Materials & Design 218 (2022) 110713

Contents lists available at ScienceDirect

Materials & Design

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

A benchmark activity on the fatigue life assessment of AlSi10Mg

components manufactured by L-PBF
S. Beretta a,⇑, L. Patriarca a, M. Gargourimotlagh a, A. Hardaker b, D. Brackett b, M. Salimian c, J. Gumpinger c,
T. Ghidini c
a
Politecnico di Milano, Dept. Mechanical Engineering, Via La Masa 1, I-20156 Milan, Italy
b
The Manufacturing Technology Centre (MTC), Coventry CV7 9JU, UK
c
ESA/ESTEC, European Space Research and Technology Center, Keplerlaan 1, NL 2200AG Noordwijk, Netherlands

h i g h l i g h t s g r a p h i c a l a b s t r a c t

A wide test campaign to compare

fatigue specimens with components
manufactured in AlSi10Mg by L-PBF.

There are several issues in the
trasnferrability of specimen fatigue
data to components.

A fracture-based fatigue assessment
is able to obtain good life predictions.

Fracture-based assessment considers
the effects of key parameters (defect-
seze, residual stresses).

a r t i c l e i n f o a b s t r a c t

Article history: One of the challenges associated with additive manufacturing (AM) is the definition of an assessment
Received 28 January 2022 route which considers the main process signatures of the AM process. To this end, this work presents
Revised 12 April 2022 a complete benchmark activity for the assessment of an AlSi10Mg component produced by a laser pow-
Accepted 30 April 2022
der bed fusion process, aimed at advancing the understanding of the fatigue resistance of AM materials
Available online 7 May 2022
with particular focus on the comparison between the fatigue performances of small coupons and demon-
strators. Four builds of AlSi10Mg specimen geometries were manufactured to: (i) determine the fatigue
Keywords:
curves for both as-built and machined conditions; (ii) measure the fatigue crack growth rate; (iii) produce
Additive Manufacturing
Laser Powder Bed Fusion
and test under fatigue a benchmark component used as a reference for the validation of the fatigue
AlSi10Mg assessment procedure. Tools and concepts of flaw tolerance were then used to perform the fatigue assess-
Defect ment of the benchmark component and were shown to be successful in the life prediction. Results
Fatigue propagation obtained from this wide database (related to internal defects and surface features) show that only a
Demonstrator fracture-based fatigue assessment is able to provide precise life estimates consistent with material crack
Assessment growth properties. Eventually, all the experimental results including specimens design, analysis of frac-
ture surfaces and raw tests’ data will be made available in a database which can be accessed and used by
the industrial and scientific communities to calibrate and validate alternative fatigue assessment proce-
dures of AM parts.
Ó 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

⇑ Corresponding author.
E-mail address: [email protected] (S. Beretta).

https://doi.org/10.1016/j.matdes.2022.110713
0264-1275/Ó 2022 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Nomenclature

a El-Haddad parameter
p0ffiffiffiffiffiffiffiffiffiffiffiffi
Symbols area0 El-Haddad parameter according to the Murakami’s
Sy yield stress equivalent crack size
Su ultimate tensile stress A0 parameter of Nasgro threshold equation
E elastic modulus A1 parameter of Nasgro threshold equation
Au ultimate tensile strain A2 parameter of Nasgro threshold equation
A, B equation parameters of the finite life region of the S/N A3 parameter of Nasgro threshold equation
diagram f crack closure factor
Dr range of stress a constraint factor
Drw endurance limit in presence of a defect C pth parameter of Nasgro threshold equation for positive R
Drw;0 theoretical endurance limit for the defect-free material Cm th parameter of Nasgro threshold equation for negative R
rres residual stress C parameter of Nasgro equation in the Paris regime
rmax ; rmin maximum and minimum stresses m parameter of Nasgro equation in the Paris regime
rmax;eff ; rmin;eff maximum and minimum effective stresses p parameter of Nasgro equation in the Paris regime
r0 flow stress Ra mean roughness value
DF range of axial force Rv maximum profile valley depth
Fmax maximum axial force F max;V cumulative probability of the largest defect a in a vol-
r11 ; r22 ; r12 residual stress components ume V
k; d ffi
pffiffiffiffiffiffiffiffiffi parameters of LEVD distribution
area Murakami’s equivalent crack size Abbreviations
a in-depth crack length MTC Manufacturing Technology Centre
c superficial crack length Polimi Politecnico di Milano
N number of cycles ESA/ESTEC European Space Agency
R stress ratio AM additive manufacturing
Reff effective stress ratio L-PBF laser powder bed fusion
RL effective load ratio RS residual stress
da=dN crack growth rate M machined
DK range of stress intensity factor AB as-built
DKth;LC range of stress intensity factor at the long crack thresh- SEB single-edge bending
old WEDM wire electric discharge machine
DKth range of stress intensity factor at the crack threshold SEM scanning electron microscope
Kmax maximum stress intensity factor CA constant amplitude
KC fracture toughness LR load reduction
DK1 range of stress intensity factor at the long threshold for SIF stress intensity factor
R!1
DK1 range of stress intensity factor at threshold for R ! 1

1. Introduction understood [15,16]. However, it has been recognised that one of

the challenges of today’s L-PBF AM techniques is the rather high
Additive manufacturing (AM), and in particular the laser- surface roughness and internal defects which strongly influence
powder bed fusion (L-PBF) process, has been analysed by acade- the fatigue performance [17–20]. The internal defects and surface
mia, agencies, and industry for more than two decades [1,2]. roughness and surface-defects can be relieved by different post
Whereas at the beginning it was considered a research field more processing techniques: the hot isostatic pressing techniques for
suitable for rapid prototyping, recently it has made the step internal defects [21–23]; post-processing techniques like laser re-
towards becoming an industrially applied process [3–9]. The melting, sand blasting, chemical or electrochemical processes for
intriguing possibilities of high design freedom, the possibility to surface features [24–27].
fully exploit topology optimisation, the integration of different It has been recognized that the typical scatter of fatigue proper-
parts or the fact that no tooling is needed make this manufacturing ties of AM materials [2] is due to the dependence of fatigue prop-
technique particularly attractive for high-tech industries like erties on the size of defects/inhomogeneities [28]. It follows that
biomedical [3], motorsports [10,11], shipbuilding [4], aviation [5] the fatigue life is strongly dependent on the position, shape and
and space [6,7,9,8,12]. These benefits are the main reasons why size of the so-called the killer defect [17] and it is the key ingredi-
AM has gained significant attention in the space industry. For this ent for the application of the damage tolerance approach to AM
industrial branch, the production volumes have traditionally been products [29]. On top of that, due to the particular processing con-
low, as typically only two or maybe three parts of the same design ditions of L-PBF machines, that involve melting, re-melting of the
are produced to cover development and qualification models and material, and constrained shrinking, high residual stresses (RSs)
the flight part. Exceptions to this are commercial constellation pro- are typically accumulated in the produced parts [30–32]. Heat
grammes like Oneweb [13], where, as of January 2021, 110 out of treatments applied after the AM process can effectively reduce
more than 6000 planned satellites have been launched [14]. internal stresses, but these are not expected to be removed com-
Certainly for aviation applications, but also for some of the pletely [33] and can be different on witness specimens and
space products, fatigue can be a driving factor for design. For tradi- components.
tionally produced materials, the fatigue behaviour has been stud- The literature on AM is rich with research works based on the
ied for decades, and influencing factors are rather well experimental characterisation of conventional coupons. Few works

2
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

have focused on the manufacturing and assessment of real compo- a rectangular section of 24 mm  10.6 mm. The notch (6 mm
nents [34–38], most of them concentrated on the application of length) was produced by means of wire electric discharge machin-
traditional fatigue approaches, and some others on the porosity/ ing (WEDM) with a wire diameter equal to 100 lm: this guaran-
defect assessment of printed parts [39–41]. However, as evidenced teed an initial sharp notch which favoured crack nucleation.
by A. Yadollahi and N. Shamsaei [15], changes in the size, geome-
try, or number of fabricated parts strongly influence the final
2.2. Geometry of the benchmark demonstrators
mechanical properties of the AM part. This means that the coupons
used for the characterisation might not be fully representative of
The benchmark components (Fig. 2) were manufactured and
the parts produced within the same build, especially for those
tested successively in the same two conditions of AB and M as
properties related to fatigue, rather than monotonic ones [42].
the cylindrical specimens. The critical locations of failure for the
It is then essential to further investigate and compare the fati-
AB benchmark components were characterised by the same rough-
gue properties of coupons and components for the sake of compo-
ness of the AB specimens, as it will be evidenced in the following.
nent qualification [43], with particular focus on the application of
The total height of the benchmark components was 180 mm. The
the assessment procedure [44] and the key features that dictate
benchmark component was designed to have two main branches
the fatigue performance such as the critical defects, the surface
that bifurcate from the top part and are connected at the bottom
roughness and the RSs.
part which is designed to accommodate a horizontal pin a with
For this purpose, this work defines a comprehensive benchmark
diameter of 7 mm. The bottom part’s width was 70 mm and the
activity to assess the fatigue performances of AlSi10Mg parts pro-
thickness 30 mm. The two parts that accommodate the horizontal
duced by the L-PBF process. This study features different sample
pin were each 20 mm thick. The top part was designed to have a
geometries and volumes (a total of 132 specimens and benchmark
threaded hole which was coaxial with the loading direction. The
components tested) which were designed and tested to generate a
benchmark components were all machined in the top and bottom
solid material database for the implementation of the life predic-
parts (region shaded light red, highlighted in Fig. 2) which were
tion procedure. A special demonstrator was designed, tested and
connected with the loading frame.
then used as a reference benchmark. This effort enabled to gener-
The comparison between the fatigue behaviour of cylindrical
ate a big database, which can be accessed by the scientific and
specimens and components requires the calculation of the local
industrial communities for the sake of verifying different fatigue
stresses at the failure location of the benchmark components. This
approaches and validation of fatigue assessment software (all the
calculation was done with a linear-elastic finite element (FE)
data generated are available through the Supplementary Material
model of the benchmark component. The finite elements adopted
of this manuscript).
were the quadratic tetrahedral elements of type C3D10. The aver-
The activity described herein dealt with different phases
age dimension of the finite elements was selected to be approxi-
explained in the following:
mately 0.5 mm in the region of the two branches and
approximately 1 mm for the top and bottom ends. The total num-

Manufacturing of L-PBF specimens and benchmark
ber of elements was 1.28106 for a total number of 1.8106 nodes.
components;
The material was modelled as linear-elastic since the yield condi-

Fatigue testing of specimens and benchmark components to
tion was not reached in any point for the maximum applied force.
provide material properties and identify the critical features
The axial force (1 kN) was applied to a reference point which
(defects/surface features) at the origin of failures;
was constrained to move only vertically and was kinematically

Set-up of a fracture mechanics-based fatigue life prediction
coupled with the internal surface of the top part of the benchmark
model;
component (Fig. 2). In the bottom part, a second reference point

Prediction of specimens and benchmark component test results.
was positioned in the x-y plane of symmetry and in the same posi-
tion as the axis of the horizontal pins passing towards the two
2. Test campaign preparation
holes. The reference point was coupled with the two surfaces of
the holes to have the same relative displacements in the Y and Z
2.1. Geometry of the specimens
directions being free to move along X and rotate around X. Finally,
the reference point was fully constrained. This allowed us to nor-
The test campaign for this benchmark activity was based on
malise the local stress values in the points of interest and simply
AlSi10Mg manufactured by L-PBF, because this alloy has been
calculate the local stresses for all the loads applied during the tests
widely adopted for AM space components [12] and its process-
(see Section 5.3).
properties-performance has been already studied at ESA [45].
The benchmark components are characterized by three highly
The geometry of the specimens adopted for the characterisation
stressed regions: P1 at the neck of the upper head, P2 at the inser-
of the fatigue properties are reported in Fig. 1. The endurance lim-
tion of the legs onto the upper head and P3 in the middle of the
its and the stress versus the number of cycles to failure (S/N) dia-
legs (at the inner surface).
grams were investigated according to the cylindrical geometry
depicted in Fig. 1a. The gauge length of these specimens was
16 mm, while the diameter of the cross-section selected was 2.3. 3D printing and manufacturing of test pieces
6 mm. The fatigue specimens were produced in two different con-
ditions: as-built (AB) and machined (M). For the AB condition, only All the specimens used in this study were manufactured by the
the two specimens’ ends were machined to obtain the proper geo- UK National Centre for AM at the Manufacturing Technology Cen-
metrical tolerances required for the fatigue tests. The machining of tre (MTC). L-PBF AM was used to manufacture the parts in
the M specimens was performed instead on the entire length to AlSi10Mg material and the specific machine was an EOS M280
completely remove the external machining stock layer, which which has a build volume of 250  250  325 mm3 (XYZ). Follow-
was 1.09 mm for the areas of interest (expected failures) and ing a brief parameter development study the machine parameters
0.09 mm for the attachment faces. that were found to offer the highest density levels were: power
The single-edge bending (SEB) specimens (Fig. 1b) were 370 W, speed 1300 mm/s, hatch distance 0.19 mm, layer thickness
designed to perform crack propagation tests under different load 30 lm, energy density 49.9 J/mm3, and using a carbon brush
ratios. The nominal dimensions were 110 mm in length and with recoater.
3
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 1. Geometry of the specimens adopted in this study: (a) cylindrical specimens for fatigue characterisation; (b) the SEB specimen for fatigue crack growth
characterisation.

