2019-High Cycle Fatigue Behavior of Hard Turned 300 M Ultra-High Strength Steel
2019-High Cycle Fatigue Behavior of Hard Turned 300 M Ultra-High Strength Steel
2019-High Cycle Fatigue Behavior of Hard Turned 300 M Ultra-High Strength Steel
High cycle fatigue behavior of hard turned 300 M ultra-high strength steel
⁎
J. Ajajaa, , W. Jomaab,1, P. Bocherb, R.R. Chromikc, M. Brochua
a
Department of Mechanical Engineering, Polytechnique Montréal, Montreal, Quebec, Canada
b
Department of Mechanical Engineering, École de technologie supérieure, ETS, Montreal, Quebec, Canada
c
Department of Materials Engineering, McGill University, Montreal, Quebec, Canada
A R T I C LE I N FO A B S T R A C T
Keywords: Although the effects of hard turning on surface integrity have been extensively studied over the past decades, the
Fatigue fatigue behavior of hard turned ultra-high strength steels is still not completely understood. This study in-
Crack initiation vestigated the influence of surface integrity characteristics generated by hard turning on the rotating bending
300M steel fatigue life of 55 HRC 300 M steel. Four finish cutting conditions were used to generate different combinations of
Hard turning
surface integrity characteristics. Resulting fatigue lives ranged between 1.01∙105 and 9.84∙106 cycles for a stress
Surface integrity
amplitude of 965 MPa. A single distribution was able to fit the entire fatigue life data showing that the finish
hard turning conditions evaluated are comparable. A fractographic analysis showed that specimens failed under
two different crack initiation mechanisms: surface feed marks and subsurface Al2O3∙CaO∙MgO inclusions.
Murakami’s Kmax equation for surface defects underestimates the severity of machined surfaces and cannot
explain the competition between the two crack initiation mechanisms.
⁎
Corresponding author.
E-mail address: [email protected] (J. Ajaja).
1
Currently affiliated with Centre technologique en aérospatiale (CTA), Cégep Édouard-Montpetit, Saint-Hubert, Quebec, Canada.
https://doi.org/10.1016/j.ijfatigue.2019.105380
Received 29 September 2019; Received in revised form 9 November 2019; Accepted 15 November 2019
0142-1123/ © 2019 Elsevier Ltd. All rights reserved.
Please cite this article as: J. Ajaja, et al., International Journal of Fatigue, https://doi.org/10.1016/j.ijfatigue.2019.105380
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Table 1
Chemical composition (wt.%) of 300 M ultra-high strength steel.
C Mn Si P S Cr Ni Mo V Cu Fe
300 M 0.36 0.84 1.53 0.008 0.001 0.73 1.76 0.39 0.07 0.25 Bal.
AMS6257E max. 0.44 0.90 1.80 0.010 0.008 0.95 2.00 0.45 0.10 0.35 Bal.
2
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Fig. 1. Microstructure of the material studied showing (a) martensite laths observed by electron channeling contrast imaging (ECCI) and (b) prior-austenitic grains.
3
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Table 4
Average surface roughness parameters.
