Engineering Failure Analysis 98 (2019) 228–239

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Review

Defects as a root cause of fatigue failure of metallic components. II:

T
Non-metallic inclusions
⁎
Zerbst U.a, , Madia M.a, Klinger C.a, Bettge D.a, Murakami Y.b
a
Bundesanstalt für Materialforschung und -prüfung (BAM), Berlin D-12205, Germany
b
Emeritus Professor, Kyushu University, Fukuoka, Japan

A R T IC LE I N F O ABS TRA CT

Keywords: This second part of the review on defects as root cause of fatigue failure comprises the origin, the
non-metallic inclusions nature and the effects of non-metallic inclusions. Topics addressed are the different kinds of
mis-match inclusions formed during the manufacturing process, various types of mis-match causing local
inclusion size stresses and, as a consequence, fatigue crack initiation, and effects of characteristics such as size,
inclusion cluster
morphology, localization, spatial distribution and orientation of the defects on the fatigue be-
statistics
havior. Methods for inclusion counting and sizing are discussed along with statistical aspects
necessary to be considered when evaluating structural components.

1. Introduction

In the first part of this review [1], basic aspects such as the propagation and the arrest of short and long cracks under cyclic
loading during heating or cooling, which are initiated at defects have been discussed. This was important insofar as the question
when a material or geometrical imperfection is a defect in the sense of a damage analysis can only be answered with that background
knowledge. Here, non-metallic inclusions resulting from the material manufacturing process will be discussed in more detail. Other
defects such as cavities, dents, corrosion pits and scratches will be addressed in the third part of this review [2].

2. Non-metallic inclusions

2.1. The origin of non-metallic inclusions

Non-metallic inclusions are chemical compounds of metals with non-metals, e.g., oxides, sulfides or nitrides but there are also
more complex species such as oxysulfides, carbonitrides or others (e.g., [3]). With respect to their origin it is distinguished between
endogenous and exogenous inclusions [4–6], see Fig. 1 for steel case.

(a) Endogenous inclusions may be formed by reactions of the liquid melt with deoxidizers such as silicon, manganese or aluminum
particles or during desulfurization but other processes are possible as well [6]. The resulting reaction products are not completely
absorbed by the slag bath but their volume fraction can be reduced by secondary metallurgy such as vacuum-arc melting [4] or
metal-melt filtration [6].
(b) Exogenous inclusions have their origin in external sources, e.g. refractory fragments, slag covering the molten metal or sand in

⁎
Corresponding author.
E-mail address: [email protected] (U. Zerbst).

https://doi.org/10.1016/j.engfailanal.2019.01.054
Received 12 June 2018; Received in revised form 8 January 2019; Accepted 10 January 2019
Available online 11 January 2019
1350-6307/ © 2019 Elsevier Ltd. All rights reserved.
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

Nomenclature Kmin minimum K in the loading cycle

Ms martensite start temperature
a crack depth, crack length N number of loading cycles
C, n parameters of the Paris law; Eq. (4) R R ratio (= σmin/σmax or Κmin/Κmax)
d diameter of cylindrical specimen/component (Fig. Rm ultimate tensile stress
8) α thermal expansion coefficient
da/dN fatigue crack propagation rate α parameter in Eqs. (1) and (2)
E modulus of elasticity (Young's modulus) ΔK range of K, cyclic K factor (Kmax – Kmin)
Ei E of the inclusion ΔKth fatigue crack propagation threshold
Em E of the matrix material σ stress
F cumulative probability (Fig. 12) σe fatigue or endurance limit (amplitude) of the ma-
HV Vickers hardness terial at R = −1
K linear elastic stress intensity factor (K factor) σmax maximum stress in the loading cycle
KI K factor for mode I loading σmin minimum stress in the loading cycle
KI,max maximum K factor along the front of a semi-el- area square root of the defect area normal to the
liptical surface crack; Eq. (3) loading direction
Kmax maximum K in the loading cycle

Fig. 1. Sources of non-metallic inclusions in steel production; figure taken from [9].

229
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

cast alloys, from where they are mechanically incorporated in the metal. Their size is usually larger than those of their en-
dogenous counterparts [5,7].

2.2. The contribution of non-metallic inclusions to fatigue crack initiation and propagation

2.2.1. The empirical evidence

Non-metallic inclusions are typical fatigue crack initiation sites, particularly in higher strength steels, wrought aluminum alloys
and Nickel-based alloys [10], see also [4,6,7,11–21]. Typical failure cases with non-metallic inclusions as the potential root cause are
reported for components such as bearings, crankshafts, gears, springs and others made of materials showing a high fatigue limit (e.g.
[12,19,20,22–29]).

2.2.2. Mechanical, thermo-mechanical and chemical mis-match between inclusion and matrix
Fatigue cracks will always initiate at the sites of the highest local strain at macro-notches such as machined holes, notches, slots,
gross-sectional transitions, etc. but the local stress/strain concentration of these will be enhanced by the presence of defects. Besides
the notch effect of the particles, this is caused by the mis-match between inclusions and the surrounding matrix material. With respect
to this, distinction has to be made between at least four mis-match types: stiffness mis-match, strength and ductility mis-match,
thermal expansion mis-match and chemical mis-match. Table 1 summarizes the modulus of elasticity (Young's modulus), the coef-
ficient of thermal expansion and the Vickers hardness (as a substitute measure of the strength) for selected types of inclusions in steel.
Note that the data are taken from different sources and have been obtained by different methods what partly explains the scatter.
Note further that the parameters also show some temperature dependency; e.g., the authors in [30] report a decrease in the modulus
of elasticity of Al2O3 inclusions from 353.1 GPa at 800 °C to 225.5 GPa at 1400 °C.