The parts listed in Table 1 were arranged as shown in Fig. 3. 2.4. Machining of the benchmark components
Four identical builds were manufactured and they are indicated
as 242, 243, 244 and 245. All parts were built directly onto the All benchmark components were machined in the AB condition
build plate with only the benchmark components requiring sup- without any heat treatment to relieve RSs. This lack of heat treat-
ports. 1 mm of stock was added to the base of all benchmark com- ment increased the risk of deformation during the machining pro-
ponents to allow the WEDM cut to remove them from the base cess which took a significant number of iterations to achieve the
plate. required tolerances. The central surfaces of the benchmark compo-
Results of the metallurgical analysis of the 4 cubes from each nents were machined in one single operation without using a con-
build (ref HC-X) to measure porosity are shown in Table 2 and give ventional strategy consisting of roughing, semi-finishing and
an indication of the material quality of the built components. The finishing as this was also found to reduce distortion during the
processing of the machine is shown to be relatively stable across machining. This method also allowed access so that both sides of
all builds with part porosities in the region of 0.1–0.2% which is the benchmark component could be machined in the same opera-
in-line with expectations from the parameter development work. tion, eliminating the need to attempt to account for distortion
The built samples were cut off the build plate by WEDM and the when flipping the benchmark component to machine the opposite
parts were cleaned in an ultrasonic bath and rinsed with iso- face. A final roughness for the M specimens of Ra <0.8 lm was
propanol to ensure they were not stored while wet. As a final reached. The AB benchmark components were also partially
check, before passing the benchmark components on to machining, machined following a similar procedure but with the first stage
one of each part type was optically scanned to ensure there was omitted and more attention paid to mounting the part given the
enough stock material to machine back to the final state. increased surface roughness in the central region, which meant tol-
erances had to be relaxed.
4
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 2. The component geometry for the benchmark activity.

Table 1 crystal lattice with strong anisotropy can be obtained. Strategies

Full list of components per build. where the scan vectors are always kept parallel from one layer to
Part Name Part Label Quantity the next can promote the formation of elongated grains at the cen-
Benchmark component fully machined BM 1–4 4
treline of the tracks, growing in build direction (z) over several lay-
Benchmark component partially machined BN 1–4 4 ers. According to the authors in [51], this anisotropic crystal lattice
Fatigue specimen fully machined FM 1–7 7 formation can be interrupted by rotating the scan vector direction
Fatigue specimen partially machined FN 1–7 7 by 90 degrees for each layer. The strategy applied in this work
Tensile specimen T 1–4 4
applies a scan vector rotation of 67 degrees between layers and
Fatigue crack growth specimens FCG 1–3 3
Metallurgy carrier samples HC 1–4 4 it is believed to interrupt the formation of long, elongated grains
over several layers. Since some of the parts produced in this work
were large in volume, the microstructure was characterised in dif-
ferent locations along the build direction to verify the homogeneity
3. Material and experiments
of the microstructure. Fig. 4 shows the micrographs of the melt
tracks for three different positions of a failed benchmark compo-
3.1. Material and microstructure
nent. The specimens for the microstructural characterization were
cut from the b-c-d positions indicated in the figure where the load-
AlSi10Mg is surely one of the most adopted metal alloys for L-
ing was not expected to induce changes in the microstructure. The
PBF [46]. AlSi10Mg is also used in traditional casting processes
etched sections show the same features in all the observed posi-
where, due to slow cooling conditions, the solid solution of silicon
tions providing the evidence of a uniform process towards the
in aluminium is transformed into an aluminium–silicon eutectic
building direction.
microstructure with dispersed primary a-Al [47]. The morphology
of AlSi10Mg obtained through L-PBF processes with the inherently
high melting and cooling rates is significantly different and can be
3.2. Experimental planning
described with macro- and micro-structural features. The
microstructure of AM AlSi10Mg is mainly driven by the applied
The experimental planning related to this wide experimental
scanning strategy [47]. For this alloy, the applied scanning strategy
campaign is summarised in Table 3. A total of 8 SEB specimens
is characterised by rotations between layers of 67 degrees. Accord-
were tested and an additional set of 5 SEB specimens from a differ-
ingly, the majority of melt pool boundaries on a vertical plane are
ent manufacturer (same nominal alloy) was added to further con-
elliptically shaped, whereas longitudinally shaped boundaries can
firm the results and support the crack growth data acquired. The
be seen on horizontal planes. AlSi10Mg can precipitate Mg2Si par-
cylindrical specimens in the AB and M conditions were produced
ticles naturally or through dedicated heat treatments, which
from the AM builts 242, 243 and 245. A total of 23 AB specimens
increases the strength [48]. Zhang et al. [49] describe that the
and 17 M specimens were used to characterise the entire S/N dia-
specific processing conditions of L-PBF lead to a microstructure
gram, from the endurance limit to the finite-life regions. 15 bench-
where some of the Si is dissolved in the face-centred cubic a-Al
mark components were produced in the AB conditions and 15 in
matrix, resulting in a cellular dendritic a-Al phase. Si and Mg2Si
the M conditions, for a total of 30 benchmark components. It
particles are assumed to precipitate in this alloy and under these
should be noted that the number of tested specimens is compliant
processing conditions [50]. The solid solution of Si was also found
with the ECSS standard that prescribes the determination of the
to decrease the lattice constant of a-Al [49]. The cell size is in the
entire S/N diagram with at least 20 specimens for each condition.
sub-micron size range and, depending on the scanning strategy, a
In this study, this requirement is fulfilled for both AB and M spec-
5
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 3. (a) Image of build layout and (b) photograph of 1 of the 4 identical manufactured builds.

imen conditions considering that the run-out specimens were suc- 3.3. Fatigue tests of cylindrical specimens
cessively re-tested at higher stress levels (test numbers in brackets
in Table 3). The axial fatigue tests on the cylindrical AB and M specimens
were performed under a uniaxial Instron Electronpulse E10000

6
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 2 and size. These characteristics influence the fatigue performance

Porosity measurements of the metallurgy samples in each builds. of the material and their quantification is important in relation
Build Number Sample Porosity Mean Pore Size to the calculation of the stress intensity factor (SIF), which is the
(%) (lm2) driving force parameter selected for the characterisation of the
Build 1 HC-1 0.146 59.3 crack propagation process.
HC-2 0.133 58.1
HC-3 0.141 54.7
HC-4 0.106 56.0
3.4. Crack propagation tests
Build 2 HC-1 0.127 61.4
HC-2 0.172 74.7
HC-3 0.157 66.0 The crack propagation tests were performed using a resonant
HC-4 0.144 63.7 Rumul loading frame. The SEB specimens were subjected to a sinu-
Build 3 HC-1 0.128 60.8
soidal bending moment at a frequency of approximately 90 Hz. The
HC-2 0.144 75.2
HC-3 0.212 74.6 crack advancement was monitored continuously by means of two
HC-4 0.176 70.8 alternative techniques: i) the specimen’s compliance measure-
Spare Build HC-1 0.144 74.6 ments or ii) surface crack measurements. A Rumul clip-on-gauge
HC-2 0.145 75.6 was used to detect the specimen’s compliance, which depends on
HC-3 0.157 72.4
the crack length; this technique was adopted for positive load
HC-4 0.148 67.6
ratios. For negative load ratios, the measurement of the crack
length was performed using two lateral resistance foils produced
by Rumul (krak-gages); in this case, the change of resistance of
machine equipped with a 10 kN load cell. The tests were conducted
the foils is correlated with the crack advancement. In both cases,
in load amplitude control at a nominal load ratio of R = 0.1 and at a
after the tests, the specimens were broken at liquid nitrogen tem-
frequency of approximately 35 Hz. A reduction of 10% stiffness was
perature and the final crack length was measured precisely on the
considered as the failure condition, while the tests that reached
fracture surfaces to correct the experimental data acquired. The
5106 cycles without failure were interrupted and classified as correction of the crack length was effective for the calculation of
run-outs. The run-out specimens were successively re-tested at the crack growth rate da=dN and for the calculation of the range
higher load levels to feed the finite-life region of the S/N diagram. of SIF DK. Before testing, the SEB specimens were pre-cracked to
After testing, all the specimens were statically broken in two sep- obtain an initial crack with limited load history effects, with the
arate parts with the aim of inducing a final brittle failure through aim of reducing the effect of early plastic wake developed with
the use of liquid nitrogen to reveal and analyse the fracture sur- positive pre-cracking loadings on the obtained long crack thresh-
faces. The images of the initial defects were captured with the olds DK th;LC [45]. The pre-cracking procedure was implemented
scanning-electron microscope (SEM) at different magnifications. pffiffiffiffiffi
with a load ratio of R = 10 and a constant DK=2.2 MPa m. An
The aim of this investigation was to precisely locate the killer
average pre-crack length from 50 lm to 100 lm was measured
defect in terms of the following characteristics: position, shape
for all the specimens.

Fig. 4. Microstructure on a vertical plane of one of the benchmark components: (a) overview, indicating three areas of interest, (b) microstructure in lower area, (c)
microstructure in intermediate area, (d) microstructure in upper area.

7
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 3 series. The RS measurements were taken in the mid axial length
Experimental planning of this study. of the specimens in four different locations along the circumfer-
Specimen Condition AM build Number of specimens ence at the center. The longitudinal component (parallel to the
Tensile Machined 242 4 specimen axis) of RSs was calculated.
243 1 The RS measurements for the benchmark components were
244 2 performed by means of a Bruker D8 Discover diffractometer
245 3 equipped with VANTEC-500 area detector with a Cu-Ka radiation
SEB Machined 243 2
244 3
at 40 kV and 50 lA and a 1 mm collimator size. The lattice strain
245 3 of 422 and 420 lattice planes at 2H = 116.56 and 2H= 137.46
Cylindrical As-Built 242 8 (1) were measured on different points of each component. The mea-
(Fatigue) 243 6 (1) surements were performed on different W angles from 45 to
245 9 (2)
45 . The Leptos software was used to calculate the RSs using sliding
Cylindrical Machined 242 5 (1)
(Fatigue) 243 7 (2) gravity and a biaxial shear stress model.
245 5 (1)
Benchmark component As-Built 242 3
3.7. Roughness measurements
243 5 (1)
244 2
245 5 (1) The surface roughness measurements on AB cylindrical fatigue
Benchmark component Machined 242 3 and benchmark components were carried out using a Keyence VK-
243 4 X1000 Confocal Microscope (Keyence Corporation, Osaka, Japan) at
244 3
245 5 (1)
15X magnification. The roughness measurements for cylindrical
fatigue specimens were done on 4 areas centred in the gauge
length on the surface with a 90 degree orientation towards each
other. The same measurements were implemented for the bench-
The crack propagation tests were conducted according to two
mark components close to the fracture surface area of the failed
different strategies: i) constant amplitude (CA), where the range
part. The following steps were taken into consideration while mea-
of applied bending moment was kept constant for the test, leading
suring the surface parameters:
to an increased DK; ii) load reduction (LR), where the range of
applied bending moment was decreased to allow the applied DK

Cylindrical specimens: three measurements with lengths of
to reach the DK th;LC . The CA is adopted to measure the crack growth
5.6 mm were taken in the vicinity of crack initiation sites as
rates in the so-called Paris region. The LR procedure, instead,
shown in Fig. 5a.
enables us to characterise the knee-region of the crack propagation

Benchmark components: a side surface(s) corresponding to the
curve and the DK th;LC . It should be noted that, before starting the
crack initiation site was chosen for surface evaluation. Three
effective LR procedure, an initial CA procedure is also implemented
measurements with lengths of 5.6 mm were taken (Fig. 5b). In
to develop an initial plastic wake and stabilise the level of crack
case of a corner crack, four measurements (two measurements
closure for the proper load ratio under investigation. The crack
on each side) were taken as shown in Fig. 5b.
advancement for this initial CA step is typically 1.5 mm. Following

Maximum Rt and Rv values of all measurements for each spec-
the tests, the crack propagation rates were corrected according to
imen or benchmark component were considered for further
the final crack length and calculated adopting a crack advancement
analysis.
of 50 lm.

4. Results
3.5. Fatigue tests of benchmark components
4.1. Tensile properties
The fatigue tests on the benchmark components were con-
ducted on two different machines: (i) for the tests requiring a max-
The tensile tests were conducted on cylindrical specimens with
imum load lower than 10 kN, an Instron ElectroPulsTM E10000
a nominal diameter of 6 mm according the ASTM E8/E8M-21 stan-
machine equipped with a 10 kN load cell was used; (ii) for the tests
dard [52] that prescribes a strain rate of approximately 0.015 mm/
requiring a load higher than 10 kN, a servo-hydraulic fatigue test-
mm/min. The number of tests was 10, as also indicated in Table 3.
ing system called Instron 8802 equipped with a 250 kN load cell
The tensile properties of the present AlSi10Mg alloy are sum-
was used. All the tests were conducted at a load ratio of R = 0.1
marised in Table 4, where both mean values and standard devia-
with a frequency of 9 Hz and 20 Hz, respectively. Those tests in
tions are given. The mean elastic modulus was measured to be
which the component did not break until 107 cycles were consid- E = 69492 MPa. The yield stress Sy =258.4 MPa was determined as
ered runouts and the components were re-tested for the higher the proof stress for a nominal plastic deformation of 0.2%. The ulti-
load levels. After failure, the fracture surfaces were then observed mate tensile stress was Su =469.1 MPa and, similar to the yield
under SEM to reveal the inhomogeneities at the origin of fatigue stress, shows a very low scatter. The elongation at fracture was
failures. A = 8.34%.