Roughness parameters Polished Hard turning conditions
C0 C1 C2 C3 C4
Ra [µm] 0.00 ± 0.00 0.35 ± 0.05 1.10 ± 0.14 0.31 ± 0.08 0.30 ± 0.06
Rt [µm] 0.06 ± 0.01 3.43 ± 0.20 5.55 ± 0.23 3.94 ± 1.74 2.44 ± 0.74
Rq [µm] 0.01 ± 0.00 0.44 ± 0.05 1.26 ± 0.15 0.42 ± 0.14 0.37 ± 0.06
Rv [µm] 0.01 ± 0.00 1.64 ± 0.06 2.16 ± 0.08 1.27 ± 0.31 1.06 ± 0.37
Rsm [µm] 2 ± 0 100 ± 31 151 ± 0 234 ± 144 133 ± 68
Rku 3.11 ± 0.19 3.43 ± 1.00 2.04 ± 0.19 5.96 ± 3.24 3.00 ± 1.00
Rsk 0.24 ± 0.09 −0.36 ± 0.13 0.50 ± 0.11 1.00 ± 1.11 −0.07 ± 0.68
recognizable in conditions C1 (Fig. 4b), C3 (Fig. 4d), and C4 (Fig. 4e). machined layer (≈200 µm) by automated mechanical polishing gen-
For these conditions, the surface profiles are dominated by many sharp erated a minimum axial surface residual stress (ASRS) of −240 MPa,
peaks and low valleys, resulting in kurtosis Rku values above 3. Ad- consistent with the value of −230 MPa reported by Bag et al. [7] in
ditionally, surface roughness profiles for conditions C1 and C4 exhibit polished 54 HRC 300 M. Hard turning generated compressive axial
blunt features with deep scratches, resulting in negative skewness va- surface residual stress (ASRS) for all tested machining conditions, ex-
lues (Rsk = −0.07; −0.36), while condition C3 generated a surface cept for condition C2. Machining conditions with low feed rates (C1,
profile with shallow valleys and high peaks, resulting in a relatively C3, and C4) generated similar levels of compressive average ASRS
high positive skewness (Rsk = 1). (−301 to −438 MPa). Results for conditions C1 and C3 show that
The profiles presented in Fig. 4 along with the surface roughness increasing cutting speed from 85 to 100 m/min increases compressive
parameters presented in Table 4 show that conventional amplitude ASRS by 45% while shifting HSRS from tensile (131 MPa) to com-
roughness parameters are not sufficient to describe the complexity of pressive (−27 MPa). As shown in Fig. 5 where a comparison between
the surface features that can possibly influence fatigue life. For in- axial and hoop residual stresses is made, HSRS values are very low
stance, although cutting conditions C1, C3 and C4 have the same level compared to ASRS. Moreover, the maximum axial residual stress be-
of Ra, their Rsm and Rsk are significantly different. Thus, in the present neath the machined surface is higher and less deep when compared to
study, in addition to the conventional amplitude parameters, spacing hoop residual stresses.
and statistical parameters including the mean spacing of the asperities The hydrostatic residual stress provides additional information on
(Rsm), the kurtosis (Rku) and the skewness (Rsk) will be used to de- the residual stress state (compressive or tensile). It characterizes the
scribe the fatigue behavior of hard turned specimens. combined influence of axial and hoop residual stresses. The residual
stress state at each measured depth is assumed to be an in-plane stress
3.1.2. Residual stress state because (i) the depth of the residual stress profile is negligible
In-depth residual stresses were measured in the axial and hoop di- with respect to the specimen’s diameter and is considered to be the
rections as shown in Fig. 5. An average of three surface residual stress near-surface and (ii) the penetration depth of the X-rays in the radial
measurements was reported for each cutting condition as well as for direction is negligible. Thus, the hydrostatic stress σh values were cal-
polished specimens in Table 5. Mechanical polishing did not reduce the culated using Equation (1) and presented in Table 5 [12]:
level of axial surface residual stresses to zero. In fact, removal of the
Fig. 4. Surface roughness profiles for polished (C0, reference) and hard cutting conditions (C1 to C4): (a) C0, (b) C1, (c) C2, (d) C3 and (e) C4. Units on X and Y axes
given in micrometers. X represents a sample profile length while Y represents the height of the profile.
4
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Fig. 5. Axial and hoop residual stress profiles for cutting condition (a) C1, (b) C2, (c) C3 and (d) C4. The dashed lines are used to connect the measured data points to
account for the continuity of the residual stress profile between the measured points.
5
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Fig. 6. Hydrostatic stress profiles for cutting condition (a) C1, (b) C2, (c) C3 and (d) C4. Red circles represent the depth at which subsurface inclusions responsible for
crack initiations are located (see Section 3.2). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)
be discussed later. sample size n of 14 and a significance level of 0.05, a critical AD* value
A two-parameter Weibull distribution was used to describe the fa- of 0.74 is obtained [14]. Considering that AD < AD*, we can conclude
tigue life data. Such distribution is commonly used to model reliability with 95% confidence that a two-parameter Weibull distribution can fit
data and provides reasonably accurate failure predictions [13]. Eq. (2) adequately the fatigue life data. Consequently, we can also conclude
describes the probability density function (PDF) of the Weibull dis- that although the four cutting conditions studied generated different
tribution whereα and β ≥ 0 and represent respectively the scale and surface integrity characteristics, particularly C2, those differences did
shape parameters of the distribution function while x refers to fatigue not result in a level of variability in the fatigue data that would ne-
life Nf . cessitate to consider those conditions as non-equivalent in terms of
fatigue life and therefore, to fit one Weibull for each one of those
β x β − 1 −( x ) β
f (x ) = ⎛ ⎞ e α conditions.