(a) Stiffness and strength/ductility mis-match

The stiffness of a material is characterized by the modulus of elasticity or Young's modulus E. Based on this parameter, the mis-
match between inclusion and matrix material can be under- and over-matching. With the indices “i” and “m” referring to inclusion
and matrix, under-matching stands for Ei < Em and over-matching for Ei > Em. In the case of a steel matrix, Al2O3 and TiN in-
clusions show stiffness over-matching (this is also the case for TiC and spinels [7,36]). In contrast, MnS inclusions are characterized
by under-matching. What does potentially happen in the case of stiffness over-matching? Under uniform stresses, the strain is higher
in the matrix material than in the inclusion, and a strain concentration zone develops next to the inclusion which can cause residual
stresses in this region [10,37,38] but can also cause the initiation of small cracks at the matrix-inclusion interface or in the matrix
material surrounding the inclusion [33,39]. Of course, crack initiation in the matrix requires a higher strength of the inclusion when
compared to the matrix.
If stiffness under-matching is present as in the case of MnS inclusions in steel, the strain concentration zone develops in the
inclusion and not in the surrounding matrix. However, because the sulfides are not only softer than the matrix but also show a better
ductility than other inclusion types such as oxides particularly at elevated working temperature [40], the probability of crack de-
velopment inside them is low. This is the reason, why oxides and nitrides are usually considered to be more harmful than sulfides
[36,41–43], Fig. 2. Note, however, that MnS inclusions, when deformed, may flatten with the consequence of an increased stress
concentration factor, i.e., the notch effect of the elongated inclusions is strengthened in case of an unfavorable macroscopic loading
direction, see Section 2.2.4.

(b) Thermal expansion mis-match

Initial cracks can already be present before any mechanical loading is applied [39]. The reason is thermal expansion mis-match
between the matrix and the inclusions. As can be seen in Table 1, the thermal expansion coefficients of Al2O3 and TiN are significantly
lower than those of the surrounding steel matrix. Thus, when the temperature is changed during heating or cooling, the effect is
similar to mechanical loading generating local strain concentrations and, as a consequence, residual stresses and microcracking occur
[5,10,20,38,45]. Temperature changes and gradients are typical for manufacturing (forging, rolling, heat treatment, welding) but
they can also occur in service, e.g. due to the frictional heat during the braking process of a wheel [8,46].

Table 1
Material data of selected inclusion types and steel. They are taken from [7] citing older sources of [31], see also [5,32–35].
Type of inclusion Modulus of elasticity E in GPa Thermal expansion coefficient α in 10−6/K Vickers hardness HV

Al2O3 252–389 8.0 900–2500

TiN 278–317 9.4 1000–4000
MnS 69–138 12.1–18.1 100
Iron/steel (for comparison) 206 12.1; 12.5a
23–25b

a
γ-α transition 850 °C ➔ room temperature.
b
850 °C ➔ Ms temperature.

230
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

Fig. 2. Relative harmfulness of various types of inclusions on fatigue life of a bearing steel; according to [44]. The ordinate scale unit is a decrease in
fatigue strength of 125 MPa at 108 loading cycles, see [11].

(c) Chemical mis-match

Chemical mis-match plays a role in the presence of a corrosive environment, i.e., if the medium has access to the inclusion site.
Matrix material and inclusions are electrochemically different, i.e., they occupy different places on the electrochemical series. E.g.,
titanium nitride inclusions are cathodic with respect to steels and, therefore cause the anodic dissolution of the latter [47].

2.2.3. The influence of the inclusion size

Besides its chemical composition, the size, shape and localization of the inclusions are important. The effect of the defect size on
the fatigue limit and of the statistical size distribution on the scatter of the fatigue limit has already been briefly addressed in the first
part of this overview (Section 1 in [1]). With respect to inclusions, there is a clear correlation in that a larger inclusion size refers to a
lower fatigue strength [11,12,36,41,48–51] as long as no other defects such as cavities, scratches, etc. take over the role as crack
initiation sites. An example is provided in Fig. 3 for high strength spring steel.
The effect of the inclusion size on the fatigue life (and, correlated with this, on the fatigue strength) can simply be explained when
the inclusion is considered as a crack such as done, e.g., in [11,53–55], see, however, the discussion provided in [56,57]. In this case,
a larger inclusion length refers to a larger crack length. This not only increases the crack driving force and hence the crack propa-
gation rate, but it also reduces the fatigue lifetime in that the number of cycles, a short crack would have needed to grow to the larger
size, is omitted. The initial stage of crack initiation at a defect is the debonding between defect and matrix [13,50,55,58–60].
Frequently this takes place during the first or the first few loading cycles [55,58,60], see also Section 2.2 in [1] and, as an example,
Fig. 4. In other cases, the crack will grow along the particle-matrix interface for a while [17], and sometimes, also the inclusion
particle will break (e.g., [61,62]).
As outlined in Section 1 of the first part of this review [1], there exists a critical defect size below which no defect size dependence
of the fatigue limit is observed any more. Examples of critical non-metallic inclusions depths in that sense have been reported as
8–10 μm [28], 10 μm [63], and 17–18 μm [25], for steel and 10–20 μm [17] for aluminum. The authors of [64] report critical values
of around 10 μm just below the surface and of about 30 μm at a position 100 μm below the surface for bearing steels loaded in rotating
bending. As mentioned above, the defect size becomes uncritical when

(a) another kind of defects takes over the role as crack initiation site, or
(b) the cracks initiated at the inclusions arrest after some extension

Fig. 3. The effect of the diameter of non-metallic inclusions on the fatigue limit of a high strength spring steel; according to [52].

231
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

Fig. 4. Inclusion as fatigue crack initiation site in the material AL5083 H321, (a) Macrograph; (b) Micrograph. Only the imprint of the inclusion(s) is
preserved which is an indicator for early debonding between inclusion and matrix; according to [55].

See also Section 4 in [1], where the authors, based on crack arrest considerations, predicted the critical defect depth of semi-
circular cracks that would arrest at a plain surface of about 10 μm and 17 μm (mean values only) for a higher and a lower strength
steel. These values fit in well into what is otherwise known, and they provide an explanation for the different critical crack sizes as an
inverse function of the material strength (e.g. [65]). In [1] it has been shown that this dependency is an immediate outcome of the
crack arrest analyses made with the cyclic R curve approach. The size of the crack at arrest becomes smaller at higher strength and,
below a certain limit, it becomes smaller than the inclusion size. Of course, this pattern must be discussed in statistical terms, since
both the arrest crack size and the inclusion size show some scatter band. An indirect indication for the restriction of the effect of non-
metallic inclusions on the fatigue limit to higher strength steels is provided in the well-known plot given in Fig. 5 from which it results
that the linear dependency between hardness and fatigue strength cease to exist above a Vickers hardness of about HV = 400.
Murakami and Endo [68], see also [11,14] proposed a simple half-empirical correlation which ever since has experienced a
widespread acceptance and application. It considers the inclusion size in the hardness-fatigue limit correlation, HV-σe, reproduced in
Fig. 5. For surface inclusions it is given by

1.43⋅(HV + 120) 1 − R α
σe = ⋅⎡ ⎤ .
( area )1/6 ⎣ 2 ⎦ (1)

For internal inclusions the factor 1.43 is replaced with 1.56. The parameter R designates the stress ratio σmin/σmax, and the

Fig. 5. Relationship between hardness and fatigue limit for steels; according to Garwod et al. [66], cit. in [11,14]. For other materials, different
curve slopes exist [67]. In cast iron, the deviation from the linear curve occurs above ca. HV = 150 presumably because of larger fatigue controlling
defects [14] which are in that case graphite nodules or cavities, not inclusions.