3.6. Residual stress measurements 4.2. Fatigue tests on cylindrical specimens

The RSs due to the manufacturing and machining processes Fig. 6 considers a comprehensive overview of the test results
were measured for the cylindrical fatigue specimens in different obtained on the AB and M cylindrical specimens tested under con-
locations for both the AB and M conditions. In addition, a total of stant amplitude fatigue cycles at load ratio R = 0.1. The number of
five crack propagation SEB specimens were also analysed. For all repetitions of the specific combination of specimen type (AB versus
these specimens, the RS measurements were carried out by means M) and Dr was chosen to be approximately 3 with some excep-
of an AST X-Stress 3000 portable X-ray diffractometer using the tions for the endurance limit region where the lowest stress range
sinð/Þ2 method. Two specimens were chosen from the AB and M was repeated only with two tests. The continuum solid lines were
8
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 5. Roughness measurements: position of the roughness measurements for the (a) cylindrical and (b) benchmark components.

Table 4 rosion pits displayed a systematic lower fatigue life. For this rea-
AlSi10Mg properties (mean and standard deviation values) obtained from the tensile son, the analysis of this data is provided considering the separate
tests.
cases of including and excluding those four data points. The fatigue
Property Mean Standard deviation tests performed on the AB and M specimens and summarised in
Ultimate Tensile Stress (Su ) 469.1 MPa 1.5 MPa Fig. 6 clearly indicate a strong influence on the fatigue perfor-
Yield Stress (Sy ) 258.4 MPa 3.8 MPa mances of the AB condition which is also relevant when analysing
Elastic Modulus (E) 69492 MPa 1511 MPa the parameters of the fitted type in bi-logarithmic scale
Elongation at fracture (Au ) 8.34 % 0.46 %
logðNÞ ¼ logðAÞ þ B  logðDSÞ and the endurance limits DrAB w and
DrMw , see Table 5.
The typical defects detected from the fracture surfaces are
depicted in Fig. 7. The fatigue strength of M specimens was con-
trolled by the presence of small defects and pores close to the sur-
face. For AB specimens failures were triggered by the presence of
the surface features observed on net-shape surfaces. The SEM
results show that, while all the cracks were initiated from the sur-
face, the types of defects are considerably different:

Semi-circular defects: pores and trapped particles appear as

defects with equal depth to length ratio which were observed
more frequently on the surface of M specimens. These defects,
which occurred prior to machining the surface, were below
the surface and brought to the surface by the process of machin-
ing (Fig. 7a-b).

Elongated defects: this type of defect is characterised by a low
depth to length ratio and observed over the circumference of
a specimen which is a combination of defect clusters and sur-
face valleys. This type of defect was mostly present in AB spec-
imens and led to several crack initiation sites (Fig. 7c-d).

Corrosion pits: this type of defect was observed on a few spec-
imens and is characterised by a relatively large defect area and a
Fig. 6. S/N diagram for the cylindrical specimens, arrows indicate specimens
retested at high Dr after run-out condition. high percentage of oxygen (in the vicinity of a corrosion pit)
with respect to the other types of defects (Fig. 7e-f). Interest-
ingly, only the specimens manufactured from AM build 242
defined according to the interpolation of the failures with the were characterised by corrosion pits. This was also confirmed
equation of the type N ¼ A  ðDSÞB by means of the least square
method (see ASTM-E739 standard [53]). The endurance limits
DrAB
w and Drw were calculated instead by applying the Dixon up
M

and down method for this case of short staircase sequence [54]. Table 5
Summary of the fatigue properties of the present AlSi10Mg alloy in the AB and M
One important thing to note from the results contained in Fig. 6
conditions.
relates to the evidence of a sub-category of the results pertaining
to the AB condition. As will be shown subsequently, some of the Condition logðAÞ B Drw (MPa)
AB specimens (yellow dots in the S/N diagram of Fig. 6) were char- As-build (all) 13.53 -4.09 48
acterised by the presence of superficial corrosion pits. In between As-built (excluding corrosion pits) 14.25 -4.47 55
Machined 19.90 -6.54 152
the AB specimen tests, those specimens characterised by these cor-
9
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 7. Defects detected on the cylindrical specimens: (a) and (b) show two examples of defects detected from the M specimens; (c) and (d) for the AB specimens, while (e)
and (f) show the origin of failure for the AB specimens which were observed to denote corrosion pits on the surface.

by the high content of oxygen (between 25% to 50%) detected in 4.4. Crack propagation tests
the proximity of these defects (points indicated with A, B and C
in Fig. 7e-f). The data acquired on the SEB specimens show that the crack
growth rate correlates with the crack driving force parameter DK
4.3. Residual stresses on cylindrical specimens under the assumption of linear elastic fracture mechanics. The long
crack SIF threshold DK th values are reported in Fig. 8a as a function
Table 6 reports surface RS measurements for two AB and M of the load ratio R. The data acquired from the current AM builds
specimens in terms of average of the 4 measurements with the are indicated with the circular solid black dots (labelled with ESA
amximum deviations. The results show tensile and compressive in the legend). The data were then integrated with additional
stresses for AB and M specimens, respectively. The presence of ten- DK th values from the PoliMi database characterised by the same
sile stresses of AB specimens is due to the combination of two fac- nominal alloy composition and the same printing direction [58].
tors: the temperature gradients in the AM process and the cooling These data points are indicated with the square blue symbols
down phase of the molten top layers [55]. On the other hand, for and they integrate the data obtained from the current AM builds.
the M specimens, the machining process lead to compressive RSs The two databases considered were used successively to fit the
[56]. NASGRO equation for the long thresholds which is indicated in
The values here obtained are in good agreement with X-ray Fig. 8a with a solid red continuum line. The equations used to fit
measurement by Sausto on AlSi10Mg specimens [57]. the DK th values are as follows:
 
f ¼ max R; A0 þ A1 R þ A2 R2 þ A3 R3
Table 6
h ið1þRCp Þ RP0 ð1Þ
Average surface RS results of AB and M specimens in four different positions. 1R
th

DK 1
1f
Specimens Sres (MPa) Deviation (MPa) DK th ¼ ð1RÞC
p
ð1A0 Þ th

M cylindrical -76 9.4

AB cylindrical 56 14.9

10
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

   1=a
A0 ¼ 0:825  0:34a þ 0:05a2 cos p2 rrmax
0

A1 ¼ ð0:415  0:071aÞ rrmax

0 ð4Þ
A3 ¼ 2A0 þ A1  1
A2 ¼ 1  A0  A1  A3

The constraint factor a = 1.9 considers the condition of dominant

plain strain for the threshold test and the ratio between the maxi-
mum stress and the flow stress is herein considered to be constant
and equal to rmax =r0 =0.3.
The data points indicated in Fig. 8b show an evident effect of the
RSs for the AB specimens. In particular, it can be noted that the
specimens FC1-243 (R = 0.7), FC3-243 (R = 0.1) and FC1-244
(R = 1) almost overlap despite the different load ratio. This result
clearly demonstrates the high level of RSs. It is also expected that,
with crack propagation, these stresses tend to relax, and this fur-
ther complicates the proper measurements of the crack propaga-
tion curves for the specific load ratio under investigation. For this
reason, we implemented some heat treatments as indicated in
the legend of Fig. 8b. The solid black continuum lines indicate
the correspondence with the NASGRO equation considering all
the crack growth data shown in the plot:
 p
  n 1  DDKKth
da 1f
¼C  q ð5Þ
dN 1R 1  KKmax
C

The parameters C, n and p are obtained from the fitting of the exper-
imental data (Table 7), while the parameter q is set to zero as it cor-
responds to the part of the NASGRO equation that describes the
unstable crack propagation that occurs when K max approaches the
fracture toughness K c .

4.5. Fatigue tests on benchmark components

The fatigue tests on the M benchmark components were per-

formed based on four load levels (Fig. 9), while the AB benchmark
components were tested at five load levels. The solid dots (blue for
M benchmark component and black for AB benchmark compo-
nents) indicate the run-outs. It is important to remark that the
benchmark component geometry was designed to have different
points with similar local maximum stress levels (P1, P2 and P3),
leading to failures occurring in different positions, and this was
also confirmed by the failure locations reported in Fig. 9 (white-
filled squared dots). The point P1 is located externally and in the
neck region of the components. P2 is still located on the external
Fig. 8. Crack propagation results and their fitting with the NASGRO equation: (a) side of the components and it corresponds approximately with
long crack thresholds; (b) crack propagation curves. The results combine two the initial position of the bifurcation of the two branches. P3 is
batches from different manufacturers.
the internal point of the branches which is the location of the max-
f ¼ A0 þ A1 R imum local stress and will be discussed subsequently. It should be
h ið1þRCm Þ noted that the failures occurred approximately in the points indi-
1R
th
R<0 ð2Þ cated within a range of 5 mm. The number of failures for each
DK 1
1f
DK th ¼ p
RC m Þ
ð1A0 Þ
ðC
th th position indicates that the M benchmark components mostly fail
at location P2, while the AB benchmark components fail at location
rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a P3 (Table 8). Interestingly, only one failure was observed in the
DK 1 ¼ DK 1 ð3Þ neck region P1 for an M benchmark component. The supplemen-
a þ a0
tary material contains all the information which pertain to the pre-
In these equations, the small crack parameter was set to cise vertical position of the failures, also considering the position
a0 ¼ 27:7lm while the fitting parameters were DK 1 ; C pth and C mth . towards the thickness. It should be borne in mind that one M
The values of f and A0;1;2;3 are provided instead with the following benchmark component was out of specification (red dot), but nev-
equations: ertheless, its result is contained in the dispersion of the data points
for the specific load level.
The defects detected on the fracture surfaces for the AB bench-
mark component indicate that all the failures originated from
superficial or sub-superficial defects (Fig. 10a-c). In particular,
some of the killer defects were observed along the side of the
11
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 7
pffiffiffiffiffi
Parameters of the Nasgro equation for the present AlSi10Mg alloy; crack growth rates in m/cycle and SIF in MPa m.

DK 1 C pth Cm
th
a0 (m) C m p

1.0741 0.5408 0.1240 27.7106 1.05109 3.51 0.66

Fig. 9. Life of the M and AB benchmark components as a function of the applied load range.

Table 8 cross-section (Fig. 10b-c), while failures also occurred close to the
Number of failures for the locations of the M/AB benchmark components as indicated corner fillets (Fig. 10a). The failures of AB benchmark components
in Fig. 9.
are characterised by a different geometry of killer defects as evi-
Number of failures denced in Fig. 10d-f. Interestingly, only AB benchmark components
Benchmark component Location P1 Location P2 Location P3 had a feature that could be labelled as a corner crack (Fig. 10d).
Machined 1 9 2
As-built 0 3 10

Fig. 10. Defects detected on the fracture surfaces of M and AB benchmark components.

12
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

4.6. Residual stresses on components where = t max is the maximum defect depth. It is important to
apply Eq. 6 for wide shallow defects (c=a >10), because these
The RSs measurements were performed on two opposite front defects are equivalent to a 2D edge crack and its SIF is
pffiffiffiffiffiffi
surfaces for two M benchmark components, namely the BM2-243 K ¼ 1:12r pt .
and the BM3-242 (see Tables 15 and 12 in the Appendix). For those
measurements a total of 11 points for each surface was selected, as The defect data were interpolated with a largest extreme value
indicated in the schematic of Fig. 11a (each point represents the distribution (LEVD) distribution and the results (LEVD parameters
mean value in the specific location, and the standard deviation of k and d together with their 95% confidence bands calculated with
those measurements is also provided in the Appendix). The loca- moment method [63]) are reported in Table 9.
tions were labelled as F1 to F5 and the normal stress r22 , which Fig. 12 compares the LEVD plots adopted for defects at fracture
is parallel to the leg axis, was analysed and plotted, as indicated origins of cylindrical specimens and components. It is interesting
in the schematic. In particular, the r22 component is the stress to see that for the M specimens the killer defect distributions show
component that was used to perform the life predictions. As for defects that are lower than the ones observed on components. This
the M cylindrical specimens, for the M benchmark components difference is due to the size effect, which can be explained with the
all the stress measurements indicate that compressive RSs are pre- different critical volume for the two test pieces [65–68] and other
sent on the front and opposite surfaces. reasons such as the more complex geometry of the benchmark
Additional measurements were performed to detect the RSs on components, the altered heat flux, the solidification conditions,
the internal and external leg sides as well, which can not be etc. Such an effect is not visible for the AB samples and compo-
accessed when the benchmark component is not sectioned nents because the surface of the specimens is comparable with
(Fig. 11b). For those measurements, one M and one AB benchmark the sum of the areas of the most stressed regions of the
components were selected and sectioned. Four surfaces were anal- components.
ysed: Z2 is the front surface (corresponding to F2 and F4 from the
first measurements), Z4 is the back surface parallel to the Z2, Z1 is 5.2. Correlation between roughness measurements and defects
the external surface and Z3 is the internal surface. It is important to
highlight that the machining of the Z1 and Z3 surfaces consisted of The aim of this section is to evaluate the correlation between
removing 1 mm in depth in one single pass initially, leaving surface features (defects) and surface roughness parameters.
0.2 mm as overstock which was then removed with a final depth Indeed, the method used here is based on a point-by-point com-
cut. For the lateral surfaces (Z2 and Z4), a 3 mm rough pass was pffiffiffiffiffiffiffiffiffiffi
parison of area with Rt (maximum height of profile) or Rv (max-
adopted and followed by a final pass of 0.5 mm. For the Z2 and imum depth of valley) as surface roughness representatives. The
Z4 surfaces the RSs were measured only in two locations, while post-processing analysis to determine the maximum height Rt
for the internal and external surfaces three points were investi- and lowest valley Rv was performed by MultiFileAnalyzer software
gated. The results are also reported in the Appendix, see Tables (Keyence Corporation) using a multi-line roughness method. The
16 and 17. For the M benchmark component, the RSs r22 varies analysis was carried out fitting the data to bi-variate gaussin distri-
from approximately 80 MPa to 18 MPa in the Z2 and Z4 surfaces butions (BGD): the best correlation between defect and roughness
confirming the presence of compressive RSs. For the lateral sur- data was found for Rv in terms of a log–log scale. Fig. 13c combines
faces Z1 and Z3, the RSs are observed to be positive, with values the results of specimen and benchmark component defect sizes
in the range of 38 to 120 MPa. These results will be extremely versus Rv indicator together with contour levels of the BGD. The
important for the life prediction calculations as the RS measured diagonal lines drawn in Fig. 13b are the approximation of the 2D
on the M cylindrical specimens (compressive) are different from crack of this type:
the ones measured on the lateral sides of the M benchmark compo-
nent (tensile). For all the four lateral surfaces, the M benchmark pffiffiffiffiffiffiffiffiffiffi a
area ¼ Rv ð7Þ
component highlights tensile RSs in the range of 65 to 138 MPa c
which are similar to values published for the same alloy and man-
ufacturing process [59–61]. where a=c is the aspect ratio of a semi-elliptical crack. For an elon-
gated shallow crack (when c=a >10), SIF can be approximated to a
5. Analysis of results 2D surface crack as [62]:
pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi
5.1. Analysis of defect distributions area ¼ 10Rv ð8Þ