α ⎝α⎠ (2) In summary, results demonstrated that different finishing cutting
The scale parameter α represents the characteristic life corre- conditions could generate similar fatigue lives. In fact, a single dis-
sponding to the number of cycles at which 63.2% of the population is tribution was effectively used to estimate the probability of failure in 55
expected to fail. The studied set of data was found to be best defined by HRC 300 M specimens hard turned using cutting conditions C1 to C4.
a α of 3.24∙106 and a β of 1.25 which describes a right-skewed dis- However, as shown in Fig. 9, the distribution was found to under-
tribution for which the data is significantly spread out. estimate the probability of failure of the material for a group of results
Fig. 9 represents the cumulative percent failure (CPF) plot showing corresponding to cutting conditions C2 and C4 (see red rectangle).
life data according to cutting conditions. To assess the goodness of the Therefore, the conclusions presented in this section are valid for the
two-parameter Weibull fit, the Anderson-Darling (AD) test statistic was small number of specimens tested only and further testing would allow
used and calculated as follows: to determine if each cutting condition would be better described by its
n
own distribution.
(2i − 1)
AD = −n − ∑ n
[lnF (Yi ) + ln(1 − F (Yn + 1 − i ))]
i=1 (3)
3.2.2. Fractographic analysis
where F represents the cumulative distribution function and Yi the To better explain the fatigue results obtained, a fractographic ana-
ordered life data. The calculated AD value of 0.285 was then compared lysis was performed on all tested specimens to determine the crack
to a critical AD value which depends on the type of distribution, sample initiation types leading to fatigue failure in hard turned 55 HRC 300 M
size and significance level. For a two-parameter Weibull distribution, a steel and to compare them with those leading to fatigue failure in the
6
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Fig. 7. SEM images showing the presence of a white layer thickness on specimens cut in the transverse direction for machining conditions (a) C1, (b) C2, (c) C3 and
(d) C4. Arrows show the distorted martensite laths.
Fig. 8. Fatigue life for polished and machined specimens. The extremes of a
typical box represent the minimum and maximum values while the top and Fig. 9. Weibull cumulative percent failure (CPF) plot based on the entire set of
bottom represent respectively the first and third quartiles. The x sign represents fatigue data and represented according to cutting conditions.
the mean value and the line the median.
7
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Fig. 10. Fatigue crack initiation mechanisms observed in polished 55 HRC 300 M tested under rotating bending fatigue at σa = 965 MPa. In order of increasing
fatigue life: (a) 6.8∙104, (b) 1.9 ∙105, (c) 3.2∙106 and (d) 8.3∙106 cycles.
mark initiation where several short cracks emanating from features in a subsurface inclusions-based initiations were characteristic of inclusions
feed length create a crack front propagating into the material. Fig. 11(b) with diameter di ranging between 16 and 23 µm and distance from the
shows a fish-eye of circular shape surrounding an Al2O3∙CaO∙MgO in- surface ds varying between 69 and 259 µm. All crack initiations in
clusion which has been commonly reported in high strength steels [15], specimens machined under cutting condition C2 occurred at feed marks
including in shot peened 54 HRC 300 M [7]. As reported in Table 6, while they occurred under both initiation mechanisms in specimens
Fig. 11. Fatigue crack initiation features observed in hard turned 55 HRC 300 M tested under rotating bending fatigue at σa = 965 MPa: (a) feed mark, Nf = 1.75∙106
cycles and (b) subsurface inclusion, Nf = 9.84∙106 cycles.
8
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Table 6
Summary of cutting conditions, crack initiation mechanisms, surface integrity characteristics and fatigue life for all machined specimens listed in order of increasing
fatigue life.
CC Initiation Type di [µm] ds [µm] Ra [µm] Rsm [µm] ASRS [MPa] Nf [cycles]
machined under cutting conditions C1, C3 and C4. As reported earlier, occur at the surface due to hard turning-induced surface alterations.
surfaces generated by C2 are characterized by tensile ASRS and the However, unlike expected, the previous analysis showed that 43% of
highest roughness amplitude parameters. fatigue failures were caused by cracks originating from material in-
Overall, results for machined specimens show that the shortest and clusions. Therefore, the following section will only consider the data
longest life times were the result of a specimen machined under C2 and representative of crack initiation at feed marks to evaluate the influence
C3 respectively. In the former case, a fatigue crack initiated at a feed of surface characteristics on fatigue life and determine if any parameter
mark while in the latter, a fatigue crack initiated at a subsurface in- can explain better the variability in the data.
