. Testing was performed in uniaxial loading

with R = -1 to evaluate the analytical models. It was found that internal defects are less

damaging than defects located at or near the surface. The six methods that were used to

16
estimate the fatigue limit were developed by El Haddad and Tanaka who modeled the

defect as a crack, Lukas and Mitchell who modeled it as a notch, and Murakami and de

Kazinczy who took into account the defect size and the material parameters. For

Murakami’s method, the projected area was measured on the fracture surface, after

failure. Also, for this method, surface and internal defects need to be considered

separately. In Table 2.4(a), relation 2 gave less than satisfactory results. Relations 1 and 2

impose an evolution of the fatigue limit as the dimensions of the defect change. When

modeling the defect as a notch, relations 3 and 4, provided better results. Relation 4 was

more accurate for smaller defects and relation 3 requires a critical defect size in order to

be used. Relation 5 provided an acceptable approximate value of the fatigue limit, but the

equation needs to be adapted for individual materials instead of just using the

macroscopic hardness. Relation 6 could not accurately account for large defects in cast

iron. Having the defect modeled as a crack gave conservative results, whereas modeling

it as a notch provided more accurate results. Non-propagating fatigue cracks were also

observed to appear when tested under the fatigue limit.

The model used in [18] was originally proposed by de Kazinczy [19] which was

used to investigate cast steel in the 800 MPa to 1040 MPa tensile strength range. The

defect size used in the equation was considered the smallest circle that enclosed the entire

defect, which was determined to be 0.05 mm. As with cast iron, surface or near surface

defects were more detrimental to cast steel than internal defects. Equation 2.5 was used to

determine the fatigue limit:

𝜎0
𝜎=
𝑆𝑦 𝑑0.5
1+ 𝐾 (2.5)

17
where σ0 is the defect free fatigue limit taken as 0.4Su, a is a constant, d is the defect size

and σ is the endurance limit, Sy is the yield strength, and K represents defect geometry. A

linear relationship between the fatigue limit and tensile strength was not found. The

fatigue limit is a function of the defect size and the constant a is proportional to the yield

strength.

Costa et al. [20] attempted to establish a formula to estimate the fatigue limit, σf,

of nodular iron and compare results by considering the internal defects as a crack, a

notch, and other geometrical defect parameters. The first model used as a basis of

comparison was the Murakami model, see Table 2.4(a) relation 5. The second model

applied was the de Kazinczy model, see Table 2.4(a) relation 6. A third model by

Mitchell was also used, see Table 2.4(a) relation 4. The model developed in [20] given by

equation 2.6 uses a normalized factor, ψi, which takes into account the areas of the

nodules and the frequency of nodule areas:

𝜎𝑓 = 1.25𝐻𝑣(1 − 0.003𝜓𝑖 ) (2.6)

The normalized factor is defined as:

𝜓𝑖 = 𝑥 𝑦
(2.7)
where x is the equivalent graphite nodule area and y is the frequency of the graphite

nodule area. The specimens were tested using a 3-point bending fatigue test machine.

Based on the R2 values, the results showed that the proposed model was a better predictor

than the other three models. The comparison of all four models can be seen in Figure 2-7.

Kogoh et al. [21] evaluated the fatigue strength and ultimate strength of smooth

spheroidal graphite cast iron specimens under axial load at elevated temperatures. Results

showed the ultimate tensile strength and fatigue strength decreased as temperatures

18
increased. The effect of heat treatment was evident when evaluating the tensile strengths

of the normalized and bainitized specimens regardless of testing temperature as shown in

Figure 2-8. Different heat treatments did not have significant effects on the fatigue limit

at elevated testing temperatures, shown in Table 2.5. For the cases of room temperature,

300 ºC, and 500 ºC the annealed material had the highest ratio of fatigue strength

normalized by the tensile strength, followed by the as-cast, normalized, and bainitzed

materials.

Maluf et al. [22] used surface rolled notched specimens made of 100-70-03 cast

iron and compared the fatigue test results with those of smooth and notched specimens

that were not surface rolled. The diameter of the notch was 6.3 mm with a notch root

radius of 1.2 mm with elastic stress concentration factor of Kt = 1.63. The matrix

equivalent micro-hardness (MEM) was found to be 304 Hv. It was found that there was a

correlation between the MEM and endurance limit of the smooth specimens. The fatigue

notch factor was determined by using Peterson’s equation [4]:

𝐾𝑡 − 1
𝐾𝑓 = 1 +
1 + 𝑎⁄𝑟 (2.8)

300 1.8
𝑎 = 0.001 ( ) 𝑆𝑢 in ksi and 𝑎 in inch
𝑆𝑢 (2.9)

where 𝑎 is a material characteristic length, r is the radius of the notch, and Su is the

ultimate strength of the material. Kf was experimentally determined to be 1.8, but

Peterson’s equation predicts Kf = 1.56. It should be noted that the Peterson model was

developed for ferrous, wrought materials. Stress-life data, shown in Figure 2-9, indicate

that surface rolling the notch increased life well beyond the smooth specimens resulting

in a fatigue limit that was 60% higher than the smooth specimen fatigue limit. It was also

19
found that surface rolling the notch caused subsurface crack nucleation at stress levels

near the fatigue limit.

Nodular cast iron specimens which were as-cast, shot-blast (having the as-cast

surface finish), and fine-ground were studied by Konečná et al. [23]. Testing consisted of

constant amplitude displacement cyclic bending. A change in 5% of the maximum stress

was defined as the boundary between crack initiation and propagation phases. Based on

fatigue tests results shown in Figure 2-10, the fine ground specimens had the best fatigue

performance and as-cast specimens had the worst fatigue performance. The shot-blast

specimens were in-between the fine-ground and as-cast. A fracture mechanics approach

was suggested to estimate the fatigue strength based on the amount of time in crack

propagation for as-cast and shot-blast specimens.

Zammit et al. [24] determined the optimal shot peen for austenitized and

austempered Cu-Ni ductile iron. Through testing, it was determined that an Almen

intensity rating greater than 0.4 reduced the fatigue life, see Figure 2-11. The surface

hardness also decreased from the edge to the interior from 535 Hv to 370 Hv. The fatigue

strength, however, increased from 250 MPa for a polished specimen to 390 MPa for the

shot peened specimen. The effect of shot-peening also delayed crack propagation and

shifted the initiation to subsurface. Fracture surfaces showed multiple crack initiation

sites, where graphite nodules arrested crack growth. It was proposed that cracks

propagate through the weak interface of copper precipitates with the ausferrite matrix.

Ochi et al. [25] studied fracture morphology and fatigue properties of a ferrite-

pearlite and pearlite ductile cast iron. The specimens were shot-peened and subjected to

rotating bending. It was observed that in the long-life regime the ferrite-pearlite matrix

20
structure exhibited a two-step S-N curve beyond 107 cycles, whereas the pearlite matrix

did not behave in the two-step manner shown, as in Figure 2-12. It was determined that

this two-step phenomenon occurs due to the location of the fracture origin. Typically, at

longer fatigue life, the fractures occurred internally. It was also observed that the effects

of a surface or near surface defect are more detrimental to the fatigue life compared to

internal defects. Figure 2-12 depicts the transition region between surface and internal

fracture.

With the increase of ductile iron being used in the automotive industry, a failure

analysis of a crankshaft used in a truck was performed by Asi [26]. The carbon content

was found to affect the fatigue and impact strengths, as it affects the size and number of

graphite nodules. Smaller nodules with a higher count increases the fatigue strength [27].

The crack path is strongly influenced by non-spheroidal graphite, which also produces a

rough fracture surface composed of brittle cleavage facets, indicating low material

toughness.

Šamec et al. [28] studied low-cycle fatigue of ductile cast iron EN-GJS-500-7

which is used for railway brake discs. Investigation was performed at room temperature,

300 °C and 400 °C under axial strain-controlled tests. At room temperature, a slight

hardening was noticed near the end of the fatigue life. At the elevated temperatures,

hardening was more pronounced at 300 °C, while the initial hardening observed at

400 °C transitioned to softening near the end of the fatigue life. Fatigue life was shown to

decrease with increasing temperatures.

21
2.2 Correlations for Cast Iron Data Based on Data from Literature

The purpose of analysis in this section was to find correlations among cast iron

data extracted from literature, see Tables A.1 and A.2. With successful correlations, only

a few tests would need to be performed to verify the predictions for a particular material.

Correlation attempts were made to determine tensile, yield, and fatigue strenghts.

Literature studies suggest that the most controlling microstructural feature on material

behavior of cast iron is the nodule size. Therefore, data in the correlations are

distinguished by nodule size.

For steels, the S-N curve can be approximated by drawing a line from tensile

strength at one cycle to 50% of the tensile strength at one million cycles [4]. Several

authors have shown that the fatigue ratio of cast iron is not constant [9, 21]. Therefore, a

new method is needed to estimate the fatigue limit of cast iron. Fatigue strength

predictions proposed in literature were evaluated to verify the applicability of the

methods for a wide range of cast iron.

2.2.1 Correlations Among Mechanical Properties Pertaining to Cast

Iron
Evaluation of the tensile and yield strength by grouping the average nodule size in

groups of 10 μm shows that the tensile and yield strength are dependent on material

hardness, rather than the nodule size. A linear fit was obtained by using all applicable

data from literature up to a hardness of 450 Hv for estimations of both Su and Sy, see

Figures 2-13 and 2-14. The R2 values for the tensile and yield strengths are 0.815 and

0.913, respectively, and are represented by the following equations:

22
𝑆𝑢 = 3.24𝐻𝑣 − 79.40 (2.10)

𝑆𝑦 = 3.56𝐻𝑣 − 382 (2.11)

Attempts to correlate fatigue strength with hardness and tensile strength were made

with data separated by nodule sizes. Figure 2-15 shows the fatigue limit as a function of

hardness indicating a correlation does not exist. Figure 2-16 shows the fatigue limit as a

function of tensile strength, also indicating a lack of correlation. These figures, however,

show that smaller average nodule diameters tend to result in higher fatigue limits. Similar

analysis using the maximum nodule size shows that smaller nodules and higher hardness

combinations result in higher fatigue limits, as shown in Figure 2-17. However, a strong

correlation between fatigue limit and nodule size could not be found. Results shown in

Figure 2-18 indicate no correlation of fatigue limit with tensile strength.

Analysis of the fatigue limit as a function of hardness and tensile strength shows

that material hardness and maximum nodule size play a role in determining the fatigue

limit. Smaller nodules and increasing hardness are beneficial to fatigue behavior,

although a mathematical relationship could not be found.

2.2.2 Methods to Predict Fatigue Limit

Four methods found in literature are discussed in this section to estimate the

fatigue limit. These include the method proposed in [14] and Murakami’s, Mitchell’s, and

de Kazinczy’s methods. The fatigue limit is defined as the fatigue strength at 106 or 107

cycles.

The first method evaluated, which was proposed by Lin and Wei [14] suggests

that the fatigue limit is dependent on both impact energy and average nodule diameter.

Application of this method to data from the literature in this study yielded a similar trend,

23
but with a much less accurate power fit of all the data. Lin and Wei used a very limited

number of materials when developing this method, which resulted in an R2 value of 0.91

(see Figure 2-4). When this method was applied to a wide range of cast irons, the R2

value decreased to a value of 0.535, see Figure 2-19.

Murakami’s method [10] was applied to the data using the coefficients for both

internal and external defects because the data obtained from literature on nodule size does

not specifically state the location of the performed microstructural analysis. The equation

for Murakami’s method can be seen in Table 2.4(a), relation 5. For internal defects

A = 1.56 and for external defects A = 1.43. Results show that when assuming the

maximum defect to be on the surface of the specimen, the calculated values are closer to

the experimental values. However, estimates are still over-predicted. The internal defect

assumption over-estimates the fatigue limit for nearly all materials analyzed. Results of

this analysis can be seen in Figure 2-20(a) for external defects and Figure 2-20(b) for

internal defects. Data are differentiated by axial or rotating bending fatigue tests.

The method proposed by de Kazinczy [19] was originally developed for cast steel,

but is applied here for the case of cast iron. Table 2.4(a), relation 6 shows de Kazinczy’s

equation. The constant K is used to account for defect geometry. Three values of K are

2550 MN (mm)-1/2 for internal microshrinkages, 1130 MN (mm)-1/2 for surface

microshrinkages, and 750 MN (mm)-1/2 for pores. As suggested, the defect free fatigue

limit was taken to be 0.4Su, and K was assumed to be 750 MN (mm)-1/2 due to the nature

of the defects. Analysis of the results, as shown in Figure 2-21, indicates that about half

of the data fall within a ±10% scatter.

24
 The method proposed by Mitchell which can be found in [20, 29] and Table 2.4(a),

relation 4. This method uses the elastic stress concentration factor Kt, the size of the

defect, and An. For nodular iron, Mitchell [29] assumes the graphite nodules as spheres,

therefore Kt = 3. However, research performed by Nadot et al. [18] assumed Kt = 2. Both

values were used in this work. Mitchell’s method was analyzed for both average and

maximum nodule size. For Kt = 2 and the use of average nodule diameter the results

shown in Figure 2-22(a) are non-satisfactory. When the value of Kt = 3 was used, the data

fell to more reasonable predictions, as shown in Figure 2-22(b) but the results are mostly

non-conservative. More than half of the data fall outside ±10% scatter bands in both

cases. When the maximum nodule diameter is used, a value of Kt = 2 resulted in more

reasonable predictions for the fatigue limit as shown in Figure 2-22(c) with about half of

the data falling within ±10% scatter bands. When using the maximum nodule diameter

and Kt = 3, the results are then conservative, as shown in Figure 2-22(d), where over 60%

of the data fall outside the lower scatter band. Using the premise of the circularity factor,

as previously described in relation to equation 2.1, perfect spherical geometry is

considered when CSF = 1, resulting in Kt = 3. If the circularity factor is not equal to 1

then the shape is irregular. Nodule geometry tends to be uniform throughout its

microstructure but the nodules are not necessarily perfect spheres. Therefore, if it is

assumed that the nodule is elliptical and the width of a nodule is one-half of its height,

then Kt = 2. This may explain why Kt = 2 results in more reasonable approximations.

2.3 Mean Stress Effects on Fatigue Behavior

Mean stress is present when loading conditions are not fully-reversed, that is the

magnitude of the maximum and minimum loads are not the same. Many mechanical

25
designs undergo cyclic loading where the loading is not symmetric. An example of this

would be a pressure vessel where there is pretension on the bolts, which is always

present, in addition to the loading caused by pressure and depressurization of the

component. Typically, compressive mean stress is beneficial for fatigue life such that

lives are longer than the fully-reversed loading condition. Conversely, tensile mean stress

is detrimental to fatigue life causing a reduction in useful life. Understanding the effects

of mean stress on material fatigue behavior is important for a design engineer.

This section contains a brief history on the study of mean stress, followed by a

review of the more common mean stress models. Then, less commonly used models are

discussed. After the models are discussed, a review of other works involving the effect of

mean stress on smooth and notched specimens for various materials are included,

followed by a section specific to cast iron.

2.3.1 A Historical Review of Mean Stress

Sendeckyj [30] wrote a detailed paper on the historical aspects of mean stress and

constant life diagrams. What follows is a summary from this paper.

In 1873, Müller published the first graphical representation of constant life data

where maximum stress was plotted against minimum stress. In 1874, Gerber published

graphical representations of fatigue data produced by Wöhler by plotting maximum

versus minimum stress. Both values were normalized by the ultimate tensile strength. In

1880, Smith plotted the maximum and minimum stress as a function of the stress ratio. In

1899, Goodman published a similar representation, where he plotted Smax/Su and Smin/Su as

the ordinates against an arbitrary abscissa, which was based on dynamic theory. Smith

plotted his original data from 1880 against the mean stress in 1910. Haigh first presented

26
the earliest example of an amplitude-mean plot in 1917 where he plotted the maximum,

minimum, and stress range against the mean stress. Wilson and Haigh presented the first

amplitude-mean plot for constant life in 1923. The development of constant life diagrams

and mean stress equations were first introduced into engineering design by bridge and

railroad engineers. Due to the nature of loading these structures undergo, maximum and

minimum stresses were the primary focus in development of these equations.

In 1870, Wöhler stated that the maximum stress is a function of the stress range.

In 1873, Launhardt developed the following equation for the maximum stress which is

known as Launhardt’s formula:

𝑆𝑚𝑎𝑥 = 𝑆0 + (𝑆𝑢 − 𝑆0 )𝑅 (2.12)

where So is the value of Smax for R = 0. However, Launhardt’s formula is only valid for

the range of 0 ≤ R ≤ 1. In 1877, Weyrauch used Launhardt’s formula, but modified it for

the case of -1 ≤ R ≤ 0, which resulted in the Weyrauch formula, given by:

𝑆𝑚𝑎𝑥 = 𝑆0 + (𝑆𝑢 − 𝑆−1 )𝑅 (2.13)

where S-1 is the value of Smax for R = -1. Both Launhardt’s and Weyrauch’s formulae are

always used together and have become known as the Launhardt-Weyrauch Formula.

Several proposed modifications of Launhardt’s formula were debated between the years

of 1885-1897. Johnson, Rankine, and Seaman all had rationale that lead to the following

representation of the Launhardt-Weyrauch formula:

𝑆𝑚𝑎𝑥 = 𝑆𝑚𝑖𝑛 + 2∆𝑆 − 𝑆𝑢 (2.14)

In 1887, Fidler published a derivation of dynamic theory for use in bridge design.

Goodman in 1899 suggested that safe operating loads could be determined by dynamic

theory. In 1917, Haigh developed the equation that is now known as the Goodman

27
equation. In 1923, Wilson and Haigh included the constant yield stress boundary line to

the amplitude-mean diagram. In 1930, Soderberg modified the Goodman formula by

replacing ultimate strength with yield strength. In 1943, Kinasoshivili proposed that

constant life diagrams can be used by using the fatigue limits at R = 0 and R = -1 by

drawing a straight line through those two points. The safe operating region is defined

within the straight line and yield line.

2.3.2 Discussion of Mean Stress Equations

The following relationships are used to relate mean stress to stress amplitude:

𝑆𝑚𝑖𝑛
𝑅=
𝑆𝑚𝑎𝑥 (2.15)

𝑆𝑚𝑖𝑛 = 𝑆𝑚 − 𝑆𝑎 (2.16)

𝑆𝑚𝑎𝑥 = 𝑆𝑚 + 𝑆𝑎 (2.17)

By substitution of equations 2.16 and 2.17 into equation 2.15 and rearranging for Sm, the

following relationship is derived:

1+𝑅
𝑆𝑚 = 𝑆
1−𝑅 𝑎 (2.18)

Equation 2.18 can then be substituted into the following mean stress equations which are

discussed next, thus resulting in an equivalent alternating stress for R = -1, which can

then be used to predict fatigue lives based on the fully-reversed S-N equation.

Some of the most common stress-based mean stress equations referenced in the

literature are the Modified Goodman, Morrow, Smith-Watson-Topper, Walker, Gerber

and Soderberg equations. The Modified Goodman equation is given by:

𝑆𝑎 𝑆𝑚
+ =1
𝑆𝑎𝑟 𝑆𝑢 (2.19)

28
It suggests that as the ultimate strength of a material increases, Sar becomes less sensitive

to mean stress. It has been shown in the literature that the Modified Goodman approach

tends to under-predict fatigue lives, thus being considered a conservative approach [31-

34]. Dowling et al. [33] noted that for tensile mean stresses the Modified Goodman line

under-predicts fatigue life; however, for compressive mean stress, it over-predicts life.

The next model was developed by Morrow, given by:

𝑆𝑎 𝑆𝑚
+ =1
𝑆𝑎𝑟 𝜎𝑓 (2.20)

Instead of using the ultimate strength in Modified Goodman equation, the monotonic

tension test value of true fracture stress is used in this equation. For materials that do not

exhibit a high degree of necking, the difference between tensile strength and true fracture

stress is rather small, thus, Morrow and Modified Goodman equations give similar

results. However, in the case of materials that exhibit a high degree of necking, the

difference between the tensile strength and true fracture stress can be rather large. Some

authors have replaced the true fracture stress by the fatigue strength coefficient [31, 32].

Dowling has shown that for various steels and aluminum materials, the Morrow equation

gives excellent results in comparison to the Goodman equation, using either the true

fracture stress or the fatigue strength coefficient [32]. However, for nonferrous metals,

using the fatigue strength coefficient resulted in non-conservative estimates [32].

The stress-based version of the Smith-Watson-Topper (SWT) approach, also used

in [31-33], is a simple model for mean stress effect, given by:

𝑆𝑎𝑟 = √𝑆𝑎 𝑆𝑚𝑎𝑥

(2.21)
This approach has been shown to give acceptable and consistent results for a wide range

of materials. Dowling [32] suggests both the stress and strain versions of the SWT

29
method for general use. It is suggested in [32] that this method is a better predictor for

aluminum alloys compared to steels.

The Walker equation, used in [31-33, 35] is considered to be highly versatile due

to the adjustable curve fitting parameter γ, given by:

1−𝛾 𝛾
𝑆𝑎𝑟 = 𝑆𝑚𝑎𝑥 𝑆𝑎 (2.22)
The parameter γ, is a mean stress sensitivity indicator that ranges from 0 to 1. The lower

the value of gamma, the more sensitive a material is to tensile mean stress. If a material’s

γ value is unknown, it is most commonly assumed as 0.5 [4]. When this value is assumed,

the Walker equation collapses to the SWT approach, equation 2.21. The procedure for

determination of γ can be found in [33]. For steel, a correlation between tensile strength

and γ exists such that as material strength increases, γ tends to decrease [33]. Burger and

Lee [31] found a good correlation of their mean stress data with the Walker relationship

and used this model as a base comparison for other models evaluated.

The Gerber method is represented by the following equation:

𝑆𝑎 𝑆𝑚 2
+( ) =1 (2.23)
𝑆𝑎𝑟 𝑆𝑢

This method tends to over-predict fatigue lives and does not account for the beneficial

effect of compressive mean stress [36]. It is stated in [34] that 90% of the results for

ductile metals are above the modified Goodman line and 2/3 of the way to the Gerber

parabola. The effect of mean stress on the fatigue life of 316 LN stainless steel was tested

in both air and mercury by Strizak and Mansur [37]. It was shown that the Gerber

equation over-predicts mean stress effects for R = 0.1 and under-predicts mean stress

effects for R-ratios higher than zero, see Figure 2-23. For R-ratios of 0.1 and 0.3, the data

fell in between the Goodman and Gerber lines.

30
 The Soderberg line equation is similar to the Modified Goodman line equation,

except ultimate strength is replaced by yield strength:

𝑆𝑎 𝑆𝑚 (2.24)
+ =1
𝑆𝑎𝑟 𝑆𝑦

Woodward et al. [38] stated, “The Soderberg line is safe for nearly all materials,

but in very many instances the line seriously over-estimates the effect of mean stress.”

The Soderberg method is considered an extremely conservative method.

2.3.3 Less Commonly Used Mean Stress Equations

Some other less commonly used mean stress models are the Bagci, Clemson, the

Quadratic-Soderberg, and the FKM method (popular in Europe). These methods are

discussed next.

The Bagci, Clemson, and Quadratic-Soderberg equations are demonstrated in

[34]. The quadratic version of the Soderberg method was shown by Hohenemser and

Prager to predict fatigue failure under the presence of a static tensile and alternating

torsion stresses. It is given by:

𝑆𝑎 2 𝑆𝑚 2
( ) +( ) =1
𝑆𝑎𝑟 𝑆𝑢 (2.25)

This method is similar to Gerber; however, it extends beyond the yielding region. The

Bagci line equation is given as:

4
𝑆𝑎 𝑆𝑚
+( ) =1
𝑆𝑎𝑟 𝑆𝑦 (2.26)

Wang and Dixon [39] used Bagci line and showed that yielding may happen when the

ratio of the fatigue strength to yield strength is greater than 0.25.

31
 The Clemson method given by equation 2.27 was shown to give good initial

approximations for ferrous ductile materials [34]:

𝑆𝑦
⁄𝑆
𝑎𝑟
𝑆𝑎 𝑆𝑚
+( ) =1 (2.27)
𝑆𝑎𝑟 𝑆𝑦

The Clemson line has a slope that is always greater than the slope of the yield line, thus

loading conditions that cause data to fall below the Clemson line do not result in any

form of yielding. This method was developed based on the Bagci criteria of “a good

failure criterion would define a stress domain free of both fatigue failure and the

possibility of yield” [34]. Results of this method compared with several other methods for

a medium carbon steel can be seen in Figure 2-24.

Burger and Lee [31] applied the FKM method [40] (a German strength

assessment of components guideline) to several steels and aluminum alloys. A low

degree of validation of this method through independent research exists, thus the aim of

the work was to compare the FKM method to the Walker equation. The standard

deviations of several mean stress models were evaluated with respect to the Walker

equation. The FKM method divides the Haigh diagram into four regions, as shown in

Figure 2-25. The first region is for R > 1, region II is for ratios between -∞ ≤ R ≤ 0.

Region III is for 0 < R < 0.5, and region IV is for R ≥ 0.5. All regions are characterized

by a slope Mσ which is valid for elevated or room temperature conditions, given by:

𝑀𝜎 = 𝑎𝑀 10−3 𝑆𝑢 + 𝑏𝑀 (2.28)

The values aM and bM for various materials are listed in Table 2.6. Determination of the

equivalent alternating stress for each region is broken down as follows:

𝑆𝑎𝑟 = 𝑆𝑎 (1 − 𝑀𝜎 )
Regions I (2.29)

32
𝑆𝑎𝑟 = 𝑆𝑎 + 𝑀𝜎 𝑆𝑚
Region II (2.30)
𝑀
𝑆𝑎 +( 𝜎⁄3)𝑆𝑚
𝑆𝑎𝑟 = (1 + 𝑀𝜎 ) 𝑀
1+( 𝜎⁄3) Region III (2.31)

3(1 + 𝑀𝜎 )2
𝑆𝑎𝑟 = 𝑆𝑎
3 + 𝑀𝜎 Region IV (2.32)

The results for axial mean stress tests of AISI 4340 steel are shown in Figure 2-26. The

results for 50CrMo4 steel and 2014-T6 aluminum were similar with respect to the

analysis by the FKM method. The FKM method was shown to exhibit reasonable

agreement with the Walker equation with the smallest standard deviation compared to the

other mean stress models evaluated.

Kwofie [41] developed an exponential stress function to predict the fatigue

strength and life in the presence of mean stress. A mean stress sensitivity parameter is

suggested such that 0 ≤ α ≤ 1. If α is zero, the material is insensitive to mean stress, but if

greater than zero, it has some degree of sensitivity. In order to predict a stress-life

relationship, the following equation was developed:

𝑆𝑎𝑟 = 𝑆𝑎 𝑒 [𝛼(𝑆𝑚 ⁄𝑆𝑢)] (2.33)

A similar method was developed in [41] for strain-controlled fatigue testing where strain-

life properties are considered. It has been suggested that α may not always be constant

and is dependent upon the material [41].