In order to analyse defect distributions in cylindrical specimens These lines highlight the fact that the data scatter is not only
and components, the defects were divided into two groups based affected by defect size, but also by the aspect ratio of the defects.
on their aspect ratios (semi-superficial length c over depth The central contour of BGD is intersected by the line in which
pffiffiffiffiffiffiffiffiffiffi
a; c=a). Most observed defects in M specimens/components were area is approximated with an aspect ratio of a=c=0.25. This con-
semi-circular pores or trapped gas (c=a=1), while the defects for firms the observed defects on fracture surfaces where the majority
AB specimens/components are rather elongated or semi-elliptical of AB defects had an aspect ratio of a=c=0.25. The upper and lower
(lack of fusion, pores clusters, surface irregularities). These defects limits defined by semi-circular and elongated approximation which
pffiffiffiffiffiffiffiffiffiffi (in this study) are a minority among defect types.
were measured in terms of Murakami’s area parameter [62] as
follows: Several studies have investigated the correlation of fatigue life
data points (defect size or number of cycles to failure) with a sur-
pffiffiffiffiffiffiffiffiffiffi face representative (Rv ; Rt ) of AM parts: last evidences show a

for a defect with an aspect ratio, c=a <10 the area was calcu-
strong correlation with profile depth [70,70]. The expression of
lated from the measured area of the feature at crack initiation; pffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffi area in terms of Rv is more consistent with the concept of a 2D

for a defect with an aspect ratio c=a >10 the effect area was
crack (whose driving force is controlled by the depth) [18], but it
calculated as:
pffiffiffiffiffiffi needs a suitable filtering of the profile [72,72]. Conversely, the
pffiffiffiffiffiffiffiffiffiffi
area ¼ tmax  10 ð6Þ assumption of Rt (or Rz ) [74–77,36] would provide simple conser-
vative estimates of defect/crack size.
13
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 11. RS measurements on the benchmark components: (a) front surface of two M benchmark components; (b) front and side surfaces from two pieces of benchmark
component legs (M and AB, respectively); (c) in-depth RSs measured on the two points P2 and P3 for the AB benchmark component.

Table 9
LEVD parameters: comparison between AB and M specimens and components. of the M benchmark components are in tension (locations Z1 and
Z3 in Fig. 11), while cylindrical specimens had compressive RSs.
k (lm) d (lm)
Conversely, the AB benchmark component results overlapped
Machined Specimens 65.7  5.1 8.3 = 1.81 almost perfectly with the results on the AB cylindrical specimens.
Components 76.5 7.1 12.1 = 1.76
At first glance, the average dimension of the killer defect is similar
As-built Specimens 191.0  46.8 82.8  60.6
Components 210.6  48.6 85.2  115 for the AB cylindrical and benchmark components (Fig. 12b). How-
ever, the level of the RSs is even higher than that of the specimens.
To overcome these limitations, an analysis based on fracture
5.3. Analysis of benchmark component fatigue tests in terms of local mechanics concepts was adopted and its comparison with a local
stress stress approach is discussed in the following section.

Fig. 14 shows the contour plot of the maximum principal stress

6. Fracture mechanics-based life prediction
on the benchmark component calculated according the finite ele-
ment model. Once the local state of stress was available, the exper-
6.1. Life prediction of cylindrical specimens
imental test data (load, failure position) on components could be
processed to obtain the local stress at each failure location. The
It is well established that the condition of failure or run-out for
S/N diagram reported in Fig. 14 compares the results obtained from
loadings under fatigue for ALSi10Mg parts produced by AM is con-
the fatigue tests on the cylindrical specimens with the results
trolled by the features associated with the killer defect and the
obtained from the local stress ranges on the failure location of
endurance limit, which is, in general, a material property that is
the benchmark components. The results of the M benchmark com-
not unique but also depends on the crack length. The El-Haddad
ponents are very dispersed and, contrary to the results in [38], they
model modified according to the Murakami approach to defects
also show a lower fatigue resistance with respect to the M cylindri-
applied to the AM AlSi10Mg has been widely discussed in
cal specimens. In fact, the lower fatigue resistance of the M bench-
[19,28]; the reader is referred to these references for further
mark components is related to the different killer defect
details. In this framework, the relationship between the fatigue
distributions, as evidenced clearly in Fig. 12a (further discussion
limit and the defect/crack size can be described by the following
in Section 9). Moreover, the RSs in the internal and external sides
equation:
14
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

on the load ratio is then defined by the Goodman model for a load
ratio higher than 1:
!1
>1 1 1 1þR
Dr ¼ þ ð11Þ
w;0
Dr1
w;0 2  Su 1  R

The effective stress ratio in the absence of RSs is equal to the stress
ratio applied during the test (RL =0.1). However, as shown previ-
ously, compressive and tensile RSs are present on the surface of
M and AB specimens, respectively. Whether the RSs are tensile or
compressive, the effective stress ratio Reff on the surface of speci-
mens can be calculated, which is relevant for the threshold condi-
tion of surface defects based on the stress intensity factors:
DK
K max;eff ¼ þ K res ð12Þ
1  RL

DK
K min;eff ¼ RL þ K res ð13Þ
1  RL
where DK is the range of SIF, K max;eff is the maximum effective SIF,
K min;eff is the minimum effective SIF, and K res is the SIF as calculated
considering the average measured surface RS reported in Table 6.
The calculation of the SIFs was performed according to the weight
functions of Wang and Lambert for low a=c < 1 [78] and high
a=c > 1 [79]. In correspondence with the endurance limit deter-
mined for the M and AB cylindrical specimens, we calculated Reff
which was 0.22 for the M condition and 0.56 for the AB condition.
Providing these Reff values, the Drw;0 and DK th;LC are then deter-
mined and used to construct the Kitagawa diagrams according to
the El-Haddad model. Table 10 compares the reference case RL =
0.1 and the calculated Kitagawa diagram parameters at Reff .
The El-Haddad models are reported in Figs. 15a-b together with
the experimental results obtained on the cylindrical specimens.
Please note that for each load ratio, three lines are plotted which
correspond to three values of theoretical endurance limit Drw;0 ,
considering a scatter of 5% on the theoretical values estimated
from the static properties. Each data point was introduced in the
pffiffiffiffiffiffiffiffiffiffi
figure according to the size in terms of area and the applied stress
range. For the M specimens, the effective load ratio was calculated
considering that the surface RS in compression decreases to zero in
a depth of 0.1 mm (as experimentally measured by Sausto et al.
[57]). Under this assumption, the deepest point of the crack front
Fig. 12. Defect distributions for the (a) M and (b) AB specimens. determines the critical condition, for this point Reff ¼ 0:22 was
calculated. The results evidence that the predictions performed
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffi ffi
for R = 0.1 are conservative, while for Reff =-0.22, they are close to
area
Drw ¼ Drw;0 pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ the experimental data. Similarly, also the AB results are estimated
area þ area0
with a high level of accuracy by the El-Haddad model considering
pffiffiffiffiffiffiffiffiffiffiffiffi the effective load ratio Reff =0.56, whereas the prediction performed
where area0 is defined as the El-Haddad parameter which is found
as the intersection point in the Kitagawa–Takahashi diagram withe the nominal load ratio of R = 0.1 does not match the exper-
between the theoretical endurance limit Drw;0 and the line repre- iments precisely. In summary, the results reported in Figs. 15a-b
senting the endurance limit determined by the long crack threshold demonstrate that the endurance limit of the cylindrical specimens
DK th;LC : can be readily predicted by the El-Haddad model and by consider-
 2 ing the effect of the RSs.
pffiffiffiffiffiffiffiffiffiffiffiffi 1 DK th;LC We now focus our attention on the prediction of the finite life
area0 ¼ ð10Þ
p Y  Drw;0 region for the cylindrical specimens. The life predictions were per-
formed considering the following assumptions:
The geometry factor selected was Y ¼ 0:65, which can be applied to
pffiffiffiffiffiffiffiffiffiffi
a superficial crack with the dimension defined by area. The DK th;LC
The initial crack size was taken as the average size of the killer
values were obtained from the NASGRO fitting, see Fig. 8a and defects detected on the fracture surfaces of specimens (Figs. 12-
pffiffiffiffiffiffiffiffiffiffi
Table 7. To estimate the theoretical endurance limit ( area ! 0) a-b).
Drw;0 , we used the monotonic properties (Table 1). We herein
For the M specimens, the crack shape assumed was semi-
denote it as theoretical in relation to the evidence that it cannot circular with aspect ratio of approximately a=c=1 suggested
be measured experimentally due to the unavoidable presence of by the shape of defects detected on the fracture surfaces (Figs. 7-
defects in the fatigue specimens. The theoretical endurance limit a-b). For the AB specimens, the semi-elliptical (elongated) shape
Drw;0 is then estimated from the Su for the load ratio R=-1 according with an aspect ratio of a=c=1/2.5 was selected (Figs. 7c-d).
Dr1
w;0 ¼ 2  ð0:4  Su Þ=375.3 MPa [77]. The dependence of the Drw;0

15
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 13. Surface roughness analysis by adopting bi-variate gaussian distribution to the data set: correlation of defect size (specimens and components) versus Rv obtained by
stylus.

Fig. 14. S/N diagram considering the local stress range on the failure location of the components and the results of the M specimens.

The RS profile was assumed to linearly decrease from the exper-
The effect of the RSs was accounted for by a contribution (pos-
imental RS value measured on the surface to zero towards a itive or negative depending on the RS profile) to the average SIFs
depth of 0.1 mm for M specs (as experimentally measured by thus changing the local effective stress ratio.
Sausto et al. [57]) and 0.5 mm for AB specs, a constant stress
The calculation of the SIFs was performed according to the
was then assumed to balance the equilibrium over the section. weight functions of Wang and Lambert for low a=c < 1 [78]
and high a=c > 1 [79].

16
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 10 formed at a variable Reff since the RSs are constant for each stress
Comparison of Kitagawa diagram parameters at applied and effective stress ratios.
level Dr and this dictates a variable Reff which is considered in
pffiffiffiffiffi
Stress ratio DK th;lc (MPa m) Drw0 (MPa) the present simulations. Additional life predictions at the nominal
Reference RL ¼ 0:1 1.20 251.3
stress ratio R = 0.1 are also shown with black lines. The results
Machined Reff ¼ 0:22 (at fatigue limit) 1.53 298.2 show that for both the M and AB specimens, the crack growth sim-
As-built Reff ¼ 0:56 (at fatigue limit) 1.04 155.2 ulations that consider the RSs are close to the experimental data
points in both finite and infinite life regions of the S/N diagrams.
On the other hand, the predictions considering the nominal stress
ratio lead to a significant underestimation of the stress ranges for
the M specimens and to non-conservative predictions for the AB
specimens.

6.2. Life prediction of benchmark components

For the purpose of predicting the life of components, the general

approach is the same as that adopted for the cylindrical specimens.

Fig. 15. Kitagawa diagrams and comparison with cylindrical specimen tests
considering the effect of the RSs in the effective load ratio: (a) M specimens, (b)
AB specimens.

The crack growth model adopted is the NASGRO equation fitted

on crack growth data reported in Fig. 8 and the parameter
pffiffiffiffiffiffiffiffiffiffiffiffi
area0 dependent on the stress ratio [80] according the Eq.
10; the failure condition corresponded to the average crack
depth at failure observed on the fracture surfaces.