232
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

exponent α is given by

α = 0.226 + HV ⋅10−4 (2)

for steels (σe in MPa; area in μm). The parameter area is the square root of the area of the defect projected to a plane perpendicular
to the maximum tensile stress. The basis of the approach is the experimentally confirmed assumption that the maximum mode I stress
intensity factor KI,max, can be determined with an accuracy of ± 10% by

KI,max ≈ 0.65⋅σ⋅ π area . (3)

for surface cracks with widely varying crack aspect ratios a/c. The possibility to apply a solution for two-dimensional cracks to three-
dimensional defects is based on the assumption that small cracks emanate from defects soon after fatigue process initiation, see
Section 2.2.2. A value of area = 1000 μm is stated as the upper limit of Eq. (1) in [14], however a value of no more than 100–200 μm
is given elsewhere for high strength steels [69]. Note that the equation loses its validity above this value as well as below a lower
bound in the order of a few grain sizes [70].

2.2.4. The influence of inclusion morphology, localization, spatial distribution and orientation
It is not only the strength, the ductility and the size of the inclusions and the matrix which has an influence on the fatigue strength
but also other features such as the chemical composition (see Section 2.2.2 (c)), the shape or morphology of the particles, their
localization with respect to the surface and the macro-notches, their spatial distribution and their orientation with respect to the
loading direction.
Inclusions are also micro-notches, and it is clear that their morphology has an effect on the stress-strain concentration they cause.
An irregular shape with sharp edges will be more detrimental than a rounded one, a small aspect ratio between the heights and length
will cause higher local stresses than a sphere. Although there are significant shape differences between the various oxides, sulfides
etc., the discussion on different morphologies is a bit academic because inclusions predominantly do not exist individually but do
exist as more or less irregular clusters, stringers etc. (e.g. [6,20,23,40,42,47,50,71–81]. Examples are shown in Figs. 6 and 7.
Inclusion clusters and stringers are generated by particle deformation and fragmentation during mechanical manufacturing such
as rolling or forging [20,47], clusters and agglomerates form also during casting [80]. The texturization of the inclusion pattern has
the consequence that the fatigue strength becomes anisotropic [40,41,50,83–86]. A consequence is that the fatigue limit, in rolled,
forged or wrought materials is higher in the longitudinal than in the transverse direction with respect to the applied loading stresses
[39,50]. The difference can be as large as 50% [40].
Inclusion fragmentation may additionally affect fatigue crack propagation in that the crack propagation rate is increased by the
interaction between adjacent defects. Cracks which have been formed independently of each other at different inclusions or other
structural items may combine in larger cracks by coalescence [5,42,77,87–91]. An indication for this kind of crack growth is an
increased Paris exponent n in
da/ dN = C⋅ΔK n (4)

Note that the theoretical value of n is 2 with usual values being between 2 and 4 [92,93]. In contrast a value of n = 5.2 for
medium carbon steel used for a connecting rod of an air compressor is reported in [94]. The authors associate this with a potential
effect of inclusions on crack propagation. Inclusions within clusters at initiation sites have also been observed to interact with each
other and with other defect types [76,95] via their micro-notch stress-strain fields. What always matters with respect to damage, is
the local plastic strain. Usually this will be higher for larger inclusions and the increasing number of inclusions in a limited volume.
Referring to [96], the strain is shown to be highest when the inclusion spacing is in the order of the inclusion size [10]. In [40] it is
reported that the coalescence of sulfides in a cluster increases the crack size rapidly during the initial cycles of fatigue loading. A last

Fig. 6. Cluster of non-metallic inclusions in the material AL5083 H321 with the latter being characterized by an increased concentration of
manganese, iron and nickel; according to [55].

233
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

Fig. 7. Axially orientated non-metallic inclusion stringers found near the crack origin in a broken high-speed railway axle; according to [26], see
also [9,82]; (a) location and orientation; (b) chemical composition; axle dimensions schematic.

effect that potentially plays a role in the present context is provided by the residual stresses next to the inclusion-matrix interface due
to the thermal expansion mis-match effect, see Section 2.2.2 (b).

2.2.5. Surface versus sub-surface crack initiation at inclusions

As a rule, surface defects have a greater effect on fatigue strength than defects in the material matrix. Note that the most harmful
defects are just below the surface [38]. This is a general statement on defects of any kind, which is also true with respect to non-
metallic inclusions [17,25,36,38,39,42,72,81,86,97]. However, under certain circumstances the crack initiation site is shifted from
the surface to the interior:

(a) A well-known exception to surface crack initiation is provided by the very high cycle fatigue regime VHCF of high strength steels
(for a review see, e.g. [14]) where fatigue cracks are nearly exclusively initiated at non-metallic inclusions in the interior at loads
well below the conventional fatigue limit. Sub-surface crack initiation is, however, not limited to VHCF. According to [98] the
transition point from surface to sub-surface initiation is in the range between 104 and 106 loading cycles, so it includes high cycle
fatigue (HCF).
(b) When inclusion stringers such as MnS are textured, e.g., due to rolling (see Section 2.2.4), the loading direction might not only
have an influence on the fatigue strength but also on the crack initiation site. E.g., the authors of [50], investigating a hot rolled
low carbon steel, found that the initiation site changed from the surface to the interior with an increase of the angle between the
loading and rolling direction.
(c) The initiation site might also be shifted to the interior when compressive residual stresses are introduced into the surface layer of
a component, e.g., by post-manufacturing treatment such as shot blasting, peening or by heat treatment such as carburization or
similar methods (e.g. [17,76,78,99]). For a review of the existing methods for post-weld treatment of welds see [100].
(d) Finally, sub-surface inclusions play a role for loading cases where the highest shear stress does not occur at the surface but is
shifted to the interior, e.g. rolling contact fatigue [101–104], see also Section 5.3.2.2 in [2].