Another proposed method by Niesłony and Böhm [42] suggests using two S-N

curves, one fully-reversed curve and another at a different R-ratio. The general case

equation is given by:

1 − 𝑅 𝜎𝑎𝑁,𝑅=−1 − 𝜎𝑎𝑁,𝑅
𝜎𝑎𝑇 = 𝜎𝑎𝑖 [1 + 𝜎𝑚 ( ) 2 ]
𝑅+1 (𝜎 ) (2.34)
𝑎𝑁,𝑅

33
where σm is the mean stress, σaN,R is the fatigue strength amplitude for a given number of

cycles, σaT is the transformed alternating stress amplitude, σai is the desired stress

amplitude, and R is the stress ratio. Acceptable results were found with this model.

However, the main limitation is the requirement that a minimum of two S-N curves are

required for fatigue calculations.

2.3.4 Effects of Mean Stress on Notched Specimens

Mean stress is a broad topic and present in many loading scenarios and analyses.

Extensive literature reviews have been performed in the past, therefore, papers

concerning mean stress effect alone are not considered in this review. In 1975, Maddox

[43] performed a literature review pertaining to mean stress with over 50 references. The

main aim of this section is to provide a narrower range of papers concerning the effect of

mean stress on notched specimens for any metallic material.

Chiandussi and Rossetto [44] used the point and line methods as proposed by

Taylor [45] to predict fatigue limits of Ti6A14V alloy using round and flat notched

specimens with various R-ratios. Both specimens contained a 60º v-notch, and the flat

specimen contained a notch on both sides in the gage section. This method was also

compared with Goodman and SWT approaches. Data used were taken from literature

with three R-ratios of 0.1, 0.5, and 0.8. The authors considered three scenarios: the first

used knowledge of ultimate strength, yield strength, and fully-reversed fatigue strength of

the material and the longitudinal normal stress distribution around the notch root. The

second scenario used the same mechanical properties, except a multi-axial stress

distribution around the notch root was considered. The third scenario used the

experimental fatigue strength and the threshold stress intensity range values for different

34
stress ratios. Analysis showed that the point and line methods were always conservative

and the line method was more accurate than the point method. In order to increase the

effectiveness of the proposed methods, an equivalent mean and alternating stress was

applied through calculations such as Sines theory. The models were further improved by

introducing the El Hadadd parameter. These proposed methods are found to be suitable

for small R-ratios. However, for high R-ratios they were less accurate. The scatter at the

high stress ratio was attributed to plastic deformation near the notch, which was not taken

into consideration. By introduction of the SWT equation, the gap between the theoretical

and experimental values for the larger R-ratio was closed. Calculations were conservative

for small notches and slightly non-conservative for large notches.

Krug et al. [46] studied the effects of mean loading and residual stresses on

notched specimen fatigue strength of 16MnCr5 steel. The specimens were case-hardened

and the notch was cut such that it resembled the root of a gear. Peened and un-peened

specimens were tested under cyclic bending with R-ratios of -1, 0.1, and 0.5.

Additionally, un-peened specimens were tested at an R-ratio of 0.8. By using a local

concept, crack initiation was determined to start partially at and below the surface. A

fracture mechanics approach was used to study the crack stop behavior as well as

determining the critical area for failure. A modified Haigh diagram was used to

characterize lifetime behavior. Results indicated no difference between mean stress and

residual stress sensitivity, which can be seen in Figure 2-27.

Vantiger et al. [47] studied the effects of high R-ratios (R = 0.8 and R = 0.9) on

notched specimens of 1045 steel that underwent three different heat treatments resulting

in Rockwell-C hardness (HRC) values of 10, 37, and 50. Results of the tension tests were

35
such that the ultimate strength of the notched specimens was higher than the smooth

specimens. Cyclic fatigue tests for the materials resulted in large scatter due to the flat

nature of the respective slopes, as shown in Figure 2-28. With respect to mean stress

equations, the Gerber and Modified Goodman equations could not be used due to the

bounding by tensile strength. Failure of the 10 and 37 HRC specimens was due to internal

mircovoid coalescence, not fatigue. Cyclic creep was also influential for those specimens,

but had insignificant influence for the 50 HRC specimens. The hardest specimens failed

by brittle fracture.

Pals and Stephens [48] studied how high R-ratios (R = 0.8 and R = 0.9) affected

the fatigue behavior of 1045 steel that underwent three different heat treatments resulting

in HRC values of 10, 37, and 50. Specimens were either smooth or contained a sharp or

mild notch. The stress concentration factor Kt was 1, 1.65, or 3.65. The results for failure

were similar to those in [47]. Fatigue life curves showed a very shallow slope indicating

that a slight change in stress results in a significant change in life. In terms of the notch

effect, fatigue resistance was generally higher than the smooth specimens. For the

specimen with a 50 HRC and Kt = 3.65, the notch was shown to greatly reduce the fatigue

behavior.

Fatemi et al. [49] studied fatigue behavior using notched specimens made from

QT and forged micro-alloyed steels. Double-notched plate geometries were used for

mean stress testing at R = 0 with stress concentration factors of 1.8 or 2.8. The Modified

Goodman equation was used to convert the test data to R = -1, results were reasonable

with the exception for specimen geometry where Kt = 1.8. Fatigue behavior was

underestimated by Modified Goodman which was believed to be because of a milder

36
notch and crack growth was a significant portion until failure. The difference between

experimental lives for both notched geometries was small. Notch effects were most

severe at long life, whereas at short life, localized plastic deformation and notch blunting

effects controlled fatigue life.

Susmel et al. [50] estimated the fatigue life of several metallic materials by

investigating the effect of notches and mean stress inherent in multiaxiality. Stage I was

considered the most important stage to be modelled correctly with the Modified Manson-

Coffin Curve Method (MMCCM), which considered elastic-plastic behavior. By

considering stage I and MMCCM, the crack initiation plane is always coincident to the

plane with the maximum shear strain. Morrow, Modified Morrow, and Smith-Watson-

Topper mean stress formulae were evaluated and then compared with the MMCCM. The

results between all four methods were similar. The method was also applied to notched

samples with various notch root radii under R-ratios of -1 and 0. Application of the

MMCCM was ideal for mean stress and notch effects.

Yu et al. [51] evaluated the effect of mean stress on center-notched 1045 steel

plates for R = 0. Notch effects were also compared to fully-reversed loading conditions. It

was found that the fatigue notch factor was higher for R = -1 compared to R = 0 and the

effect of the fatigue notch factor was most significant at long life compared to short life.

A computational method based on Nueber’s method for predicting the fatigue notch

factor, Kf, was used for specimens subjected to R = 0 loading because of significant

plasticity at the notch root. For sharp notches, predictions were not as accurate as for

blunt notches. For R = -1 however, predictions were accurate for sharp and blunt notches.

37
2.3.5 Mean Stress and Notch Effects Related to Cast Iron
This section discusses several papers where authors studied the effects of mean

stress on both smooth and/or notched specimens made from cast iron.

Fash and Socie [52] studied fatigue behavior and mean effects in grey cast iron

under constant amplitude load-controlled and strain-controlled tests. Cyclic stability was

not observed in either strain-controlled or load-controlled tests. In strain-control a

continual decrease in the tensile load was noticed, while the compression load remained

constant. Strain was monitored in load-control and it was observed that both compressive

and tensile strains increased with the tensile strain increasing at a faster rate compared to

the compressive strain. The authors also performed a variety of strain-controlled tests

with varying strain amplitude and mean strain. They were able to incorporate the Smith-

Watson-Topper (SWT) equation for all loading conditions which provided accurate

results in a life regime of one million or less cycles. Figure 2-29 shows the experimental

results with the SWT equation.

Lin and Pai [53] investigated low-cycle strain-controlled fatigue behavior of

austempered ductile iron subjected to various strain ratios. Material was austempered at

300 °C and 360 °C; both materials had shorter fatigue lives as the strain ratio increased

due to the presence of a mean tensile stress. When the testing conditions were fully-

reversed, a compressive mean stress developed and no stable hysteresis response was

observed, regardless of the strain ratio. Austempering at 300 °C provided a material with

a higher fatigue resistance compared to austempering at 360 °C, but the strengthening

effect was lost at R-ratios of 0 and 0.5 and the latter material had a better fatigue

performance. It was noticed that austenite in the matrix attributed to a lower fatigue

38
strength for low-cycle fatigue behavior. Life predictions were made using Morrow and

Smith-Watson-Topper (SWT) equations. Both parameters overestimated the fatigue life

for the iron austempered at 300 °C, but provided better predictions for mean stress effects

for the iron austempered at 360 °C. Each approach overestimated the fatigue life of the

ADIs austempered at 360 ºC because they cannot predict the sensitivity of mean stress for

these ADIs. Poor life predictions were also apparent for the ADIs austempered at 300 ºC.

This may be explained by the smaller amount of ductility compared to the ADIs

austempered at 360 ºC. A modified version of the SWT approach gave reasonable life

predictions with results shown in Figure 2-30 and the SWT equation given as:
−0.248
𝜎𝑚𝑎𝑥 𝜀𝑎 = 19.7(𝑁𝑓 ) (2.35)
Stokes et al. [54] studied fatigue crack initiation and propagation in austempered

ductile iron based on the effects of graphite nodules. Three-point bending fatigue tests

were performed with R = 0.1 and R = 0.5 with the fatigue life results shown in Figure 2-

31. Fractography showed that the dominating fatigue cracks initiated from pores at or

near the surface with further crack initiation happening ahead of the propagating crack

tip. The dominate crack advanced by the coalescing of micro-cracks, and the number of

micro-cracks increased with the applied stress level. From a fracture mechanics

viewpoint, it was found that the dominate crack passes through the greatest number of

nodules when ΔK > 20 MPa (m)1/2.

Korkmaz [55] successfully applied the Smith-Watson-Topper equation to account

for mean stress of GJS-600 cast iron. Collini et al. [56] studied how casting-defects

affected crack initiation and the fatigue limit of GJS-400 ductile cast iron. Non-standard

smooth specimens were used in order to maximize the volume of material subjected to

39
fatigue and maximized the surface-volume ratio to promote fatigue crack initiation from

an internal defect. A staircase method was applied with a global fatigue notch factor

Kf = 1.6 at R = 0. It was noticed that fatigue failures originated from internal defects such

as porosities and micro-shrinkages. Prediction techniques such as equivalent crack or

equivalent notch models developed for steels did not accurately predict fatigue behavior

and generally overestimate the fatigue limit of ductile iron. Figure 2-32 shows the results

of the different methods used suggesting that the nodules should be considered a

geometrical defect.

Fatigue properties of various ductile iron matrices alloyed with up to 0.11 wt% of

Ti were investigated under tension-tension loading at R = 0.1 [57]. The increase of

fatigue resistance was reported to be caused by an increase in material strength for ferritic

and pearlite microstructures. The fatigue limit as well as tensile strength increased with

nodularity. It was reported in [14, 58, 59] that no simple correlation between fatigue

strength and ultimate strength existed. However, Luo et al. [57] showed that the fatigue

limit can be correlated with the ultimate strength, because the ratio of fatigue limit to

tensile strength remained constant when graphite nodularity was high, see Table 2.7. It

was found that Ti tends to decrease the nodule count, but adverse effects on ferritic and

pearlite microstructures were not seen on S-N behavior with Ti content less than 0.1 wt%.

Fatigue failures tended to happen from surface defects at high stress levels, but from

internal defects at lower stress levels. The critical maximum stress level at which crack

initiation transitions from surface to subsurface was determined to be 0.76Sy. A bilinear

behavior was observed, as shown in Figure 2-33.

40
Table 2.1: Estimated fatigue limit (σ ′w) as predicted by Murakami’s modified area method compared to the actual fatigue limit
(σ wo) (where the fatigue limit was considered as fatigue strength at run-out for ≥ 107 cycles) at R = -1 for JIS
FCD400, JIS FCD500, and JIS FCD700 ductile iron [9].

Number ∗
√𝑎𝑟𝑒𝑎𝑅,𝑚𝑎𝑥 √𝑎𝑟𝑒𝑎𝐷,𝑚𝑎𝑥 √𝑎𝑟𝑒𝑎𝑒𝑓𝑓 𝜎𝑤′ 𝜎𝑤0 𝜎𝑤′ ⁄𝜎𝑤0
of
(μm) (μm) (MPa) (MPa)
specimens (μm)
n=1 538 493 1031 178 275 0.65
n = 10 664 1050 1714 145 275 0.53
n = 20 703 1548 2251 126 275 0.46
The experimental fatigue limit σw0 of smooth specimen was obtained from fatigue tests conducted using 8 specimens and
n is the number of specimens considered in a statistical analysis

Table 2.2: Experimental and predicted fatigue limits as predicted by Murakami’s area method for cast irons with different
microstructures including vacuum quenched (VQ) and subzero treatment under rotary bending loads [13].

Material Maximum casting Experimental Predicted fatigue 𝜎𝑤𝑝

Vickers hardness, HV
defect size, √𝑎𝑟𝑒𝑎𝑚𝑎𝑥 fatigue limit, σwe limit, σwp ⁄𝜎𝑤𝑒
(kgf/mm2)
(μm) (MPa) (MPa)
FM 470 218 350 339 0.969
HM 348 258 270 262 0.970
MFM-1 644 117 500 487 0.974
MFM-2 629 491 400 376 0.940
VQ 663 515 300 390 1.300
Subzero 762 302 100 480 4.800

41
Table 2.3: Non-constant fatigue ratios shown for different matrix microstructures of
an alloyed ductile iron where the fatigue limit is defined at 10 7 cycles at
R = -1 [15].

42
Table 2.4: (a) Various models used to estimate the fatigue limit (defined as fatigue
strength at 107 cycles) of nodular cast iron with microstructural defects at
R = -1 and (b) input parameters for models listed in (a) [18].

(a) Comparison of different approaches: evolution of the fatigue limit with defect size
Defect size
Author Relation Fatigue limit as a function of defect size
parameter
El Haddad (1) 𝜎𝐷 = 𝐾𝑡ℎ ⁄𝑄√𝜋(𝑙 − 𝑙0 ) 𝑙0 = 1⁄𝜋(𝐾𝑡ℎ ⁄𝑄𝜎𝐷0 )2 l

Tanaka (2) 𝜎𝐷 = 𝐾𝑐𝑚 ⁄√𝜋𝑏 + 2⁄𝜋(𝜎 ∗ 𝑎𝑟𝑐𝑐𝑜𝑠(𝑎⁄𝑏)) a

Lukas (3) 𝜎𝐷 = 𝜎𝐷0⁄𝐾𝑡 √1 + 4.5(𝑙1 ⁄𝜌) ρ

Mitchell (4) 𝜎𝐷 = 𝜎𝐷0⁄1 + (𝐾𝑡 − 1)⁄(1 + 𝐴𝑛 ⁄𝜌) ρ

1/6
Murakami (5) 𝜎𝐷 = 𝐴(𝐻𝑣 + 120)⁄√𝑎𝑟𝑒𝑎 √𝑎𝑟𝑒𝑎
de (6) 𝜎𝐷 = 𝜎𝐷0⁄1 + (𝑅𝑝0.2⁄𝐾𝐷𝐾 ) √𝑎 a
Kazinczy

(b) Input data for the different models tested

Nodular cast
Data Symbol Cast steel Unit
iron
Grain size D 50 50 μm
Effective threshold ΔKeffth 3.75 3 MPa m1/2
Yield strength Rp0.2 380 590 MPa
Ultimate tensile strength Rm 510 800 MPa
Hardness Hv 170 270
Defect free fatigue limit σD0 300 320 MPa
Crack shape factor α 2/Π 2/Π
Stress concentration factor Kt 2 2
(surface half-spherical void)
Friction stress σ* 114 114 MPa
Microscopic stress intensity Kcm 1.98 1.98 MPa m1/2
factor
Material constant An 250 250
Material constant A 1.43 1.43
Slip band zone size (Wo = d/2) W0 25 25 μm
Material constant KDK 36 36
Type of defect Artificial Microshrinkage
hole

43
Table 2.5: Fatigue limit, tensile strength, and fatigue ratio for R = 0.1 of spheroidal
graphite cast iron with different heat treatments at room temperature, 300
°C, and 500 °C [21].

Material Fatigue limits (MPa) Tensile Strengths (MPa) σf/Su

RT 300 ºC 500 ºC RT 300 ºC 500 ºC RT 300 ºC 500 ºC
As-cast 114 95 85 503 452 360 0.227 0.210 0.236
Annealed 110 91 77 462 396 288 0.238 0.230 0.267
Normalized 123 100 90 798 718 540 0.154 0.139 0.167
Bainitized 123 103 92 905 834 680 0.136 0.124 0.135

Table 2.6: Values of aM and bM adapted from the FKM guideline [40].

Values of aM and bM for various materials.

Material aM bM
Steel 0.35 -0.10
Steel Casting 0.35 0.05
Ductile Irons 0.35 0.08
Malleable Cast Iron 0.35 0.13
Grey Cast Iron 0.00 0.50
Wrought Aluminum Alloys 1.00 -0.04
Cast Aluminum Alloys 1.00 0.20
Titanium 0.45 -0.04
Mσ = aM10-3 Su + bM

44
Table 2.7: Material properties of the ductile irons containing Ti showing a non-constant fatigue ratio between material
microstructures for R = 0.1 where the fatigue limit is defined as the fatigue strength at 2 x 10 7 cycles [57].

Structure
Ti Ferrite Count (V + VI) σ0.2 σb σf
-2
Iron Material (Wt %) (%) (mm ) (%) (MPa) (MPa) (MPa) σf/σ0.2 σf/σb
F1 ferritic ductile iron 0.006 97.3 337 85.0 318 448 333 1.05 0.74
F2 0.006 97.3 337 85.0 321 448 333 1.04 0.74
F3 0.035 97.8 280 89.4 322 453 329 1.02 0.73
F4 0.063 98.1 253 83.0 338 463 338 1.00 0.73
F5 0.100 96.6 275 77.5 328 456 325 0.99 0.71
F6 0.100 96.8 224 82.5 327 455 342 1.04 0.75
F7 0.085 90.4 156 71.7 330 468 338 1.02 0.72
F8 93.4 293 90.1 336 467 343 1.02 0.73
P1 pearlitic ductile iron 0.006 18.5 295 90.8 465 781 428 0.92 0.55
P2 0.063 20.0 225 87.4 462 768 430 0.93 0.56
P3 0.100 15.9 210 86.0 465 779 433 0.93 0.56
A1 ADI 0.005 NA 240 90.1 788 1061 555 0.70 0.52
A2 0.064 NA 195 84.9 753 984 535 0.71 0.54
A3 0.110 NA 168 71.4 705 926 477 0.68 0.52
σ0.2: 0.2% offset yield strength, σb: ultimate tensile strength, σf: fatigue limit
V, VI: Type of graphite as defined by French standard NF EN ISO 945

45
 320
Nf, conv.

σa [MPa]
Nf, calc. (εa,p)
Nf, calc. (ΔR)
Nf, calc. (ΔT)

240

200

160
103 104 105 Nf 106

Figure 2-1: Conventionally determined fatigue lives compared to S-N curves based on
plastic strain amplitude, change in temperature and change in electrical
resistance of the specimen for cast iron EN-GJV-400 at R = -1 [5].

Figure 2-2: Fatigue life scatter of pearlitic gray cast iron due to differences in
microstructure from different foundries tested at R = -1 [12].

46
Figure 2-3: Fatigue limit under rotary bending testing conditions of austempered ductile
irons as a function of casting block leg thickness where the fatigue limit is
defined as the fatigue strength at 107 cycles [14].

Figure 2-4: Relationship of the rotating bending fatigue limit under rotary bending
conditions with impact toughness and nodule diameter of austempered ductile
iron where the fatigue limit is defined as the fatigue strength at 107 cycles
[14].

47
Figure 2-5: Fatigue limit as a function of maximum and mean nodule diameters for
spheroidal graphite cast iron under fully-reversed bending where the fatigue
limit is defined as the fatigue strength at 107 cycles [16].

Figure 2-6: Fully-reversed bending fatigue limit as a function of tensile strength for
austempered ductile iron compared to steel where the fatigue limit is defined
as the fatigue strength at 107 cycles [16].

48
Figure 2-7: Comparison of predicted fatigue limit for nodular cast iron based on several
methods with experimental fatigue limit defined as the fatigue strength at
106 cycles obtained from 3-point bending tests at R = 0.3 [20].

Figure 2-8: Ultimate tensile strength of cast iron containing spheroidal graphite with
various heat treatments at different temperatures [21].

49
 1000

Sa=889.6(Nf)-0.0409

Sa=927.1(Nf)-0.0745
Sa, MPa

Sa=1187.9(Nf)-0.1294
Surface rolled notch
Smooth
Notched
100
104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07 108
1.E+08
Nf (Number of cycles for failure)

Figure 2-9: Comparisons of S-N behavior of smooth, notched, and surface rolled notched
specimens for 100-70-03 ductile cast iron under rotating bending conditions
with the fatigue limit defined at 4 x 106 cycles [22].

200
Stress amplitude σa (MPa)

180

160

140

120

100
103 104 105 106 107
Cycles to Failure (Nf)

Figure 2-10: Comparisons of plae bending fatigue test results of as-cast, shot blast, and
fine ground specimens of nodular cast iron at R = 0 where the fatigue limit
was defined at 2 x 106 cycles [23].

50
Figure 2-11: Effects of Almen intensity on smooth rotating bending fatigue life of Cu-Ni
austempered ductile iron where the fatigue limit was defined at 107 cycles
[24].

Figure 2-12: Two-step diagram for ductile cast iron under rotating bending depicting the
transition region between surface fracture and internal fracture investigated
for shot-peened conditions [25].

51
 1800
Avg. Nodular

Ultimate Tensile Strength, Su ( MPa)

1600 Size Ranges (μm)
20-30
1400 30-40
40-50
50-60
1200 60-70

1000

800

600
Su = 3.24Hv - 79.40
400
150 200 250 300 350 400 450 500
Vickers Hardness, Hv

Figure 2-13: Tensile strength vs. Vickers hardness with cast iron data obtained from
literature grouped by average nodule size showing a linear relationship.

1400
Avg. Nodular
Size Ranges (μm)
1200
20-30
Yield Strength, Sy (MPa)

30-40
1000 40-50
50-60
800 60-70

600

400
Sy = 3.56Hv - 382
200
150 200 250 300 350 400 450 500
Vickers Hardness, Hv

Figure 2-14: Yield strength vs. Vickers hardness with cast iron data obtained from
literature grouped by average nodule size showing a linear relationship.

52
 450
Avg. Nodular
400 Size Ranges (μm)
Fatigue Limit, Sf (MPa) 20-30
350 30-40
40-50
300 50-60
60-70
250

200

150

100
150 250 350 450 550
Vickers Hardness, Hv

Figure 2-15: Fatigue limit defined as fatigue strength at 107 cycles vs. Vickers hardness
grouped by average nodule size showing no correlation.

425
Avg. Nodular
375 Size Ranges (μm)
Fatigue Limit, Sf (MPa)

20-30
30-40
325 40-50
50-60
275 60-70

225

175

125
450 650 850 1050 1250 1450 1650
Ultimate Tensile Strength, Su (MPa)

Figure 2-16: Fatigue limit defined as fatigue strength at 107 cycles vs. tensile strength
with data from literature grouped by average nodule size showing no
correlation.

53
 600
Max Nodular
Size Ranges (μm)
Fatigue Limit, Sf (MPa) 500 10-35
35-90
100-120
400
120-200

300

200

100
150 250 350 450 550 650
Vickers Hardness, Hv

Figure 2-17: Fatigue limit defined as the fatigue strength between 106 and 107 cycles vs.
hardness with cast iron data obtained from literature grouped by maximum
nodule size.

600
Max Nodular
550 Size Ranges (μm)
10-35
Fatigue Limit, Sf (MPa)

500
35-90
450
100-120
400 120-200
350
300
250
200
150
100
400 600 800 1000 1200 1400
Ultimate Tensile Strength, Su (MPa)

Figure 2-18: Fatigue limit defined as the fatigue strength between 106 and 107 cycles vs.
tensile strength with cast iron data obtained from literature grouped by
maximum nodule size.

54
 1000

Experimental Fatigue Limit, Sf (MPa)

Sf = 285.28 (I0.31d-0.4) 0.7131
R² = 0.535

100
0.1 1 10
I0.31d-0.4

Figure 2-19: Correlation of fatigue limit data from literature with the parameter used in
[14] based on impact strength and mean nodule size (also see Figure 2-4).

55
 550

Predicted Fatigue Limit, Sf (MPa)

500

450

400

350

300

250

200 Axial
Rotating Bending
150
150 250 350 450 550
Experimental Fatigue Limit, Sf (MPa)

550

500
Predicted Fatigue Limit, Sf (MPa)

450

400

350

300

250
Axial
200
Rotating Bending
150
150 250 350 450 550
Experimental Fatigue Limit, Sf (MPa)

Figure 2-20: Fatigue limit of ductile cast iron predicted by Murakami’s method using
maximum nodule size vs. experimental fatigue limit (defined as fatigue
strength at 106 to 107 cycles) with cast iron data from literature and
assuming (a) external defects, (b) internal defects.

56
 400
𝜎0
𝑆𝑓 =
𝑆𝑦
Predicted Fatigue Limit, Sf (MPa) 350 1+ 𝐾 𝑑

300

250

200

150
150 200 250 300 350 400
Experimental Fatigue limit, Sf (MPa)

Figure 2-21: Fatigue limit as predicted by de Kazinczy’s method vs. experimental fatigue
limit with cast iron data from literature assuming defects as pores.

57
 500 500
450 450
Predicted Fatigule Limit,

Predicted Fatigue Limit,

400 400
350 350

Sf (MPa)
Sf (MPa)

300 300
250 250
200 𝜎0 200
𝑆𝑓 =
𝐾 −1
150 1+ 𝑡 150
1+𝐴𝑛 /r
100 100
100 200 300 400 500 100 200 300 400 500
(a) Experimental Fatigue Limit, Sf (MPa) (b) Experimental Fatigue Limit, Sf (MPa)
500 500

Predicted Faituge Limit,

450 450
Predicted Fatigue Limit, Sf

400 400
350 350

Sf (MPa)
300 300
(MPa)

250 250
200 200
150 150
100 100
100 200 300 400 500 100 200 300 400 500
(c) Experimental Fatigue Limit, Sf (MPa) (d) Experimental Fatigue Limit, Sf (MPa)

Figure 2-22: Fatigue limit as predicted by Mitchell’s method with average nodule size vs. experimental fatigue limit with cast iron
data obtained from literature assuming (a) Kt = 2, and (b) Kt = 3. Reapplication of Mitchell’s method using maximum
nodule size assuming (c) Kt = 2, and (d) Kt = 3.

58
Figure 2-23: Comparison of Goodman and Gerber mean stress predictions for mean
stress fatigue data of 316 LN stainless steel in air and mercury
environments [37].

Figure 2-24: Comparison of experimental data with various mean stress models for a
medium carbon steel [34].