Life predictions for the AB and M cylindrical are compared with

the number of cycles to failure detected (see Fig. 16). The continu-
ous grey lines refer to the interpolation of the experimental data
Fig. 16. Life predictions for the (a) M specimens and (b) AB specimens. The S/N
points and the dashed lines refer to the 95% scatter bands. The pre-
diagrams were predicted considering the stress ratio as the nominal load ratio
dictions are indicated in the same figures with the blue and black applied to the specimens and the effective stress ratio considering the measured
lines. It is important to highlight that these calculations were per- superficial RSs.

17
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

In particular, a set of predictions were performed considering the contained in ECSS standards for the qualification of AM materials
most stressed position of the components, as calculated from the [81] and the validation of a probabilistic fatigue assessment soft-
finite element simulation (P1, P2, and P3, see Fig. 14). In this ware for components with defects [82] that will be addressed in
regard, three life predictions will be considered which correspond a forthcoming paper.
to the three potential failure locations P1, P2 and P3. As for the The first result of this wide-ranging activity deals with the fati-
specimens, to consider the stress gradient towards the thickness gue assessment of the components by adopting the fatigue proper-
of the components, the Wang and Lambert SIF functions [79,79] ties obtained from the M and AB specimens. The two scenarios for
were taken into consideration. transferability of fatigue data from specimens to components can
The effective stress ratio at any stress level for components was be summarised in the schematic of Fig. 18. The simple idea that
calculated by superimposing the RS profile with the stress distribu- specimen data could be directly transferred to components (with
tion along the thickness of the above mentioned failure positions. a suitable multiaxial criterion as in [83]), worked well for AB
Determine the RS profile at each failure position is time consuming benchmark components. Conversely, for M components, the analy-
and complicated, thus, a general pattern was considered for the M sis based on the local stresses clearly showed that fatigue proper-
benchmark components in which the surface RS until the depth of ties of M benchmark components were lower than those of
100 lm equals the measured average surface RS and linearly goes specimens, with approximately 15% lower fatigue limit.
to zero through the thickness of the components. Then, the SIF was To understand the factors behind this lack of direct transferabil-
obtained by the Wang and Lambert weight function using the ity, it is worth remembering that fatigue properties of AM materi-
superimposed stress distributions. For the AB benchmark compo- als are inherently controlled by the size of manufacturing defects
nents, the stress profile was taken as the one measured experimen- and that they could be modelled adopting suitable short-cracks
tally (see Fig. 11c). models (i.e. models in which the crack growth properties depend
Figs. 17a-b show the life prediction results for the M and AB on the crack size) [28,29]. Many papers have shown the success
benchmark components, respectively. The life predictions were of this concept for describing fatigue properties of Al alloys
performed considering the nominal load ratio (R = 0.1, green lines). [19,84,85], Ti6Al4V [76,86,87], stainless steels [89–91] and Ni-
In addition, the life predictions were also performed with the effec- based superalloys [93–95] manufactured by AM processes.
tive load ratio Reff , which considers the effect of the RSs, and are When the problem of fatigue of AM materials is correctly
reported in red. As expected, the life predictions performed with addressed in terms of crack growth (or threshold for the fatigue
Reff for the AB benchmark components show a lower fatigue resis- strength) assessment, then the relevant factors (for a given mate-
tance due to the presence of tensile RSs (Fig. 11a). For the M bench- rial) are [96]: i) the defect/flaw size and ii) the stress ratio, which
mark components, the life predictions performed adopting the in our case is also influenced by the RSs.
nominal stress ratio are slightly conservative (especially at the
endurance limit region), as indicated by the green lines in 7.1. Defect size and ”Size effect”
Fig. 17b. It should be noted that for the M benchmark components
it is more difficult to introduce the effect of the RSs as these stres- Concerning the first factor, it is clear that the defects detected at
ses strictly depend on the failure position. In fact, Fig. 11b indicates the origin of fatigue failures of M benchmark components are lar-
that the two front surfaces are characterised by compressive RSs. ger than the ones detected on the M specimens. The larger critical
On the other hand, the failures are mostly located on the internal defect observed for the benchmark components is related to the
(position P3) and external (position P2) sides on the components ”size effect” [65–68]. Considering two different volumes V 1 and
where slight tensile stresses were measured. For the life predic- V 2 , the cumulative probability of the largest defects in V 2 could
tions, the most conservative assumption (tensile stresses) was be derived from [97]:
adopted and, as highlighted in Fig. 17, this choice determined con-
V 2 =V 1
servative life predictions. However, it should be noted that some F max;V 2 ðaÞ ¼ F max;V 1 ðaÞ ð14Þ
failures were experienced at the corners where the RS pattern
where a is the defect size. Eq. 14 indicates that the larger is the vol-
should be discontinuous. This is also evidenced by the fact that
ume V 2 , the larger will be the critical defect that can be found in it.
the life predictions adopting the nominal stress ratio are close to
Analysis of defects through Eq. 14 is the approach usually adopted
the experimental values. When comparing the life predictions of
for considering the scale effect due to internal defects in life predic-
the three typical failure locations, it is evident that the most severe
tion of AM materials [93,75] and processing CT-scan data
location is P3, while the predictions performed for P1 display the
[39,98,99,86,100,85].
longest lives. If we compare the failures observed experimentally,
The transformation of Eq. (16) produces, if the distribution of
Table 8 indicates that for the AB benchmark components, the fail-
maximum defects in V 1 is LEVD, that largest defects on V 2 should
ure position that occurs in the majority of cases is P3. The life pre-
be described by a LEVD with parameters:
dictions performed for P3 (continuum solid lines in Fig. 17a) are
very close to the experimental values when the RSs are considered kV 2 ¼ kV 1 þ d1  log ðV 2 =V 1 Þ
in the simulation. The failure positions observed for the M bench- ð15Þ
d2 ¼ d1
mark components indicate that the most critical point is P2
(Table 8). For the life predictions the difference between the curves It is possible to apply Eq. (15) considering for V 2 the 80% volume of
comparing P2 and P3 can be estimated to be less than 10% in terms specimens and components (the material volume in which
of predicted failure load for a specific expected life. In this case, the 0:8rmax < r < rmax ), as suggested in [82], and take the specimen
prediction based on point P3 would provide a reasonably conserva- volume for V 1 . The confidence bands of the estimated distribution
tive estimate. can be simply obtained from the estimates for ðk1 ; d1 Þ [102,63].
The application of this analysis to the distribution of defects in
machined samples to estimate the distribution on M benchmark
7. Discussion components is shown in Fig. 19a. The defects detected on the
benchmark components are at the upper bound of the 95 % confi-
This study addressed the manufacturing, characterisation and dence of the estimated distribution. As it can be expected, the vol-
fatigue assessment of AM demonstrators produced by the L-PBF ume of the printed part, the inter-layer time intervals and thermal
process with the final aim to verify the applicability of the concepts dissipation during the AM process have an influence [15], but nev-
18
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Fig. 17. Life prediction for: (a) M benchmark components, (b) AB benchmark components.

ertheless the estimates are quite good. This confirms that a suitable 14 would lead to a good estimation of defects to be considered
analysis (considering the confidence intervals of the estimated val- for AB benchmark components if the data of corrosion pits would
ues) of fatigue specimen data can provide a first estimate of the size be omitted (they are not relevant for benchmark components).
effect for a correct analysis of components. (See Table 11).
Similarly, the results of AB surface defects show that the two
7.2. Residual stresses and stress ratio
distributions are similar even if the most stressed region of the sur-
face area of the components (once again defined for surface region
The second main factor that dictates the fatigue resistance of
for which 0:8rmax < r < rmax ) is approximately twice the lateral
AM parts is the presence of RSs. The AB specimens and AB bench-
surface of cylindrical specimens. In this case, the adoption of Eq.
mark components were both characterised by tensile superficial
19
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 11
Relevant surfaces and volumes for ’size effect’.

80% Volume [mm3] 80% Surface [mm2]

Specimens 452.5 301.6
Components 818.0 740.8

RSs, although the levels are different considering the AB speci-

mens, 56 MPa, and the AB benchmark components, 92.5 MPa. It
should be considered that the RSs measured in this study are com-
parable with values already reported in literature for the same
manufacturing process and alloy [55]. From the modelling perspec-
tive, the presence of tensile RSs increases the effective load ratio as
clearly demonstrated by the Kitagawa diagram reported in Fig. 15b
where the El-Haddad model enables to precisely capture the con-
Fig. 18. Scheme of the results about transferrability of specimen fatigue tests data
dition of propagation (failure) or non-propagation (run-out) of
to components.
the AB specimens when calculated according the effective load
ratio.
Similarly, the fatigue life predictions performed on the AB spec-
imens (Fig. 16b) and components (Fig. 17b) with the effective load
ratio are in excellent agreement with the experimental results,
while the simulations performed with the nominal load ratio of
R = 0.1 are not conservative, especially in the endurance limit
region as also shown by W. Schneller and co-authors in [61].
Moreover, another important factor for the fatigue performance
of M benchmark components is the heterogeneity of the RSs mea-
sured at different locations (Figs. 11a-b). In all the front surfaces
(F1 to F5 in Fig. 11a, Z2 and Z4 in Fig. 11b), we measured compres-
sive RSs down to 150 MPa, while on the two lateral surfaces (Z1
and Z3 in Fig. 11b) tensile RSs were detected. This evidences that it
is difficult to precisely associate the failure location with the
proper residual stress profile since some failures were detected
close to corners (Fig. 7a), where the RSs (due to machining or print-
ing process) vanish. Correspondingly, the life predictions per-
formed on the M benchmark components are particularly
meaningful (Fig. 17a) as they clearly demonstrate that the present
assessment approach is capable of accounting for the variability
associated with the RSs.
It is interesting to notice that for the M cylindrical specimens
we measured lower RSs (-76 MPa, see Table 6). On one hand it is
expected that the residual stress profile should be almost uniform
along the cylindrical specimen surface, on the other hand it indi-
cates that the effective stress ratio at the failure location could
be significantly different considering specimens and components.
This evidence further suggests that the RSs might provide an addi-
tional contribution to the difference in the fatigue performance
between M specimens and benchmark components.

7.3. Fracture mechanics-based assessment

The results of the local stress analysis provide a clear indication

that an approach based on fracture-mechanics concepts [19] might
be more appropriate for the qualification of components manufac-
tured by AM [43,44]. In detail, the adoption of a fracture-based
approach enables to take into account the dependence of fatigue
properties on defect size, thus allowing to consider the component
volume (or size effect) and to clearly establish the lower bound of
fatigue properties by adopting an upper bound of the defect size.
Fig. 19. Comparison between defects on benchmark components and estimates The effect of the RSs, which is relevant for the AM materials such
from specimen data (Eqs. (16)–(17)): (a) M benchmark components; (b) AB as Al-Si-Mg and modern Al-based alloys [102] that are subjected
benchmark components (corrosion pits were excluded from the analysis because to ageing treatments not able to completely relieve RSs, can be
they are not relevant for demonstrators).
easily determined by superimposing the RSs and the stresses from
the external loadings (the superposition works only until the max-
imum stress is lower than material yield limit and assumptions of
20
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

thermal treatment does not reach conditions of full relieve of RSs

due to manufacturing). In particular, the crack growth experiments
require special attention since the fracture properties have to be
measured under specific load ratio and this can be accomplished
only by testing stress-relieved specimens (or specimens whose
location on the built plate minimises the RSs).

8. Conclusions

The present study addressed the manufacturing, characteriza-

tion and fatigue assessment of AM demonstrators produced by
the L-PBF process with the final aim to verify the applicability of
the concepts contained in ECSS standards for the qualification of
AM components.
To this aim four builds of AlSi10Mg specimen and components
were manufactured to: (i) determine the fatigue curves for both AB
and M conditions; (ii) measure the fatigue crack growth rate; (iii)
Fig. 20. Effect of the superficial RSs on the endurance limits of AB specimens and AB test under fatigue a benchmark component used as a reference
benchmark components. for the validation of the fatigue assessment procedure; (iv) perform
a detailed investigation of experimental test data (fractographies,
RSs) to support a detailed fatigue analysis.
The following conclusions are drawn:
elastic-shakedown have to be adopted exceeding this limit
[58,103,104]).
The transferability of fatigue performances from AM specimens
If we consider the average surface defect detected on AB speci- to components can be accomplished only with a fracture-based
pffiffiffiffiffiffiffiffiffiffi
mens and components ( area=260lm) and the RSs at P2 and P3 fatigue assessment due to the size effect of the defect distribu-
locations, the estimated fatigue limit for different load ratios is tion and the heterogeneity of the RSs;
plotted in Fig. 20 with and without (nominal) RSs comparing the
The fatigue performances of AB specimens are comparable with
result for specimens. It can be seen how the fatigue strength is those of the AB demonstrators, while a significant difference is
strongly affected by the RSs and how the effective stress ratio Reff observed for the M conditions;
is different from the applied load ratio RL . In fact, due to the high
The fatigue assessments based on the fracture mechanics are in
level of tensile RSs, for both the AB specimens and AB benchmark good agreement with the experimental fatigue performances of
components the effective load (stress) ratio is always positive even both specimens and demonstrators in AB and M conditions,
though the nominal load ratio RL ¼ 2. If the average residual because it allows the designer to properly consider the defect
stress of AB specimens was considered, then the strength would dimension (through extreme value statistics) and the residual
be almost the same: this is due to the flattening of DK th curve for stress profile (or to consider a conservative scenario for the
high stress ratios and it explains completely the overlap between assessment);
S-N diagrams of AB specimens and components. It is also interest-
A fracture mechanics-based fatigue assessment requires a pre-
ing to observe that fatigue strength predictions on AB specimens cise measurement of material crack growth rates and thresh-
and components were quite good considering the RSs at the defect olds, together with residual stresses on specimens and
location (at the surface). Once the fatigue limit is determined, then components.
a log–log linear diagram properly describes well the fatigue life in
terms of the driving force [90,105]. Along the same trend, fatigue
life of the AB benchmark components does not depend very much 9. Data Availability
on the RSs deeper than 0.5 mm and the simple constant stress pro-
file provides good results [106]. All the data of this benchmark test campaign are available
If fracture mechanics-based fatigue assessment can overcome through the Supplementary Material of this manuscript.
the problems in the transferability of the S-N curves derived from
specimens printed with components, on the other side it is clear
that it requires an additional effort in terms of experiments for: Declaration of Competing Interest
a) analysing fatigue specimens for determining the size of inhomo-
geneities at the fracture origin; b) determining crack growth rates The authors declare that they have no known competing finan-
and thresholds; c) the RSs on specimens and components (if the cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

Table 12
Residual stress measurements for M benchmark component BM2-243, Side 1.