clusion. However, as stated earlier, 86% of the machined specimens
failed at a number of cycles ranging between 106-107 cycles regardless 3.2.3. Influence of surface integrity on fatigue life
of the type of crack initiation. Therefore, the two crack initiation types Fig. 13 shows the influence of all surface integrity parameters
observed in finish hard turned 55HRC 300 M steel generated compar- characterized earlier on fatigue life. Results show a large dispersion in
able fatigue lives. Also, although a comparison with the reference ma- the data which makes difficult any attempt to assess the existence of
terial allows to conclude that the removal of the machined layer by trends between surface roughness parameters and/or residual stresses
mechanical polishing causes fatigue cracks to initiate mostly from and fatigue life. Therefore, to determine which characteristics among
surface inclusions, results for specimens machined under conditions C1, surface integrity features are most relevant to fatigue life, a multivariate
C3 and C4, show that the presence of compressive surface residual analysis based on the nonparametric Spearman’s rank-order correlation
stresses did not prevent the occurrence of surface crack initiations. Last, coefficient was performed.
as shown in Fig. 12 where fatigue life is plotted against crack initiation The Spearman’s correlation coefficient measures the strength of the
position, there is more dispersion in fatigue life when crack initiation monotonic association between two variables. Unlike the Pearson’s
occurs at the surface than when it occurs in the subsurface layers which correlation coefficient, it does not assume linearity of the relationship
may explain the larger dispersion in fatigue life for polished specimens between two variables [16] and it is robust against the presence of
illustrated in Fig. 8. outliers [17]. Its value ranges between ± 1, a value of 0 corresponding
Since 300 M steel (AMS6257E) is vacuum remelted to ensure its to no correlation, and its sign provides the direction of the relationship.
cleanliness, it was initially expected that crack initiation would only For instance, the coefficient is negative if one variable tends to increase
while the other decreases. For two variables x and y, the Spearman
correlation coefficient ρ is determined by calculating the Pearson cor-
relation coefficient between the ranks of two given variables:
n
∑i = 1 (Xi − X¯ )(Yi − Y¯ )
ρ=
(n − 1) SX SY (4)
where X̄ and Ȳ represent the sample mean rank of the first and second
variable, SX and S Y their standard deviation and n the number of ob-
servations. The Spearman correlation coefficients are listed in Table 7
along with their 95% confidence interval and corresponding p-values.
The relationship between two variables is considered to be statistically
significant if the correspond p-value is below or equal to the 0.05 alpha
level.
The interpretation of the Spearman correlation coefficients must not
be done without the 95% confidence intervals reported because each
interval contains the true population’s correlation coefficient for the
parameter evaluated at the confidence level chosen. The parameters
showing a statistically significant association with fatigue life at a 95%
Fig. 12. Fatigue life as a function of crack initiation position for each cutting confidence interval are surface roughness parameters Ra, Rq and Rsm.
condition studied (C1 to C4) and the reference material (C0). Crack initiations The corresponding correlation coefficients calculated are respectively
at feed marks are attributed a value ds = 0 µm while crack initiations at in- −0.59, −0.56 and 0.54 and thus, while fatigue life decreases with
clusions are attributed a value ds corresponding to the position of the center of increasing Ra and Rq, it increases with increasing Rsm. This result
the inclusion with respect to the surface of the specimen. shows that amplitude and spacing surface roughness parameters have
9
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Fig. 13. Influence of surface roughness parameters, surface residual stresses on fatigue life: (a) Ra, (b) Rv, (c) Rt, (d) Rq, (e) Rsk, (f) Rku, (g) Rsm, (h) ASRS and (i)
HSRS and (j) σh. Empty circles represent crack initiation at feed marks while filled circles represent crack initiation at subsurface inclusions.
an opposite effect on fatigue life although not all amplitude parameters between Ra and Nf. However, at a 95% confidence level, the calculated
were found to have a statistically significant association with fatigue Spearman coefficient can lie anywhere between −0.87 and −0.04
life. Further, the confidence intervals reported in Table 7 confirm the showing that we cannot conclude on the strength of the relationship
direction of the trends established by the small specimen size evaluated. between Ra and Nf with such a confidence level.