59
 σa
R = -1
R = -∞
R=0
II
Mσ = 0
- Mσ
σar -Mσ/3 III
R>1 R = 0.5
1
I
Mσ = 0
IV

- + σm
0

Figure 2-25: Divisions of the Haigh diagram for use with the FKM mean stress method
[31, 40].

Figure 2-26: Comparison of the FKM mean stress parameter with several other mean
stress parameters for AISI 4340 steel [31].

60
Figure 2-27: Stress amplitude vs. combination of mean stress and residual stress for shot
peened and unpeened notched 16MnCr5 steel specimens where the fatigue
limit was defined as the fatigue strength at 107 cycles [46].

Figure 2-28: S-N diagrams for high R-ratios of notched (Kt = 1.65) 1045 steel with
Rockwell C hardness of 10, 37 and 50 with the fatigue limit defined as the
fatigue strength at 5 x 106 cycles [47].

61
 1
Strain-control tests
Load-control tests

∆𝜀
2
10-1
𝜎𝑚𝑎𝑥

∆𝜀 −0.25
𝜎𝑚𝑎𝑥 = 1.82(𝑁𝑓 )
2

10-2
101 102 103 104 105 106

Figure 2-29: Correlation of fatigue test results using the Smith-Watson-Topper mean
stress model to account for the development of mean stress due to the
differences between tensile and compressive strengths of grey cast iron
under strain and load-controlled testing [52].

Figure 2-30: Correlation of fatigue test results of austempered ductile irons with
different heat treatments and R-ratios based on the modified Smith-
Watson-Topper parameter as proposed by Fash and Socie [53].

62
Figure 2-31: Maximum stress vs. number of cycles to failure for different R-ratios and
heat-treatment conditions for austempered ductile irons where the fatigue
limit is defined as the fatigue strength at 106 cycles [54].

Figure 2-32: Fatigue limit (defined as the fatigue strength at 2 x 106 cycles) correlation
of ductile cast iron at R = 0 with a geometrical defect parameter as
compared to notch or crack equivalence methods [56].

63
Figure 2-33: The stress transition level at which failure stems from surface to subsurface
for ductile cast iron at 0.76Sy for (a) iron A1, (b) iron A2, and (c) iron A3
(see Table 2.7) at R = 0.1 where the fatigue limit is defined as the fatigue
strength at 2 x 107 cycles [57].

64
Chapter 3

Experimental Program

This chapter discusses the experimental procedures used to generate fatigue data

in this study. A description of the material as provided by Eaton is included along with

specimen geometry. Discussion of the testing procedures, types of tests performed, along

with the motivation behind performing those particular tests follow. Analyses of all

experimental results are presented and discussed in chapter 4.

3.1 Material Description

The experimental program was performed on both smooth and notched specimens

made of 120-90-02 ductile cast iron machined and provided by Eaton Corporation.

Specimens were heat treated in an IPSEN furnace with an atmospheric vacuum

maintained at a level of 50 to 400 microns of atmospheric pressure throughout the heating

cycles. The baking temperature was between 650 °F to 950 °F for 2 to 3 hours. Heat

treatment was performed by a 3rd party designated by Eaton.

Microstructural analysis, certified by Dura-Bar, lists the average chemical

makeup of the material as C:3.660%, Si:2.540%, Mn: 0.250%, P:0.023%, S:0.012%, and

residual amounts of Ni, Cr, Cu, Mo, and Sn. The hardness of the material ranged from

65
215 BHN to 237 BHN. The microstructure consisted of 69% Pearlite and 31% Ferrite at

the edge of the casting piece, and 79% Pearlite and 21% Ferrite at the center of the

casting piece. Nodule count was determined to be 100 nodules per mm2 with a nodule

percent of 99%. The nodule size falls in the class range of 4 (0.16 mm) to 6 (0.04 mm)

[60]. The material conforms to ASTM A-536 standard [61].

3.2 Specimen Description

Smooth and notched specimens were provided for fatigue testing, and separate

specimens were provided for tension tests by Eaton with specimen geometries shown in

Figure 3-1. The smooth specimens were polished on a commercial round-specimen

polishing machine with three different grades of aluminum oxide lapping film. The

grades, coarsest to finest, were 30μ, 9μ, and 3μ. The 3μ grit film was used as the final

polish, and polishing marks coincided with the longitudinal direction of the specimen.

Specimens were then carefully inspected to ensure machining marks were removed.

Finite Element Analysis was performed on the notched geometry using Abaqus.

The specimen was meshed with a quadratic tetrahedron element of type C3D10, which

contained 10 nodes per element. A very fine mesh was used at notch root radius to

increase the accuracy and homogeneity of the results, see Figure 3-2. There were 37,870

elements in the notch where the largest length of an element was 0.12 μm and the shortest

length was 0.05 μm. Results of the local notch stress to nominal stress indicate that the

elastic stress concentration factor is Kt = 3.

66
3.3 Testing Equipment
An MTS closed-loop servo-controlled hydraulic axial load frame, shown in

Figure 3-3(a), in conjunction with an Instron Fast-Track digital servo-controller was used

to conduct the tests. The load cell used had a capacity of 100 kN. Hydraulically operated

grips were employed to secure the specimens' ends in series with the load cell, as shown

in Figure 3-3(b).

For the tension and incremental step tests, which were performed under strain-

control, total strain was controlled using extensometers rated as ASTM class B2 [62].

The calibration of the extensometers was verified using a displacement apparatus

containing a micrometer barrel with divisions of 0.000254 mm. Extensometers with gage

lengths of 7.62 mm for fatigue specimens and 12.7 mm for tensile specimens were used.

Both extensometers are capable of measuring strains up to 15 %.

In order to protect the specimens' surface from the knife-edges of the

extensometer, ASTM Standard E606 [63] recommends the use of transparent tape or

epoxy to 'cushion' the attachment. For this study, transparent tape strips were used.

Significant effort was put forth to align the load train (load cell, grips, specimen,

and actuator). Misalignment can result from both tilt and offset between the central lines

of the load train components. In order to align the machine, a round strain-gage bar with

two arrays of four strain gages per array were arranged at the upper and lower ends of the

uniform gage section was used. This was done in accordance with ASTM Standard

E1012 [64]. The alignment of the machine produced acceptable results in reference to the

allowable bending strain in the axial direction of 5% or less, which is in compliance with

ASTM Standard E606 [63].

67
3.4 Test Methods and Procedures

3.4.1 Monotonic Tension and Compression Tests

All monotonic tension tests in this study were performed using test methods

specified by ASTM Standard E8 [65]. Two specimens were used to obtain the monotonic

tensile properties.

INSTRON Blue-hill software was used for the monotonic tension tests. A strain

rate of 0.006 mm/mm/min was used for the strain-controlled tests. Attempts were made

to protect the extensometer by removing the device after ultimate strength was reached

and then switched to displacement control. Two specimens were tested following the

above procedure, one specimen failed in strain control and the other failed in

displacement control.

One compression test was performed using a fatigue specimen. The specimen was

subjected to a compressive strain rate of 0.006 mm/mm/min, and the test was stopped

shortly past yielding to prevent buckling. The purpose of this test was to obtain

appropriate compressive stress levels for fatigue testing under compressive mean stress.

3.4.2 Incremental Step Test

One incremental step test was performed on a fatigue specimen in order to obtain

the cyclic stress-strain relationship, which in turn facilitates obtaining the strain-life

properties. An extensometer with a 7.62 mm gage length was used to measure and record

strain values. The step test was performed under load-control with stress levels ranging

from 300 MPa to 700 MPa. At each stress level, sufficient cycles were applied to reach

cyclic deformation stability.

68
3.4.3 Constant Amplitude Smooth and Notched Fatigue Tests

All constant amplitude, load-controlled fatigue tests in this study were performed

according to ASTM Standard E606 [63], where appropriate. It is recommended by this

standard that at least ten specimens be used to generate the fatigue properties. For this

study, 24 specimens at five different stress amplitudes ranging from 300 MPa to 600 MPa

were utilized for smooth specimens and 26 specimens at six different stress amplitudes

ranging from 115 MPa to 400 MPa were used for notched specimens at R = -1. Failure

was defined as fracture of the specimen. Based on the small degree of scatter, one or two

specimens were tested at several stress amplitude levels with R-ratios of -∞,-7, -3, -1, 0,

1/3, 0.5, and 0.75 to evaluate mean stress effect in tension and compression for both

smooth and notched specimens. A modified load history with some tension cycles was

necessary for R = -∞ testing to allow for crack growth. One block of load history for

R = -∞ can be seen in Figure 3-4.

INSTRON SAX software was used in all tests. Testing frequencies ranged from

0.5 Hz to 30 Hz, depending on the stress amplitude level. A triangular waveform was

used for all tests below 15 Hz. Higher frequency tests were performed with a sinusoidal

waveform.

69
 (a)

(b)

(c)

Figure 3-1: Geometry of the (a) smooth and, (b) notched axial fatigue specimens, and (c)
monotonic tension specimen. Dimensions are in mm. Drawings based on
Eaton Corporation’s specifications.

70
Figure 3-2: Finite element mesh of the notch.

71
 (a) (b)

Figure 3-3: (a) MTS load frame with Instron controller (not shown) used for testing and
(b) close up of specimen in gripped position.

400

200 1000 Cycles

-200
100 Cycles
Sa (MPa)

-400

-600

-800

-1000

-1200
0 2 4 6 8 10
Time (s)

Figure 3-4: Load block used for testing at R = -∞.

72
Chapter 4

Experimental Results and Discussion

4.1 Monotonic Tension and Compression Test Results

The following properties were determined from two monotonic tension tests:

modulus of elasticity, E, yield strength, Sy, ultimate tensile strength, Su, strength

coefficient, K, strain-hardening exponent, n, percent elongation, %EL, true fracture strain

εf, and true fracture stress, σf. The yield strength was determined from the 0.2% offset

method. True stress, , true strain, , and true plastic strain, p, were calculated from

engineering stress, S, and engineering strain, e, according to the following relationships,

which are based on a constant volume assumption:

𝜎 = 𝑆(1 + 𝑒)
(4.1)
𝜀 = ln(1 + 𝑒)
(4.2)
𝜎
𝜀𝑝 = 𝜀 − 𝜀𝑒 = 𝜀 −
𝐸 (4.3)

The true stress - true strain curve is often represented by the Ramberg-Osgood equation:
1
𝜎 𝜎 𝑛
𝜀 = 𝜀𝑒 + 𝜀𝑝 = + ( ) (4.4)
𝐸 𝐾

This study utilized two monotonic specimens for tension tests and one fatigue

specimen for a compression test. Table 4.1 shows the results of the tension and

73
compression tests. The yield strength in tension is lower than that in compression because

the stress concentration effects of the nodules are reduced. It was also observed that the

stiffness is lower in compression than tension. Superimposed results of the tension and

compression tests are shown in Figure 4-1.

The strength coefficient, K, and strain-hardening exponent, n, are the intercept

and slope of the best-fit line to true stress versus true plastic strain data in log-log scale:

𝜎 = 𝐾(𝜀𝑝 )𝑛 (4.5)

In accordance with ASTM Standard E739 [66], when performing the least squares fit, the

true plastic strain was the independent variable and the stress was the dependent variable.

The best-fit lines for the two tests conducted are shown in Figure 4-2. To generate the K

and n values, the range of data used in this figure was chosen according to ASTM

Standard E646 [67]. Therefore, the valid data range occurred between the yield point and

the strain at maximum load. Stress and strain data at equal intervals of 0.5% engineering

strain were used to determine the K and n values.

4.2 Analysis of the Incremental Step Test Results

Data were obtained from one constant amplitude, load-controlled, incremental

step test to determine cyclic-induced change in the deformation resistance of the material.

The data from the incremental step test are shown in Table 4.2. Change in the strain

amplitude and the stable strain amplitudes can be seen in Figure 4-3. Cyclic stability was

reached within the first few cycles at low stress-amplitude levels. At higher stress levels

where plastic deformation was present, cyclic stabilization was not immediately reached.

The properties determined from the steady-state hysteresis loops were the cyclic

strength coefficient, K', and cyclic strain hardening exponent, n'. These properties are

74
used to develop the cyclic stress-strain curve. Hysteresis loops and data, taken from the

last cycle before increasing the stress level, were used to obtain the stable cyclic

properties. A composite plot of the steady state hysteresis loops can be observed from

Figure 4-4.

Similar to monotonic behavior, the cyclic true stress-strain behavior can be

characterized by the Ramberg-Osgood type equation:

1
∆𝜀 ∆𝜀𝑒 ∆𝜀𝑝 ∆𝜎 ∆𝜎 𝑛′
= + = ′ + ( ′) (4.6)
2 2 2 2𝐸 2𝐾

Typically, in equation 4.6, the monotonic value of E is used because for most engineering

materials, E is similar in both tension and compression. For ductile iron, the difference

between the tensile and compressive monotonic modulus values is significant. It is found

that using the cyclic modulus, E′ = 162 GPa, fits the data better and this can be attributed

to E′ being an 'average' between the monotonic tension and compression modulus values.

The cyclic strength coefficient, K', and cyclic strain hardening exponent, n', are

the intercept and slope of the best-fit line to stress amplitude versus plastic strain

amplitude data in log-log scale:

′
∆𝜎 ∆𝜀𝑝 𝑛
= 𝐾′ ( ) (4.7)
2 2

In accordance with ASTM Standard E739 [66], when performing the least squares fit, the

plastic strain amplitude was the independent variable and the stress amplitude was the

dependent variable. The plastic strain amplitude was calculated by the following

equation:

∆𝜀𝑝 ∆𝜀 ∆𝜎
= − (4.8)
2 2 2𝐸 ′

75
The plot is shown in Figure 4-5. Because K′ and n′ are used to characterize stress during

plastic deformation, the range of data used to generate these values requires applied stress

levels with plastic strain. Therefore K′ and n′ were determined from data where the

calculated plastic strain amplitude was greater than or equal to 0.00020 mm/mm.

The cyclic stress-strain curve reflects the resistance of a material to cyclic

deformation and can be vastly different from the monotonic stress-strain curve. In Figure

4-6, superimposed plots of monotonic and cyclic curves are shown. As seen in this figure,

120-90-02 ductile cast iron slightly cyclically softens.

4.3 Fully-Reversed Fatigue Test Results of Smooth and Notched

Specimens
Constant amplitude load-controlled fatigue tests were performed to determine the

stress-life curve for both smooth and notched specimens. The following equation relates

the stress amplitude to the fatigue life:

𝑆𝑎 = 𝐴(𝑁𝑓 )𝐵 (4.9)

where  is the coefficient of the S-N line, B is the exponent of the S-N line, and Nf is the

number of cycles to failure. The coefficient, , and the exponent, B, of the S-N line are

the intercept and slope of the best-fit line through the fatigue data in log-log scale,

respectively. In accordance with ASTM Standard E739 [66], when performing the least

squares fit, the stress amplitude, Sa, was the independent variable and the cycles to

failure, Nf, was the dependent variable. To generate the  and B values, the range of data

used were chosen for stress levels of 400 MPa and greater for smooth specimens and 125

MPa and greater for notched specimens, as these levels did not contain any run-out tests.

The fatigue limit was defined at 106 cycles. In this study 24 smooth and 26 notched

76
specimens were used to develop fully-reversed (R = -1) stress-life curves. Tables 4.3 and

4.4 contain individual test results of the fully-reversed, load-controlled fatigue data for

smooth and notched specimens, respectively.

Results of the fully-reversed fatigue tests, for both smooth and notched data and

best-fit lines are shown in Figure 4-7 The effect of the notch is the most detrimental at

long life compared to short life. The experimental fatigue notch factor determined at 106

cycles results in Kf = 2.44. Equations 4.10 and 4.11 are the best-fit S-N line equations for

smooth and notched data, respectively:

−0.120
𝑆𝑎 = 1608(𝑁𝑓 ) for smooth specimens (4.10)
−0.204
𝑆𝑎 = 2129(𝑁𝑓 ) for notched specimens (4.11)

4.4 Prediction of Fully-Reversed Notched Fatigue Behavior

When it is not possible to generate a notched S-N line from experimental data, it is

often convenient to estimate notched behavior from smooth specimen behavior. Two

methods of predicting notched fatigue behavior from smooth behavior are based on

nominal stress or local stress and strain. Peterson’s method and Neuber’s method are

examined for the nominal stress-based approach. The nominal-based approach involves

reducing the fatigue strength by the fatigue notch factor Kf. Methods developed by

Peterson and Neuber can be used to predict Kf with a material constant and the elastic

stress concentration factor Kt. The elastic stress concentration factor may be found from

engineering handbooks or FEA analysis.

There are three common strain-based approaches which can be used to predict

local stresses and strains from nominal stress and the geometry of the notch. These

77
include the linear rule, Neuber rule, and the Glinka or the Strain Energy Density (SED)

rule. Finite element analysis was also performed to obtain the local stress and strain

values at the notch root. Results from FEA were also used to determine the stress state

under cyclic loading. Local stress and strain values obtained from FEA were then

compared with predictions from Neuber and SED rules to determine which local rule best

predicts the notched behavior.

4.4.1 Finite Element Analysis of the Notched Geometry

Before notched fatigue life prediction can be performed, knowledge of the notch

stress concentration and deformation is required. Both nominal and local based

approaches require knowledge of Kt, which is easily obtained from FEA. Elastic-plastic

analysis was performed. Input requirements for the elastic region were E′ and ν.

According to the “Ductile Iron Society”, Poisson’s ratio for ductile iron is commonly

accepted as 0.275. The cyclic Ramberg-Osgood stress-strain curve was used as the input

for material behavior in the plastic region. The applied stress levels for FEA analysis

were the same levels used during experimental testing. Data were obtained from the

midpoint at the edge of the notch (one node) which corresponded to the maximum stress

location.

The elastic stress concentration factor is the ratio of the local, uniaxial stress in

the loading direction to the nominal stress. Determination of Kt is obtained when both

local and nominal stresses are elastic. Results from FEA indicate that Kt equals 3.

The stress state at the surface in the presence of a notch may be either a plane

stress or a plane strain condition. Which condition exists can be evaluated by considering

the ratio of the second principal stress or strain to the first principal stress or strain. If the

78
strain ratio, α, is equal to Poisson’s ratio and the stress ratio β is equal to 0 then a plane

stress condition exists. If α is equal to zero, a plane strain condition exists [68]. Table 4.5

shows the first and second principal stresses and strains from FEA and includes

calculations of α and β. Based on Table 4.5, a plane strain condition exists at the notch

root. In order to account for the multiaxial stress state developed in plane strain, the von

Mises criterion was used to obtain the equivalent stresses and strains, as can be observed

in Figure 4-8.

4.4.2 Nominal Stress-Based Approach

The nominal stress approach uses the fatigue notch factor, Kf, to account for the

notch effect. There are two common methods for calculating Kf, Neuber’s method and

Peterson’s method. Neuber’s method requires the knowledge of Kt, the notch root radius,

r, and the characteristic length, ρ. For fully-reversed loading, Kf values according to

Neuber’s approach can be estimated as follows:

𝐾𝑡 − 1
𝐾𝑓 = 1 +
1 + √𝜌⁄𝑟 (4.12)

The characteristic length used in Neuber’s equation can be obtained from [69], which was

developed for steels. Because material properties of ductile cast iron and steel are similar,

the characteristic length developed for steels was used. Based on the tensile strength of

120-90-02 cast iron, ρ was found to be 0.034 mm, resulting in Kf = 2.54, which is similar

to the experimental fatigue notch factor.

Peterson developed a similar empirical approach which defines Kf as follows:

𝐾𝑡 − 1
𝐾𝑓 = 1 +
1 + 𝑎⁄𝑟 (4.13)

where a (developed for steels) is the material characteristic length which is defined as:

79
 2070 1.8
𝑎 = 0.0254 ( ) with Su in MPa and a in mm
𝑆𝑢 (4.14)

Through substitution of a = 0.093 mm into Peterson’s equation, Kf was calculated to be

2.61.

Both Peterson’s and Neuber’s methods are reasonable (within a factor of 3 of

fatigue life) and conservative in predicting fatigue behavior. To estimate the notched

fatigue behavior, fatigue strength at one cycle was considered the same for both smooth

and notched data. The smooth fatigue strength at 106 cycles was reduced by Kf = 2.54. A

straight line was drawn connecting the point at one cycle to the reduced fatigue strength

at 106 cycles, after which a horizontal line was drawn as fatigue limit. Figure 4-9 shows

the smooth and notched data with the corresponding fits and Neuber’s notched behavior

prediction. The predicted notched S-N line is given as follows:

−0.187 (4.15)
𝑆𝑎 = 1608(𝑁𝑓 )

Another parameter to consider is the notch sensitivity factor, q. This quantifies

how sensitive a material is to a notch. When q = 0, there is no notch sensitivity and when

q = 1 there is full notch sensitivity. The value of q is calculated from:

𝐾𝑓 − 1 (4.16)
𝑞=
𝐾𝑡 + 1
Using equation 4.16, the notch sensitivity of the material is 0.72. Therefore, 120-90-02

ductile iron is fairly notch sensitive, as expected.

4.4.3 Local-Based Approaches

The local approaches involve obtaining the local stress and strain based on

nominal stress input and the geometry of the notch. In order to estimate the notched

fatigue behavior from smooth behavior using the local strain approach, the strain-life

80
curve and properties need to be generated. Because the experimental work performed in

this study was in load-control, the stress-life data were converted to strain-life data by

applying the cyclic stress-strain relation obtained from the incremental step test of the

smooth specimen under fully-reversed loading, as shown in Figure 4-8. The following

equation relates the total strain amplitude to the fatigue life:

∆𝜀 ∆𝜀𝑒 ∆𝜀𝑝 𝜎𝑓′ 𝑏 𝑐

= + = ′ (2𝑁𝑓 ) + 𝜀𝑓′ (2𝑁𝑓 ) (4.17)
2 2 2 𝐸

where σf′ is the fatigue strength coefficient, b is the fatigue strength exponent, εf' is the

fatigue ductility coefficient, c is the fatigue ductility exponent, and 2Nf is the number of

reversals to failure. Typically, equation 4.17 uses monotonic E; however, E′ is used

because of the asymmetry between the tensile and compressive monotonic E values.

The fatigue strength coefficient, σf′, and fatigue strength exponent, b, are the

intercept and slope of the best-fit line to stress amplitude versus reversals to failure data

in log-log scale:

𝑏
𝑆𝑎 = 𝜎𝑓′ (2𝑁𝑓 ) (4.18)
This is the same as equation 4.9, except with 2Nf rather than Nf and σf′ = 2-bA and b = B.

The fatigue ductility coefficient, εf', and fatigue ductility exponent, c, are the

intercept and slope of the best-fit line to calculated plastic strain amplitude versus

reversals to failure data in log-log scale:

∆𝜀𝑝 𝑐
( ) = 𝜀𝑓′ (2𝑁𝑓 )
2 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 (4.19)

In accordance with ASTM Standard E739 [66], when performing the least squares fit, the

calculated plastic strain amplitude was the independent variable and the reversals to

failure was the dependent variable. The calculated plastic strain amplitude was

81
determined from equation 4.8. The data range used to generate εf' and c values was

chosen such that the calculated plastic strain was greater than or equal to 0.00020

mm/mm.

Figure 4-10 displays the strain amplitude versus reversals to failure, where the

strain-life curve, the elastic strain line, the plastic strain line, and superimposed fatigue

data are shown. Table 4.6 shows the strain-life fatigue and cyclic deformation properties,

including the results from the compatibility equations for K′ and n′ which are as follows:

′
𝜎𝑓′
𝐾 = 𝑏
(𝜀𝑓′ ) 𝑐 (4.20)

𝑏
𝑛′ =
𝑐 (4.21)

Results based on compatibility equations are acceptable as the calculated percent

difference is less than 5% for both K′ and n′. The elastic and plastic strain-life behavior is,

therefore, well represented by log-log fits [4]. The total strain-life curve is given by:

∆𝜀 1748 −0.12 −1.06 (4.22)

= (2𝑁𝑓 ) + 14.2(2𝑁𝑓 )
2 162000

The local strain-based method assumes that if the local strain at the notch in the

notched specimen is applied to a smooth specimen, the resulting fatigue life will be the

same for both specimens. Linear rule, Neuber’s rule and the SED rule are used to relate

notch root strain to nominal stress. The linear rule represents an extreme case of plane

strain and it is defined as:

∆𝑆
∆𝜀 = 𝐾𝑡 ∆𝑒 = 𝐾𝑡 (4.23)
𝐸′

The second equality in this equation is based on the assumption that the nominal loading

is elastic. In this case, the nominal strain range, Δe, can be replaced by ΔS/E′. The linear

82
rule has been found to be in agreement with measurements taken for plane strain

conditions, such as a circumferentially grooved shaft in tension. Reported from literature

in [68], estimations of strain ranges from the linear rule were up to 50% lower than FEA

predictions.

Neuber’s rule best represents the case of plane stress [4]. The product of the local

stress-strain range is defined as:

∆𝜀∆𝜎 = 𝐾𝑡2 ∆𝑒∆𝑆

(4.24)
If the nominal behavior is elastic, the product of the local stress-strain range can be

rewritten as:

(𝐾𝑡 ∆𝑆)2
∆𝜀∆𝜎 = (4.25)
𝐸′

Equation 4.25 is combined with the Ramberg-Osgood equation as shown in equation 4.6

to determine the local stress, Δσ, and strain, Δε, ranges:

1⁄
(∆𝜎)2 ∆𝜎 𝑛′ (𝐾𝑡 ∆𝑆)2
+ 2∆𝜎 ( ′ ) = (4.26)
𝐸′ 2𝐾 𝐸′

Neuber’s rule has been found to provide accurate estimates of the notch strain for plates

and thin sheets as these geometries usually result in plane stress conditions [68].

The Strain Energy Density (SED) or Glinka’s rule is a more recent method to

predict notch deformation behavior and the results typically fit in between the extreme

cases of plane stress and plane strain. The SED rule is given as follows:
1⁄
(∆𝜎)2 4∆𝜎 ∆𝜎 𝑛′ (𝐾𝑡 ∆𝑆)2
+ ′ ( ) = (4.27)
𝐸′ 𝑛 + 1 2𝐾 ′ 𝐸′

where the difference between equations 4.26 and 4.27 is an added factor of 2/(n′+1) to the

second term [4].

83
 Plots of the local stresses and strains verses the nominal stress can provide insight

as to which strain prediction rule would best estimate the notched behavior, as compared

to the FEA results. Plots of local stress and local strain verses nominal stress are shown

superimposed with FEA results in Figure 4-11. Because plane strain conditions exist at

the notch root, von Mises stresses and strains from FEA are used to compare the rules.

For both notch stress and strain, FEA results are closest to the SED rule. Therefore, it is

expected that the SED rule would also best predict notched fatigue behavior.