Z1 Z2 Z3 Z4 Z5
Point 1 Point 2 Point 1 Point 2 Point 1 Point 2 Point 3 Point 1 Point 2 Point 1 Point 2
r11 (MPa) 122 120.8 122.8 126.3 104.7 95.4 101.9 122.3 115.3 113.5 111.1
SD (MPa) 13.6 11.7 8.4 7.9 7.8 7.4 7.6 8.1 8 11.1 11.8
r22 (MPa) 71.6 114.5 107 103.7 84.5 80.6 77.1 100.7 100.2 106.5 80.5
SD (MPa) 13.7 11.7 9.4 7.9 9.4 9 9.3 8.1 8 11.1 11.8
r12 (MPa) 25.6 4.1 11.3 14.8 13.4 1.4 6.2 17.3 14.5 7.2 1.2
SD (MPa) 11.9 10.2 7.4 6.9 7.1 6.7 7 7 7 9.6 10.2

21
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 13
Residual stress measurements for M benchmark component BM2-243, Side 2.

Z1 Z2 Z3 Z4 Z5
Point 1 Point 2 Point 1 Point 2 Point 1 Point 2 Point 3 Point 1 Point 2 Point 1 Point 2
r11 (MPa) 121 123.8 125.7 121.3 109.7 101.7 107.5 117.9 116.8 127 126.1
SD (MPa) 9.8 9.1 8 8.1 7.5 7.9 7.6 8.1 8.2 8 7.8
r22 (MPa) 106.2 99.5 124.1 114.7 127.9 133.5 137.5 120.8 123.6 133.7 125.5
SD (MPa) 9.8 9.1 8 8.1 9.2 9.5 9.3 8.1 8.2 8 7.8
r12 (MPa) 14.6 26.8 9.8 8.6 2.4 0 99.5 18.5 27.7 9.1 8
SD (MPa) 8.6 8 6.9 7.1 6.9 7.2 7 7.1 7.1 7 6.8

Table 14
Residual stress measurements for M benchmark component BM3-242, Side 1.

Z1 Z2 Z3 Z4 Z5
Point 1 Point 2 Point 1 Point 2 Point 1 Point 2 Point 3 Point 1 Point 2 Point 1 Point 2
r11 (MPa) 110.4 106.8 99.9 98.4 118.5 110.4 110.4 87.8 115.7 93.6 93.6
SD (MPa) 13.2 14.4 6.8 6.9 10.1 10.1 10.1 6.5 9.8 10.1 10.1
r22 (MPa) 113.8 50.9 119.5 110.3 131.1 129.3 123.4 103.7 140.8 106.7 106.7
SD (MPa) 13.2 14.4 6.9 6.9 12.3 12.2 12.3 6.5 10.4 10.1 10.1
r12 (MPa) 19.9 41.6 9.3 13.8 12 7.8 0.8 7.4 3.5 3.3 3.3
SD (MPa) 11.5 12.5 6 6 9.2 9.2 9.2 5.7 9.1 8.8 8.8

Table 15
Residual stress measurements for M benchmark component BM3-242, Side 2.

Z1 Z2 Z3 Z4 Z5
Point 1 Point 2 Point 1 Point 2 Point 1 Point 2 Point 3 Point 1 Point 2 Point 1 Point 2
r11 (MPa) 118.6 114.9 108.4 110.3 116.5 108.8 127.9 115.4 103.9 118.9 114
SD (MPa) 12.2 12.9 10.2 10.2 10.3 10.4 15 10.4 10.4 13.6 15
r22 (MPa) 114.9 110.1 111.8 109.8 128.7 127.3 204.2 103.1 100.7 115.8 144.6
SD (MPa) 12.2 12.9 10.2 10.2 12.6 12.7 18.3 10.5 10.4 13.6 15
r12 (MPa) 1.8 3.7 12 15.8 12 8 31.7 12.9 9.3 14.6 20.8
SD (MPa) 10.6 11.2 8.9 8.9 9.5 9.5 13.8 9.1 9.1 11.8 13

Table 16
Residual stress measurements on the cut M benchmark component.

Z1 Z2 Z3 Z4
Point 1 Point 2 Point 3 Point 1 Point 2 Point 1 Point 2 Point 3 Point 1 Point 2
r11 (MPa) 17.8 28.1 5.9 -87.7 -70.6 20 58 19.2 -43.8 -54.6
SD (MPa) 15.6 15.8 15.7 14.9 15.2 15.3 16.3 15.6 15.7 15.2
r22 (MPa) 46 48.9 38.2 -79.5 -34.8 100.2 119.5 76.4 -18.2 -58.4
SD (MPa) 15.6 15.4 15.9 14.9 15.4 15.7 16.4 15.6 16.2 16
r12 (MPa) 9.9 16.6 1.4 13.1 4.9 1.6 -5.5 -2.8 -24.1 -16.8
SD (MPa) 14 14 14 13.2 13.6 13.8 14.3 13.8 14 13.5

Table 17
Residual stress measurements on the cut AB benchmark component.

Z1 Z2 Z3 Z4
Point 1 Point 2 Point 3 Point 1 Point 2 Point 1 Point 2 Point 3 Point 1 Point 2
r11 (MPa) 76.5 94 81.8 66.7 66.8 63.4 82.6 70.2 70.3 62.5
SD (MPa) 22.9 23.9 23.1 23.8 24.4 23 22.7 22.9 22.4 22.6
r22 (MPa) 119.9 137 127.2 136 90.3 84.9 91.8 65.4 82.5 138.3
SD (MPa) 22.6 24.4 25.7 25.5 26.1 23.5 23.4 23.8 22.4 23.9
r12 (MPa) 3.3 -11.2 -9.9 -4.3 -9.1 3.7 -7 -17.8 -5.9 -31.1
SD (MPa) 19.8 20 20.8 21.2 22 19.9 19.7 20 19.5 20.2

Acknowledgements the industrial side and produced under well-controlled conditions

the investigated specimens and benchmark components and car-
The present work was carried out within a call-off order of ESA’s ried out the necessary post-machining and metrology operations.
‘‘Space Additive Manufacturing Benchmarking Center”, contract Polimi carried out material testing, fractographies, analyses, frac-
number 4000120221-17- NL-LvH , hosted by the MTC in Coventry. ture mechanics calculations and assessment. ESA/ESTEC performed
The consortium was composed of the Manufacturing Technology complementary materials tests (XRD, roughness) and fatigue tests
Center (MTC), Politecnico di Milano (Polimi), and the European on the benchmark components. Italian authors acknowledge sup-
Space Agency (ESA/ESTEC). The MTC managed the contract from port provided by MIUR Italian Ministry of Education, University

22
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

Table 18 and Research through the Project ”Department of Excellence LIS4.0

Roughness measurements for the AB cylindrical specimens. - Lightweight and Smart Structures for Industry 4.0”.
Stylus
pffiffiffiffiffiffiffiffiffiffi
Specimen Number Rt Rv area
Appendix A
(lm) (lm) (lm)
FN1-245 131 79 230
A.1. Residual stress measurements on Machined benchmark
FN2-245 156 109 315
FN3-245 135 70 226 components
FN4-245 129 74 184
FN5-245 112 68 130 Tables 12–17.
FN7-245 160 100 232
FN3-243 122 77 223 pffiffiffiffiffiffiffiffiffiffi
FN4-243 161 119 204 A.2. Surface roughness measurements and area of specimens and
FN5-243 127 71 100 components
FN6-243 140 105 105
FN7-243 120 80 257
Tables 18–20.

Appendix B. Supplementary material

Supplementary data associated with this article can be found, in

Table 19
Roughness measurements for the AB benchmark components.
the online version, at https://doi.org/10.1016/j.matdes.2022.
110713.
Stylus Confocal microscope
pffiffiffiffiffiffiffiffiffiffi
Specimen Number Rt Rv Rv area
(lm) (lm) (lm) (lm) References
BN2-242 120 63 97 217 [1] N. Guo, M.C. Leu, Additive manufacturing: technology, applications and
BN4-242 163 109 75 207 research needs, Front. Mech. Eng. 8 (3) (2013) 215–243.
BN1-243 101 58 117 142 [2] W.E. Frazier, Metal additive manufacturing: A review (Apr 2014).
BN2-243 136 79 119 177 doi:10.1007/s11665-014-0958-z.
BN3-243 132 80 73 212 [3] S.L. Sing, J. An, W.Y. Yeong, F.E. Wiria, Laser and electron-beam powder-bed
BN4-243 122 81 - 314 additive manufacturing of metallic implants: A review on processes,
130 70 - 240 materials and designs, J. Orthop. Res. 34 (3) (2016) 369–385, https://doi.
BN2-244 153 92 76 167 org/10.1002/jor.23075.
181 129 115 188 [4] A. Tasßdemir, S. Nohut, An overview of wire arc additive manufacturing
BN1-245 126 86 97 253 (WAAM) in shipbuilding industry, Ships and Offshore Structures (2020) 1–18,
BN2-245 136 75 - 276 https://doi.org/10.1080/17445302.2020.1786232.
[5] A. Gisario, M. Kazarian, F. Martina, M. Mehrpouya, Metal additive
BN3-245 106 62 - 165
manufacturing in the commercial aviation industry: A review (oct 2019).
113 63 - 238
doi:10.1016/j.jmsy.2019.08.005.
BN4-245 159 73 - 245 [6] A. Uriondo, M. Esperon-Miguez, S. Perinpanayagam, The present and future of
additive manufacturing in the aerospace sector: A review of important
aspects, Proc. Inst. Mech. Eng., Part G: J. Aerospace Eng. 229 (11) (2015) 2132–
2147, https://doi.org/10.1177/0954410014568797.
[7] L. Nickels, AM and aerospace: An ideal combination, Met. Powder Rep. 70 (6)
(2015) 300–303, https://doi.org/10.1016/j.mprp.2015.06.005.
[8] R. Liu, Z. Wang, T. Sparks, F. Liou, J. Newkirk, Aerospace applications of laser
Table 20 additive manufacturing, in: Laser Additive Manufacturing: Materials, Design,
Surface roughness measurements (multiline method) for the AB benchmark compo- Technologies, and Applications, Elsevier Inc., 2017, pp. 351–371. doi:10.1016/
nents near the fracture surface with corresponding failure location. B978-0-08-100433-3.00013-0.
[9] S.C. Joshi, A.A. Sheikh, 3D printing in aerospace and its long-term
Specimen Number Ra Rt Rv sustainability, Virtual and Physical Prototyping 10 (4) (2015) 175–185,
(lm) (lm) (lm) https://doi.org/10.1080/17452759.2015.1111519.
[10] D.E. Cooper, M. Stanford, K.A. Kibble, G.J. Gibbons, Additive Manufacturing for
BN1-243 (A) 20  4.6 140  31.2 68.7  24.1
product improvement at Red Bull Technology, Mater. Des. 41 (2012) 226–
BN1-243 (B) 19.0  2.9 126.5  23.8 62.1  19.9
230, https://doi.org/10.1016/j.matdes.2012.05.017.
BN2-242 (A) 18.7  3.5 120  25.2 63.4  17
[11] F. Cucinotta, M. Raffaele, F. Salmeri, A Topology Optimization of a Motorsport
BN2-243 (C) 22.8  4.5 155.4  32.9 73.4  22.7 Safety Device (2020) 400–409.
BN2-244 (A) 18.2  4.3 136  36 67.4  23.9 [12] B. Blakey-Milner, P. Gradl, G. Snedden, M. Brooks, J. Pitot, E. Lopez, M. Leary, F.
BN2-244 (B) 17.5  2.2 115.7  15.5 53.2  8.6 Berto, A. du Plessis, Metal additive manufacturing in aerospace: A review,
BN2-244 (B0 ) 16.4  3.3 104.8  23 50  13.1 Materials & Design 209 (2021) 110008.
BN3-242 (C0 ) 14.9  2.9 93.7524.95 53.1822.95 [13] Airbus website. https://www.airbus.com/space/telecommunications-
BN3-242 (D0 ) 14.9  2.9 93.8  25 53.2  23 satellites/oneweb-satellites-connection-for-people-all-over-the-globe.html.
BN3-243 (C) 18.3  3.3 122.5  252 65.9  17.8 [14] Spacenews website. https://spacenews.com/oneweb-slashes-size-of-future-
BN4-242 (C) 20.7  3.9 123.9  25 52.4  10.1 satellite-constellation/.
BN4-242 (D) 17.8  3.1 106.4  20.2 51.5  11.7 [15] A. Yadollahi, N. Shamsaei, Additive manufacturing of fatigue resistant
BN1-245 (B) 14.8  2.3 84.1  12.2 46.9  11.7 materials: Challenges and opportunities, Int. J. Fatigue 98 (2017) 14–31,
https://doi.org/10.1016/j.ijfatigue.2017.01.001.
BN1-245 (C) 19.1  3.3 119.5  23.8 61.8  17.6
[16] A. Mostafaei, C. Zhao, Y. He, S.R. Ghiaasiaan, B. Shi, S. Shao, N. Shamsaei, Z.
BN3-245 (C) 15.2  2.7 89.9  13.1 43.4  11.3
Wu, N. Kouraytem, T. Sun, et al., Defects and anomalies in powder bed fusion
BN3-245 (D) 17.2  2.8 107.8  19.7 61.4  18
metal additive manufacturing, Curr. Opin. Solid State Mater. Sci. 26 (2) (2022)
BN3-245 (C’) 22.4  3.8 137.8  23 62.8  16.9 100974.
BN1-244 (C) 23.2  4.5 152.3  25.4 70.6  14.2 [17] N. Sanaei, A. Fatemi, Defects in additive manufactured metals and their effect
BN4-243(B0 ) 21.5  4.2 140.3  29.7 65.2  25.1 on fatigue performance: A state-of-the-art review, Prog. Mater Sci. 117 (2021)
BN4-243(C0 ) 20.9  4.2 137.6  24.2 63.3  16.2 100724, https://doi.org/10.1016/J.PMATSCI.2020.100724.
BN4-243(C) 20.4  3.5 133.5  26.8 67.5  19.6 [18] A. du Plessis, S. Beretta, Killer notches: The effect of as-built surface
BN2-245 (C0 ) 20.9  2.9 138  22.3 75.6  16.8 roughness on fatigue failure in AlSi10Mg produced by laser powder bed
BN2-245 (D0 ) 16.1  2.5 100.7  19.1 52.9  15.7 fusion, Additive Manufacturing 35 (2020) 101424, https://doi.org/10.1016/j.
BN4-245 (C) 21.1  4.0 137.8  24 61.8  13 addma.2020.101424.
BN4-245 (D) 16.9  2.2 112.7  22.6 58.4  21.7 [19] S. Romano, A. Brückner-Foit, A. Brand ao, J. Gumpinger, T. Ghidini, S. Beretta,
Fatigue properties of AlSi10Mg obtained by additive manufacturing: Defect-