However, the wide range of values representative of the 95% con- In summary, results from the Spearman correlation analysis allow to
fidence intervals shows that this analysis do not allow concluding on conclude on the existence of a statistically significant correlation be-
the strength of the associations between Ra/Rq/Rsm and Nf. For in- tween surface roughness parameters Ra, Rq, Rsm and fatigue life at a
stance, a p-value of 0.03 indicates that there is a strong relationship 95% confidence interval. However, although the direction of the
10
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
Table 7 a a 2 a 3
Spearman coefficient for each parameter with respect to fatigue life Nf.
areaR = 2b·⎡2.97 ⎛ ⎞ − 3.51 ⎛ ⎞ − 9.74 ⎛ ⎞ ⎤
⎢
⎣ ⎝ 2b ⎠ ⎝ 2b ⎠ ⎝ 2b ⎠ ⎥
⎦ (9)
Parameter Spearman coefficient (p-value) 95% confidence interval
Because the Spearman’s analysis demonstrated the absence of a
Rv −0.19 (0.52) (−0.66, 0.39) statistically significant correlation between Rt and Nf, the calculation of
Rt −0.19 (0.45) (−0.66, 0.39)
areaR was performed using Ra for which a statistically significant
Ra −0.59 (0.03) (−0.87, −0.04)
Rq −0.56 (0.04) (−0.85, 0.00) correlation was found with Nf. This approach is in line with results
Rsk −0.13 (0.66) (−0.62, 0.43) reported in Itoga et al. [19] where fatigue limits for varying surface
Rku 0.34 (0.23) (−0.25, 0.75) roughness levels were calculated using Murakami’s equation based on
Rsm 0.54 (0.05) (−0.03, 0.85)
areaR . The authors used both Ra and Rt to calculate areaR and
ASRS −0.28 (0.33) (−0.71, 0.31)
HSRS −0.15 (0.61) (−0.63, 0.42) showed that the use of Ra resulted in better fatigue limit predictions.
σh −0.35 (0.22) (−0.75, 0.24) They suggested that a plausible explanation to this result is the over-
estimation of areaR with Rt because the value calculated represents
the area of a surface roughness profile with periodic cracks of depths
relationships between the different variables was established, no con- equal to Rt. Considering that a real surface roughness profile is con-
clusions could be drawn on their strength. stituted of randomly distributed features with depths mostly smaller
than Rt, it can be concluded that Ra provides a better estimation of
3.2.4. Influence of the maximum stress intensity factor Kmax on fatigue life areaR [19].
In a final attempt to identify a parameter that unifies fatigue results The highest Kmax values (3.04–3.89 MPa∙m½) were obtained for
generated with varying surface integrity characteristics, the maximum crack initiations at subsurface inclusions (SBI). Those inclusions are
stress intensity factor Kmax was calculated for all tested specimens using characterized with a size areainc varying between 15.1 and 25.4 μm½
Eqs. (5) and (6) [18] for respectively surface feed mark and subsurface and a position ds varying between 69 and 259 μm. Fig. 14(a) shows that
inclusions initiations where Kmax , surf , Kmax , sub , σnet , areaR and areainc the calculated Kmax , surf values for crack initiations at feed marks were
represent respectively the maximum surface intensity factor, maximum found to be lower than all the Kmax , sub values obtained. Since the reverse
subsurface intensity factor, net applied stress, Murakami’s equivalent trend is expected, this result implies that either no big enough inclusion
square root area function for surface roughness and Murakami’s square was found within a relatively small distance ds (maximum ds ob-
root area function for an inclusion. served = 259 µm) in specimens where failure was caused by a crack
originating from a feed mark feature, either the factor 0.65 in equation
Kmax , surf = 0.65·σnet · π areaR (5) (5) underestimates the severity of a surface defect such as surface
roughness. For comparison purposes, Kmax , surf was also calculated using
Kmax , sub = 0.50·σnet · π areainc (6) a = Rt as shown in Fig. 14(b). In this case, the values calculated show
that cutting condition C2, which is the only condition with tensile axial
The net stress represents the net applied stress at the crack initiation surface residual stresses, are greater than the obtained Kmax , sub values.