Typically, notch deformation rules such as Neuber and SED rules assume one Kf

value obtained from long life fully-reversed experimental data. However, Kf changes with

fatigue life based on varying degrees of plastic deformation. The variation of Kf with

fatigue life is shown in Table 4.7(b). The fatigue notch factor, Kf, used in the life

prediction methods may also be calculated by dividing equation 4.10 by 4.11 to obtain Kf

as a function of fatigue life. Solving the Kf equation as a function of fatigue life with

equation 4.11 relating fatigue life of notched specimens to the applied stress amplitude,

Kf can be determined as a function of applied stress amplitude as follows:

𝐾𝑓 = 17.75(𝑆𝑎 )−0.412 (4.28)

If experimental data are not available, then Kf = 1 at one cycle with a stress amplitude of

1,608 MPa (assuming the same fatigue strength of smooth and notched specimens) and Kf

= 2.54 with a stress amplitude of 121 MPa at 106 cycles can be used. The equation based

on this estimate gives similar results as equation 4.28 and given as:

𝐾𝑓 = 14.3(𝑆𝑎 )−0.36 (4.29)

Equation 4.28 accounts for the change in the fatigue notch factor with stress amplitude or

fatigue life and is used with the life prediction methods.

84
 Strain-life predictions based on notch strain from Neuber, SED, and FEA are

shown in Figure 4-12. The resulting strains obtained from all three methods were used

with the strain-life equation to predict fatigue life. For life predictions with the SED or

Neuber rule Kt = 3 has been replaced with Kf = 2.54 or calculated from equation 4.28 to

account for the stress gradient effect [4]. By accounting for the stress gradient effect, life

prediction results between SED or Neuber, and FEA diverge, but strain-life curves based

on these rules become closer to the experimental curve. As observed from Figure 4-12(a),

the SED rule with Kf = 2.54 best predicts the fatigue behavior of the notched specimens.

Figure 4-12(b) shows a scatter plot with factors of two, five, and ten lower bounds. It can

be observed that all methods are conservative, and the modified SED rule provides the

closest prediction to experimental life. Figure 4-12(c) shows the same rules used, except

Kf was calculated from equation 4.28. Both Neuber and SED make satisfactory and

similar predictions in this case and are only slightly non-conservative for short life, as

shown in Figure 4-12(d).

The Fatemi-Socie and SWT mean stress parameters were also used in this study

to predict smooth and notched specimen fatigue behavior in the presence of mean stress.

Both rules use strain-amplitudes, maximum stress, and strain-life fatigue properties to

characterize fatigue behavior. The Fatemi-Socie parameter, which is a critical plane

approach found in [4] is given as follows:

𝜎𝑛,𝑚𝑎𝑥 𝜎𝑓′ 𝑏 ′ 𝑐 𝜎𝑓′ 𝑏

𝛾𝑎 (1 + 𝑘 ) = [1.275 ′ (2𝑁𝑓 ) + 1.5𝜀𝑓 (2𝑁𝑓 ) ] [1 + 𝑘 (2𝑁𝑓 ) ] (4.30)
𝑆𝑦 𝐸 2𝑆𝑦

The shear strain amplitude γa and maximum stress σn,max act on the maximum shear plane

and for uniaxial loading can be obtained from the following equations:

85
 (4.31)
𝛾𝑎 = (1 + 𝜐𝑒𝑓𝑓 )𝜀𝑎
∆𝜀 ∆𝜀𝑝
𝜐𝑒 ( 2𝑒 ) + 𝜐𝑝 ( 2 )
(4.32)
𝜐𝑒𝑓𝑓 =
𝜀𝑎
𝜎𝑚𝑎𝑥
𝜎𝑛,𝑚𝑎𝑥 = (4.33)
2

The constant k is determined from axial and torsional fatigue data, but when sufficient

data are unavailable it can be approximated as 1. Throughout this study, the value of k

used was 4. A value of 2 was also used, but differences in predictions between the two

values was small. In equation 4.32, the elastic Poisson’s ratio υe = 0.275 and the plastic

Poisson’s ratio υp = 0.5. The maximum stress and strain amplitude values used in

equations 4.31 through 4.33 are obtained from Neuber and SED rules.

The strain-based SWT method is based on the product of the local strain

amplitude and maximum local stress and given as follows:

2
𝜎𝑓′ 2𝑏 𝑏+𝑐
𝜀𝑎 𝜎𝑚𝑎𝑥 = ′ (2𝑁𝑓 ) + 𝜀𝑓′ 𝜎𝑓′ (2𝑁𝑓 ) (4.34)
𝐸

Figures 4-13(a) and 4-13(b) show results of the Fatemi-Socie and SWT mean stress

parameters, respectively, with local stresses and strains obtained from the SED rule with

Kf = 2.54. Figure 4-13(c) and 4-13(d) show similar results, except with Kf calculated from

equation 4.28. Figures 4-14(a) through 4-14(d) show similar plots with Neuber’s rule. It

is observed that the SED rule predicts fatigue behavior of notched specimens better than

Neuber’s rule when considering Kf = 2.54. The difference between Neuber and SED rules

and FS and SWT parameters with Kf calculated from equation 4.28 is negligible. Table

4.7 shows the calculated stress and strain amplitudes from Neuber and SED as well as the

FS and SWT parameter values.

86
 It is also convenient to use the local stress which was obtained simultaneously

with local strain from the Neuber, SED, and FEA for life prediction. By obtaining the

local stress for the notched specimens, the smooth S-N curve can be used to estimate

notched fatigue behavior. Figure 4-15(a) shows the results between the two models (i.e.

Neuber and SED) with Kf = 2.54 and FEA. Fatigue lives predicted by Neuber and SED

are similar, as shown in Figure 4-15(b). Similar to the local strain approach, the modified

version of the SED rule based on Kf rather than Kt best predicts fatigue behavior. Figures

4-15(c) and 4-15(d) show notched fatigue predictions with Kf calculated from equation

4.28. Similar to the local strain approach with Kf calculated from this equation, the

difference between Neuber and SED are small and predictions are within a factor of two

fatigue life. Figure 4-16 shows a flow chart with the steps to determine the notched

specimen behavior for the S-N and local approaches.

Another method to determine notched fatigue behavior is to use the Theory of

Critical Distances (TCD). This method accounts for the stress gradient at the notch and

has been reviewed by Susmel and Taylor [70, 71]. The Theory of Critical Distances has

shown promising results to predict fatigue behavior of engineering materials. This

method is derived from linear-elastic fracture mechanics and relies on FEA results. It also

requires the use of the fatigue limit and crack growth threshold stress intensity factor as

material properties. In the case of ductile iron however, the TCD method applied by

Baicchi et al. resulted in over-estimation of the fatigue life [72].

4.5 Mean Stress Effects on Smooth and Notched Fatigue Behavior

To study mean stress effects on fatigue behavior, other than fully-reversed loading

(R = -1), six additional R-ratios were selected for smooth specimens and five additional

87
R-ratios were chosen for the notched specimens. The R-ratios for the smooth specimens

were R = -7, -3, 0, 1/3, 0.5, and 0.75. The notched R-ratios were R = -∞ -7, -3, 0, and 1/3.

Results for the smooth specimens at all R-ratios and the respective best-fit lines

are shown in Figure 4-17(a). The fully-reversed data from Figure 4-7 for the smooth

specimens are also included for comparison. Tabulated results for the smooth specimen

mean stress tests are shown in Table 4.8. As shown in Figure 4-17(a), the slopes of all the

S-N lines are nearly parallel (with the exception of R = -7), which indicates that the effect

of mean stress is similar in all fatigue life regimes for the selected R-ratios. The best-fit

line for R = 0.5 was estimated due to the one run-out test at 175 MPa and only one other

stress level with finite lives. The effects of higher compressive stress at R = -7 is more

beneficial compared to R = -3, but due to damage caused by excessive plasticity at a

stress amplitude of 600 MPa resulting in a minimum stress of -1050 MPa, the fatigue life

was only increased by a factor of 2. By decreasing the stress amplitude to 500 MPa, the

damaging effect of plasticity decreased, which resulted in one run-out. A best-fit line

could not be drawn for R = 0.75 due to material strength limitations. Stress amplitude

levels beyond 113 MPa at this R-ratio would have resulted in a maximum stress larger

than 900 MPa and fracture would result. The following equations represent the S-N lines

under the various mean stress conditions:

−0.054 (4.35)
𝑆𝑎𝑟 = 1071(𝑁𝑓 ) for R = -7

−0.125 (4.36)
𝑆𝑎𝑟 = 2173(𝑁𝑓 ) for R = -3

−0.130
𝑆𝑎𝑟 = 1256(𝑁𝑓 ) for R = 0 (4.37)

−0.123
𝑆𝑎𝑟 = 986(𝑁𝑓 ) for R = 1/3 (4.38)

88
 −0.124
𝑆𝑎𝑟 = 850(𝑁𝑓 ) for R = 0.5 (4.39)

Results of the notched specimens at different R-ratios, including the respective

best-fit lines, are shown in Figure 4-17(b). The fully-reversed data from Figure 4-7 for

the notched specimens are also included for comparison. Tabulated results for the

notched specimen testing are shown in Table 4.9. As shown in Figure 4-17(b), the slopes

of all the S-N lines are similar. The best-fit line for R = -7 has a more shallow slope. Two

data points were generated for R = -∞, however, and a best-fit line could not be generated

due to limited data. Full compression loading inherently will not allow cracks to grow,

which caused the need for the small number of tensile cycles in each load block. The

complication with the tensile cycles is the requirement that the tensile stress amplitude

needs to be at a high enough level to cause crack growth while not causing crack

initiation damage. This constraint and the high value of Smin limiting the allowable

applied stress amplitude were the cause of limited data for R = -∞. The following

equations are the best-fit lines for each R-ratio tested:

−0.157
𝑆𝑎𝑟 = 2387(𝑁𝑓 ) for R = -7 (4.40)

−0.188
𝑆𝑎𝑟 = 2536(𝑁𝑓 ) for R = -3 (4.41)

−0.223
𝑆𝑎𝑟 = 1841(𝑁𝑓 ) for R = 0 (4.42)

−0.212
𝑆𝑎𝑟 = 1430(𝑁𝑓 ) for R = 1/3 (4.43)

Table 4.10 shows for all R-ratios, the experimental values of A and B for the smooth and

notched S-N lines with the respective fatigue limits.

89
4.6 Correlation of Experimental Data with Mean Stress Parameters

4.6.1 Predicting Mean Stress Notched Fatigue Behavior from Smooth

Fatigue Behavior
Similar to the need for predicting notched fatigue behavior from smooth behavior

for fully-reversed loading, it is sometimes necessary to predict notched fatigue behavior

under mean stress loading conditions. Modified Goodman mean stress equation has often

been used to predict notched fatigue behavior under various mean stress conditions. This

equation for estimating notched fatigue behavior in the presence of mean stress is given

as follows:

𝑆𝑎 𝑆𝑚
+ =1
𝑆𝑓 ⁄𝐾𝑓 𝑆𝑢 (4.44)

Equation 2.18 can be used with equation 4.44 for various R-ratios to estimate the nominal

stress amplitude which corresponds to the fatigue limit for the given R-ratio. The fully-

reversed, smooth specimen fatigue limit, Sf, is reduced by the fatigue notch factor Kf to

account for the notch effect. Equation 4.44 was applied to experimental data and the

results can be seen in Figure 4-18.

Unlike the experimental S-N line slopes shown in Figure 4-17(b), the slopes of the

predicted lines for each R-ratio are not parallel. Modified Goodman predicts very little

difference between compressive mean stress and fully-reversed loading conditions for the

notched fatigue behavior, contrary to the experimental observations. The difference is

greater for tensile mean stress and the slopes between R = 0 and R = 1/3 clearly diverge

as the fatigue life increases. Modified Goodman is overly conservative for compressive

mean stress data and non-conservative for tensile mean stress data of notched specimens.

A contributing factor for the discrepancy between experimental and predicted lives based

90
on this equation is due to prediction of the notched S-N line for fully-reversed loading. As

observed in Figure 4-18, the notched line prediction is conservative with respect to

experimental data for R = -1 notched condition.

4.6.2 Nominal Stress Approach for Smooth and Notched Specimens

Four mean stress models were chosen to evaluate the fatigue behavior of smooth

and notched specimens and were compared against each other and with experimental

data. The Modified Goodman, stress version of Smith-Watson-Topper (SWT), FKM, and

Fatemi-Socie (FS) methods were chosen. The Modified Goodman and SWT parameters

were chosen because they are popular methods used to account for mean stress. The

FKM method was chosen because it is a comprehensive mean stress parameter, although

it has not received much attention in the United States. The Fatemi-Socie parameter was

chosen because of its success in predicting fatigue life under complex stress states,

including those with mean stress. The Gerber parabola was not included in analysis as it

lacks capability to account for compressive mean stress and is highly non-conservative.

Walker equation (equation 2.22) was not considered, because the gamma value is near

0.5 for all R-ratios, which results in the same equation as the SWT mean stress parameter.

A perfect mean stress parameter would convert the S-N data at different R-ratios

such that they fall on the S-N line of R = -1. Figure 4-19 shows the smooth mean stress

correlation results and Figure 4-21 shows the notched mean stress correlation results with

these four parameters. Green symbols in these figures designate tensile mean stress, while

blue represents compressive mean stress.

The Modified Goodman model is as follows:

91
𝑆𝑎 𝑆𝑚
+ =1
𝑆𝑎𝑟 𝑆𝑢 (4.45)

Data correlations of the smooth specimens, shown in Figure 4-19(a) under the presence

of compressive mean stress are good. This method is inaccurate for estimating the fatigue

behavior under the presence of tensile mean stress, as most of the tensile mean stress data

are away from the predicted line. For notched specimens, shown in Figure 4-21(a) under

the presence of compressive mean stress, Modified Goodman is very conservative. As

both fatigue life and mean stress increase, the degree of conservatism increases. Modified

Goodman is non-conservative for tensile mean stress, with the exception of the shorter

life data. Figure 4-23(a) shows a scatter plot with both smooth and notched specimen data

with upper and lower bounds of 3, 5, and 10. For notched and mean stress data, 69% fall

within a factor of ±3, 80% fall within a factor of ±5, and 86% fall within a factor of ±10.

A high degree of scatter is also shown for all the mean stress data. Modified Goodman

predicts infinite life for some notched specimens under tensile mean stress where failure

was observed. Modified Goodman is also very conservative for high mean stress in both

tension and compression, as shown for R = -∞, 0.5, and 0.75.

The stress-based version of the Smith-Watson-Topper (SWT) model is defined as

follows:

𝑆𝑎𝑟 = √𝑆𝑎 𝑆𝑚𝑎𝑥

(4.46)
For smooth specimen data, shown in Figure 4-19(b), under the presence of tensile mean

stress, SWT accurately predicts the fatigue behavior. For smooth specimens under high

compressive mean stress (R = -7), SWT is non-conservative, as data fall below the line

and fatigue limit. This parameter predicts the fatigue behaviors of tensile and

compressive mean stress data for the notched specimens very well, except for R = -∞, see

92
Figure 4-21(b). SWT is unable to account for fully compressive loading (R = -∞) because

the maximum stress is zero resulting in an undefined equivalent stress amplitude from

equation 4.46. Figure 4-23(b) shows that the majority of fatigue data fall within a life

factor of 3. For notched and mean stress data, 90% fall within a factor of ±3, and 91% fall

within both a factor of ±5 and a factor of ±10. A few exceptions lie with smooth

specimens tested under high compressive mean stress (R = -7). The SWT parameter

predicts infinite life for this case, while failures were observed. This method tends to be

conservative at long life regardless of specimen type or R-ratio.

The FKM method resulted in comparable results to the SWT approach. As

described in chapter 2, this method accounts for several ranges of mean stress differently.

Equations 2.30 through 2.32 are shown below with aM = 0.35 and bM = 0.08 values

substituted in from Table 2.6 for ductile iron:

𝑆𝑎𝑟 = 𝑆𝑎 + 0.432𝑆𝑚 for -∞ ≤ R ≤ 0 (4.47)

𝑆𝑎𝑟 = 1.25𝑆𝑎 + 0.18𝑆𝑚 for 0 < R < 0.5 (4.48)
𝑆𝑎𝑟 = 1.79𝑆𝑎 for R ≥ 0.5 (4.49)

For smooth specimen data shown in Figure 4-19(c), the FKM method results in

predictions which are very similar to SWT, with the exception that compressive mean

stress was better accounted for. The prediction for R = 0.5, which is the transition to

region IV (R ≥ 0.5) shows very good results. However, as the tensile mean stress

increases, the results are non-conservative (i.e. R = 0.75), implying infinite life. For

notched specimens, see Figure 4-21(c), FKM accurately predicts the fatigue behavior for

specimens subjected to tensile mean stress. The results are conservative for compressive

mean stress and increase in conservatism as the compressive mean stress increases.

Figure 4-23(c) shows that FKM accounted for most tensile mean stress data of smooth

93
and notched specimens within a factor of three. For notched and mean stress data, 89%

fall within a factor of ±3, 94% fall within a factor of ±5, and 96% fall within a factor of

±10. FKM predicts infinite life for very high tensile mean stress (R = 0.75) where failure

was observed and is very conservative for high compressive mean stress (R = -∞).

The Fatemi-Socie (FS) model, which was introduced as equation 4.30, was

modified to a stress-based version to obtain the equivalent fully-reversed stress as

follows:

𝑘 𝑘 𝐵 𝑘 2 2𝐵
𝑆𝑎𝑟 (1 + 𝑆𝑎𝑟 ) = 𝑆𝑎 (1 + (𝑆𝑎 + 𝑆𝑚 )) = 𝐴(𝑁𝑓 ) + 𝐴 (𝑁𝑓 )
2𝑆𝑦 2𝑆𝑦 2𝑆𝑦 (4.50)

where A and B are the intercept and slope of the fully-reversed S-N line. As mentioned

earlier, k = 4 was used for the following analysis with this parameter. For smooth

specimen data shown in Figure 4-19(d), the FS model tends to be somewhat non-

conservative for tensile mean stress at long fatigue lives. Figure 4-20 shows the

parameter-life plot for smooth specimens. The results show a similar representation of the

data in Figure 4-19(d). For notched behavior, shown in Figure 4-21(d), the method

predictions are different between tensile and compressive mean stress. Under the

presence of tensile mean stress, the FS model is somewhat non-conservative, but it is

conservative for compressive mean stress. As the compressive mean stress increases,

predictions become more conservative. Similar to the smooth specimens, Figure 4-22

shows the parameter-life plot for notched specimens, with a similar representation to

Figure 4-21(d). Correlations for compressive mean stress tend to fall within a

conservative factor of 10. Figure 4-23(d) shows a scatter plot for the FS parameter. For

notched and mean stress data, 77% fall within a factor of ±3, 90% fall within a factor of

±5, and 99% fall within a factor of ±10.

94
 Haigh, or constant life, diagrams were constructed for both smooth and notched

specimens. Figures 4-24 and 4-27 show the smooth and notched experimental data,

respectively, with each mean stress parameters for fatigue lives of 103, 104, 105, and ≥ 106

cycles. Figures 4-25 and 4-28 compile all mean stress parameters for smooth and notched

specimens, respectively, for a given life in order to show differences and similarities

between parameters. Data points plotted on the constant life diagrams have been obtained

from the best-fit equations for each respective R-ratio. Data that fall above a given line

for each mean stress model results in a conservative median fatigue life equal to or less

than the predicted life from the mean stress model. Conversely, data which fall below a

mean stress model line, predicts a longer, non-conservative median fatigue life.

In order to quantify the differences between mean stress parameters, the percent

error between the experimental stress amplitude and predicted stress amplitude were also

calculated. A positive percent error represents non-conservatism and a negative percent

error represents conservatism in the predictions. The experimental stress amplitude was

determined from the respective S-N equations and the predicted stress amplitude was

determined from the intersection of the mean stress parameter and R-ratio lines. The

experimental stress amplitude was used as the basis of calculating percent errors which

are shown in Figures 4-26 and 4-29 for smooth and notched data, respectively.

The constant life diagram for Modified Goodman and smooth specimens, shown

in Figure 4-24(a), indicates conservative estimates for tensile mean stress for fatigue lives

of 104 and 105 cycles. Predictions tend to become slightly non-conservative for

intermediate values of tensile mean stress and a fatigue life of ≥ 106 cycles, however, for

high tensile mean stress (R = 0.75), predictions are conservative. Predictions increase in

95
conservatism as the compressive mean stress value increases for high cycle fatigue life,

see Figure 4-26. Figure 4-27(a) for notched specimens shows that under tensile mean

stress at short life predictions are conservative, but become non-conservative as the life

increases. The compressive mean stress, however, is similar to smooth fatigue behavior.

The degree of conservatism increases with life and the level of mean stress. Full-

compression loading (R = -∞) is shown to be the most conservative.

Figure 4-24(b) shows that for tensile mean stress, Smith-Watson-Topper

correlations are very good. Nearly all the data fall on the prediction lines for each fatigue

life for smooth specimens. The predictions are non-conservative for compressive mean

stress. Figure 4-26 shows the percentage error between predicted and experimental stress

amplitudes. Notched fatigue behavior, represented in Figure 4-27(b), is slightly worse for

tensile mean stress compared to smooth fatigue behavior. Compressive mean stress is

better accounted for compared to the smooth fatigue behavior. For low compressive mean

stress (R = -3), nearly all the data fall on the line. As the level of compressive mean stress

increases, the predictions become non-conservative.

The FKM method for smooth fatigue behavior is represented in Figure 4-24(c).

The results are similar to SWT for tensile mean stress, except data correlations using

FKM are non-conservative for tensile mean stress. For compressive mean stress, FKM

accurately predicts the fatigue behavior. Figure 4-26 shows that for very high tensile

mean stress (R = 0.75), FKM is highly non-conservative. For low tensile mean stress (R =

0), and compressive mean stress, FKM is conservative. Figure 4-27(c) shows that

notched specimen fatigue predictions based on this method are similar to smooth

96
specimen fatigue predictions for tensile mean stress. Figure 4-29 shows FKM is

conservative for nearly all R-ratios and fatigue lives of notched specimens.

The Fatemi-Socie parameter for smooth specimens is shown in Figure 4-24(d).

The method is mostly non-conservative for tensile mean stress. Compressive mean stress

is predicted similarly to the FKM method, except all predictions are conservative at long

life (see Figure 4-26). Notched specimen fatigue behavior is shown in Figure 4-27(d) and

is similar to smooth behavior. Results are mostly non-conservative for tensile mean

stress. Similar to smooth behavior, notched fatigue behavior under compressive mean

stress is conservative. The degree of conservatism increases as the level of compressive

mean stress increases (see Figure 4-29).

Figures 4-25 and 4-28 compare all mean stress parameters for a given fatigue life

for smooth and notched specimens, respectively. Figures 4-25(a) through 4-25(c) for

short to intermediate lives show similar and reasonable predictions for FKM, SWT, and

FS for -3 ≤ R ≤ 0.5. For R-ratios less than -3 and larger than 0.5 the differences between

these models increase. Modified Goodman diverges from the other models, particularly at

high and very low R-ratios and is the most conservative model. Figure 4-25(d) for high

cycle fatigue (Nf ≥ 106 cycles) shows more separation of the mean stress models at low

and high R-ratios. Fatemi-Socie and FKM are close to each other at all R-ratios.

Experimental data fall on or between the SWT and Fatemi-Socie curves for both tensile

and compressive mean stresses.

Figure 4-28 shows similar representations of the mean stress models for notched

specimens. Figure 4-28(a) through 4-28(d) show that Modified Goodman, SWT, FS and

FKM are similar for all fatigue lives for -3 ≤ R ≤ 0. For R-ratios less than -3 and greater

97
than 0, differences between the models increases. Similar to smooth specimen results,

Modified Goodman diverges from the other models at low and high R-ratios. SWT shows

greatest separation from other models as life increases for low compressive mean stress.

FS and FKM models are similar to each other for all fatigue lives except when R = -∞.

Experimental data mostly fall on or between SWT and FKM models for tensile and

compressive mean stresses.

4.6.3 Local Stress Approach for Predicting Notched Behavior

The local approach was also used to determine stresses and strains at the notch

root and predict fatigue life in the presence of mean stress. Neuber and SED rules were

used to predict stresses and strains at the notch under the presence of mean stress. Both

rules were used to obtain the local stress amplitude, which was then converted to the

equivalent fully-reversed stress using SWT and FS mean stress parameters. If the

absolute value of the largest nominal stress exceeded 0.8Sy, equation 4.24 was used

where the nominal strain, Δe, was calculated for the nominal stress, ΔS, from the

Ramberg-Osgood equation 4.6. Table 4.11(a) shows all relevant local stress values

obtained from Neuber and SED rules for notched specimens subjected to mean stress

loading for Kf = 2.54. Table 4.11(b) shows the same information, except Kf is a function

of Sa. Calculations of the Fatemi-Socie and SWT parameters are also included in these

tables.

To obtain the equivalent fully-reversed stress using SWT, the local stress

amplitude and local maximum stress obtained from Neuber and SED rules were used

with equation 4.46. Figures 4-30(a) and 4-30(b) show the results between Neuber and

SED rules, respectively, using the SWT parameter. For very high compressive mean

98
stress (R = -7), low cycle and high cycle fatigue data have relatively small differences in

equivalent stress amplitudes but with vastly different fatigue lives. This is attributed to

variation of Kf with fatigue life due to varying plastic deformation and the stress gradient

effect. Overall, the notched data are predicted conservatively. Figures 4-30(c) and 4-30(d)

show improved results by using Kf as a function of Sa. Figure 4-31 shows scatter plots of

the data for both rules with constant Kf and with variable Kf. By using SWT with

Neuber’s rule and constant Kf, 43% of notched and mean stress data are correlated within

a factor of ±3, 66% of data within a factor of ±5, and 86% within a factor of ±10. Using

SWT with SED and constant Kf, 56% of notched and mean stress data are correlated

within a factor of ±3, 81% within a factor of ±5, and 87% within a factor of ±10. Figures

4-31(c) and 4-31(d) show the same rules applied, except Kf was calculated as a function

of stress amplitude which resulted in improved correlations for both Neuber and SED

rules. Neuber correlated 80% and SED correlated 81% of notched and mean stress data

within a factor ±3, 87% (Neuber) and 89% (SED) of data within a factor of ±5, and 93%

(Neuber and SED) of data within a factor ±10.

The stress-based Fatemi-Socie parameter, equation 4.50, is shown in Figure 4-32.