23
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

based modelling and prediction of fatigue strength, Engineering Fracture [42] M. Seifi, M. Gorelik, J. Waller, N. Hrabe, N. Shamsaei, S. Daniewicz, J.J.
Mechanics 187 (2018) 165–189. doi:10.1016/j.engfracmech.2017.11.002. Lewandowski, Progress Towards Metal Additive Manufacturing
[20] J.J. Lewandowski, M. Seifi, Metal additive manufacturing: a review of Standardization to Support Qualification and Certification, JOM 2017 69:3
mechanical properties, Annual review of materials research 46 (2016) 151– 69 (3) (2017) 439–455. doi:10.1007/S11837-017-2265-2.
186. [43] M. Gorelik, Additive manufacturing in the context of structural integrity, Int.
[21] W. Tillmann, C. Schaak, J. Nellesen, M. Schaper, M.E. Aydinöz, K.P. Hoyer, Hot J. Fatigue 94 (2017) 168–177, https://doi.org/10.1016/J.
isostatic pressing of IN718 components manufactured by selective laser IJFATIGUE.2016.07.005.
melting, Additive Manufacturing 13 (2017) 93–102, https://doi.org/10.1016/j. [44] J. Mochache, R.M. Taylor, A review of fatigue and damage tolerance life
addma.2016.11.006. prediction methodologies toward certification of additively manufactured
[22] J. Benzing, N. Hrabe, T. Quinn, R. White, R. Rentz, M. Ahlfors, Hot isostatic metallic principal structural elements, in: AIAA Scitech 2021 Forum, 2021, p.
pressing (HIP) to achieve isotropic microstructure and retain as-built 1509.
strength in an additive manufacturing titanium alloy (Ti-6Al-4V), Mater. [45] E. Beevers, A.D. Brand ao, J. Gumpinger, M. Gschweitl, C. Seyfert, P. Hofbauer,
Lett. 257 (2019) 126690, https://doi.org/10.1016/j.matlet.2019.126690. T. Rohr, T. Ghidini, Fatigue properties and material characteristics of
[23] A. Du Plessis, E. Macdonald, Hot isostatic pressing in metal additive additively manufactured AlSi10Mg – Effect of the contour parameter on the
manufacturing: X-ray tomography reveals details of pore closure, Additive microstructure, density, residual stress, roughness and mechanical
Manufacturing 34 (2020) 101191. properties, Int. J. Fatigue 117 (2018) 148–162. doi:10.1016/j.
[24] E. Yasa, J. Deckers, J.P. Kruth, The investigation of the influence of laser re- ijfatigue.2018.08.023.
melting on density, surface quality and microstructure of selective laser [46] N.T. Aboulkhair, M. Simonelli, L. Parry, I. Ashcroft, C. Tuck, R. Hague, 3D
melting parts, Rapid Prototyping Journal 17 (5) (2011) 312–327, https://doi. printing of Aluminium alloys: Additive Manufacturing of Aluminium alloys
org/10.1108/13552541111156450. using selective laser melting (dec 2019). doi:10.1016/j.pmatsci.2019.100578.
[25] S. Bagherifard, N. Beretta, S. Monti, M. Riccio, M. Bandini, M. Guagliano, On [47] F. Trevisan, F. Calignano, M. Lorusso, J. Pakkanen, A. Aversa, E. Ambrosio, M.
the fatigue strength enhancement of additive manufactured AlSi10Mg parts Lombardi, P. Fino, D. Manfredi, On the Selective Laser Melting (SLM) of the
by mechanical and thermal post-processing, Materials & Design 145 (2018) AlSi10Mg Alloy: Process, Microstructure, and Mechanical Properties,
28–41, https://doi.org/10.1016/J.MATDES.2018.02.055. Materials 10 (1) (2017) 76, https://doi.org/10.3390/ma10010076.
[26] S.M. Ahmadi, R. Kumar, E.V. Borisov, R. Petrov, S. Leeflang, Y. Li, N. Tümer, R. [48] G.E. Totten (Ed.), ASM Handbook Volume 4E: Heat Treating of Nonferrous
Huizenga, C. Ayas, A.A. Zadpoor, V.A. Popovich, From microstructural design Alloys, ASM International, 2016.
to surface engineering: A tailored approach for improving fatigue life of [49] L. Lam, D. Zhang, Z. Liu, C. Chua, Phase analysis and microstructure
additively manufactured meta-biomaterials, Acta Biomater. 83 (2019) 153– characterisation of AlSi10Mg parts produced by Selective Laser Melting,
166, https://doi.org/10.1016/J.ACTBIO.2018.10.043. Virtual and Physical Prototyping 10 (4) (2015) 207–215, https://doi.org/
[27] E. Maleki, S. Bagherifard, M. Bandini, M. Guagliano, Surface post-treatments 10.1080/17452759.2015.1110868.
for metal additive manufacturing: Progress, challenges, and opportunities, [50] D. Manfredi, F. Calignano, M. Krishnan, R. Canali, E. Ambrosio, E. Atzeni, From
Additive Manufacturing 37 (2021) 101619. Powders to Dense Metal Parts: Characterization of a Commercial AlSiMg
[28] S. Beretta, S. Romano, A comparison of fatigue strength sensitivity to defects Alloy Processed through Direct Metal Laser Sintering, Materials 6 (3) (2013)
for materials manufactured by am or traditional processes, International 856–869, https://doi.org/10.3390/ma6030856.
Journal of Fatigue 94 (2017) 178–191, fatigue and Fracture Behavior of [51] L. Thijs, K. Kempen, J.P. Kruth, J. Van Humbeeck, Fine-structured aluminium
Additive Manufactured Parts. doi: 10.1016/j.ijfatigue.2016.06.020. products with controllable texture by selective laser melting of pre-alloyed
[29] U. Zerbst, G. Bruno, J.-Y. Buffiere, T. Wegener, T. Niendorf, T. Wu, X. Zhang, N. AlSi10Mg powder, Acta Mater. 61 (5) (2013) 1809–1819, https://doi.org/
Kashaev, G. Meneghetti, N. Hrabe, Damage tolerant design of additively 10.1016/j.actamat.2012.11.052.
manufactured metallic components subjected to cyclic loading: State of the [52] E8/E8M-21 Standard Test Methods for Tension Testing of Metallic Materials
art and challenges, Progress in materials science 121 (2021). (2021).
[30] P. Mercelis, J.P. Kruth, Residual stresses in selective laser sintering and [53] ASTM, E739-10 Standard Practice for Statistical Analysis of Linear or
selective laser melting, Rapid Prototyping Journal 12 (5) (2006) 254–265, Linearized Stress-Life (S-N) and Strain-Life (-N) Fatigue Data, Tech. rep.,
https://doi.org/10.1108/13552540610707013. American Sosciety for Testing and Materials, West Conshohocken, PA (2015).
[31] T. DebRoy, H.L. Wei, J.S. Zuback, T. Mukherjee, J.W. Elmer, J.O. Milewski, A.M. doi:https://doi.org/10.1520/E0739-10R15.
Beese, A. Wilson-Heid, A. De, W. Zhang, Additive manufacturing of metallic [54] W.J. Dixon, The up-and-down method for small samples, Journal of the
components – Process, structure and properties, Prog. Mater Sci. 92 (2018) American Statistical Association 60 (312) (1965) 967–978.
112–224, https://doi.org/10.1016/j.pmatsci.2017.10.001. [55] A. Salmi, E. Atzeni, Residual stress analysis of thin AlSi10Mg parts produced
[32] K. Carpenter, A. Tabei, On Residual Stress Development, Prevention, and by Laser Powder Bed Fusion, Virtual and Physical Prototyping 15 (1) (2020)
Compensation in Metal Additive Manufacturing, Materials 2020, Vol. 13, Page 49–61, https://doi.org/10.1080/17452759.2019.1650237.
255 13 (2) (2020) 255. doi:10.3390/MA13020255. [56] E. Capello, Residual stresses in turning: Part i: Influence of process
[33] N.T. Aboulkhair, I. Maskery, C. Tuck, I.A. Ashcroft, N.M. Everitt, Improving the parameters, Journal of materials processing technology 160 (2) (2005) 221–
fatigue behaviour of a selectively laser melted aluminium alloy: Influence of 228.
heat treatment and surface quality, Mater. Des. 104 (2016) 174–182, https:// [57] F. Sausto, S. Beretta, P. Carrion, N. Shamsaei, Fatigue failure mechanisms for
doi.org/10.1016/j.matdes.2016.05.041. AlSi10Mg manufactured by L-PBF under axial and torsional loads: the role of
[34] P. Edwards, A. O’Conner, M. Ramulu, Electron Beam Additive Manufacturing defects and residual stresses, Int. J. Fatigue (accepted manuscript) (2022).
of Titanium Components: Properties and Performance, Journal of [58] S. Beretta, M. Gargourimotlagh, S. Foletti, A. du Plessis, M. Riccio, Fatigue
Manufacturing Science and Engineering 135 (6) (dec 2013). doi:10.1115/ strength assessment of ”as built” AlSi10Mg manufactured by SLM with
1.4025773. different build orientations, Int. J. Fatigue 139 (2020) 105737, https://doi.org/
[35] S. Leuders, S. Meiners, L. Wu, A. Taube, T. Tröster, T. Niendorf, Structural 10.1016/j.ijfatigue.2020.105737.
components manufactured by Selective Laser Melting and Investment [59] L. Wang, X. Jiang, Y. Zhu, X. Zhu, J. Sun, B. Yan, An approach to predict the
Casting—Impact of the process route on the damage mechanism under residual stress and distortion during the selective laser melting of AlSi10Mg
cyclic loading, J. Mater. Process. Technol. 248 (2017) 130–142, https://doi. parts, The International Journal of Advanced Manufacturing Technology 97
org/10.1016/J.JMATPROTEC.2017.04.026. (9–12) (2018) 3535–3546.
[36] R. Molaei, A. Fatemi, N. Sanaei, J. Pegues, N. Shamsaei, S. Shao, P. Li, D.H. [60] A. Salmi, E. Atzeni, History of residual stresses during the production phases
Warner, N. Phan, Fatigue of additive manufactured Ti-6Al-4V, Part II: The of AlSi10Mg parts processed by powder bed additive manufacturing
relationship between microstructure, material cyclic properties, and technology, Virtual and Physical Prototyping 12 (2) (2017) 153–160,
component performance, Int. J. Fatigue 132 (2020) 105363, https://doi.org/ https://doi.org/10.1080/17452759.2017.1310439.
10.1016/j.ijfatigue.2019.105363. [61] W. Schneller, M. Leitner, S. Pomberger, F. Grün, S. Leuders, T. Pfeifer, O.
[37] P. Li, D.H. Warner, N. Phan, Predicting the fatigue performance of an Jantschner, Fatigue strength assessment of additively manufactured metallic
additively manufactured Ti-6Al-4V component from witness coupon structures considering bulk and surface layer characteristics, Additive
behavior, Additive Manufacturing 35 (2020) 101230, https://doi.org/ Manufacturing 40 (2021) 101930, https://doi.org/10.1016/j.
10.1016/J.ADDMA.2020.101230. addma.2021.101930.
[38] K. Schnabel, J. Baumgartner, B. Möller, M. Scurria, Fatigue assessment of [62] Y. Murakami, Metal fatigue: effects of small defects and nonmetallic
additively manufactured AlSi10Mg structures using effective stress concepts inclusions, Academic Press, 2019.
based on the critical distance approach, Welding in the World 65 (11) (2021) [63] S. Beretta, Y. Murakami, Statistical analysis of defects for fatigue strength
2119–2133, https://doi.org/10.1007/S40194-021-01153-9/FIGURES/18. prediction and quality control of materials, Fatigue Fract. Eng. Mater. Struct.
[39] S. Romano, A. Brand ao, J. Gumpinger, M. Gschweitl, S. Beretta, Qualification 21 (9) (1998) 1049–1065, https://doi.org/10.1046/j.1460-2695.1998.00104.x.
of AM parts: Extreme value statistics applied to tomographic measurements, [64] G. Schweiger, K. Heckel, Size effect in randomly loaded specimens, Int. J.
Materials and Design 131 (2017) 32–48. doi:10.1016/j.matdes.2017.05.091. Fatigue 8 (4) (1986) 231–234, https://doi.org/10.1016/0142-1123(86)90026-
[40] A. du Plessis, S.G. le Roux, Standardized x-ray tomography testing of 5.
additively manufactured parts: A round robin test, Additive Manufacturing [65] H. Bomas, T. Linkewitz, P. Mayr, Application of a weakest-link concept to the
24 (2018) 125–136. fatigue limit of the bearing steel SAE 52100 in a bainitic condition, Fatigue &
[41] S. Senck, M. Happl, M. Reiter, M. Scheerer, M. Kendel, J. Glinz, J. Kastner, Fracture of Engineering Materials & Structures 22 (9) (1999) 733–741,
Additive manufacturing and non-destructive testing of topology-optimised https://doi.org/10.1046/J.1460-2695.1999.T01-1-00211.X.
aluminium components, Nondestruct. Test. Eval. 35 (3) (2020) 315–327.