site position and can be calculated using Eqs. (7) and (8) for rotating This result demonstrates either that (i) Eq. (9) underestimates the
bending loading conditions where σapp (ds ) , σRS (ds ) and r represent re- areaR values characterizing a surface roughness profile with non-
spectively the applied stress at the position of the crack initiation site, uniform machining marks as reported by Bag et al. [20], or that (ii) the
the axial residual stress at the position of the crack initiation site and factor 0.65 in Eq. (5) underestimates the severity of a surface defect
the specimen’s radius. such as surface roughness, or both (i) and (ii). Since the threshold for
σnet = σapp (ds ) + σRS (ds ) crack propagation at R = −1 is in the range of 4.2 MPa∙m½, as reported
(7)
in Bag et al. [20], (i) calculated Kmax , sub values should correspond to the
d threshold, considering that material inclusions are the limiting factor
σapp (ds ) = 965 ⎛1 − s ⎞ for crack propagation when no more severe defect can be found at the
⎝ r ⎠ (8)
surface, and (ii) Kmax , surf values should fall above the threshold for crack
Murakami’s equivalent square root area function for surface propagation. This supports the hypothesis that Eq. (5) underestimates
roughness is represented by Eq. (9) [19] where a and b represent sur- the real Kmax values.
face roughness parameters Rt and Rsm respectively. In summary, Kmax was used to gather together fatigue data resulting
Fig. 14. Influence of the maximum stress intensity factor Kmax on fatigue life calculated using (a) a = Ra and (b) a = Rt. SBI and FM represent subsurface inclusions
and feed mark crack initiations respectively.
11
J. Ajaja, et al. International Journal of Fatigue xxx (xxxx) xxxx
from the propagation of cracks from two different features. Both defect interests or personal relationships that could have appeared to influ-
size and residual stresses were taken into account in the calculation of ence the work reported in this paper.
this factor. Results show that cutting conditions C1, C3 and C4 gener-
ated surfaces with similar Kmax levels while C2 generated surfaces with Acknowledgements
higher severity. Considering that a fatigue crack should only propagate
from a material inclusion if there is no surface defect with higher se- This work was supported by the Consortium for Research and
verity, results either demonstrate the limitation of Eq. (5) in char- Innovation in Aerospace in Quebec (CRIAQ), MITACS, the Natural
acterizing accurately Kmax values for machined surfaces, or that the Sciences and Engineering Research Council (NSERC), Héroux-Devtek
propagation of short cracks cannot be described directly using a Kmax and Pratt & Whitney Canada. The authors would like to acknowledge
value. Results depicted in Fig. 14 also demonstrate that finish cutting the technical assistance received from Cristina del Vasto (Héroux-
conditions C1, C3 and C4 cannot be optimized further for fatigue life for Devtek) and Nicolas Brodusch (McGill University) during this project.
the material studied considering that they all generate failures from
subsurface inclusions. Thus, if the need be, efforts to optimize fatigue References
life should be invested into making the material cleaner prior to any
attempt in optimizing further the machining process. [1] Dickinson GR. The influence of machining on the performance of ultra high strength
steels. ASME 1970 international gas turbine conference and products show.
American Society of Mechanical Engineers; 1970.
4. Conclusions [2] Matsumoto Y, Magda D, Hoeppner DW, Kim TY. Effect of machining processes on
the fatigue strength of hardened AISI 4340 steel. J Manuf Sci Eng Trans ASME
The main objective of the current study was to better understand the 1991;113(2):154–9.
[3] Siqueira C, Pereira C, Nascimento M, Voorwald H. Effects of nitriding and shot
fatigue behavior of finish hard turned 55 HRC 300 M steel. A total of peening treatments and stress concentration on the fatigue strength of AISI 4340
four cutting conditions were selected to generate a range of surface steel. SAE Tech Pap 2001.
characteristics representative of production finish hard turning cutting [4] Buddy Damm E. 4340 Steel fatigue initiation and performance in rotating bending
fatigue as a function of surface finish and melt practice. Mater Sci Technol 2005.
conditions. Specimens were tested under rotating bending fatigue under [5] Pistochini TE, Hill MR. Effect of laser peening on fatigue performance in 300M steel.
a stress amplitude of 965 MPa and a target of 106 cycles. Results from Fatigue Fract Eng Mater Struct 2011;34(7):521–33.
the analysis of the fatigue life and fractographic data allowed drawing [6] Hill MR, Pistochini TE, DeWald AT. Fatigue performance of laser peened materials.