The results were not significantly different compared to SWT when considering Kf =

2.54. As with SWT, the SED rule was better than Neuber’s rule with the Fatemi-Socie

parameter. Figure 4-33(a) shows that Neuber’s rule with Kf = 2.54 correlates 43% of

notched and mean stress data within a factor of ±3, 66% of data within a factor of ±5, and

87% of data within a factor of ±10. The SED rule, Figure 4-33(b), correlates 57% of

notched and mean stress data within a factor of ±3, 79% of data within a factor of ±5, and

89% of data within a factor of ±10. Figures 4-32(c) and 4-32(d) use Kf calculated as a

99
function of Sa with Neuber and SED rules. Results are significantly better compared to

using a constant Kf. The percent of data correlated within each factor are similar between

Neuber and SED rules. The resulting correlations are as follows: 83% (Neuber) and 84%

(SED) of data are correlated within a factor of ±3, 93% (Neuber) and 91% (SED) of data

correlated within a factor of ±5, and 99% (Neuber and SED) of data within a factor of

±10 as shown in Figures 4-33(c) and 4-33(d). Significant difference between the Fatemi-

Socie and SWT mean stress parameters is apparent for R = -∞. The SWT mean stress

model predicts R = -∞ outside a non-conservative factor of 5 for both Neuber and SED

rules with Kf calculated from equation 4.28. The Fatemi-Socie parameter, however,

predicts R = -∞ on the conservative side, within a factor of 3.

4.6.4 Local Strain Approaches for Predicting Notched Behavior

The strain-based version of the Smith-Watson-Topper and Fatemi-Socie models

were evaluated with local strains obtained from Neuber’s rule and the SED rule. The

local strains were obtained simultaneously with the local stresses explained in the

previous section. Table 4.11(a) shows all relevant local strain values obtained from

Neuber and SED rules for notched specimens subjected to mean stress loading for Kf =

2.54. Table 4.11(b) shows the same information, except Kf is a function of Sa.

Calculations of the Fatemi-Socie and SWT parameters are also included in these tables.

Parameter-life plots of the SWT parameter are shown in Figure 4-34. As with the

local stress approach, the SED rule best characterized fatigue behavior of notched

specimens when Kf = 2.54. Tensile mean stress data were much better correlated

compared to the compressive mean stress data. However, using Kf calculated from

equation 4.28, with Neuber and SED rules, with the SWT model, notched data were more

100
closely correlated to the smooth prediction line. Figures 4-35(a) and 4-35(b) show scatter

plots of the SWT parameter with both Neuber and SED rules used with Kf = 2.54.

Notched and mean stress data used with Neuber and SWT are correlated to 49% of data

within a factor of ±3, 61% within a factor of ±5, and 84% of data within a factor of ±10.

The SED rule allowed for better correlations with 59% of data correlated within a factor

of ±3, 76% within a factor of ±5, and 84% of data within a factor of ±10. Using Kf

calculated as a function of Sa with Neuber and SED rules resulted in nearly identical

correlations as shown in Figures 4-35(c) and 4-35(d), respectively. For Neuber’s rule,

87% of data are correlated within a factor of ±3, 90% within a factor of ±5, and 99% of

data within a factor of ±10. For the SED rule, 87% of data are correlated within a factor

of ±3, 91% within a factor of ±5, and 99% of data within a factor of ±10. Unlike the

nominal based approach, SWT was able to account for R = -∞ because a local maximum

stress was present due to residual stress during unloading.

The Fatemi-Socie parameter, equation 4.30, was used with equations 4.31 through

4.33 to characterize mean stress effects for smooth and notched specimens. The Fatemi-

Socie parameter is slightly non-conservative for smooth, tensile mean stress data.

However, it is highly conservative (similar to SWT) for all compressive mean stress data,

for both Neuber and SED rules when Kf = 2.54, shown in Figures 4-36(a) and 4-36(b).

This is most likely because cracks grow very quickly under tensile mean stress once a

crack is initiated. However, for compressive mean stress, once a crack is initiated, it may

spend a significant portion of its life growing the crack to failure. If crack initiation

defined failure, the data would shift closer to the FS parameter line. The variable Kf

causes the most significant change for notched, compressive mean stress data compared

101
to notched, tensile mean stress data, because of a larger shift in data towards the

prediction line. Figures 4-37(a) and 4-37(b) show scatter plots of the smooth and notched

data where Neuber and SED rules are used with Kf = 2.54, respectively. Neuber’s rule

correlates 46% of notched and mean stress data within a factor of ±3, 63% of data within

a factor of ±5, and 84% of data within a factor of ±10. Strain energy density rule

correlates 54% of notched and mean stress data within a factor of ±3, 76% of data within

a factor of ±5, and 86% of data within a factor of ±10. As with the SWT parameter, using

Kf calculated form equation 4.28 resulted in similar predictions between Neuber and SED

rules. Figure 4-37(c) shows that Neuber’s rule correlated 81% of notched and mean stress

data within a factor of ±3, 89% of data within a factor of ±5, and 99% of data within a

factor of ±10. Figure 4-37(d) shows that the SED rule correlated 81% of notched and

mean stress data within a factor of ±3, 90% of data within a factor of ±5, and 99% of data

within a factor of ±10. Data for SWT and FS parameters also fall in amplitude scatter

bands within a factor of two.

4.7 Fatigue Notch Factor as a Function of R-ratio and Life

Plots of Kf vs. Sa, Kf vs. R-ratio and Kf vs. Nf were developed. Fatigue lives of 103,

104, 105, and ≥ 106 cycles were considered for each R-ratio. Figure 4-38(a) shows Kf as a

function of applied stress amplitude for the notched specimens. Figure 4-38(b) shows Kf

as a function of the R-ratio for each individual fatigue life and Figure 4-38(c) shows Kf as

a function of fatigue life. As shown in Figure 4-38(a), the fatigue notch factor decreases

with increasing stress amplitude. Stress amplitude values shown coincide with stress

levels applied to the notched specimens.

102
 The fatigue notch factor is the largest for ≥ 106 cycles, and decreases in value as

the life decreases. The severity of the notch decreases at short life because of higher

degrees of local plastic deformation controlling the fatigue process. The effect of the

notch was minimized under high compressive stress increased up to R = 0. Local yielding

at the notch is present under high compressive mean stress, therefore, reducing the effect

of the notch. Beyond R = 0, the severity of the notch tends to decrease which is believed

to be due to higher maximum load and, therefore, higher plastic deformation.

Figure 4-38(c) shows Kf as a function of fatigue life for individual R-ratios. As the

amount of compressive mean stress increases, the severity of the notch decreases. The

slopes of the lines tended to be similar with the exception at R = -3. Notch severity is

similar for R = 0 and -1 for life regimes between 104 to 106 cycles.

4.8 Fatigue Fracture Surfaces

One representative picture of the fatigue fracture surface at each stress amplitude

level and R-ratio of smooth specimens is shown in Figure 4-39 and for notched

specimens shown in Figure 4-40. Cracking behavior was similar for the smooth

specimens at all R-ratios as a semielliptical crack growth region was observed on the

fracture surface. The crack initiation sites are noticeable within the crack growth region.

Under compressive mean stress, three different regions are visible: crack initiation with

crack growth up to a boundary line, followed by another lighter area of crack growth

followed by final fracture. The darkest regions indicate the initial crack initiation and

growth. Once a crack reaches a critical length, the specimen is near failure but fracture

has not occurred, which is the transition to the lighter crack growth region. This second

lighter region can be considered the final crack growth region before fracture occurs. The

103
three different shades are most noticeable for R = -7 and less distinguishable for R = -3.

This can be attributed to smaller tensile stress amplitudes during cyclic loading causing

slower macrocrack growth.

Under the presence of tensile mean stress, the crack initiation sites were more

pronounced compared to fully-reversed or compressive mean stress loading. As with

compressive mean stress, higher stress amplitudes resulted in smaller, darker crack

initiation and growth regions compared to the lower stress amplitude levels which had

larger and lighter crack initiation and growth regions. Overall, it can be observed that

crack growth, up to final fracture has the largest area on the fracture surface for the

highest values of compressive mean stress and decreases in area as the mean stress

increases in the tensile direction. Higher applied stress amplitudes resulted in the smallest

crack initiation and growth regions and increased in area as the stress amplitude was

decreased.

Notched specimen fracture surfaces also exhibited similar features. Almost all

specimens contained a circular region indicating final fracture region where the fatigue

cracks grew towards. This circle was most prominent for high to mid stress levels applied

at each R-ratio. For the highest stress amplitude levels, the circle tended to be more

concentric in the cross section. As the stress level decreased, the circle was more

noticeably offset from the center of the cross section. At the lowest applied stress level,

the offset circle shifted to a prominent initiation location at the edge of the specimen,

which was semielliptical in shape, similar to the smooth specimens. Also similar to the

smooth specimens, the crack growth region was largest under high compressive mean

stress and decreased in area with increasing tensile mean stress. As the stress amplitude

104
decreased, the crack growth region increased. Almost all fracture surfaces contained a

rough surface, most noticeable by ratchet marks.

105
Table 4.1: Summary of monotonic tension and compression properties of 120-90-02
ductile cast iron.

Do Lo E Sy Su K εf σf
Specimen ID n %EL
(mm) (mm) (GPa) (MPa) (MPa) (MPa) (%) (MPa)

I-T3-03 6.33 12.70 177 650 1,022 1,776 0.169 10.4% 6.22% 1088

I-T3-02 6.37 12.70 171 638 990 1,614 0.155 7.87% 5.27% 1009

I-S3-21
6.30 12.70 157 700 - - - - - -
(compression)
Average
Tensile 174 644 1,006 1,695 0.162 9.13% 5.74% 1049
Values

Table 4.2: Elastic and plastic strain amplitudes obtained from the incremental cyclic
deformation step test in load control. E′ was used in elastic or plastic
strain calculations, E′ = 162 GPa.

∆𝜀 ∆𝜀𝑒 ∆𝜀𝑝 ∆𝜀𝑝 Applied

Sa
(MPa) 2 2 2 2 Cycles
(measured) (calculated) (calculated) (measured) (N)

700 0.850% 0.432% 0.418% 0.334% 108

600 0.494% 0.370% 0.124% 0.122% 600
500 0.329% 0.309% 0.020% 0.031% 500
400 0.251% 0.247% 0.004% 0.010% 108
350 0.218% 0.216% 0.001% 0.008% 108
300 0.185% 0.185% 0.000% 0.006% 102

106
Table 4.3: Summary of fully-reversed (R = -1) constant amplitude fatigue tests of
smooth specimens.

Cycles to
dmin Pa Fq Sa Sm δa
Spec. ID Failure
(mm) (kN) (Hz) (MPa) (MPa) (mm)
(Nf)

I-S3-06 6.32 18.82 1 600 -0.02 0.197 3,148

I-S3-15 6.32 18.82 1 600 -0.05 0.197 3,180
I-S3-17 6.31 18.76 1 600 0.11 0.198 3,224
I-S3-10 6.32 18.82 1 600 0.29 0.194 3,847
I-S3-07 6.33 18.88 1 600 0.06 0.195 4,200
I-S3-13 6.31 15.64 2 500 -0.05 0.156 16,071
I-S3-14 6.33 15.74 2 500 0.05 0.156 16,198
I-S3-16 6.32 15.69 2 500 -0.02 0.156 19,541
I-S3-09 6.32 15.69 2 500 0.29 0.156 21,960
I-S3-02 6.30 15.59 2 500 0.05 0.154 23,903
I-S3-37 6.33 12.59 4 400 0.16 0.123 97,294
I-S3-08 6.32 12.55 4 400 -0.18 0.121 102,350
I-S3-18 6.35 12.67 4 400 0.22 0.123 102,846
I-S3-03 6.30 12.47 4 400 -0.08 0.122 104,169
I-S3-36 6.32 12.55 4 400 0.08 0.122 115,066
I-S3-01 6.31 10.95 8 350 0.00 0.106 246,205
I-S3-04 6.30 10.91 8 350 0.00 0.104 294,698
I-S3-20 6.29 10.88 8 350 0.05 0.106 351,029
I-S3-46 6.34 11.05 8 350 -0.21 0.108 433,836
I-S3-05 6.30 10.91 8, 25 350 -0.10 0.105 > 5,000,000
I-S3-24 6.32 9.41 30 300 -0.16 0.095 > 5,000,000
I-S3-48 6.35 9.50 30 300 0.32 0.092 > 5,000,000
I-S3-22 6.29 9.32 30 300 0.00 0.092 > 5,000,000
I-S3-44 6.32 9.41 30 300 0.10 0.092 > 5,000,000

107
Table 4.4: Summary of fully-reversed (R = -1) constant amplitude fatigue tests of
notched specimens.

Cycles to
dmin Pa Fq Sa Sm δa
Spec. ID Failure
(mm) (kN) (Hz) (MPa) (MPa) (mm)
(Nf)

I-N3-02 6.38 12.79 0.5 400 0.02 0.111 3,225

I-N3-06 6.38 12.79 0.5 400 0.00 0.111 3,413
I-N3-05 6.38 12.79 0.5 400 0.02 0.111 3,582
I-N3-25 6.35 12.67 0.5 400 -0.03 0.110 3,598
I-N3-12 6.38 12.79 0.5 400 0.30 0.112 3,718
I-N3-11 6.38 12.79 0.5 400 -0.02 0.112 3,780
I-N3-09 6.38 9.59 1 300 0.03 0.084 13,015
I-N3-08 6.38 9.59 1 300 0.06 0.082 13,685
I-N3-01 6.38 9.59 1 300 0.09 0.082 14,220
I-N3-13 6.35 9.50 1 300 0.03 0.084 14,471
I-N3-03 6.38 9.59 1 300 0.23 0.083 15,045
I-N3-32 6.35 9.50 1 300 0.05 0.084 16,663
I-N3-23 6.35 6.65 3 210 0.30 0.058 64,303
I-N3-28 6.35 6.65 3 210 -0.17 0.057 79,173
I-N3-07 6.38 6.71 3 210 0.00 0.060 79,999
I-N3-47 6.35 6.65 3 210 -0.08 0.057 84,718
I-N3-04 6.38 6.71 2 210 0.02 0.062 98,025
I-N3-38 6.35 6.65 3 210 0.03 0.059 103,017
I-N3-16 6.35 4.43 15 140 -0.03 0.039 495,367
I-N3-26 6.35 4.43 15 140 0.11 0.038 566,493
I-N3-19 6.35 4.43 15 140 0.05 0.039 715,277
I-N3-10 6.40 4.50 15 140 -0.05 0.041 877,381
I-N3-22 6.32 3.92 25 125 -0.08 0.032 863,899
I-N3-30 6.35 3.96 25 125 0.11 0.034 986,882
I-N3-33 6.32 3.61 30 115 0.22 0.032 > 5,000,000
I-N3-42 6.35 3.64 30 115 -0.09 0.033 > 5,000,000

108
Table 4.5: Notch root stresses and strains obtained from FEA.

Sa σ1 σ2 ε1 ε2
β α
(MPa) (MPa) (MPa) (mm/mm) (mm/mm)
115 360 98 0.272 0.002055 -0.000008 -0.004
125 392 107 0.272 0.002234 -0.000009 -0.004
140 439 119 0.272 0.002502 -0.000010 -0.004
210 614 180 0.294 0.003846 -0.000029 -0.007
300 735 255 0.348 0.005923 -0.000053 -0.009
400 809 318 0.393 0.008776 -0.000118 -0.013

Table 4.6: Strain-life fatigue and cyclic deformation properties of 120-90-02 ductile
iron.

Compatibility
Best-Fits
Equations
σf′ (MPa) 1748
b -0.120
εf′ 14.2
c -1.06
K' (MPa) 1266 1293
n' 0.110 0.114
Sy' (MPa) 640
E' (GPa) 162

109
 Table 4.7: Calculations of Fatemi-Socie (with k = 4) and SWT mean stress
parameters for notched specimens under the fully-reversed (R = -1) cyclic
loading condition with notch root stresses and strains obtained from
Neuber and SED rules for (a) Kf = 2.54 and (b) Kf as a function of Sa.

(a)
σa
R Sa εa FS SWT
(MPa)
Ratio (MPa)
Neuber/SED Neuber/SED Neuber/SED Neuber/SED
-1 400 0.0091/0.0071 703/668 0.040/0.030 6.37/4.77
-1 300 0.0057/0.0050 631/605 0.023/0.019 3.58/3.00
-1 210 0.0034/0.0033 512/502 0.012/0.011 1.76/1.67
-1 140 0.0022/0.0022 355/354 0.0059/0.0059 0.78/0.78
-1 125 0.0020/0.0020 317/317 0.0050/0.0050 0.62/0.62
-1 115 0.0018/0.0018 292/292 0.0044/0.0044 0.53/0.53

(b)
σa
R Sa εa FS SWT
Kf (MPa)
Ratio (MPa)
Neuber/SED Neuber/SED Neuber/SED Neuber/SED
-1 400 1.50 0.0040/0.0038 557/541 0.014/0.013 2.23/2.04
-1 300 1.69 0.0032/0.0032 493/485 0.011/0.010 1.59/1.53
-1 210 1.96 0.0026/0.0026 409/407 0.0074/0.0074 1.05/1.04
-1 140 2.32 0.0020/0.0020 324/324 0.0051/0.0051 0.65/0.65
-1 125 2.43 0.0019/0.0019 303/303 0.0046/0.0046 0.57/0.57
-1 115 2.51 0.0018/0.0018 289/289 0.0043/0.0043 0.52/0.52

110
Table 4.8: Summary of constant amplitude mean stress fatigue tests at different R-
ratios for smooth specimens.

Cycles to
R dmin Pa Fq Sa Sm Smax Smin δa
Spec. ID Failure
Ratio (mm) (kN) (Hz) (MPa) (MPa) (MPa) (MPa) (mm)
(Nf)

HT3-S-22 -7 6.26 18.47 2 600 -450 150 -1050 0.186 39,385

HT3-S-25 -7 6.30 18.70 1 600 -450 150 -1050 0.189 58,406
HT3-S-26 -7 6.17 16.44 3 550 -413 137 -963 0.166 237,240
HT3-S-20 -7 6.20 16.60 2 550 -413 137 -963 0.166 246,438
I-S3-27 -7 6.32 16.24 10, 30 518 -388 130 -906 0.158 > 5,000,000
I-S3-41 -3 6.35 19.00 2 600 -300 300 -900 0.195 33,346
I-S3-29 -3 6.33 16.52 5 525 -263 262 -788 0.164 83,411
I-S3-47 -3 6.32 16.47 5 525 -263 262 -788 0.162 84,970
I-S3-33 -3 6.32 14.12 10 450 -225 225 -675 0.135 333,047
I-S3-26 -3 6.35 12.67 30 400 -200 200 -600 0.122 > 5,000,000
I-S3-50 0 6.30 13.25 0.75 425 425 850 0 0.140 4,649
I-S3-38 0 6.30 10.91 1 350 350 700 0 0.112 16,217
I-S3-11 0 6.35 11.08 1 350 350 700 0 0.113 19,289
I-S3-32 0 6.34 9.47 5 300 300 600 0 0.095 63,994
I-S3-39 0 6.33 9.44 3 300 300 600 0 0.095 69,221
I-S3-12 0 6.35 7.92 5, 10, 30 250 250 500 0 0.078 > 5,000,000
I-S3-40 0 6.32 7.84 30 250 250 500 0 0.078 > 5,000,000

I-S3-28 1/3 6.32 9.41 1 300 600 900 300 0.099 15,800
I-S3-45 1/3 6.32 9.41 1 300 600 900 300 0.098 17,979
I-S3-49 1/3 6.32 7.84 5 250 500 750 250 0.082 56,160
I-S3-35 1/3 6.29 6.21 10 200 400 600 200 0.061 217,936
I-S3-43 1/3 6.33 6.29 10 200 400 600 200 0.064 254,014
I-S3-42 1/3 6.32 6.27 10, 30 200 400 600 200 0.064 1,569,271
I-S3-23 1/3 6.28 5.11 30 165 330 495 165 0.051 > 5,000,000
HT3-S-2 0.5 6.20 6.34 3 210 630 840 420 0.068 82,334
I-S3-51 0.5 6.33 6.61 3 210 630 840 420 0.068 84,675
I-S3-34 0.5 6.33 5.51 7 175 525 700 350 0.056 280,219
HT3-S-25 0.5 6.23 5.44 7, 20 175 525 700 350 0.055 > 5,000,000
I-S3-31 0.75 6.34 3.55 10, 20 113 788 901 675 0.037 2,058,774
I-S3-30 0.75 6.30 3.51 10, 20 113 788 901 675 0.037 2,114,297

111
Table 4.9: Summary of constant amplitude fatigue tests at different R-ratios for
notched specimens.

Cycles to
R dmin Pa Fq Sa Sm Smax Smin δa
Spec. ID Failure
Ratio (mm) (kN) (Hz) (MPa) (MPa) (MPa) (MPa) (mm)
(Nf)

I-N-32 -∞ 6.38 16.78 1, 2 525 -525 0 -1050 0.143 188,000

I-N-33 -∞ 6.38 16.78 1 525 -525 0 -1050 0.143 206,000

I-N3-48 -7 6.38 19.18 0.5 600 -450 150 -1050 0.169 6,598
I-N3-31 -7 6.38 15.98 0.5 500 -375 125 -875 0.137 21,175
I-N3-49 -7 6.38 12.79 3 400 -300 100 -700 0.11 100,929
I-N3-37 -7 6.38 9.59 15 300 -225 75 -525 0.082 535,623
I-N3-44 -7 6.38 8.63 25 270 -203 67 -473 0.075 1,976,940

I-N3-39 -3 6.38 15.98 0.5 500 -250 250 -750 0.138 5,085
I-N3-41 -3 6.38 11.99 1 375 -188 187 -563 0.104 27,785
I-N3-20 -3 6.35 8.71 3 275 -138 137 -413 0.075 163,668
I-N3-36 -3 6.38 6.71 15 210 -105 105 -315 0.058 488,308
I-N3-45 -3 6.35 6.02 20 190 -95 95 -285 0.053 1,201,446

I-N3-50 0 6.35 8.87 0.5 280 280 560 0 0.078 4,597

I-N3-21 0 6.35 6.65 1 210 210 420 0 0.059 16,867
I-N3-18 0 6.35 4.75 5 150 150 300 0 0.043 84,587
I-N3-24 0 6.35 3.17 20 100 100 200 0 0.027 453,015
I-N3-17 0 6.38 3.2 20 100 100 200 0 0.027 482,507
I-N3-34 0 6.35 2.69 30 85 85 170 0 0.023 > 5,000,000

I-N3-27 1/3 6.35 7.6 0.5 240 480 720 240 0.066 4,345
I-N3-40 1/3 6.35 5.7 1 180 360 540 180 0.05 19,234
I-N3-46 1/3 6.35 4.12 4 130 260 390 130 0.038 79,082
I-N3-43 1/3 6.38 2.72 10 85 170 255 85 0.024 808,435
I-N3-29 1/3 6.35 2.53 30 80 160 240 80 0.023 637,051
I-N3-35 1/3 6.35 2.38 20 75 150 225 75 0.023 > 5,000,000

Table 4.10: Intercepts and slopes of the best-fit S-N lines including fatigue limits
defined at 10 6 cycles for each R-ratio.

Smooth Notched
R-ratio
A (MPa) B Sf (MPa) A (MPa) B Sf (MPa)
-7 1071 -0.054 500 2387 -0.157 274
-3 2173 -0.125 389 2536 -0.188 189
-1 1608 -0.120 307 2129 -0.204 126
0 1256 -0.130 250 1841 -0.223 85
1/3 986 -0.123 180 1430 -0.212 77
0.5 850 -0.124 153 - - -

112
Table 4.11(a): Equivalent fully-reversed (R = -1) local stresses and strains obtained with Neuber and SED methods with Kf =
2.54 and calculations of the Fatemi-Socie (with k = 4) and SWT mean stress parameters for notched mean stress
data.

σar σar
R Sa Sm (MPa) (MPa) εa σa (MPa) σm (MPa) FS SWT
Spec ID
ratio (MPa) (MPa) SWT/FS SWT/FS Neuber/SED Neuber/SED Neuber/SED Neuber/SED Neuber/SED
Neuber SED
I-N-32 -∞ 525 -525 460/536 441/514 0.016/0.011 771/731 -496/-465 0.042/0.029 4.32/2.98
I-N-33 -∞ 525 -525 460/536 441/514 0.016/0.011 771/731 -496/-465 0.042/0.029 4.32/2.98
I-N3-48 -7 600 -450 542/602 515/574 0.023/0.016 814/770 -453/-425 0.071/0.046 8.29/5.38
I-N3-31 -7 500 -375 584/621 558/593 0.014/0.010 758/719 -307/-286 0.048/0.034 6.34/4.42
I-N3-49 -7 400 -300 605/626 579/599 0.0092/0.0072 704/670 -184/-169 0.033/0.025 4.77/3.60
I-N3-37 -7 300 -225 557/574 540/555 0.0057/0.0050 632/606 -140/-125 0.019/0.016 2.80/2.38
I-N3-44 -7 270 -203 524/542 511/527 0.0048/0.0044 601/579 -144/-127 0.016/0.014 2.21/1.98
I-N3-39 -3 500 -250 665/683 631/649 0.014/0.010 758/719 -175/-164 0.057/0.039 8.23/5.67
I-N3-41 -3 375 -188 635/646 608/618 0.0082/0.0066 689/656 -103/-93 0.032/0.025 4.84/3.70
I-N3-20 -3 275 -138 550/563 535/546 0.0050/0.0045 606/584 -106/-93 0.017/0.015 2.49/2.19
I-N3-36 -3 210 -105 441/460 441/457 0.0034/0.0033 513/502 -133/-115 0.010/0.0094 1.30/1.28
I-N3-45 -3 190 -95 394/416 399/418 0.0030/0.0030 473/467 -145/-126 0.0079/0.0079 1.00/1.02
I-N3-50 0 280 280 695/679 663/648 0.0051/0.0046 612/588 179/160 0.024/0.020 4.04/3.41
I-N3-21 0 210 210 604/585 584/566 0.0034/0.0033 512/502 201/176 0.014/0.013 2.45/2.25
I-N3-18 0 150 150 489/462 479/453 0.0024/0.0024 380/379 252/227 0.0089/0.0087 1.49/1.42
I-N3-24 0 100 100 354/322 351/319 0.0016/0.0016 254/254 239/231 0.0051/0.0050 0.77/0.76
I-N3-17 0 100 100 354/322 351/319 0.0016/0.0016 254/254 239/231 0.0051/0.0050
00 0.77/0.76
I-N3-34 0 85 85 304/272 303/272 0.0013/0.0013 216/216 212/209 0.0040/0.0039 0.57/0.57
I-N3-27 1/3 240 480 714/685 683/656 0.0041/0.0038 562/545 344/310 0.020/0.013 3.71/2.42
I-N3-40 1/3 180 360 592/561 574/546 0.0029/0.0028 451/446 329/292 0.013/0.012 2.23/2.09
I-N3-46 1/3 130 260 479/441 468/431 0.0020/0.0020 330/330 367/334 0.0082/0.0080 1.42/1.35
I-N3-43 1/3 85 170 354/309 349/305 0.0013/0.0013 216/216 366/347 0.0048/0.0047 0.48/0.75
I-N3-29 1/3 80 160 338/292 333/289 0.0012/0.0012 203/203 358/342 0.0044/0.0043 0.70/0.68
I-N3-35 1/3 75 150 320/275 316/272 0.0012/0.0012 190/190 348/335 0.0040/0.0039 0.63/0.62

113
Table 4.11(b): Equivalent fully-reversed (R = -1) local stresses and strains obtained with Neuber and SED methods with Kf as a
function of Sa and calculations of the Fatemi-Socie (with k = 4) and SWT mean stress parameters for notched
mean stress data.