24
S. Beretta, L. Patriarca, M. Gargourimotlagh et al. Materials & Design 218 (2022) 110713

[66] M. Shirani, G. Härkegård, Fatigue life distribution and size effect in ductile [86] Y. Hu, S. Wu, P. Withers, J. Zhang, H. Bao, Y. Fu, G. Kang, The effect of
cast iron for wind turbine components, Eng. Fail. Anal. 18 (1) (2011) 12–24, manufacturing defects on the fatigue life of selective laser melted ti-6al-4v
https://doi.org/10.1016/J.ENGFAILANAL.2010.07.001. structures, Materials & Design 192 (2020) 108708.
[67] D.S. Paolino, Very high cycle fatigue life and critical defect size: Modeling of [87] E. Akgun, X. Zhang, T. Lowe, Y. Zhang, M. Doré, Fatigue of laser powder-bed
statistical size effects, Fatigue & Fracture of Engineering Materials & fusion additive manufactured Ti-6Al-4V in presence of process-induced
Structures 44 (5) (2021) 1209–1224, https://doi.org/10.1111/FFE.13424. porosity defects, Eng. Fract. Mech. 259 (2022) 108140.
[68] D. El Khoukhi, F. Morel, N. Saintier, D. Bellett, P. Osmond, V.D. Le, Probabilistic [88] A. Yadollahi, M. Mahtabi, A. Khalili, H. Doude, J. Newman Jr, Fatigue life
modeling of the size effect and scatter in High Cycle Fatigue using a Monte- prediction of additively manufactured material: Effects of surface roughness,
Carlo approach: Role of the defect population in cast aluminum alloys, Int. J. defect size, and shape, Fatigue & Fracture of Engineering Materials &
Fatigue 147 (2021) 106177, https://doi.org/10.1016/j.ijfatigue.2021.106177. Structures 41 (7) (2018) 1602–1614.
[69] J. Gockel, L. Sheridan, B. Koerper, B. Whip, The influence of additive [89] K. Solberg, S. Guan, S.M.J. Razavi, T. Welo, K.C. Chan, F. Berto, Fatigue of
manufacturing processing parameters on surface roughness and fatigue life, additively manufactured 316l stainless steel: The influence of porosity and
Int. J. Fatigue 124 (2019) 380–388, https://doi.org/10.1016/j. surface roughness, Fatigue & Fracture of Engineering Materials & Structures
ijfatigue.2019.03.025. 42 (9) (2019) 2043–2052.
[70] S. Lee, B. Rasoolian, D.F. Silva, J.W. Pegues, N. Shamsaei, Surface roughness [90] S. Romano, P. Nezhadfar, N. Shamsaei, M. Seifi, S. Beretta, High cycle fatigue
parameter and modeling for fatigue behavior of additive manufactured parts: behavior and life prediction for additively manufactured 17–4 ph stainless
A non-destructive data-driven approach, Additive Manufacturing (2021) steel: Effect of sub-surface porosity and surface roughness, Theoret. Appl.
102094. Fract. Mech. 106 (2020) 102477.
[71] T. Persenot, A. Burr, R. Dendievel, J.Y. Buffière, E. Maire, J. Lachambre, G. [91] P. Merot, F. Morel, L.G. Mayorga, E. Pessard, P. Buttin, T. Baffie, Observations
Martin, Fatigue performances of chemically etched thin struts built by on the influence of process and corrosion related defects on the fatigue
selective electron beam melting: Experiments and predictions, Materialia 9 strength of 316l stainless steel manufactured by laser powder bed fusion (l-
(2020), https://doi.org/10.1016/j.mtla.2020.100589. pbf), Int. J. Fatigue 155 (2022) 106552.
[72] L. Barricelli, S. Beretta, Analysis of prospective sif and shielding effect for [92] Y. Yamashita, T. Murakami, R. Mihara, M. Okada, Y. Murakami, Defect analysis
cylindrical rough surfaces obtained by l-pbf, Eng. Fract. Mech. 256 (2021) and fatigue design basis for ni-based superalloy 718 manufactured by
107983. selective laser melting, Int. J. Fatigue 117 (2018) 485–495.
[73] D. Greitemeier, C.D. Donne, F. Syassen, J. Eufinger, T. Melz, Effect of surface [93] A. Yadollahi, M. Mahtabi, A. Khalili, H. Doude, J. Newman, Fatigue life
roughness on fatigue performance of additive manufactured Ti-6Al-4V, prediction of additively manufactured material: Effects of surface roughness,
Mater. Sci. Tech. 32 (2015) 629–634. defect size, and shape, Fatigue & Fracture of Engineering Materials &
[74] D. Greitemeier, F. Palm, F. Syassen, T. Melz, Fatigue performance of additive Structures 41 (7) (2018) 1602–1614, https://doi.org/10.1111/ffe.12799.
manufactured TiAl6V4 using electron and laser beam melting, Int. J. Fatigue [94] J.-R. Poulin, A. Kreitcberg, P. Terriault, V. Brailovski, Fatigue strength
94 (2017) 211–217. prediction of laser powder bed fusion processed inconel 625 specimens
[75] H. Masuo, Y. Tanaka, S. Morokoshi, T. Uchida, Y. Yamamoto, Y. Murakami, with intentionally-seeded porosity: Feasibility study, Int. J. Fatigue 132
Influence of defects, surface roughness and HIP on the fatigue strength of Ti- (2020) 105394.
6Al-4V manufactured by additive manufacturing, Int. J. Fatigue 117 (2018) [95] L. Sheridan, J.E. Gockel, O.E. Scott-Emuakpor, Stress-defect-life interactions of
163–179. fatigued additively manufactured alloy 718, Int. J. Fatigue 143 (2021) 106033.
[76] M. Nakatani, H. Masuo, Y. Tanaka, Y. Murakami, Effect of Surface Roughness [96] J. Schijve, Fatigue of structures and materials, Springer Science & Business
on Fatigue Strength of Ti-6Al-4V Alloy Manufactured by Additive Media, 2001.
Manufacturing, Procedia Structural Integrity 19 (2019) 294–301, https:// [97] S. Beretta, More than 25 years of Extreme Value Statistics for defects:
doi.org/10.1016/j.prostr.2019.12.032. Fundamentals, historical developments, recent applications, Int. J. Fatigue
[77] N. Dowling, Mechanical behaviour of materials. engineering methods for 151 (2021) 106407, https://doi.org/10.1016/J.IJFATIGUE.2021.106407.
deformation, fracture, and fatigue. fourth (2013). [98] N. Sanaei, A. Fatemi, N. Phan, Defect characteristics and analysis of their
[78] X. Wang, S.B. Lambert, Stress intensity factors for low aspect ratio semi- variability in metal l-pbf additive manufacturing, Materials & Design 182
elliptical surface cracks in finite-thickness plates subjected to nonuniform (2019) 108091.
stresses, Eng. Fract. Mech. 51 (4) (1995) 517–532, https://doi.org/10.1016/ [99] J.W. Pegues, S. Shao, N. Shamsaei, N. Sanaei, A. Fatemi, D.H. Warner, P. Li, N.
0013-7944(94)00311-5. Phan, Fatigue of additive manufactured Ti-6Al-4V, Part I: The effects of
[79] X. Wang, S. Lambert, Stress intensity factors and weight functions for high powder feedstock, manufacturing, and post-process conditions on the
aspect ratio semi-elliptical surface cracks in finite-thickness plates, Eng. resulting microstructure and defects, Int. J. Fatigue 132 (2020) 105358,
Fract. Mech. 57 (1) (1997) 13–24, https://doi.org/10.1016/S0013-7944(97) https://doi.org/10.1016/J.IJFATIGUE.2019.105358.
00018-0. [100] V. Sandell, T. Hansson, S. Roychowdhury, T. Månsson, M. Delin, P. Åkerfeldt,
[80] L. Patriarca, M. Filippini, S. Beretta, Short-crack thresholds and propagation in M.-L. Antti, Defects in electron beam melted ti-6al-4v: Fatigue life prediction
an aisi 4340 steel under the effect of sp residual stresses, Fatigue & Fracture of using experimental data and extreme value statistics, Materials 14 (3) (2021)
Engineering Materials & Structures 41 (6) (2018) 1275–1290. 640.
[81] ECSS-Q-ST-70-80C, Processing and quality assurance requirements for [101] R.-D. Reiss, M. Thomas, R. Reiss, Statistical analysis of extreme values (1997).
metallic powder bed fusion technologies for space applications, European [102] M. Muhammad, P. Nezhadfar, S. Thompson, A. Saharan, N. Phan, N. Shamsaei,
Space Agency (July 2021). A comparative investigation on the microstructure and mechanical
[82] S. Romano, S. Miccoli, S. Beretta, A new FE post-processor for probabilistic properties of additively manufactured aluminum alloys, Int. J. Fatigue 146
fatigue assessment in the presence of defects and its application to AM parts, (2021) 106165.
Int. J. Fatigue 125 (2019) 324–341, https://doi.org/10.1016/J. [103] W.T. Koiter, General theorems for elastic plastic solids, in: Progress in Solid
IJFATIGUE.2019.04.008. Mechanics, North Holland Press, 1960, pp. 167–221.
[83] A. Fatemi, R. Molaei, S. Sharifimehr, N. Phan, N. Shamsaei, Multiaxial fatigue [104] A. Ponter, A general shakedown theorem for elastic/plastic bodies with work
behavior of wrought and additive manufactured ti-6al-4v including surface hardening, in: 3rd SMiRT Conference London, IASMiRT, 1975, p. L 5/2.
finish effect, Int. J. Fatigue 100 (2017) 347–366, https://doi.org/10.1016/j. [105] Y. Murakami, T. Takagi, K. Wada, H. Matsunaga, Essential structure of s-n
ijfatigue.2017.03.044. curve: Prediction of fatigue life and fatigue limit of defective materials and
[84] J.N.D. Ngnekou, Y. Nadot, G. Henaff, J. Nicolai, W.H. Kan, J.M. Cairney, L. nature of scatter, Int. J. Fatigue 146 (2021).
Ridosz, Fatigue properties of alsi10mg produced by additive layer [106] S. Beretta, F. Sausto, C. Tezzele, Fracture-based fatigue life prediction, based
manufacturing, Int. J. Fatigue 119 (2019) 160–172. on tests results of AlSiMg10 wishbones, with special emphasis on the effect
[85] Z. Wu, S. Wu, J. Bao, W. Qian, S. Karabal, W. Sun, P.J. Withers, The effect of of residual stress, Tech. Rep. n. 1–82/21CR, Politecnico di Milano, Dept.
defect population on the anisotropic fatigue resistance of AlSi10Mg alloy Mechanical Engineering (2021).
fabricated by laser powder bed fusion, Int. J. Fatigue 151 (2021) 106317.

25