2.3. Statistical aspects

It is well known from empirical investigations that the fatigue strength of a given material is both dependent on the size of the

Fig. 8. Fatigue limits determined on standard specimens and full-scale railway axles for various batches of 30NiCrMoV12 steel; according to [105].

234
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

cross section and on the loading type (tension/compression, rotary bending, etc.). An example is shown in Fig. 8.
One of the mechanisms yielding the significantly lower fatigue limit of the full-scale axles compared to the small-scale specimens
of Fig. 8 is a statistical size effect of the non-metallic inclusions which act as crack initiation sites (with respect to the other me-
chanisms see the discussion in Section 3.2 in [9]). The inclusion size shows a scatter band. If a weakest link mechanism is assumed
such that the largest inclusion in the highly stressed volume controls the fatigue life and strength [12] (see also Section 2.2.3 and
Fig. 9 as an experimental example), then the statistical size effect of the larger cross section consists in a higher probability to have
larger inclusions. A numerical example for this, which refers to the two geometries of Fig. 8 is provided in Fig. 10.
Consequently, a methodology is needed for finding the maximum inclusion size in a sufficiently large material volume. This
includes the experimental determination of the scatter of the inclusion size as well as its statistical processing. Non-destructive testing
is usually unsuitable for this, except for cases with rather large defects. E.g., inclusion clusters larger than 0.5 mm have been detected
by ultrasonic immersion technique in [26].
Common methods for inclusion sizing are:

(a) Fatigue tests are performed on a larger number of specimens. The fracture surfaces are then analyzed post mortem to identify the
crack initiation sites and to measure the sizes of the inclusions associated with them (e.g. [55,78,86,107,108]). Two examples are
shown in Fig. 11. The identification of the inclusions can be improved by a pre-treatment of the samples, e.g., by annealing [52].
(b) Metallographic sections are made for a number of control areas [11,78,86,110]. An example of inclusions near a weld toe
determined this way is provided in Fig. 12. The advantage over the fractographic method is that many defects are identified
whilst the latter gives only one defect per test. However, because many inclusions are not cut along their largest axes, mostly
apparent maximum particle sizes are measured instead of true sizes (see Section 6.6.2 in [11]. One should, therefore, not expect
to get identical statistical size distributions by the fractographic and the metallographic method. Note, however, that this is not
really the goal since it is not the complete distribution but its high-end tail which is of interest. But even at this level there might
be significant differences in that the metallographic method underestimates the real maximum inclusion size [78]; see also
Figs. 10 and 11 in [86].

The discrepancy is aimed to be avoided by two measures. The first uses the statistic of extremes and emphasizes the upper tail of
the distribution function [11,12,64,86,110,112–116]. The second measure refers to the increase of the control area of the me-
tallographic investigation or to increase the number of areas. E.g., ASTM standard ASTM E2283–08 [117] bases the analysis on the
largest inclusion of 24 different polished areas [110].

(c) For completeness it should be mentioned that further methods have become available in more recent times for the sizing of
defects. A method that is increasingly used is micro-computed tomography (CT) [118], however, rather for larger defects such as
pores. Atkinson and Shi [64], in their review, discuss further methods such as inclusion concentration and the chemical extraction
of inclusions which are only to be mentioned here.

Excursus: Crack initiation at reinforcement fibers

Failure mechanisms of composites constitute a very wide field which shall not be discussed here. The reason why they are,
nevertheless, mentioned here is that there might be parallels with the fatigue crack initiation at inclusions due to the mis-match
phenomena addressed in Section 2.2.2. An example is provided in Fig. 13 where the fatigue failure of a titanium matrix composite
reinforced with continuous SiC fibers loaded at temperatures up to 550 °C is reproduced [119]. Whilst the thermal expansion mis-
match between fiber and matrix material relaxes at higher temperature, the stiffness mis-match at room temperature is significant
and implies potential damage during start-up and shut-down events of aircraft engines. In the specific case, corrosion effects also
seem to have played a role.

Fig. 9. Overall size distribution of inclusions compared to those acting as crack nucleation sites, in 2024-T3 aluminum alloy; The numbers above the
black bars refer to the broken specimens; according to [106].

235
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

Fig. 10. Scale effect on the maximum inclusion size on steel batch 4 of Fig. 8; according to [105].

Fig. 11. Cumulative distribution functions of the size of fractographically identified inclusions at fatigue crack initiation sites. (a) high strength
4340 steel, according to [109]; (b) Al5083 H321 aluminum alloy, according to [55].

Fig. 12. Non-metallic inclusions in the weld toe areas of butt welds of S355NL steel. (a) Metallographic section; (b) histogram of the area equivalent
diameter of a circle; (c) extreme value statistics of inclusion diameters; according to [111].

3. Summary

In this second section of the three-part overview on defects as the root cause of fatigue failure non-metallic inclusions are
addressed with respect to their origin, their characteristics and their effect on fatigue crack initiation and propagation. Inclusions due
to the manufacturing process are typical material defects. Different types of inclusions are discussed with respect to their stiffness,
strength and ductility, thermal expansion and chemical mis-match compared to the matrix material. The influence of inclusions on
the fatigue behavior depends on factors such as their size, their orientation with respect to the loading direction, their morphology

236
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

Fig. 13. Fracture at SiC fibers in a reinforced titanium alloy; (a) Macrosdcopic cross section; (b) Crack initiation sites; (c) and (d) Subsequent fatigue
crack propagation; according to [119].

and spatial distribution including inclusion stringers and clusters and their proximity to the surface of the component. With respect to
these parameters but also to other defects such as cavities, dents, corrosion pits and scratches there exists a competitive situation in
controlling the fatigue behavior. This issue will be further addressed in Part III of the review [2].