σar σar
R Sa Sm (MPa) (MPa) εa σa (MPa) σm (MPa) FS SWT
Spec ID Kf
ratio (MPa) (MPa) SWT/FS SWT/FS Neuber/SED Neuber/SED Neuber/SED Neuber/SED Neuber/SED
Neuber SED
I-N-32 -∞ 525 -525 1.34 287/395 292/391 0.0054/0.0048 623/598 -491/-454 0.010/0.009 0.72/0.69
I-N-33 -∞ 525 -525 1.34 287/395 292/391 0.0054/0.0048 623/598 -491/-454 0.010/0.009 0.72/0.69
I-N3-48 -7 600 -450 1.27 392/467 384/454 0.0070/0.0058 667/636 -436/-404 0.017/0.014 1.62/1.35
I-N3-31 -7 500 -375 1.37 417/468 413/460 0.0051/0.0045 611/587 -326/-298 0013/0.011 1.44/1.32
I-N3-49 -7 400 -300 1.50 431/465
\ 431/460 0.0040/0.0038 559/543 -227/-201 0.011/0.010 1.34/1.30
I-N3-37 -7 300 -225 1.69 387/417
54 393/419 0.0032/0.0032 493/485 -190/-166 0.0081/0.0081 0.98/1.01
I-N3-44 -7 270 -203 1.77 359/392
54
4 368/396 0.0030/0.0030 468/462 -193/-169 0.0072/0.0073 0.83/0.87
I-N3-39 -3 500 -250 1.37 499/526 488/512 0.0051/0.0045 611/587 -203/-182 0.015/0.0014 2.07/1.84
I-N3-41 -3 375 -188 1.54 469/488 465/481 0.0038/0.0036 545/530 -141/-123 0.011/0.011 1.54/1.48
I-N3-20 -3 275 -138 1.75 393/415 397/417 0.0030/0.0030 472/466 -146/-128 0.0079/0.0079 0.99/1.01
I-N3-36 -3 210 -105 1.96 321/349 328/354 0.0026/0.0025 409/407 -158/-142 0.0058/0.0059 0.64/0.67
I-N3-45 -3 190 -95 2.04 297/327 304/331 0.0024/0.0024 387/385 -159/-146 0.0052/0.0053 0.55/0.57
I-N3-50 0 280 280 1.74 582/559 564/543 0.0031/0.0030 477/470 235/206 0.013/0.012 2.19/2.05
I-N3-21 0 210 210 1.96 517/491 505/481 0.0026/0.0025 409/407 245/218 0.010/0.010 1.67/1.59
I-N3-18 0 150 150 2.25 448/418 440/412 0.0021/0.0021 337/337 259/238 0.0076/0.0074 1.24/1.20
I-N3-24 0 100 100 2.66 369/337 365/334
58469 0.0016/0.0016 266/266 245/235 0.0054/0.0054 0.84/0.82
I-N3-17 0 100 100 2.66 369/337 365/334 0.0016/0.0016 266/266 245/235 0.0054/0.0054 0.84/0.82
I-N3-34 0 85 85 2.85 338/306 336/305 0.0015/0.0015 242/242 232/325 0.0047/0.0047 0.71/0.70
I-N3-27 1/3 240 480 1.86 610/573 590/557 0.0028/0.0028 440/437 404/362 0.013/0.012 2.35/2.20
I-N3-40 1/3 180 360 2.09 526/490 512/478 0.0023/0.0023 375/374 364/327 0.010/0.009 1.72/1.63
I-N3-46 1/3 130 260 2.39 461/420 450/412 0.0019/0.0019 310/310 374/342 0.0076/0.0074 1.31/1.25
I-N3-43 1/3 85 170 2.85 387/342 379/336 0.0015/0.0015 242/242 376/351 0.0056/0.0054 0.92/0.89
I-N3-29 1/3 80 160 2.91 376/332 369/326 0.0014/0.0014 233/233 374/351 0.0053/0.0052 0.87/0.84
I-N3-35 1/3 75 150 3.00 366/321 359/316 0.0014/0.0014 225/225 370/349 0.0050/0.0049 0.83/0.80

114
 1200

Engineering Stress (MPa) 1000 x

800

600

400

200 I-S3-21 Compression

I-T3-03 Tension
I-T3-02 Tension
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
Engineering Strain (%)

Figure 4-1: Monotonic tension and compression curves for 120-90-02 ductile cast iron.

2000
I-T3-03
I-T3-02
True Stress (MPa)

Top to Bottom:
σ =1776(εp) 0.1692
σ =1614(εp) 0.1550

200
0.1 1.0 10.0
True Plastic Strain (%)

Figure 4-2: Plot of true stress vs. true plastic strain for determining monotonic K and n
values.

115
 1
0.9
700 MPa
Strain Amplitude (%) 0.8
0.7
0.6
600 MPa
0.5

400 MPa
350 MPa
300 MPa

0.4 500 MPa

0.3
0.2
0.1
0
0 500 1000 1500 2000
Applied Cycles (N)

Figure 4-3: Strain amplitude verse applied number of cycles during load-controlled
incremental step test showing when cyclic stabilization occurs.

800

600
Engineering Stress (MPa)

400

200

0
700 MPa
-200 600 MPa
500 MPa
-400
400 MPa
-600 350 MPa
-800
-0.75 -0.25 0.25 0.75 1.25
Engineering Strain (%)

Figure 4-4: Composite plot of hysteresis loops from the incremental step test.

116
1000

Stress Amplitude (MPa)

0.110
∆𝜎 ∆𝜀𝑝
= 1266
2 2
100
0.01 0.10 1.00
Plastic Strain Amplitude (%)

Figure 4-5: Plot of engineering stress amplitude vs. plastic strain amplitude (calculated
from E′) used to determine K′ and n′.

800

700

600
True Stress (MPa)

500

400

300

200 Monotonic Tension Curve

Cyclic Curve
100
Cyclic Data
0
0.0 0.5 1.0 1.5
Engineering Strain (%)

Figure 4-6: Composite plot of cyclic and monotonic true stress-strain curves and data.

117
1000
Smooth Fatigue Data
Notched Fatigue Data
(5) Smooth Best Fit
(5)
Notched Best Fit
(6) (5)
(4)
Sa (MPa)

(6)
(4)
(6)

Top to Bottom:
Sa =1608(Nf) -0.120
Sa =2129(Nf) -0.204 (2)
100
103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07
Cycles to Failure (Nf)

Figure 4-7: S-N curves of fully-reversed smooth and notched data of 120-90-02 ductile
cast iron.

800

700

600
True Stress (MPa)

500

400

300

200
Cyclic Curve (input)
100 Uniaxial (FEA output)
von Mises (FEA output)
0
0.0 0.5 1.0 1.5
True Strain (%)

Figure 4-8: Input cyclic stress-strain curve for FEA analysis with comparison of uniaxial
and von Mises output curves showing the effect of a multiaxial stress state.

118
1000
Smooth Best Fit
Notched Best Fit
Notched Prediction
Smooth Data
Notched Data
Sa (MPa)

100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Cycles to Failure (Nf)

Figure 4-9: Prediction of notched behavior and comparison with experimental data using
the nominal S-N approach.

1.00%
1.00
∆𝜀 1748 −0.12 −1.06
= 2𝑁𝑓 + 14.2 2𝑁𝑓
Engineering Strain Amplitude (%)

2 162000

0.10%
0.10

Total Strain
Elastic Strain
Plastic Strain
0.01
0.01%
103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07 108
1.E+08
Reversals to Failure (2Nf)

Figure 4-10: Strain-life curves and data including elastic, plastic, and total strain of
smooth specimens.

119
 800
Neuber
700 SED
600 FEA
σa (MPa)

500

400

300

200

100

0
0 100 200 300 400 500
Sa (MPa)
(a)
1.2
Neuber
SED
1.0
FEA

0.8
εa (%)

0.6

0.4

0.2

0.0
0 100 200 300 400 500
Sa (MPa)
(b)

Figure 4-11: Plots of local (a) stress and (b) strain amplitudes vs. nominal stress
amplitude for Neuber’s rule, SED rule, and FEA results.

120
 1.0
1.00%
Strain-Life Curve (smooth)

Engineering Strain Amplitude (%)

FEA Prediction (notched)
Neuber Prediciton (notched)
SED Prediction (notched)

0.1 3
0.10% 4 5 6 7 8
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 10
1.E+08
Reversals to Failure (2Nf)
(a)

107
1.E+07
Neuber
106 FEA
1.E+06
SED
Predicted Life (Nf)

105
1.E+05

104
1.E+04

103
1.E+03

102 2
1.E+02 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(b)

121
 1.0
1.00%
Strain-Life Curve (smooth)

Engineering Strain Amplitude (%)

FEA Prediction (notched)
Neuber Prediciton (notched)
SED Prediction (notched)

0.10%
0.1 3 4 5 6 7 8
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 10
1.E+08
Reversals to Failure (2Nf)
(c)

107
1.E+07
Neuber
106
1.E+06 FEA
SED
Predicted Life (Nf)

105
1.E+05

104
1.E+04

103
1.E+03

102 2
1.E+02 3 4 5 6 7
1.E+02
10 1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10
Experimental Life (Nf)
(d)

Figure 4-12: Strain-life predictions and scatter plots of fully-reversed (R = -1) notched
fatigue behavior using the local strain approach with Neuber’s rule, SED
rule, and finite element analysis using Kf = 2.54 (a, b), and Kf as a function
of Sa (c, d).

122
 0.100
R = -1 Notched Data
R = -1 Smooth Data
Fatemi-Socie Parameter
FS Parameter

0.010

0.001 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 108
1.E+08
Reversals to Failure (2Nf)
(a)

10.0
R = -1 Notched Data
R = -1 Smooth Data
SWT Parameter
2 𝑚𝑎𝑥

1.0
𝜎
∆𝜀

0.1 3 4 5 6 7
10 10 10 10 10 108
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
Reversals to Failure (2Nf)
(b)

123
 0.1
R = -1 Notched Data
R = -1 Smooth Data
Fatemi-Socie Parameter
FS Parameter

0.01

0.001 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 108
1.E+08
Reversals to Failure (2Nf)
(c)

10.0
R = -1 Notched Data
R = -1 Smooth Data
SWT Parameter
2 𝑚𝑎𝑥

1.0
𝜎
∆𝜀

0.1
3 4 5 6 7
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 108
1.E+08
Reversals to Failure (2Nf)
(d)

Figure 4-13: Prediction of fully-reversed (R = -1) notched data from the SED rule using
the strain-based version of (a) Fatemi-Socie with k = 4 and Kf = 2.54, (b)
SWT with Kf = 2.54, (c) Fatemi-Socie with k = 4 and Kf varies with Sa, and
(d) SWT with Kf as a function of Sa.

124
 0.100
R = -1 Notched Data
R = -1 Smooth Data
Fatemi-Socie Parameter
FS Parameter

0.010

0.001 3 4 5 6 7
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 108
1.E+08
Reversals to Failure (2Nf)
(a)
10.0
R = -1 Notched Data
R = -1 Smooth Data
SWT Parameter
2 𝑚𝑎𝑥
𝜎

1.0
∆𝜀

0.1
3 4 5 6 7
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 108
1.E+08
Reversals to Failure (2Nf)
(b)

125
 0.100
R = -1 Notched Data
R = -1 Smooth Data
Fatemi-Socie Parameter
FS Parameter

0.010

0.001 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 108
1.E+08
Reversals to Failure (2Nf)

(c)
10.0
R = -1 Notched Data
R = -1 Smooth Data
SWT Parameter
2 𝑚𝑎𝑥

1.0
𝜎
∆𝜀

0.1 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 108
1.E+08
Reversals to Failure (2Nf)
(d)

Figure 4-14: Prediction of fully-reversed (R = -1) notched data from Neuber’s rule using
the strain-based version of (a) Fatemi-Socie with k = 4 and Kf = 2.54,
(b) SWT with Kf = 2.54, (c) Fatemi-Socie with k = 4 and Kf varies with Sa,
and (d) SWT with Kf as a function of Sa.

126
1000

Stress Amplitude (MPa)

S-N Curve (smooth)

Neuber Prediction (notched)
SED Prediction (notched)
FEA Prediction (notched)
100 3 4 5 6 7
1,000
10 10,000
10 100,000
10 1,000,000
10 10,000,000
10
Cycles to Failure (Nf)
(a)
7
10
10000000
Neuber
FEA
6
1000000
10 SED
Predicted Life (Nf)

5
10
100000

4
10
10000

3
10
1000

2
10
100 2 3 4 5 6 7
10
100 10
1000 10
10000 10
100000 10
1000000 10
10000000
Experimental Life (Nf)
(b)

127
1000

Stress Amplitude (MPa)

S-N Curve (smooth)

Neuber Prediction (notched)
SED Prediction (notched)
FEA Prediction
100 3 4 5 6 7
10
1,000 10
10,000 10
100,000 10
1,000,000 10
10,000,000
Cycles to Failure (Nf)
(c)
7
10000000
10
Neuber
6 FEA
1000000
10
SED
Predicted Life (Nf)

5
100000
10

4
10000
10

3
10
1000

2
10
100
2 3 4 5 6 7
100
10 1000
10 10000
10 100000
10 1000000
10 10000000
10
Experimental Life (Nf)
(d)

Figure 4-15: Stress-life predictions and scatter plots of fully-reversed (R = -1) notched
fatigue behavior using the local stress approach with Neuber’s rule, SED
rule, and finite element analysis using Kf = 2.54 (a, b), and Kf as a function
of Sa (c, d).

128
 Fully-Reversed Smooth Fatigue Curve

Nominal
S-N Approach Local Approach

Use cyclic material stress-strain curve and

Obtain Kt. Neuber’s rule, SED, or FEA for a given nominal
stress to obtain local stresses (σa) and strains (εa).

Calculate Kf by Neuber
or Peterson rules.

Use the local stress or strain in smooth specimen

S-N or strain-life curve to find fatigue life.
Assume the same fatigue strength as smooth
at cycle 1. Divide smooth Sf by Kf at 106
cycles. Draw a line connecting the two points
to generate notched S-N curve.

For a given nominal stress, find fatigue life of

notched specimen from notched S-N curve.

Figure 4-16: Flow chart to estimate notched specimen behavior.

129
1000
R = -7 R = -3
R = -1 R=0
R = 1/3 R = 0.5
R = 0.75

(2)
Sa (MPa)

(4)
Top to Bottom: (2)
Sa =1071(Nf) -0.054
Sa =2173(Nf) -0.125
Sa =1608(Nf) -0.120
Sa =1256(Nf) -0.130
Sa =986(Nf) -0.123
Sa =850(Nf) -0.124 (2)
100
10 3 10 4 10 5 106 107
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
(a) Cycles to Failure (Nf)

700
R = -∞ R = -7
R = -3 R = -1
R=0 R = 1/3

280
Sa (MPa)

210
Top to Bottom:
140 Sa = 2387(Nf)-0.157
Sa = 2536(Nf)-0.188
Sa = 2129(Nf)-0.204 (2)
Sa = 1841(Nf)-0.223
Sa = 1430(Nf)-0.212
70
10 3 10 4 10 5 6
10 107
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
(b) Cycles to Failure (Nf)

Figure 4-17: Best-fit lines and data at different R-ratios for (a) smooth and (b) notched
specimens.

130
 700

Sa (MPa)

R = -1 Smooth
R = -7 Notched
R = -3 Notched
R = -1 Notched
R = 0 Notched
R = 1/3 Notched
70 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Cycles to Failure (Nf)

Figure 4-18: Prediction of notched fatigue behavior for different R-ratios from fully-
reversed, smooth specimen fatigue behavior using Modified Goodman
showing predicted S-N lines and experimental notched data.

131
 1000 1000
Sar (MPa)

Sar (MPa)
R = -1 R = -7 R = -1 R = -7
R = -3 R=0 R = -3 R=0
R = 1/3 R = 0.5 R = 1/3 R = 0.5
R = 0.75 R = 0.75
100 100
103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07 103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07
Cycles to Failure (Nf) Cycles to Failure (Nf)
(a) (b)
1000 1000
Sar (MPa)

R = -1 R = -7 Sar (MPa) R = -1 R = -7
R = -3 R=0 R = -3 R=0
R = 1/3 R = 0.5 R = 1/3 R = 0.5
R = 0.75 R = 0.75
100 3 4 5 6 7 100
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07
Cycles to Failure (Nf) Cycles to Failure (Nf)
(c) (d)
Figure 4-19: Correlations of experimental data at different R-ratios for smooth specimens using (a) Modified Goodman, (b) SWT, (c)
FKM, and (d) Fatemi-Socie (with k = 4) mean stress models.

132
2000

FS Parameter

R = -1 R = -7
R = -3 R=0
R = 1/3 R = 0.5
R = 0.75
200 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Cycles to Failure (Nf)

Figure 4-20: Correlation of smooth specimen mean stress data with the stress-based
version of the Fatemi-Socie parameter (with k = 4) corresponding to the
equivalent stress approach in Figure 4-19 (d).

133
 500
500 500
500

Sar (MPa)
Sar (MPa)

150 150

100 100
R = -1 R = -∞ R = -1 R = -7
R = -7 R = -3 R = -3 R=0
R=0 R = 1/3 R = 1/3
50
50 50
50
103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07 103
1.E+03 104
1.E+04
5
10
1.E+05
6
10
1.E+06
7
10
1.E+07
Cycles to Failure (Nf) Cycles to Failure (Nf)
(a) (b)
500
500 500
500
Sar (MPa)

Sar (MPa)
150 150

100 100
R = -1 R = -∞ R = -1 R = -∞
R = -7 R = -3 R = -7 R = -3
R=0 R = 1/3 R=0 R = 1/3
50
50 50
50
103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07 103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06 107
1.E+07
Cycles to Failure (Nf) Cycles to Failure (Nf)
(c) (d)
Figure 4-21: Correlations of experimental data at different R-ratios for notched specimens using (a) Modified Goodman, (b) SWT, (c)
FKM, and (d) Fatemi-Socie (with k = 4) mean stress models.

134
1000

FS Parameter

R = -1 R = -∞
R = -7 R = -3
R=0 R = 1/3
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Cycles to Failure (Nf)

Figure 4-22: Correlation of notched specimen mean stress data with the stress-based
version of the Fatemi-Socie Parameter (with k = 4) corresponding to the
equivalent stress approach in Figure 4-21(d).

135
 7
1000000010
R Smooth Notched
-∞
6 -7
100000010 -3
-1

Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75

4
1000010

3
100010 Factor % data
±3 69
±5 80
2 ±10 86
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(a)
10000000107
R Smooth Notched
-7
6 -3
100000010 -1
0
Predicted Life (Nf)

1/3
5 0.5
10000010 0.75

4
1000010

3 Factor % data
100010
±3 90
±5 91
±10 91
100102 2 3 4 5 6 7
1.E+02
10 1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10
Experimental Life (Nf)
(b)

136
 10000000107
R Smooth Notched
-∞
6 -7
100000010 -3
-1

Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75

4
1000010

3 Factor % data
100010 ±3 89
±5 94
±10 96
2
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)

(c)
10000000107
R Smooth Notched
-∞
-7
6
100000010 -3
-1
Predicted Life (Nf)

0
1/3
100000105 0.5
0.75

4
1000010

3 Factor % data
100010 ±3 77
±5 90
±10 99
2
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(d)

Figure 4-23: Predicted versus experimental fatigue lives at different R-ratios for smooth
and notched specimens by using the nominal stress-based approach and (a)
Modified Goodman, (b) SWT, (c) FKM, and (d) Fatemi-Socie (with k = 4)
mean stress models.

137
 800 -3 800
-7 0 -7 -3
0
600 600
1/3
Sa (MPa)

Sa (MPa)
1/3
400 400
0.5
Nf = 103
1.E+03 Nf = 103
1.E+03 0.5
200 Nf = 104
1.E+04 200 Nf = 104
1.E+04
Nf = 105
1.E+05 0.75 Nf = 105
1.E+05
f ≥ 10
6
f ≥ 10 ≥N1000000
6
≥N1000000 0.75
0 0
-1000 -500 0 500 1000 -1000 -500 0 500 1000
Sm (MPa) Sm (MPa)
(a) (b)
800 800
-3 0 -7 -3
-7 0
600 600

Sa (MPa)
Sa (MPa)

1/3 1/3
400 400
0.5 0.5
Nf = 103
1.E+03 1.E+03
Nf = 103
200 Nf = 104
1.E+04 200 Nf = 104
1.E+04
Nf = 105
1.E+05 Nf = 105
1.E+05
0.75
f ≥ 10
≥N1000000 ≥ f ≥ 10
N1000000
6 6 0.75
0 0
-1000 -500 0 500 1000 -500-1000 0 500 1000
Sm (MPa) Sm (MPa)
(c) (d)
Figure 4-24: Constant life diagrams and experimental data of smooth specimens with predictions from (a) Modified Goodman, (b)
SWT, (c) FKM, and (d) Fatemi-Socie mean stress models.

138
 800 800
-7 -3 0 -7 -3 0

600 600
Sa (MPa)

1/3

Sa (MPa)
1/3
400 400
MG 0.5
0.5 MG
SWT SWT
200 FS 0.75 200 FS
FKM 0.75
FKM
Data Data
0 0
-1000 -500 0 500 1000 -1000 -500 0 500 1000
Sm (MPa) Sm (MPa)
(a) (b)
800 800
-7 -3 0 -7 -3 0

600 600
Sa (MPa)

Sa (MPa)
1/3 1/3
400 400
MG 0.5 MG 0.5
SWT SWT
200 FS 200 FS
FKM 0.75
FKM
Data Data 0.75
0 0
-1000 -500 0 500 1000 -1000
-500 0 500 1000
Sm (MPa) Sm (MPa)
(c) (d)
Figure 4-25: Constant life diagrams and experimental data of smooth specimens comparing all mean stress models for lives of (a) 103,
(b) 104, (c) 105, and (d) ≥ 106 cycles.

139
 15% 50%
R=0 R=0
10% 40% R = 1/3
R = 1/3
Percent Error

R = 0.5

Percent Error
5% 30%
R = -3
0% 20%
R = -7
-5% 10%
-10% 0%
-15% -10%
-20% -20%
Modified SWT FKM FS Modified SWT FKM FS
(a) Goodman (b) Goodman
60%
R=0
50% R = 1/3
40% R = 0.5
Percent Error

30% R = 0.75
R = -3
20%
R = -7
10%
0%
-10%
-20%
-30%
Modified SWT FKM FS
(c) Goodman

Figure 4-26: Percentage errors between experimental and predicted stress amplitude based on Modified Goodman, SWT, FKM, and
Fatemi-Socie (with k = 4) mean stress parameters for smooth specimen fatigue lives of (a) 104, (b) 105, and (c) ≥106
cycles.

140
 800 800
-∞ -3 Nf = 103
1.E+03 -∞ -7 -3
-7
Nf = 104
1.E+04 0
600 Nf = 105
1.E+05 600
Nf ≥ 106
1.E+06
Sa (MPa)

Sa (MPa)
1/3
0
400 1/3 400
1.E+03
Nf = 10
3

200 1.E+04 4
200 Nf = 10
N 5
f = 10
1.E+05
f ≥ 10
N 6
1.E+06
0 0
-1000 -500 0 500 1000 -1000 -500 0 500 1000
Sm (MPa) Sm (MPa)
(a) (b)
800 800 3
-∞ -3 1.E+03
N f = 10
3
-∞ -3 N f = 10
1.E+03
-7 -7 N 4
f = 10
1.E+04
N 4
f = 10 1.E+04
N 5
f = 10
5
600 N f = 10
1.E+05 600 1.E+05
f ≥ 10 f ≥ 10
6 N 6
N
1.E+06 1.E+06
Sa (MPa)

Sa (MPa)
0 0
400 1/3 400 1/3

200 200

0 0
-1000 -5000 500 1000 -1000 -500 0 500 1000
Sm (MPa) Sm (MPa)
(c) (d)
Figure 4-27: Constant life diagrams and experimental data of notched specimens with predictions from (a) Modified Goodman,
(b) SWT, (c) FKM, and (d) Fatemi-Socie (with k = 4) mean stress models.

141
 800 800
-∞ -∞ -7 -3 0
-7 0
600 -3 600

Sa (MPa)
Sa (MPa)

1/3 1/3
400 400
MG MG
FKM FKM
200 SWT 200 SWT
FS FS
Data Data
0 0
-1000 -500 0 500 1000 -1000 -500 0 500 1000
Sm (MPa) Sm (MPa)
(a) (b)
800 800
-∞ -7 -3 MG -∞ -7 -3 MG
FKM FKM
600 SWT 600 SWT
FS FS
Sa (MPa)

Sa (MPa)
Data 0 Data
0
400 400
1/3 1/3

200 200

0 0
-1000 -500 0 500 1000 -500 -1000
0 500 1000
Sm (MPa) Sm (MPa)
(c) (d)
Figure 4-28: Constant life diagrams and experimental data of notched specimens comparing all mean stress models for lives of
(a) 103, (b) 104, (c) 105, and (d) ≥ 106 cycles.

142
 10% 15%
R=0 R = 1/3
5% 10%
R = -3 R = -7
5%
Percent Error

Percent Error
0%
0%
-5%
-5%
-10%
R=0 -10%
-15% -15%
R = 1/3
-20% R = -3 -20%
-25% -25%
Modified SWT FKM FS Modified SWT FKM FS
(a) Goodman (b) Goodman
100% 50%
R=0 R = 1/3 R=0 R = 1/3
80% R = -3 R = -7
R = -3 R = -7 30%
60%

Percent Error
Percent Error

R = -∞
40% 10%
20%
0% -10%
-20%
-30%
-40%
-60% -50%
Modified SWT FKM FS Modified SWT FKM FS
(c) Goodman (d) Goodman

Figure 4-29: Percentage errors between experimental and predicted stress amplitudes based on Modified Goodman, SWT, FKM, and
Fatemi-Socie (with k = 4) mean stress parameters for notched specimen fatigue lives of (a) 103, (b) 104, (c) 105 and (d)
≥106 cycles.

143
1000

σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(a)

1000
σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10
Experimental Life (Nf)
(b)

144
1000

σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(c)

1000
σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(d)

Figure 4-30: Correlations of all experimental data for smooth and notched specimens
using the SWT mean stress parameter with equivalent, fully-reversed notch
stress amplitudes obtained from (a) Neuber’s rule (with Kf = 2.54), (b) the
SED rule with (Kf = 2.54), (c) Neuber’s rule (with Kf as a function of Sa),
and (d) SED rule (with Kf as a function of Sa).