2005; 41928: 203–7.
the following conclusions:
[7] Bag A, Delbergue D, Bocher P, Lévesque M, Brochu M. Statistical analysis of high
cycle fatigue life and inclusion size distribution in shot peened 300M steel. Int J
– A single distribution can be used to estimate the probability of Fatigue 2019;118:126–38.
failure in finish hard turned 55 HRC 300 M steel. This demonstrates [8] Ajaja J, Jomaa W, Bocher P, Chromik RR, Songmene V, Brochu M. Hard turning
multi-performance optimization for improving the surface integrity of 300M ultra-
that although cutting condition C2 generated surface integrity high strength steel. Int J Adv Manuf Technol 2019.
characteristics that are significantly different (tensile ASRS and [9] Petropoulos GP, Pandazaras CN, Davim JP. Surface texture characterization and
higher Ra) when compared to cutting conditions C1, C3 and C4 evaluation related to machining, in surface integrity in machining. London: Springer
London; 2010. p. 37–66.
(compressive ASRS, lower Ra), resulting fatigue lives do not reflect [10] Varela PI, Rakurty CS, Balaji AK. Surface integrity in hard machining of 300M steel:
this difference. effect of cutting-edge geometry on machining induced residual stresses. Procedia
– The removal of the machined layer by mechanical polishing causes CIRP 2014.
[11] Petropoulos GP, Pandazaras CN, Vaxevanidis NM, Antoniadis A. Multi-parameter
fatigue cracks to initiate mostly from surface inclusions in polished identification and control of turned surface textures. Int J Adv Manuf Technol
specimens resulting in more dispersed fatigue lives. 2006;29(1):118–28.
– No simple statistical model was able to describe the relationship [12] Pusavec F, Hamdi H, Kopac J, Jawahir IS. Surface integrity in cryogenic machining
of nickel based alloy—Inconel 718. J Mater Process Technol 2011;211(4):773–83.
between surface integrity characteristics generated by finish hard [13] Abernethy RB. Weibull handbook: reliability and statistical analysis for predicting
turning cutting conditions and fatigue life in samples where cracks life, safety, supportability, risk, cost and warranty claims. Dr. Robert B. Abernethy;
originated from surface feed marks. However, a Spearman correla- 2004.
[14] Evans JW, Johnson RA, Green DW. Two-and three-parameter Weibull goodness-of-
tion analysis demonstrates the existence of a statistically significant
fit tests vol. 493. US Department of Agriculture, Forest Service, Forest Products
correlation between surface roughness parameters Ra, Rq, Rsm and Laboratory; 1989.
fatigue life with 95% confidence. [15] Murakami Y, Usuki H. Quantitative evaluation of effects of non-metallic inclusions
– Finish cutting conditions C1, C3 and C4 cannot be optimized further on fatigue strength of high strength steels. II: Fatigue limit evaluation based on
statistics for extreme values of inclusion size. Int J Fatigue 1989;11(5):299–307.
for fatigue life because results demonstrate that fatigue cracks in [16] Hauke J, Kossowski T. Comparison of values of Pearson's and Spearman's correla-
specimens machined under those conditions propagated from sub- tion coefficients on the same sets of data. Quaestiones geographicae
surface inclusions which is the material’s limiting factor. 2011;30(2):87–93.
[17] Schober P, Boer C, Schwarte LA. Correlation coefficients: appropriate use and in-
– The maximum surface intensity factor Kmax was not effective in terpretation. Anesth Analg 2018;126:1.
unifying the fatigue life data resulting from two different crack in- [18] Murakami Y, Kodama S, Konuma S. Quantitative evaluation of effects of non-me-
itiation mechanisms. Kmax , surf values calculated using Murakami’s tallic inclusions on fatigue strength of high strength steels. I: Basic fatigue me-
chanism and evaluation of correlation between the fatigue fracture stress and the
equation demonstrated that the equation underestimates the se- size and location of non-metallic inclusions. Int J Fatigue 1989;11(5):291–8.
verity of a surface defect such as surface roughness. [19] Itoga H, Tokaji K, Nakajima M, Ko H-N. Effect of surface roughness on step-wise S-N
characteristics in high strength steel. Int J Fatigue 2003;25(5):379–85.
[20] Bag A, Delbergue D, Ajaja J, Bocher P, Lévesque M, Brochu M. Effect of different
Declaration of Competing Interest
shot peening conditions on the fatigue life of 300M steel submitted to high stress
amplitudes. Int J Fatigue 2020:130.
The authors declare that they have no known competing financial
12