References

[1] U. Zerbst, M. Madia, C. Klinger, D. Bettge, Y. Murakami, Defects as the root cause of fatigue failure of metallic components. Part I: Basic aspects, Eng. Fail. Anal.
(2018) xxxxxxxx this issue.
[2] U. Zerbst, M. Madia, C. Klinger, D. Bettge, Y. Murakami, Defects as the root cause of fatigue failure of metallic components. Part III: Cavities, dents, corrosion
pits, scratches, Eng. Fail. Anal. (2018) xxxxxxxx this issue.
[3] T. Gram, A. Vickerfält, Characterization of Non-Metallic Inclusions According to Morphology and Composition. A Comparison of Two Different Steels before
and after Turning, KTH Royal Institute of Technology Report, Stockholm, 2015.
[4] J. Scutti, Manufacturing aspects of failure end prevention, In: ASM Handbook, Vol. 11: Failure Analysis and Prevention. American Society for Metals, 2nd ed.,
ASM International, 2002, pp. 79–223.
[5] C. Mapelli, Non-Metallic Inclusions and Clean Steel, La Metallurgia Italiana, June 2008.
[6] D. Krewerth, T. Lippmann, A. Weidner, H. Biermann, Influence of non-metallic inclusions on fatigue life in the very high cycle fatigue regime, Int. J. Fatigue 84
(2016) 40–52.
[7] P. Juvonen, Effects of Non-Metallic Inclusions on Fatigue Properties of Calcium Treated Steels, PhD Thesis Helsinki Univ. of Technology (2004).
[8] U. Zerbst, K. Mädler, H. Hintze, Fracture mechanics in railway applications – an overview, Eng. Fract. Mech. 72 (2005) 163–194.
[9] U. Zerbst, S. Beretta, G. Köhler, A. Lawton, M. Vormwald, H.Th. Beier, C. Klinger, I. Cerny, J. Rudlin, T. Heckel, D. Klingbeil, Safe life and damage tolerance of
railway axles – a review, Eng. Fract. Mech. 98 (2013) 214–271.
[10] K.S. Chan, Roles of microstructure in fatigue crack initiation, Int. J. Fatigue 32 (2010) 1428–1447.
[11] Y. Murakami, Metal Fatigue. Effects of Small Defects and Nonmetallic Inclusions, Elsevier, Oxford, 2002.
[12] J.R. Yates, G. Shi, H.V. Atkinson, C.M. Sellars, C.W. Anderson, Fatigue tolerant design of steel components based on the size of large inclusions, Fatigue Fract.
Eng. Mater. Struct. 25 (2002) 667–676.
[13] S. Suresh, Fatigue of Materials, 2nd ed., Cambridge University Press, Cambridge, 2003 Chapter 4.9.
[14] Y. Murakami, High and ultrahigh cycle fatigue, in: R.O. Ritchie, Y. Murakami (Eds.), Comprehensive Structural Integrity; Volume 4: Cyclic loading and
Fracture, Elsevier, 2003, pp. 129–164.
[15] Z.G. Yang, G. Yao, G.Y. Li, S.X. Li, Z.M. Chu, W.J. Hui, H. Dong, Y.Q. Weng, The effect of inclusions on the fatigue behavior of fine-grained high strength
42CrMoVNb steel, Int. J. Fatigue 26 (2004) 959–966.
[16] L.T. Lu, J.W. Zhang, K. Shiozawa, Influence of inclusion size on S-N curve characteristics of high-strength steels in the giga-cycle fatigue regime, Fatigue Fract.
Eng. Mater. Struct. 32 (2009) 647–655.
[17] S.A. Barter, L. Molent, R.J.H. Wanhill, Typical fatigue-initiating discontinuities in metallic aircraft structures, Int. J. Fatigue 41 (2012) 11–22.
[18] D. Spriestersbach, P. Grad, E. Kerscher, Influence of different non-metallic inclusion types on the crack initiation in high-strength steels in the VHCF regime, Int.
J. Fatigue 64 (2014) 114–120.
[19] J. Maciejkewski, The effect of sulfide inclusions on mechanical properties and failures of steel components, J. Fail. Anal. Prev. 15 (2015) 169–178.
[20] A. Sattar, M. Abbas, H.J. Hasham, Y. Baig, Experimental and analytical investigation of steel bolts failed after isothermal heat treatment, J. Fail. Anal. Prev. 15
(2015) 327–333.
[21] Y. Yamashita, Y. Murakami, Small crack growth model from low to very high cycle fatigue regime for internal fatigue failure of high strength steel, Int. J.
Fatigue 93 (2016) 406–414.
[22] R.K. Pandey, Failure of diesel-engine crackshafts, Eng. Fail. Anal. 10 (2003) 165–175.
[23] A. Ray, S.K. Dhua, K.B. Mishra, S. Jha, Metallurgical investigation of a premature failed roller bearing used in the support and tilting system of a steel making
converter used in an integrated steel plant, ASME Int, Pract. Fail. Anal. 3 (2003) 89–97.
[24] M.E. Stevenson, J.L. McDougall, R.D. Bowman, R.L. Herman, Failure analysis of a high-speed pinion shaft, ASME Int. Pract. Fail. Anal. 5 (2005) 48–54.
[25] C. Baldizzone, A. Gruttadauria, C. Mapelli, D. Mombelli, Investigation of failure in a crankpin of a motorcycle engine, J. Fail. Anal. Prev. 12 (2012) 123–129.
[26] C. Klinger, D. Bettge, Axle failure of an ICE3 high speed train, Eng. Fail. Anal. 35 (2013) 66–81.
[27] C.-P. Bork, Failure of a motor vessel's crankshaft 1968, Eng. Fail. Anal. 38 (2014) 66–80.
[28] A.I.Z. Farahat, A. Hamid, N. Gomaa, Failure analysis of train vehicles engagement arm, J. Fail. Anal. Prev. 15 (2015) 576–582.