145
 7
1000000010
R Smooth Notched
-∞
6 -7
100000010 -3

Predicted Life (Nf)

-1
0
5 1/3
10000010 0.5
0.75

10000104

Factor % data
1000103 ±3 43
±5 66
±10 86
100 2
101.E+02
2 1.E+03
3 1.E+04
4 1.E+05
5 1.E+06
6 1.E+07
7
10 10 10 10 10 10
Experimental Life (Nf)
(a)

10000000107
R Smooth Notched
-∞
6 -7
100000010 -3
-1
Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75

4
1000010

3
100010 Factor % data
±3 56
±5 81
2 ±10 87
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(b)

146
 7
1000000010
R Smooth Notched
-∞
6 -7
100000010 -3
-1
Predicted Life (Nf) 0
5 1/3
10000010 0.5
0.75

4
1000010

3
Factor % data
100010 ±3 80
±5 87
2 ±10 93
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(c)
10000000107
R Smooth Notched
-∞
-7
1000000106 -3
Predicted Life (Nf)

-1
0
1/3
100000105 0.5
0.75

10000 4
10

Factor % data
1000 3
10 ±3 81
±5 89
±10 93
100 2 2 3 4 5 6 7
101.E+02
10 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(d)

Figure 4-31: Predicted vs. experimental fatigue life of all experimental data using the
SWT mean stress parameter with local stress obtained for notched
specimens from (a) Neuber’s rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c)
Neuber’s rule (Kf varies with Sa), and (d) SED rule (Kf varies with Sa).

147
 1000

σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(a)
1000
σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(b)

148
 1000

σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(c)
1000
σar (MPa)

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
100 3 4 5 6 7
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(d)

Figure 4-32: Correlations of all experimental data for smooth and notched specimens
using the Fatemi-Socie mean stress parameter with equivalent, fully-
reversed notch stress amplitudes obtained from (a) Neuber’s rule (Kf =
2.54), (b) the SED rule (Kf = 2.54), (c) Neuber’s rule (Kf varies with Sa), and
(d) the SED rule (Kf varies with Sa).

149
 7
1000000010
R Smooth Notched
-∞
6 -7
100000010 -3
-1

Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75

4
1000010

3
Factor % data
100010 ±3 43
±5 66
2 ±10 87
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(a)
7
1000000010
R Smooth Notched
-∞
6 -7
100000010 -3
-1
Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75
4
1000010

3 Factor % data
100010 ±3 57
±5 79
2 ±10 89
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(b)

150
 10000000107
R Smooth Notched
-∞
6 -7
100000010 -3
-1

Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75

4
1000010

3
Factor % data
100010 ±3 83
±5 93
2 ±10 99
10010 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(c)
7
1000000010
R Smooth Notched
-∞
6 -7
100000010 -3
-1
Predicted Life (Nf)

0
5 1/3
10000010 0.5
0.75
4
1000010

3 Factor % data
100010 ±3 84
±5 91
±10 99
2
10010 2 3 4 5 6 7
1.E+02
10 1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10
Experimental Life (Nf)
(d)

Figure 4-33: Predicted vs. experimental fatigue life of all experimental data using the
Fatemi-Socie mean stress parameter with local stress obtained for notched
specimens from (a) Neuber’s rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c)
Neuber’s rule (Kf varies with Sa), and (d) SED rule (Kf varies with Sa).

151
 10
10.000

2 𝑚𝑎𝑥

1.0
1.000
𝜎

R Smooth Notched
∆𝜀

-∞
-7
-3
-1
0
1/3
0.5
0.75
0.10 3
0.100 4 5 6 7 8
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 10
1.E+08
Reversals to Failure (2Nf)
(a)

10
10
2 𝑚𝑎𝑥
𝜎

1.0
1
∆𝜀

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
0.10
0 3 4 5 6 7 8
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 1.E+08
10
Reversals to Failure (2Nf)
(b)

152
 10
10.000
2 𝑚𝑎𝑥

1.0
𝜎

1.000 R Smooth Notched

∆𝜀

-∞
-7
-3
-1
0
1/3
0.5
0.75
0.10
0.100 3 4 5 6 7 8
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 1.E+08
10
Reversals to Failure (2Nf)
(c)

10.000
10
2 𝑚𝑎𝑥
𝜎

1.000
1.0
∆𝜀

R Smooth Notched
-∞
-7
-3
-1
0
1/3
0.5
0.75
0.10
0.100 3 4 5 6 7 8
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 10
1.E+08
Reversals to Failure (2Nf)
(d)

Figure 4-34: Smith-Watson-Topper parameter plots of smooth and notched mean stress
data with notch stress and strain values obtained from (a) Neuber’s rule (Kf
= 2.54), (b) SED rule (Kf = 2.54), (c) Neuber’s rule (Kf varies with Sa), and
(d) SED rule (Kf varies with Sa).

153
 7
10
1.E+07
R Smooth Notched
-∞
6 -7
10
1.E+06 -3
-1
Predicted Life (Nf)
0
5 1/3
10
1.E+05 0.5
0.75
4
10
1.E+04

3
10
1.E+03 Factor % data
±3 49
±5 61
2 ±10 84
10 2
1.E+02 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(a)
7
1.E+07
10
R Smooth Notched
-∞
6
10 -7
1.E+06 -3
Predicted Life (Nf)

-1
0
5
10
1.E+05 1/3
0.5
0.75
4
10
1.E+04

3
10
1.E+03 Factor % data
±3 59
±5 76
2 ±10 84
1.E+02
10 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(b)

154
 7
10
1.E+07
R Smooth Notched
-∞
6 -7
10
1.E+06 -3
-1
Predicted Life (Nf) 5
0
1/3
10
1.E+05 0.5
0.75
4
10
1.E+04

3
Factor % data
10
1.E+03 ±3 87
±5 90
±10 99
2
1.E+02
10 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(c)
7
10
1.E+07
R Smooth Notched
-∞
6
10 -7
1.E+06 -3
-1
Predicted Life (Nf)

0
5
10 1/3
1.E+05 0.5
0.75
4
10
1.E+04

3 Factor % data
10
1.E+03 ±3 87
±5 91
2
±10 99
1.E+02
10 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(d)

Figure 4-35: Predicted verse experimental fatigue lives at different R-ratios for smooth
and notched specimens using the strain-based version of SWT parameter
for notched specimens incorporating (a) Neuber’s rule (Kf = 2.54), (b) SED
rule (Kf = 2.54), (c) Neuber’s rule (Kf varies with Sa), and (d) SED rule (Kf
varies with Sa).

155
 0.100

FS Parameter

0.010 R Smooth Notched

-∞
-7
-3
-1
0
1/3
0.5
0.75
0.001 3 4 5 6 7 8
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 1.E+08
10
Reversals to Failure (2Nf)
(a)

0.100
FS Parameter

0.010 R Smooth Notched

-∞
-7
-3
-1
0
1/3
0.5
0.75
0.001 3 4 5 6 7 8
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 10
1.E+08
Reversals to Failure (2Nf)
(b)

156
 0.100

FS Parameter

0.010 R Smooth Notched

-∞
-7
-3
-1
0
1/3
0.5
0.75
0.001
3 4 5 6 7 8
1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10 1.E+08
10
Reversals to Failure (2Nf)
(c)

0.100
FS Parameter

0.010 R Smooth Notched

-∞
-7
-3
-1
0
1/3
0.5
0.75
0.001 3 4 5 6 7 8
10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07 10
1.E+08
Reversals to Failure (2Nf)
(d)

Figure 4-36: Fatemi-Socie parameter plots with k = 4 for smooth and notched mean
stress data with notch stress and strain values obtained from (a) Neuber’s
rule (Kf = 2.54), (b) SED rule (Kf = 2.54), (c) Neuber’s rule (Kf varies with
Sa) and (d) SED rule (Kf varies with Sa).

157
 7
10
1.E+07
R Smooth Notched
-∞
6 -7
10
1.E+06 -3
-1
Predicted Life (Nf) 0
5 1/3
10
1.E+05 0.5
0.75
4
10
1.E+04

3 Factor % data
10
1.E+03 ±3 46
±5 63
2
±10 84
10 2
1.E+02 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(a)
7
1.E+07
10
R Smooth Notched
-∞
6 -7
10
1.E+06 -3
-1
Predicted Life (Nf)

0
5 1/3
1.E+05
10 0.5
0.75
4
10
1.E+04

3
10 Factor % data
1.E+03 ±3 54
±5 76
2 ±10 86
1.E+02
10 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(b)

158
 7
10
1.E+07
R Smooth Notched
-∞
6 -7
10
1.E+06 -3
-1
Predicted Life (Nf) 5
0
1/3
1.E+05
10 0.5
0.75
4
1.E+04
10

3 Factor % data
1.E+03
10 ±3 81
±5 89
±10 99
2
1.E+02
10 2 3 4 5 6 7
10
1.E+02 10
1.E+03 10
1.E+04 10
1.E+05 10
1.E+06 10
1.E+07
Experimental Life (Nf)
(c)
7
1.E+07
10
R Smooth Notched
-∞
6 -7
10
1.E+06 -3
-1
Predicted Life (Nf)

0
5 1/3
1.E+05
10 0.5
0.75
4
1.E+04
10

3 Factor % data
1.E+03
10 ±3 81
±5 90
±10 99
2
1.E+02
10 2 3 4 5 6 7
1.E+02
10 1.E+03
10 1.E+04
10 1.E+05
10 1.E+06
10 1.E+07
10
Experimental Life (Nf)
(d)

Figure 4-37: Predicted verse experimental fatigue lives at different R-ratios for smooth
and notched specimens using the strain-based version of the Fatemi-Socie
parameter for notched specimens incorporating (a) Neuber rule (Kf = 2.54),
(b) SED rule (Kf = 2.54), (c) Neuber’s rule (Kf varies with Sa), and (d) SED
rule (Kf varies with Sa).

159
 2.75

2.5

2.25
Kf

1.75

1.5

1.25
100 Stress Amplitude (MPa) 1000
(a)

2.6
Nf = 103
1.E+03
2.4 Nf = 104
1.E+04
Nf = 105
1.E+05
2.2
1.E+06
Nf = 106
2.0
Kf

1.8

1.6

1.4

1.2

1.0
-8 -7 -6 -5 -4 -3 -2 -1 0 1
R-Ratio
(b)

160
 2.6

2.4

2.2 R = 1/3

2.0
R=0
Kf

1.8

1.6

1.4

1.2

1.0
103
1.E+03 104
1.E+04 105
1.E+05 106
1.E+06
Cycles to Failure (Nf)
(c)

Figure 4-38: Fatigue notch factor as a function of (a) fully-reversed (R = -1) stress
amplitude, (b) R-ratio and (c) fatigue life.

161
R

-7

600 MPa 550 MPa

39,385 237,240

-3

600 MPa 525 MPa 450 MPa

33,346 84,970 333,047

-1

600 MPa 500 MPa 400 MPa 350 MPa

3,847 23,903 104,169 294,698

425 MPa 350 MPa 300 MPa

4,649 19,289 69,221

162
 1/3

300 MPa 250 MPa 200 MPa

15,800 56,160 254,014

0.5

210 MPa 175 MPa

84,675 280,219

0.75

112.5 MPa
2,058,774

Figure 4-39: Fracture surfaces of smooth specimens for each R-ratio and stress amplitude
level tested. Stress amplitude and cycles to failure are shown below each
photo.

163
R

-∞

525 MPa
188,000

-7

600 MPa 500 MPa 400 MPa 300 MPa 270 MPa
6,598 21,175 100,929 535,623 1,976,940

-3

500 MPa 375 MPa 275 MPa 210 MPa 190 MPa
5,085 27,785 163,668 488,308 1,201,446

164
 -1

400 MPa 300 MPa 210 MPa 140 MPa 125 MPa
3,225 13,015 84,718 566,493 863,899

280 MPa 210 MPa 150 MPa 100 MPa

4,597 16,867 84,587 453,015

1/3

240 MPa 180 MPa 130 MPa 85 MPa 80 MPa

4,345 19,234 79,082 808,435 637,051
Figure 4-40: Fracture surfaces of notched specimens for each R-ratio and stress amplitude level tested. Stress amplitude and cycles to
failure are shown below each photo.

165
Chapter 5

Summary and Conclusions

This study consisted of two parts. Part one introduced a literature review on

fatigue of cast iron, which aided in acquisition of data for attempting to generate

predictive methods to determine mechanical properties based on hardness and/or

microstructural properties. Estimation of fatigue behavior based on correlations with

tensile and/or microstructural properties were also attempted.

The second part of this study focused on experimental methods to characterize

tensile and fatigue behaviors of smooth and notched specimens made of 120-90-02

ductile cast iron under fully-reversed, as well as with tensile or compressive mean stress

conditions. Nominal and local-based approaches, as well as FEA were used to predict

fatigue behavior. Fatigue behavior of smooth and notched specimens under various mean

stress conditions were analyzed based on Modified Goodman, Smith-Watson-Topper,

FKM, and Fatemi-Socie mean stress models.

Based on data analysis and discussions from the previous chapters, the following

conclusions can be drawn:

1. A correlation between ultimate strength and yield strength exists as a function of

hardness for cast iron. The yield strength, however, shows a stronger correlation

166
 to hardness than tensile strength with R2 values of 0.913 for yield strength and

0.815 for tensile strength. In both cases, microstructural parameters do not appear

to have a significant influence over the tensile and yield strength correlations.

Based on these correlations, yield and tensile strengths can be estimated knowing

only the hardness of the material up to a hardness of 450 Hv.

2. Fatigue strength of cast iron could not be correlated to one mechanical property.

The fatigue strength is also not solely dependent on microstructural parameters.

The general trend is that the fatigue strength typically increases with tensile

strength and hardness. Fatigue strength also tends to increase as the average

nodule size decreases. The maximum nodule sizes in conjunction with hardness

are the controlling factors in fatigue strength. As the maximum nodule size

decreases and hardness increases, the fatigue strength increases although, a

quantitative correlation could not be found when considering a large number of

cast iron data.

3. Both Mitchell’s method based on nodule size a defect parameter, K, and de

Kazinczy’s method based on Kt, and An were found to be suitable for fatigue limit

approximation of cast iron as about half of the data fall within a 10% scatter band.

Mitchell’s method was most accurate when the maximum nodule size and Kt = 2

were considered.

4. Finite element analysis of the notched geometry indicated the specimen was under

plane strain conditions for cyclic loading. Therefore, a multiaxial stress state was

present at the notch.

167
5. Local stress and strain amplitudes obtained from Neuber and SED rules plotted

against nominal stress amplitudes of the notched specimen and compared with

FEA results indicate that the SED rule best predicts the stresses and strains at the

notch root.

6. Both local stress and strain analysis show that prediction of notched behavior is

conservative in excess of a factor of two under the fully-reversed condition for

Neuber and FEA predictions. Modification of the SED rule with Kf = 2.54

resulted in conservative life predictions within a factor of five. Considering a

variable Kf as a function of applied stress amplitude, resulted in much more

reasonable predictions with negligible difference between Neuber and SED rules.

7. Strain-based versions of the Fatemi-Socie and SWT mean stress parameters

resulted in conservative estimates of fully-reversed notched specimen data for

both Neuber and SED rules when Kf = 2.54. When using a variable Kf, differences

between Fatemi-Socie and SWT were small and predictions of notched data were

similar to the strain-life approach.

8. The effect of mean stress on both smooth and notched specimens was similar for

almost all R-ratios, except for smooth data at R = -7. The slope of the S-N line was

very shallow compared to the other R-ratios, indicating more similar fatigue

behavior between low and high cycle fatigue.

9. By replacing Sf with Sf/Kf in the Modified Goodman equation, notched fatigue

behavior under the presence of mean stress could not be predicted from smooth

specimen fatigue data. Contrary to experimental results, predictions of notched

mean stress effects from smooth fatigue data with Modified Goodman equation

168
 show that the effect of mean stress is not similar between the selected R-ratios.

This is because the slopes of the predicted S-N lines are not parallel like the

experimental data S-N lines.

10. The SWT mean stress parameter best correlates mean stress fatigue for both

smooth and notched specimens using the nominal-based approach. SWT was able

to correlate 89% of notched and mean stress data within a factor of ±3 followed

by FKM with 89%, FS with 77% and Modified Goodman with 69%.

11. For full compression, notched testing, none of the mean stress parameters

accurately predicted the fatigue behavior for R = -∞ using the nominal-based

approach. Only the FS mean stress model was able to correlate R = -∞ within a

conservative factor of 10.

12. Local, fully-reversed equivalent stress amplitudes were obtained with Neuber and

SED rules (with Kf = 2.54) and applied to the stress version of SWT mean stress

parameters. Using Neuber’s rule, only 43% of notched and mean stress data were

correlated within a factor of ±3 and the SED rule correlated 56% of data within a

factor of ±3.

13. Local, fully-reversed equivalent stress amplitudes were obtained with Neuber and

SED rules (with Kf = 2.54) and applied to the stress version of FS mean stress

parameters. Using Neuber’s rule, only 43% of notched and mean stress data were

correlated within a factor of ±3 and the SED rule correlated 57% of data within a

factor of ±3.

14. Using a variable Kf with Neuber and SED rules in conjunction with stress-based

versions of SWT and FS resulted in accurate predictions with little difference

169
 between Neuber and SED. Correlations of notched and mean stress data resulted

in 80% (SWT with Neuber) and 81% (SWT with SED) of data falling within a

factor of ±3. Similarly for the FS model 83% (Neuber) and 84% (SED) of data lie

within a factor of ±3.

15. The strain-based versions of SWT and FS mean stress models were used with

both Neuber and SED rule. The fatigue notch factor was taken as Kf = 2.54 and

was also varied based on applied stress. Similar to the local stress approach, more

accurate predictions were made when Kf was not considered constant. SWT was

slightly better correlating notched and mean stress data compared to FS. When Kf

was varied, SWT correlated 87% of data within a factor of ±3 using Neuber and

SED rules, while FS correlated 81% of data within a factor of ±3 using Neuber

and SED rules.

16. The fatigue notch factor was most detrimental at low stress levels compared to

high stress levels for fully-reversed loading. The effect of the notch was

minimized under compressive loading and increased in severity up to R = 0.

beyond R = 0, the effect of the notch decreased most likely due to excessive local

plastic deformation.

170
References

[1] Callister, W.D., "Materials Science And Engineering: An Introduction," John

Wiley & Sons Inc, 2007.

[2] Budynas, R. and Nisbett, J.K., "Shigley's Mechanical Engineering Design,"

8th ed., McGraw-Hill Education Education, 2006.

[3] http://www.ductile.org/didata/Section7/7intro.htm, "Section VII. Heat Treatment"

[4] Stephens, R.I., Fatemi, A., Stephens, R.R., and Fuchs, H.O., "Metal Fatigue in
Engineering," 2nd ed., John Wiley & Sons, 2007.

[5] Germann, H., Starke, P., Kerscher, E., and Eifler, D., "Fatigue behaviour and
lifetime calculation of the cast irons EN-GJL-250, EN-GJS-600 and EN-GJV-
400," Procedia Engineering, Vol. 2, 2010, pp. 1087-1094.

[6] Salman, S., Fındık, F., and Topuz, P., "Effects of various austempering
temperatures on fatigue properties in ductile iron," Materials & Design, Vol. 28,
2007, pp. 2210-2214.

[7] Petrenec, M., Beran, P., Dluhoš, J., Zouhar, M., and Ševčík, M., "Analysis of
fatigue crack initiation in cycled austempered ductile cast irons," Procedia
Engineering, Vol. 2, 2010, pp. 2337-2346.

[8] Čanžar, P., Tonković, Z., and Kodvanj, J., "Microstructure influence on fatigue
behaviour of nodular cast iron," Materials Science and Engineering: A, Vol. 556,
2012, pp. 88-99.

[9] Yamabe, J. and Kobayashi, M., "Influence of casting surfaces on fatigue strength
of ductile cast iron," Fatigue and Fracture of Engineering Materials and
Structures, Vol. 29, 2006, pp. 403-415.

[10] Murakami, Y., "Metal Fatigue: Effects of Small Defects and Nonmetallic
Inclusions," Elsevier, 2002.

[11] Tayanç, M., Aztekin, K., and Bayram, A., "The effect of matrix structure on the
fatigue behavior of austempered ductile iron," Materials and Design, Vol. 28,
2007, pp. 797-803.

171
[12] Collini, L., Nicoletto, G., and Konečná, R., "Microstructure and mechanical
properties of pearlitic gray cast iron," Materials Science and Engineering: A, Vol.
488, 2008, pp. 529-539.

[13] Tokaji, K., Uematsu, Y., Horie, T., and Takahashi, Y., "Fatigue behaviour of cast
irons with spheroidal vanadium carbides dispersed within martensitic matrix
microstructure," Materials Science and Engineering: A, Vol. 418, 2006, pp. 326-
334.

[14] Lin, C.K. and Wei, J.Y., "High-cycle fatigue of austempered ductile irons in
various-sized Y-block casting," Materials Transactions, Vol. 38 1997, pp. 682-
691.

[15] Toktaş, G., Toktaş, A., and Tayanç, M., "Influence of matrix structure on the
fatigue properties of an alloyed ductile iron," Materials and Design, Vol. 29,
2008, pp. 1600-1608.

[16] Tanaka, Y., Yang, Z., and Miyamoto, K., "Evaluation of fatigue limit of
spheroidal graphite cast iron," Materials Transactions, Vol. 36, 1995, pp. 749-
749.

[17] Lin, C.K. and Lee, W.J., "Effects of highly stressed volume on fatigue strength of
austempered ductile irons," International Journal of Fatigue, Vol. 20, 1998, pp.
301-307.

[18] Nadot, Y., Mendez, J., and Ranganathan, N., "Influence of casting defects on the
fatigue limit of nodular cast iron," International Journal of Fatigue, Vol. 26,
2004, pp. 311-319.

[19] de Kazinczy, F., "Effect of small defects on the fatigue properties of medium-
strength cast steel," Journal of the Iron and Steel Institute, Vol. 208, 1970, pp.
851-855.

[20] Costa, N., Machado, N., and Silva, F.S., "A new method for prediction of nodular
cast iron fatigue limit," International Journal of Fatigue, Vol. 32, 2010, pp. 988-
995.

[21] Kogoh, S., Ogino, H., Nakagawa, T., Kobayashi, M., and Asami, K., "Fatigue
strength of spheroidal graphite cast iron at elevated temperatures under axial
load," International Journal of Fatigue, Vol. 12, 1990, pp. 199-205.

[22] Maluf, O., Milan, M.T., and Spinelli, D., "Effect of surface rolling on fatigue
behavior of a pearlitic ductile cast iron," Journal of Materials Engineering and
Performance, Vol. 13, 2004, pp. 195-199.

172
[23] Konečná, R., Kokavec, M., and Nicoletto, G., "Surface conditions and the fatigue
behavior of nodular cast iron," Procedia Engineering, Vol. 10, 2011, pp. 2538-
2543.

[24] Zammit, A., Mhaede, M., Grech, M., Abela, S., and Wagner, L., "Influence of
shot peening on the fatigue life of Cu–Ni austempered ductile iron," Materials
Science and Engineering: A, Vol. 545, 2012, pp. 78-85.

[25] Ochi, Y., Masaki, K., Matsumura, T., and Sekino, T., "Effect of shot-peening
treatment on high cycle fatigue property of ductile cast iron," International
Journal of Fatigue, Vol. 23, 2001, pp. 441-448.

[26] Asi, O., "Failure analysis of a crankshaft made from ductile cast iron,"
Engineering Failure Analysis, Vol. 13, 2006, pp. 1260-1267.

[27] Feng, H.P., Lee, S.C., Hsu, C.H., and Ho, J.M., "Study of high cycle fatigue of
PVD surface-modified austempered ductile iron," Materials Chemistry and
Physics, Vol. 59, 1999, pp. 154-161.

[28] Šamec, B., Potrč, I., and Šraml, M., "Low cycle fatigue of nodular cast iron used
for railway brake discs," Engineering Failure Analysis, Vol. 18, 2011, pp. 1424-
1434.

[29] Mitchell, M., "Review of the mechanical properties of cast steels with emphasis
on fatigue behavior and the influence of microdiscontinuities," Fracture Control
Program, University of Illinois at Urbana-Champaign, Report 15, 1977.

[30] Sendeckyj, G.P., "Constant life diagrams — a historical review," International

Journal of Fatigue, Vol. 23, 2001, pp. 347-353.

[31] Burger, R. and Lee, Y.L., "Assessment of the mean-stress sensitivity factor
method in stress-life fatigue predictions," Journal of Testing and Evaluation, Vol.
41, 2013.

[32] Dowling, N.E. "Mean stress effects in stress-life and strain-life fatigue," 2nd SAE
Brasil International Conference on Fatigue, São Paulo, Brazil. 2004.

[33] Dowling, N.E., Calhoun, C.A., and Arcari, A., "Mean stress effects in stress-life
fatigue and the Walker equation," Fatigue and Fracture of Engineering Materials
and Structures, Vol. 32, 2009, pp. 163-179.

[34] Wang, S. J., Dixon, M.W., Huey Jr, C.O., and Chen, S.C., "The Clemson limit
stress diagram for ductile parts subjected to positive mean fatigue loading,"
Journal of Mechanical Design, Vol. 122, 2000, pp. 143-146.

173
[35] Ince, A. and Glinka, G., "A modification of Morrow and Smith–Watson–Topper
mean stress correction models," Fatigue and Fracture of Engineering Materials
and Structures, Vol. 34, 2011, pp. 854-867.

[36] Wehner, T. and Fatemi, A., "Effects of mean stress on fatigue behaviour of a
hardened carbon steel," International Journal of Fatigue, Vol. 13, 1991, pp. 241-
248.

[37] Strizak, J. and Mansur, L., "The effect of mean stress on the fatigue behavior of
316 LN stainless steel in air and mercury," Journal of Nuclear Materials, Vol.
318, 2003, pp. 151-156.

[38] Woodward, A., Gunn, K., and Forrest, G. "The effect of mean stress on the
fatigue of aluminum alloys," International Conference on Fatigue of Metals,
1956.

[39] Wang, S.J., and Dixon M. W., "A new design criterion for ductile parts subjected
to non-zero mean and alternating stresses," Predictive Technology Symposium
and Exhibition, Orlando, FL, American Defense Preparedness Association,
Arlington, VA, 1993, pp. 220-234.

[40] Hanel, B., Haibach, E., Seeger, T., Wirthgen, G., and Zenner, H., "Analytical
Strength Assessment of Components in Mechanical Engineering," 5th edition,
2003.

[41] Kwofie, S., "An exponential stress function for predicting fatigue strength and life
due to mean stresses," International Journal of Fatigue, Vol. 23, 2001, pp. 829-
836.

[42] Niesłony, A., "Mean stress effect correction using constant stress ratio S-N
curves," International Journal of Fatigue, Vol. 52, 2013, pp. 49-56.

[43] Maddox, S.J., "The effect of mean stress on fatigue crack propagation. A
literature review," International Journal of Fracture, Vol. 11, 1975, pp. 389-408.

[44] Chiandussi, G. and Rossetto, M., "Evaluation of the fatigue strength of notched
specimens by the point and line methods with high stress ratios," International
Journal of Fatigue, Vol. 27, 2005, pp. 639-650.