237
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

[29] S. Papaefthymiou, A. Vazdirvanidis, G. Pantazopoulos, C. Goulas, Fatigue fracture of a high-resistance structural steel component destined to sustain severe
loads under service conditions, J. Fail. Anal. Prev. 17 (2017) 79–85.
[30] A. Gupta, S. Goyal, K.A. Padmanabhan, A.K. Singh, Inclusions in steel: micro-macro modelling approach to analyse the effects of inclusions on the properties of
steel, Int. J. Adv. Manif. Technol. 77 (2015) 565–572.
[31] D. Brooksbank, K.W. Andrews, Stress fields around inclusions and their relation to mechanical properties, J. Iron Steel Inst. 210 (1972) 246–255.
[32] M. Merkel, K.-H. Thomas, Taschenbuch der Werkstoffe, 7th ed., Fachbuchverl, Leipzig, 2008 (in German).
[33] H.-L. Yu, X.-H. Liu, H.-Y. Bi, L.Q. Chen, Deformation behavior of inclusions in stainless steel strips during multi-pass cold rolling, J. Mater. Process. Technol. 209
(2009) 455–461.
[34] N. Matsuoka, M. Terano, T. Ishiguro, E. Abe, N. Yukawa, T. Ishikawa, Y. Ueshima, K. Yamamoto, K. Isobe, Computer simulation of deformation behaviour of
non-metallic inclusion in hot rolling, Procedia Eng. 81 (2014) 120–125.
[35] F. Wang, J. Xu, J. Li, X. Li, H. Wang, Fatigue crack initiation and propagation in A356 alloy reinforced with in situ TiB2 particles, Mater. Des. 33 (2012)
236–241.
[36] B. Alfredsson, E. Olsson, Multi-axial fatigue initiation at inclusions and subsequent crack growth in a bainitic high strength roller bearing steel at uniaxial
experiments, Int. J. Fatigue 41 (2012) 130–139.
[37] R. Puff, R. Barberi, Effect of non-metallic inclusions on the fatigue strength of helical spring wire, Eng. Fail. Anal. 44 (2014) 441–454.
[38] A. Pineau, S.D. Antolovich, Probabilistic approaches to fatigue with special emphasis on initiation from inclusions, Int. J. Fatigue 93 (2016) 422–434.
[39] A.K. Jha, K. Sreekumar, Failure analysis of stainless steel stem used in explosive transfer assembly (ETA) of high performance solid propulsion pystem igniter, J.
Fail. Anal. Prev. 10 (2010) 77–81.
[40] C. Temmel, B. Karlsson, N.-G. Ingesten, Fatigue crack initiation in hardened medium carbon steel due to manganese sulphide inclusion clusters, Fatigue Fract.
Eng. Mater. Struct. 31 (2008) 466–477.
[41] L.E.K. Holappa, A.S. Helle, Inclusion control in high-performance steels, J. Mater. Process. Technol. 53 (1995) 177–186.
[42] J.T. Tchuindjang, J. Lecomte-Beckers, Fractography survey on high cycle fatigue failure: Crack origin characterisation and correlations between mechanical
tests and microstructure in Fe–C–Cr–Mo–X alloys, Int. J. Fatigue 29 (2007) 713–728.
[43] M.S. Moghaddam, F. Sadeghi, K. Paulson, N. Weinzapfel, M. Correns, V. Bakolas, M. Dinkel, Effect of non-metallic inclusions on butterfly wing initiation, crack
formation, and spall geometry in bearing steels, Int. J. Fatigue 80 (2015) 203–215.
[44] J.Y. Cogne, B. Heritier, J. Monnot, Cleanness and fatigue life of bearing steels, Clean Steel 3, Balatonfured, Hungary, The Institute of Metals, 2–4 June 1986, pp.
26–31.
[45] U. Krupp, Fatigue Crack Propagation Ion Metals and Alloys. Microstructural Aspects and Modelling Concepts, Wiley-VCH Verl. Gmbh & Co. KGaA, Weinheim,
2007, pp. 99–133 Chapter 5.
[46] C.G. He, Y.Z. Chen, Y.B. Huang, Q.Y. Liu, M.H. Zhu, W.J. Wang, On the surface scratch and thermal fatigue damage of wheel material under different braking
speed conditions, Eng. Fail. Anal. 79 (2017) 889–901.
[47] M.B. Leban, R. Tsiu, The effect of TiN inclusions and deformation-induced martensite on the corrosion properties of AISI 321 stainless steel, Eng. Fail. Anal. 33
(2013) 430–438.
[48] Q.Y. Wang, C. Bathias, N. Kawagoishi, Q. Chen, Effect of inclusion on subsurface crack initiation and gigacycle fatigue strength, Int. J. Fatigue 24 (2002)
1269–1274.
[49] H. Bomas, M. Schleicher, Application of the weakest-link concept to the endurance limit of notched and multiaxially loaded specimens of carburized steel
16MnCrS5, Fatigue Fract. Eng. Mater. Struct. 28 (2005) 983–995.
[50] J. Ma, B. Zhang, D. Xu, E.-H. Han, W. Ke, Effects of inclusion and loading direction on the fatigue behavior of hot rolled low carbon steel, Int. J. Fatigue 32
(2010) 1116–1125.
[51] E. Mikkola, G. Marquis, J. Solin, Mesoscale modelling of crack nucleation from defects in steel, Int. J. Fatigue 41 (2012) 64–71.