[45] Taylor, D., "Geometrical effects in fatigue: a unifying theoretical model,"

International Journal of Fatigue, Vol. 21, 1999, pp. 413-420.

[46] Krug, T., Lang, K.H., Fett, T., and Löhe, D., "Influence of residual stresses and
mean load on the fatigue strength of case-hardened notched specimens,"
Materials Science and Engineering: A, Vol. 468, 2007, pp. 158-163.

174
[47] Vantiger, T.R., Stephens, R., and Karadag, M., "The influence of high R ratio on
notched fatigue behaviour of 1045 steel with three different heat treatments,"
International Journal of Fatigue, Vol. 24, 2002, pp. 1275-1284.

[48] Pals, T. and Stephens, R., "The influence of high R ratio on mild and sharp
notched and unnotched fatigue behavior of 1045 steel with three different heat
treatments," International Journal of Fatigue, Vol. 26, 2004, pp. 651-661.

[49] Fatemi, A., Zeng, Z., and Plaseied, A., "Fatigue behavior and life predictions of
notched specimens made of QT and forged microalloyed steels," International
Journal of Fatigue, Vol. 26, 2004, pp. 663-672.

[50] Susmel, L., Atzori, B., Meneghetti, G., and Taylor, D., "Notch and mean stress
effect in fatigue as phenomena of elasto-plastic inherent multiaxiality,"
Engineering Fracture Mechanics, Vol. 78, 2011, pp. 1628-1643.

[51] Yu, M.T., DuQuesnay, D.L., and Topper, T.H., "Notch fatigue behaviour of
SAE1045 steel," International Journal of Fatigue, Vol. 10, 1988, pp. 109-116.

[52] Fash, J. and Socie, D.F., "Fatigue behaviour and mean effects in grey cast iron,"
International Journal of Fatigue, Vol. 4, 1982, pp. 137-142.

[53] Lin, C.K. and Pai, Y.L., "Low-cycle fatigue of austempered ductile irons at
various strain ratios," International Journal of Fatigue, Vol. 21, 1999, pp. 45-54.

[54] Stokes, B., Gao, N., and Reed, P.A.S., "Effects of graphite nodules on crack
growth behaviour of austempered ductile iron," Materials Science and
Engineering: A, Vol. 445–446, 2007, pp. 374-385.

[55] Korkmaz, S., "A methodology to predict fatigue life of cast iron: uniform material
law for cast iron," Journal of Iron and Steel Research, International, Vol. 18,
2011, pp. 42-45.

[56] Collini, L., Pirondi, A., Bianchi, R., Cova, M., and Milella, P.P., "Influence of
casting defects on fatigue crack initiation and fatigue limit of ductile cast iron,"
Procedia Engineering, Vol. 10, 2011, pp. 2898-2903.

[57] Luo, J., Harding, R.A., and Bowen, P., "Evaluation of the fatigue behavior of
ductile irons with various matrix microstructures," Metallurgical and Materials
Transactions A, Vol. 33, 2002, pp. 3719-3730.

[58] Lin, C.K. and Hung, T.P., "Influence of microstructure on the fatigue properties
of austempered ductile irons—II. Low-cycle fatigue," International Journal of
Fatigue, Vol. 18, 1996, pp. 309-320.

175
[59] Lin, C.K., Lai, P.K., and Shih, T.S., "Influence of microstructure on the fatigue
properties of austempered ductile irons—I. High-cycle fatigue," International
Journal of Fatigue, Vol. 18, 1996, pp. 297-307.

[60] ASTM Standard A247-67 (Reapproved 1990), "Standard Test Method for
Evaluating the Microstructure of Graphite in Iron Castings," Annual Book of
ASTM Standards, American Society for Testing and Materials, West
Conshohocken, PA, Vol. 01.02, 1990, pp. 129-130.

[61] ASTM Standard A536-84 (Reapproved 2009), "Standard Specification for Ductile
Iron Castings," Annual Book of ASTM Standards, American Society for Testing
and Materials, West Conshohocken, PA, Vol. 01.02, 2009, pp. 1-5.

[62] ASTM Standard E83-06, "Standard Practice for Verification and Classification of
Extensometer Systems," Annual Book of ASTM Standards, American Society for
Testing and Materials, West Conshohocken, PA, Vol. 03.01, 2007 pp. 277-290.

[63] ASTM Standard E606-04, "Standard Practice for Strain-Controlled Fatigue

Testing," Annual Book of ASTM Standards, American Society for Testing and
Materials, West Conshohocken, PA, Vol. 03.01, 2007, pp. 656-671.

[64] ASTM Standard E1012-05, "Standard Practice for Verification of Specimen

Alignment Under Tensile and Compressive Axial Force Application," Annual
Book of ASTM Standards, American Society for Testing and Materials, West
Conshohocken, PA, Vol. 03.01, 2007, pp. 825-835.

[65] ASTM Standard E8-04, "Standard Test Methods for Tension Testing of Metallic
Materials," Annual Book of ASTM Standards, American Society for Testing and
Materials, West Conshohocken, PA, Vol. 03.01, 2007, pp. 65-88.

[66] ASTM Standard E739-91, "Standard Practice for Statistical Analysis of Linear or
Linearized Stress-Life (S-N) and Strain-Life (e-N) Fatigue Data," Annual Book of
ASTM Standards, American Society for Testing and Materials, West
Conshohocken, PA, Vol. 03.01, 2007, pp. 737-743.

[67] ASTM Standard E646-00, "Standard Test Method for Tensile Strain-Hardening
Exponents (n-values) of Metallic Sheet Materials," Annual Book of ASTM
Standards, American Society for Testing and Materials, West Conshohocken, PA,
Vol. 03.01, 2007, pp. 684-691.

[68] Zeng, Z. and Fatemi, A., "Elasto-plastic stress and strain behaviour at notch roots
under monotonic and cyclic loadings," The Journal of Strain Analysis for
Engineering Design, Vol. 36, 2001, pp. 287-300.

[69] Kuhn, P. and Hardrath, H.F., "An engineering method for estimating notch size
effect in fatigue," NACA TN 2805, Vol. 1952.

176
[70] Susmel, L., "The theory of critical distances: a review of its applications in
fatigue," Engineering Fracture Mechanics, Vol. 75, 2008, pp. 1706-1724.

[71] Taylor, D., "Analysis of fatigue failures in components using the theory of critical
distances," Engineering Failure Analysis, Vol. 12, 2005, pp. 906-914.

[72] Baicchi, P., Collini, L., and Riva, E., "A methodology for the fatigue design of
notched castings in gray cast iron," Engineering Fracture Mechanics, Vol. 74,
2007, pp. 539-548.

[73] Bahmani, M., Elliott, R., and Varahram, N., "The relationship between fatigue
strength and microstructure in an austempered Cu-Ni-Mn-Mo alloyed ductile
iron," Journal of Materials Science, Vol. 32, 1997, pp. 5383-5388.

[74] Colombo, R.L. and Bonello, L., "Relationships between fatigue resistance and
spheroidization of graphite in nodular cast iron," Metallography, Vol. 16, 1983,
pp. 327-330.

[75] Shanmugam, P., Rao, P.P., Udupa, K.R., and Venkataraman, N., "Effect of
microstructure on the fatigue strength of an austempered ductile iron," Journal of
Materials Science, Vol. 29, 1994, pp. 4933-4940.

[76] Taylor, D., Hughes, M., and Allen, D., "Notch fatigue behaviour in cast irons
explained using a fracture mechanics approach," International Journal of Fatigue,
Vol. 18, 1996, pp. 439-445.

[77] Agha, H.Y., Béranger, A.S., Billardon, R., and Hild, F., "High-cycle fatigue
behaviour of spheroidal graphite cast iron," Fatigue and Fracture of Engineering
Materials and Structures, Vol. 21, 1998, pp. 287-296.

[78] Fuller, A.G., "Effect of graphite form on fatigue properties of pearlitic ductile
irons," AFS Transactions, Vol. 85, 1978, pp. 527-536.

[79] Sofue, M., Okada, S., and Sasaki, T. "High-quality ductile cast iron with
improved fatigue strength," AFS Transactions, Vol. 86, 1978, pp 173-182

177
Appendix A: Tables of Data Compiled from Literature
Used in Chapter 2 Correlations

Table A.1: Tensile properties from literature used to make correlations in chapter 2.

E Su Sy Ductility
Material Classification Hv* Ref
[GPa] [MPa] [MPa] [%]
CuNiMnMo 0 As-cast [73]
CuNiMnMo 1 ADI [73]
CuNiMnMo 2 ADI [73]
CuNiMnMo 3 ADI [73]
CuNiMnMo 4 ADI [73]
CuNiMnMo 5 ADI [73]
CuNiMnMo 6 ADI 1016 [73]
CuNiMnMo 7 ADI [73]
CuNiMnMo 8 ADI [73]
CuNiMnMo 9 ADI [73]
CuNiMnMo 10 ADI [73]
CuNiMnMo 11 ADI [73]
CuNiMnMo 12 ADI [73]
CuNiMnMo 13 ADI [73]
CuNiMnMo 14 ADI 945 [73]
CuNiMnMo 15 ADI [73]
CuNiMnMo 16 ADI 1288 [73]
Fully
Nodular Iron [74]
Spheroidized
Incompletely
Nodular Iron [74]
Spheroidized
PVD No coating (A) ADI [27]
PVD TiN (A) ADI [27]
PVD TiCN (A) ADI [27]
PVD No coating (B) ADI [27]
PVD TiN (B) ADI [27]
PVD TiCN (B) ADI [27]
EN-GJV-400 CGI 238 [5]
Nodular
EN-GJS-600 247 [5]
Graphite
ADI-A2 ADI 1026 840 3.68 322 [59]
ADI-A2 ADI 1241 1010 2.23 389 [59]
ADI-A2 ADI 1245 1023 2.06 382 [59]
ADI-A1 ADI 1509 1272 1.33 451 [59]
ADI-A1 ADI 896 885 0.19 503 [59]
ADI-A1 ADI 952 902 0.17 500 [59]
ADI-B1 ADI 1388 1235 0.56 444 [59]
ADI-B1 ADI 1373 1250 0.4 445 [59]
ADI-B1 ADI 1529 1268 1.69 452 [59]

178
 E Su Sy Ductility
Material Classification Hv* Ref
[GPa] [MPa] [MPa] [%]
ADI-B2 ADI 1089 890 3.33 349 [59]
ADI-B2 ADI 1073 905 2.7 352 [59]
ADI-B2 ADI 1095 913 2.98 354 [59]
ADI-C2 ADI 1125 901 4.42 350 [59]
ADI-C2 ADI 1116 921 3.41 355 [59]
ADI-C2 ADI 1126 924 3.69 354 [59]
ADI-C1 ADI 1380 1249 0.42 451 [59]
ADI-C1 ADI 1543 1313 1.42 462 [59]
ADI-C1 ADI 1602 1339 2.57 447 [59]
ADI-D2 ADI 924 778 1.11 339 [59]
ADI-D2 ADI 883 813 0.43 338 [59]
ADI-D2 ADI 997 821 1.31 326 [59]
ADI-D1 ADI 1329 1108 0.72 445 [59]
ADI-D1 ADI 1197 1181 0.24 438 [59]
ADI-D1 ADI 1359 1195 0.64 440 [59]
As-cast-A As-cast 741 428 4.26 238 [59]
As-cast-B As-cast 815 501 4.27 250 [59]
As-cast-C As-cast 735 457 4.49 240 [59]
As-cast-D As-cast 785 501 2.81 248 [59]
573K, Sm Cast Bot. ADI 171 1333 1320 0.2 480 [14]
573K, Mid Cast Bot ADI 157 1246 975 0.13 484 [14]
573K, Lg Cast Bot ADI 167 1147 1117 0.1 494 [14]
573K, Lg Cast Top ADI 139 1005 962 0.04 513 [14]
633K, Sm Cast Bot ADI 167 1036 901 1.64 376 [14]
633K, Mid Cast Bot ADI 164 1011 898 0.98 379 [14]
633K, Lg Cast Bot ADI 173 1022 900 1.45 371 [14]
633K, Lg Cast Top ADI 153 1025 910 1.28 385 [14]
ADI ADI 170 1340 1191 0.97 447 [17]
ADI ADI 161 1094 914 3.86 372 [17]
ADI ADI 170 1340 1191 0.97 447 [17]
ADI ADI 161 1094 914 3.86 372 [17]
Ductile Cast Iron Shot Peened [25]
Ductile Cast Iron Shot Peened [25]
ADI 1 ADI 1024 481 [75]
ADI 2 ADI 763 399 [75]
ADI 3 ADI 700 379 [75]
ADI 4 ADI 674 349 [75]
ADI 6 ADI 787 385 [75]
ADI 7 ADI 634 313 [75]
ADI 8 ADI 511 375 [75]
ADI 9 ADI 524 387 [75]
ADI 11 ADI 877 403 [75]
ADI 12 ADI 712 343 [75]
ADI 13 ADI 675 283 [75]
ADI 14 ADI 627 271 [75]
ADI 1 ADI [11]
ADI 2 ADI [11]
ADI 3 ADI [11]
ADI 4 ADI [11]
ADI 5 ADI [11]
As-cast As-cast [11]
Spheroidal
Grade 17 [76]
Graphite
Spheroidal
Grade 17 [76]
Graphite
Vanadium Carbide
Ductile Iron 1035 470 [13]
Iron FM
Vanadium Carbide
Ductile Iron 764 348 [13]
Iron HM

179
 E Su Sy Ductility
Material Classification Hv* Ref
[GPa] [MPa] [MPa] [%]
Vanadium Carbide
Ductile Iron 644 [13]
Iron MFM-1
Vanadium Carbide
Ductile Iron 629 [13]
Iron MFM-2
Ferritic ADI 502 345 10.94 175 [15]
Pearlitic/Ferritic ADI 635 470 8.37 237 [15]
Pearlitic ADI 682 490 7.1 268 [15]
Tempered
ADI 1147 934 5.32 326 [15]
Martensitic
Lower Ausferritic ADI 1025 704 13.12 292 [15]
Upper Ausferritic ADI 625 489 3.96 347 [15]
Spheroidal
Ductile Cast Iron 185 510 350 185 [77]
Graphite
Cu-Ni ADI ADI 158 1012 737 7 336 [24]
Nodular
1 ASTM Type IV 171 907 286 [78]
Graphite
Nodular
2 ASTM Type IV 170 905 [78]
Graphite
Nodular
3 ASTM Type IV 168 868 283 [78]
Graphite
Nodular
4 ASTM Type IV 164 784 [78]
Graphite
Nodular
5 ASTM Type IV 162 641 [78]
Graphite
Nodular
6 ASTM Type IV 160 561 284 [78]
Graphite
D1 Ferritic Ductile Iron [79]
D2 Ferritic Ductile Iron [79]
D3 Sorbitic Ductile Iron [79]
D4 Sorbitic Ductile Iron [79]
D5 Bainitic Ductile Iron [79]
D6 Sorbitic Ductile Iron [79]
D7 Sorbitic Ductile Iron [79]
D8 Sorbitic Ductile Iron [79]
P1 Ductile Iron [79]
P2 Ductile Iron [79]
P3 Ductile Iron [79]
P4 Ductile Iron [79]
P1 Ductile Iron [79]
P2 Ductile Iron [79]
P3 Ductile Iron [79]
P4 Ductile Iron [79]
1 Ductile Iron [79]
2 Ductile Iron [79]
3 Ductile Iron [79]
4 Ductile Iron [79]
5 Ductile Iron [79]
6 Ductile Iron [79]
7 Ductile Iron [79]
8 Ductile Iron [79]
9 Ductile Iron [79]
10 Ductile Iron [79]
11 Ductile Iron [79]
12 Ductile Iron [79]
13 Ductile Iron [79]
14 Ductile Iron [79]
15 Ductile Iron [79]
16 Ductile Iron [79]
*Hardness values converted based on ASTM E140-97

180
Table A.2: Fatigue and microstructural data compiled from literature for correlations
in chapter 2.

Avg Nd Nd Nodule Avg

Sf Test Nf for
Material Class Nodularity Avg Max Count/ Nodule Ref
[MPa] Style Sf
(%) (μm) (μm) mm2 Count
CuNiMnMo 0 As-cast 345 RTB 108 79 90-100 95 [73]
CuNiMnMo 1 ADI 406 RTB 108 98 [73]
CuNiMnMo 2 ADI 418 RTB 108 98 [73]
CuNiMnMo 3 ADI 415 RTB 108 98 [73]
8
CuNiMnMo 4 ADI 383 RTB 10 98 [73]
8
CuNiMnMo 5 ADI 397 RTB 10 98 [73]
CuNiMnMo 6 ADI 420 RTB 108 98 [73]
8
CuNiMnMo 7 ADI 421 RTB 10 98 [73]
CuNiMnMo 8 ADI 400 RTB 108 98 [73]
CuNiMnMo 9 ADI 362 RTB 108 98 [73]
CuNiMnMo 10 ADI 380 RTB 108 98 [73]
8
CuNiMnMo 11 ADI 387 RTB 10 98 [73]
8
CuNiMnMo 12 ADI 384 RTB 10 98 [73]
CuNiMnMo 13 ADI 392 RTB 108 98 [73]
8
CuNiMnMo 14 ADI 385 RTB 10 98 [73]
8
CuNiMnMo 15 ADI 415 RTB 10 98 [73]
8
CuNiMnMo 16 ADI 376 RTB 10 98 [73]
Fully 6
Nodular Iron 201 RTB 5x10 67 [74]
Spheroidized
Incompletely
Nodular Iron 165 RTB 5x106 64 [74]
Spheroidized
PVD No
ADI 251 RTB 107 51 70 70 [27]
coating (A)
7
PVD TiN (A) ADI 292 RTB 10 51 70 70 [27]
7
PVD TiCN (A) ADI 306 RTB 10 51 70 70 [27]
PVD No 7
ADI 200 RTB 10 81 30 30 [27]
coating (B)
7
PVD TiN (B) ADI 214 RTB 10 81 30 30 [27]
PVD TiCN (B) ADI 217 RTB 107 81 30 30 [27]
EN-GJV-400 CGI 160 Axial 2x106 39 [5]
Nodular
EN-GJS-600 220 Axial 2x106 74 [5]
Graphite
ADI-A2 ADI 235 RTB 107 80 50 100 100 [59]
ADI-A2 ADI 235 RTB 107 80 50 100 100 [59]
ADI-A2 ADI 235 RTB 107 80 50 100 100 [59]
ADI-A1 ADI 180 RTB 107 80 50 100 100 [59]
ADI-A1 ADI 180 RTB 107 80 50 100 100 [59]
ADI-A1 ADI 180 RTB 107 80 50 100 100 [59]
ADI-B1 ADI 258 RTB 107 90 40 110 110 [59]
ADI-B1 ADI 258 RTB 107 90 40 110 110 [59]
ADI-B1 ADI 258 RTB 107 90 40 110 110 [59]
ADI-B2 ADI 250 RTB 107 90 40 110 110 [59]
ADI-B2 ADI 250 RTB 107 90 40 110 110 [59]
ADI-B2 ADI 250 RTB 107 90 40 110 110 [59]
ADI-C2 ADI 290 RTB 107 90 34 150 150 [59]
ADI-C2 ADI 290 RTB 107 90 34 150 150 [59]
ADI-C2 ADI 290 RTB 107 90 34 150 150 [59]
ADI-C1 ADI 252 RTB 107 90 34 150 150 [59]

181
 Avg Nd Nd Nodule Avg
Sf Test Nf for
Material Class Nodularity Avg Max Count/ Nodule Ref
[MPa] Style Sf
(%) (μm) (μm) mm2 Count
ADI-C1 ADI 252 RTB 107 90 34 150 150 [59]
ADI-C1 ADI 252 RTB 107 90 34 150 150 [59]
ADI-D2 ADI 320 RTB 107 90 24 200 200 [59]
ADI-D2 ADI 320 RTB 107 90 24 200 200 [59]
ADI-D2 ADI 320 RTB 107 90 24 200 200 [59]
ADI-D1 ADI 250 RTB 107 90 24 200 200 [59]
ADI-D1 ADI 250 RTB 107 90 24 200 200 [59]
ADI-D1 ADI 250 RTB 107 90 24 200 200 [59]
As-cast-A As-cast 150 RTB 107 80 50 100 100 [59]
As-cast-B As-cast 175 RTB 107 90 40 110 110 [59]
As-cast-C As-cast 190 RTB 107 90 34 150 150 [59]
As-cast-D As-cast 220 RTB 107 90 24 200 200 [59]
573K, Sm Cast
ADI 350 RTB 107 90 34 87 90 90 [14]
Bot.
573K, Mid
ADI 312 RTB 107 80 49 136 45 45 [14]
Cast Bot
573K, Lg Cast
ADI 218 RTB 107 75 56 119 40 40 [14]
Bot
573K, Lg Cast
ADI 195 RTB 107 70 65 200 35 35 [14]
Top
633K, Sm Cast
ADI 410 RTB 107 90 34 87 90 90 [14]
Bot
633K, Mid
ADI 340 RTB 107 80 49 136 45 45 [14]
Cast Bot
633K, Lg Cast
ADI 312 RTB 107 75 56 119 40 40 [14]
Bot
633K, Lg Cast
ADI 244 RTB 107 70 65 200 35 35 [14]
Top
7
ADI ADI 331 RTB 10 75 34.5 65-90 77.5 [17]
ADI ADI 363 RTB 107 75 34.5 65-90 77.5 [17]
ADI ADI 380 RTB 107 75 34.5 65-90 77.5 [17]
ADI ADI 400 RTB 107 75 34.5 65-90 77.5 [17]
Ductile Cast
Shot Peened 330 RTB 108 44 [25]
Iron
Ductile Cast
Shot Peened 315 RTB 108 38 [25]
Iron
ADI 1 ADI 243 RTB 106 [75]
ADI 2 ADI 279 RTB 106 [75]
ADI 3 ADI 299 RTB 106 40 [75]
ADI 4 ADI 310 RTB 106 [75]
ADI 6 ADI 242 RTB 106 [75]
ADI 7 ADI 242 RTB 106 [75]
ADI 8 ADI 247 RTB 106 44 [75]
ADI 9 ADI 245 RTB 106 [75]
ADI 11 ADI 279 RTB 106 [75]
ADI 12 ADI 294 RTB 106 [75]
ADI 13 ADI 315 RTB 106 60 [75]
ADI 14 ADI 340 RTB 106 [75]
ADI 1 ADI 90 RTB 107 [11]
ADI 2 ADI 225 RTB 107 [11]
ADI 3 ADI 293 RTB 107 [11]
ADI 4 ADI 248 RTB 107 [11]
ADI 5 ADI 218 RTB 107 [11]
As-cast As-cast 233 RTB 107 [11]
Spheroidal
Grade 17 464 CB - [76]
Graphite
Spheroidal
Grade 17 392 CB - [76]
Graphite
Vanadium
Carbide Iron Ductile Iron 350 RTB - 33 [13]
FM

182
 Avg Nd Nd Nodule Avg
Sf Test Nf for
Material Class Nodularity Avg Max Count/ Nodule Ref
[MPa] Style Sf
(%) (μm) (μm) mm2 Count
Vanadium
Carbide Iron Ductile Iron 270 RTB - 19 [13]
HM
Vanadium
Carbide Iron Ductile Iron 500 RTB - 14 [13]
MFM-1
Vanadium
Carbide Iron Ductile Iron 400 RTB - 29 [13]
MFM-2
Ferritic ADI 170 RTB 107 95.9 32 111 135.6 135.6 [15]
Pearlitic/Ferritic ADI 225 RTB 107 95.5 37 106 119.3 119.3 [15]
Pearlitic ADI 235 RTB 107 92.4 30 108 146.4 146.4 [15]
Tempered
ADI 255 RTB 107 92.6 32 123 118.8 118.8 [15]
Martensitic
Lower 7
ADI 290 RTB 10 95.2 34 61 100.6 100.6 [15]
Ausferritic
Upper
ADI 220 RTB 107 96.3 29 88 155.3 155.3 [15]
Ausferritic
Ductile Cast Spheroidal
220 RTB 107 44 [77]
Iron Graphite
7
Cu-Ni ADI ADI 250 RTB 10 66 [24]
1 ASTM Type Nodular 7
278 RTB 2x10 98.25 [78]
IV Graphite
2 ASTM Type Nodular 7
286 RTB 2x10 96.75 [78]
IV Graphite
3 ASTM Type Nodular
263 RTB 2x107 94.75 [78]
IV Graphite
4 ASTM Type Nodular
239 RTB 2x107 77 [78]
IV Graphite
5 ASTM Type Nodular
217 RTB 2x107 44 [78]
IV Graphite
6 ASTM Type Nodular
208 RTB 2x107 50 [78]
IV Graphite
7
D1 Ferritic Ductile Iron 251 RTB 10 [79]
D2 Ferritic Ductile Iron 260 RTB 107 [79]
D3 Sorbitic Ductile Iron 281 RTB 107 [79]
D4 Sorbitic Ductile Iron 209 RTB 107 [79]
D5 Bainitic Ductile Iron 311 RTB 107 [79]
D6 Sorbitic Ductile Iron 360 RTB 107 [79]
D7 Sorbitic Ductile Iron 343 RTB 107 [79]
D8 Sorbitic Ductile Iron 352 RTB 107 [79]
P1 Ductile Iron 339 RTB 107 86 [79]
P2 Ductile Iron 319 RTB 107 74 [79]
P3 Ductile Iron 289 RTB 107 56 [79]
P4 Ductile Iron 239 RTB 107 32 [79]
P1 Ductile Iron 248 RTB 107 86 [79]
P2 Ductile Iron 220 RTB 107 74 [79]
P3 Ductile Iron 200 RTB 107 56 [79]
P4 Ductile Iron 170 RTB 107 32 [79]
1 Ductile Iron 191 RTB 107 89 [79]
2 Ductile Iron 215 RTB 107 82 [79]
3 Ductile Iron 240 RTB 107 55 [79]
4 Ductile Iron 225 RTB 107 32 [79]
5 Ductile Iron 209 RTB 107 36 [79]
6 Ductile Iron 239 RTB 107 89 [79]
7 Ductile Iron 269 RTB 107 82 [79]
8 Ductile Iron 239 RTB 107 55 [79]
9 Ductile Iron 219 RTB 107 32 [79]

183
 Avg Nd Nd Nodule Avg
Sf Test Nf for
Material Class Nodularity Avg Max Count/ Nodule Ref
[MPa] Style Sf
(%) (μm) (μm) mm2 Count
10 Ductile Iron 259 RTB 107 89 [79]
11 Ductile Iron 304 RTB 107 82 [79]
12 Ductile Iron 289 RTB 107 55 [79]
13 Ductile Iron 230 RTB 107 32 [79]
14 Ductile Iron 310 RTB 107 89 [79]
15 Ductile Iron 341 RTB 107 82 [79]
16 Ductile Iron 249 RTB 107 55 [79]

184