[52] S. Rödling, J. Fröschl, M. Hück, M. Decker, Einfluss nichtmetallischer Einschlüsse auf zulässige HCF-Bemessungskennwerte, Deutscher Verband für
Materialforschung und -prüfung (DVM), Report No. 137 (2010), pp. 135–145 (In German).
[53] M.H. El Haddad, T.H. Topper, K.N. Smith, Prediction of non-propagating cracks, Eng. Fract. Mech. 11 (1979) 573–584.
[54] K. Tanaka, Y. Nakai, M. Yamashita, Fatigue growth threshold of small cracks, Int. J. Fract. 1 (1981) 519–533.
[55] U. Zerbst, M. Madia, D. Hellmann, An analytical fracture mechanics model for estimating of S-N curves of metallic alloys containing large second phase
particles, Eng. Fract. Mech. 82 (2012) 115–134.
[56] N. Nadot, J. Mendez, N. Ranganathan, A.S. Beranger, Fatigue life of nodular cast iron containing casting defects, Fatigue Fract. Eng. Mater. Struct. 22 (1999)
289–300.
[57] N. Costa, N. Machado, F.S. Silva, A new method for prediction of nodular cast iron fatigue limit, Int. J. Fatigue 32 (2010) 989–995.
[58] J. Lankford, Inclusion-matrix debonding and fatigue crack initiation in low alloy steel, Int. J. Fract. 12 (1976) 155–156.
[59] G. Murtaza, R. Akid, Modelling short fatigue crack growth in a heat-treated low-alloy steel, Int. J. Fatigue 17 (1995) 207–214.
[60] Y.-Z. Wang, R. Akid, A. Clark, J.D. Atkinson, Further observations of early fatigue crack development, Fatigue Fract. Eng. Mater. Struct. 19 (1996) 623–627.
[61] J. Payne, G. Welsh, R.J. Christ Jr., J. Nardiello, J.M. Papazian, Observations of fatigue crack initiation in 7075-T651, Int. J. Fatigue 32 (2010) 247–255.
[62] B. Wisner, A. Kontsos, Investigation of particle fracture during fatigue of aluminium 2024, Int. J. Fatigue 111 (2018) 33–43.
[63] W.E. Duckworth, E. Ineson, The effects of externally introduced alumina particles on the fatigue life of En24 steel, Clean Steel, Iron Steel Inst. Sp. Rep. 77
(1963) 87–103.
[64] H.V. Atkinson, G. Shi, Characterization of inclusions in clean steels: a review including the statistics of extreme methods, Prog. Mat. Sci. 48 (2003) 457–520.
[65] C. Vallellano, M.R. Mariscal, A. Navarro, J. Dominguez, A micromechanical approach to fatigue in small notches, Fatigue Fract. Eng. Mater. Struct. 28 (2005)
1035–1045.
[66] M.F. Garwood, H.H. Zurburg, M.A. Erickson, Correlation of laboratory tests and service performance, interpretation of tests and correlation with service,
American Society for Metals (ASM), 1951, pp. 1–77.
[67] D. Rennert, E. Kullig, M. Vormwald, A. Esderts, D. Siegele, Analytical strength assessment of components made of steel, cast iron and aluminium materials in
mechanical engineering, FKM Guideline, 6th ed., Forschungskuratorium Maschinenbau (FKM), Frankfurt/Main, Germany, 2013.
[68] Y. Murakami, T. Endo, Effects of defects, inclusions and inhomogeneities on fatigue strength, Int. J. Fatigue 16 (1994) 163–182.
[69] M.D. Chapetti, Prediction of threshold for very high cycle fatigue (N > 107 cycles), Fatigue 2010, Procedia Eng. 2 (2010) 257–264.
[70] K. Tanaka, Fatigue crack propagation, in: R.O. Ritchie, Y. Murakami (Eds.), Comprehensive Structural Integrity; Volume 4: Cyclic loading and Fracture,
Elsevier, 2003, pp. 95–127.
[71] J.J. Scutti, R.M. Plloux, R. Fuquen-Moleno, Fatigue behaviour of a rail steel, Fatigue Fract. Eng. Mater. Struct. 7 (1984) 121–135.
[72] K.P. Balan, Failure analysis of a wire rope, ASME Int, Pract. Fail. Anal. 2 (2002) 71–74.
[73] J.M. Zhang, S.X. Li, Z.G. Yang, G.Y. Li, W.J. Hui, Y.Q. Weng, Influence of inclusion size on fatigue behavior of high strength steels in the gigacycle fatigue
regime, Int. J. Fatigue 29 (2007) 765–771.
[74] T. Makino, The effect of inclusion geometry according to forging ratio and metal flow direction on very high-cycle fatigue properties of steel bars, Int. J. Fatigue
30 (2007) 1409–1418.
[75] D. Ghosh, H. Roy, A.K. Shukla, Investigation of the probable cause of premature cracking of downcomer nozzle of heat recovery steam generator, J. Fail. Anal.
Prev. 9 (2009) 517–521.
[76] R. Prasannavenkatesan, J. Zhang, D.L. McDowell, G.B. Olson, H.-J. Jou, 3D modeling of subsurface fatigue crack nucleation potency of primary inclusions in
heat treated and shot peened martensitic gear steels, Int. J. Fatigue 31 (2009) 1176–1189.
[77] N.S. Cyril, A. Fatemi, Experimental evaluation and modeling of sulfur content and anisotropy of sulfide inclusions on fatigue behavior of steels, Int. J. Fatigue
31 (2009) 526–537.

238
U. Zerbst et al. Engineering Failure Analysis 98 (2019) 228–239

[78] R. Thumser, S. Kleemann, J.W. Bergmann, A. Kleemann, Investigation on defect distribution and its statistical evaluation for case hardened material states, Int.
J. Fatigue 41 (2012) 52–56.
[79] G. Hua, B. Zhao, Z. Zhao, C. Liu, A cluster of inclusions on Al–Si–Cu die casting cylinder block, Eng. Fail. Anal. 55 (2015) 370–375.
[80] D. Krewerth, T. Lippmann, A. Weidner, H. Biermann, Application of full-surface view in situ thermography measurements during ultrasonic fatigue of cast steel
G42CrMo4, Int. J. Fatigue 80 (2015) 459–467.
[81] B.Y. He, K.A. Soady, B.G. Mellor, G. Harrison, P.A.S. Reed, Fatigue crack growth behaviour in the LCF regime in a shot peened steam turbine blade material, Int.
J. Fatigue 82 (2016) 280–291.
[82] U. Zerbst, C. Klinger, D. Klingbeil, Structural assessment of railway axles – a critical review, Eng. Fail. Anal. 35 (2013) 54–65.
[83] C. Temmel, B. Karlsson, N.-G. Ingesten, Fatigue anisotropy in cross-rolled, hardened medium carbon steel resulting from MnS inclusions, Metal. Mater. Trans A
37A (2006) 2995–3007.
[84] E. Pessard, F. Morel, A. Morel, D. Bellet, Modelling the role of non-metallic inclusions on the anisotropic fatigue behaviour of forged steel, Int. J. Fatigue 33
(2011) 568–577.
[85] E. Pessard, F. Morel, D. Bellett, A. Morel, A new approach to model the fatigue anisotropy due to non-metallic inclusions in forged steels, Int. J. Fatigue 41
(2012) 168–178.
[86] A. Roiko, H. Hänninen, H. Vuorikari, Anisotropic distribution of non-metallic inclusions in a forged steel roll and its influence on fatigue limit, Int. J. Fatigue 41
(2012) 158–167.
[87] Y. Ochi, A. Ishii, S.K. Sasaki, An experimental and statistical investigation of surface fatigue crack initiation and growth, Fatigue Fract. Eng. Mater. Struct. 8
(1985) 327–339.
[88] A. Vasek, J. Polak, Low cycle fatigue damage accumulation in Armco-iron, Fatigue Fract. Eng. Mater. Struct. 14 (1991) 193–204.
[89] S. Meyer, E. Diegele, A. Brückner-Foit, A. Möslang, Crack interaction modelling, Fatigue Fract. Eng. Mater. Struct. 23 (2000) 315–323.
[90] G. Eisenmeier, B. Holzwarth, H.W. Höppel, H. Mugrabi, Cyclic deformation and fatigue behaviour of the magnesium alloy AZ91, Mater. Sci. Eng. A319-321
(2001) 578–582.
[91] S.P. Zhu, S. Foletti, S. Beretta, Evaluation of size effect in low cycle fatigue for Q&T rotor steel, Procedia Struct. Integr. 7 (2017) 368–375.
[92] R.O. Ritchie, Mechanisms of fatigue-crack propagation in ductile and brittle solids, Int. J. Fatigue 100 (1999) 55–83.
[93] J. Weertman, Dislocation crack tip shielding and the Paris exponent, Mater. Sci. Eng. A A468-470 (2007) 59–63.
[94] M.N. Ilman, R.A. Barizy, Failure analysis and fatigue performance evaluation of a failed connecting rod of reciprocating air compressor, Eng. Fail. Anal. 56
(2015) 142–149.
[95] S. Benhaddad, G. Lee, Role of microstructure in sucker rod string failures in oil well production, ASME Int, Pract. Fail. Anal. 1 (2001) 47–54.
[96] J. Fan, D.L. McDowell, M.F. Horstemeyer, K. Gall, Cyclic plasticity at pores and inclusions in cast Al–Si alloys, Eng. Frat. Mech. 70 (2003) 1281–1302.
[97] M.W. Tofique, J. Bergström, K. Svensson, Very high cycle fatigue of cold rolled stainless steels, crack initiation and formation of the fine granular area, Int. J.
Fatigue 100 (2017) 238–250.
[98] P. Grad, E. Kerscher, Reason for the transition of fatigue crack initiation site from surface to subsurface inclusions in high-strength steels, Fatigue Fract. Eng.
Mater. Struct. 40 (2017) 1718–1730.
[99] Y.-K. Gao, X.-B. Li, Q.-X. Yang, M. Yao, Influence of surface integrity on fatigue strength of 40CrNi2Si2MoVA steel, Mater. Lett. 61 (2007) 466–469.
[100] K.J. Kirkhope, R. Bell, L. Caron, R.I. Basu, K.-T. Ma, Weld detail fatigue life improvement techniques, Part I: Rev. Marine Struct. 12 (1999) 447–474.
[101] E.S. Alley, R.W. Neu, Microstructure-sensitive modeling of rolling contact fatigue, Int. J. Fatigue 32 (2010) 841–850.
[102] A.S. Pandkar, N. Arakere, G. Subhash, Microstructure-sensitive accumulation of plastic strain due to ratcheting in bearing steels subject to Rolling Contact
Fatigue, Int. J. Fatigue 63 (2014) 191–202.
[103] I. Corni, N. Symonds, C.E. Birrell, O.L. Katsamenis, A. Wasenczuk, D. Vincent, Characterization and mapping of rolling contact fatigue in rail-axle bearings, Eng.
Fail. Anal. 82 (2017) 617–630.
[104] J.J. Rydel, I. Toda-Caraballo, G. Guetard, P.E.J. Revera-Diaz-del-Castillo, Understanding the factors controlling rolling contact fatigue damage in VIM-VAR M50
steel, Int. J. Fatigue 198 (2018) 68–78.
[105] S. Beretta, A. Ghidini, F. Lombardo, Fracture mechanics and scale effects in the fatigue of railway axles, Eng. Fract. Mech. 72 (2005) 195–208.
[106] P.J. Laz, B.M. Hillberry, Fatigue life prediction from inclusion initiated cracks, Int. J. Fatigue 20 (1998) 263–270.
[107] J. Ekengren, J. Bergström, Extreme value distributions of inclusions in six steels, Extremes 15 (2012) 257–265.
[108] S. Saberifar, A.R. Mashreghi, A novel method for the prediction of critical inclusion size leading to fatigue failure, Metal. Mater. Trans B 43 (2012) 603–608.
[109] J.C. Newman Jr., E.P. Phillips, R.A. Everett Jr., Fatigue analyses under constant- and variable-amplitude loading using small-crack theory, NASA Report NASA/
TM-1999-209329; ARL-TR-2001 (1999).
[110] A.B. Schmiedt, H.H. Dickert, W. Bleck, U. Kamps, Evaluation of maximum non-metallic inclusion sizes in engineering steels by fitting a generalized extreme
value distribution based on vectors of largest observations, Acta Mater. 95 (2015) 1–9.
[111] B. Schork, P. Kucharzcyk, M. Madia, U. Zerbst, J. Hensel, J. Bernhard, D. Tchuindjang, M. Kaffenberger, M. Oechsner, The effect of the local and global weld
geometry as well as material defects on crack initiation and fatigue strength, Eng. Fract. Mech. 198 (2018) 103-122.
[112] Y. Murakami, T. Toriyama, E.M. Couders, Instructions for a new method of inclusion rating and correlation with the fatigue limit, J. Testing Eval. 22 (1994)
318–326.
[113] S. Beretta, C. Anderson, Y. Murakami, Extreme value models for the assessment of steels containing multiple types of inclusion, Acta Mater. 54 (2006)
2277–2289.
[114] S. Beretta, Y. Murakami, Largest extreme-value distribution analysis of multiple inclusions types in determining steel cleanliness, Metal. Mater. Trans. B 32B
(2001) 517–523.
[115] Y. Kanbe, A. Karasev, H. Todorki, P.G. Jönsson, Application of extreme value analysis for two and three-dimensional determinations of the largest inclusion in
metal samples, ISIJ Int. 51 (2011) 593–602.
[116] A. Cetin, A. Naess, G. Härkegard, A physically based extreme value characterization of material fatigue, Int. J. Fatigue 47 (2013) 216–221.
[117] ASTM, E2283, Standard Practice for Extreme Value Analysis of Nonmetallic Inclusions in Steel and Other Microstructural Features, American Society for
Testing and Materials (ASTM) West Conshohocken, (2003).
[118] J.-Y. Buffiere, Fatigue crack initiation and propagation from defects in metals: is 3D characterization important? Procedia Struct. Integr. 7 (2017) 27–32.
[119] D. Bettge, B. Günther, W. Wedell, P.D. Portella, J. Hemptenmacher, P.W.M. Peters, B. Skrotzki, Mechanical behavior and fatigue damage of a titanium matrix
composite reinforced with continuous SiC fibers, Mater. Sci. Eng. A 452-453 (2007) 536–544.

239