International Journal of Fatigue 124 (2019) 380–388

Contents lists available at ScienceDirect

International Journal of Fatigue

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

The influence of additive manufacturing processing parameters on surface T

roughness and fatigue life
⁎
Joy Gockela, , Luke Sheridana,b, Brittanie Koerpera, Bo Whipa
a
Wright State University, United States
b
The Air Force Research Laboratory, United States

A R T I C LE I N FO A B S T R A C T

Keywords: Additive manufacturing (AM) is used to create metal components with complex geometries where the internal
Defects surfaces cannot easily be post machined. Fatigue behavior in additively manufactured metals is largely driven by
Surface roughness defects in the material such as porosity and surface roughness. This work investigates the processing-structure-
Additive manufacturing property-performance (PSPP) relationship for varied contour processing parameters, and the resulting surface
Flaw size
roughness features for alloy 718 fabricated using the laser powder bed fusion AM process. Two nondestructive
Low cycle fatigue
characterization methods are used to determine the relationships between contour parameters and surface
roughness characteristics. The benefits of each characterization technique and metrics required for an accurate
representation of the additively manufactured surfaces are discussed. Axial fatigue testing is performed to
connect the entire PSPP relationship. An understanding of the PSPP relationship with respect to surface
roughness features and fatigue life can be used to guide future process parameter optimization and to inform the
qualification strategy for AM components.

1. Introduction speed [4]. The contour parameter set traces around the outside of the
component on each layer in contrast to the bulk parameter set that fills
Additive Manufacturing (AM) allows for the creation of complex in the internal portion to form solid material. Optimization of the
geometries and internal features that cannot easily be produced using contour processing parameters is typically completed using average line
other manufacturing processes. In fields where complexity and custo- (i.e. arithmetic average (Ra)) surface roughness values [5,6] or simply
mization are key advantages, AM is increasingly being used for final through qualitative observations. There have been several studies that
part applications. For example, AM production is already suitable for question the validity of Ra as a representation of AM surfaces, but the
the high-performance engineering materials currently used in the improved roughness metrics are contradictory and have not been re-
aerospace and medical industries [1]. However, there is currently a lack lated to the mechanical performance [6,7]. As-built surfaces show a
of clear understanding of the surface features caused by metal powder fully melted surface underneath a layer of powder particles stuck to the
bed fusion AM processes and the manufacturing process capabilities of surface. Typical measurements of the surface finish used for process
those processes [2]. optimization are performed aerially and average values of roughness
Rough surfaces are known to be a detriment to the fatigue perfor- are calculated [1]. These non-contact measurements of the AM surface
mance of metallic materials. Irregularities in the surface play a key role are taken from above and are obscured by powder particles attached to
in fatigue failure mechanisms, such as crack initiation, therefore a the surface, thus providing a superficial measurement increase [8].
better understanding of surface roughness will aid in an understanding However, the protrusions from the surface carry little stress and are
of the fatigue failure mechanisms. Several factors can influence AM expected to have little bearing on the fatigue performance [9]. It is
surface quality. For example, according to Barari et al., one of the major known that surface notches concentrate stress and that Ra values are
factors that affects the surface roughness of AM parts is the “staircase not reliable when trying to predict material properties or behavior [10].
effect,” which occurs due to the layer-by-layer powder buildup of AM Hybrid metrics combining several surface roughness metrics have been
processes [3]. created that relate to the stress concentration factor [11]. When com-
In terms of AM processing parameters, the surface roughness is paring surface features to fatigue performance, nondestructive char-
controlled by the contouring parameters such as laser power and laser acterization methods are required.

⁎
Corresponding author.
E-mail address: [email protected] (J. Gockel).

https://doi.org/10.1016/j.ijfatigue.2019.03.025
Received 31 October 2018; Received in revised form 6 March 2019; Accepted 18 March 2019
Available online 19 March 2019
0142-1123/ © 2019 Elsevier Ltd. All rights reserved.
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

Complex designs often involve internal channels that cannot be Table 1

easily machined after fabrication. It is well known that AM induces Contour processing parameters used for as-built fatigue bars.
defects that are detrimental to fatigue behavior and have recently Power (W) Speed (mm/sec)
drawn interest in the fatigue research community. Particularly for
complex components with internal features, the surface finish cannot be A 120 560
B 100 560
post machined and will likely be a dominating factor in a fatigue
C 80 560
failure. Greitemeier et al. has shown that the surface finish dominates D 80 700
the fatigue failure in Ti-6Al-4V. Even with a reduction of the internal E 80 820
porosity, there was no improvement in fatigue life [12]. Fatemi et al.
also concluded that the as-built surface roughness of the component is
the driving factor in multiaxial fatigue [13]. Pegues et al. showed that Increasing laser power puts more energy into the material and increases
crack initiation occurs at the rougher downward facing surface when the melt pool size. The increase in laser power also moves toward the
fatigue bars are built at an angle and, in general, the rougher surfaces keyholing regime which results in a fundamental change of melt pool
(measured by average roughness, Ra) generally had a lower fatigue life physics and a gas pore near the bottom of the melt pool. Increasing laser
[14]. speed results in less energy being put into the material and decreases
Fracture mechanics theory assumes that the largest defect will be the melt pool size. An increase in laser speed could also results in un-
the location that initiates fatigue failure. Beretta et al. showed that stable melt conditions and balling which causes an inconsistent weld
correlating the fatigue strength to the defect size at the failure origin bead. The variation of contour processing parameters could be inter-
reduces the scatter [15]. Masuo et al. used the maximum defect ob- preted relative to a linear energy density parameter, Power/Speed,
served in a metallographic cross section and extreme value statistical where a larger number is more energy being put into the material.
methods to somewhat successfully predict the fatigue behavior [16]. However, caution should be taken in this interpretation when para-
Useful information has been gained from prior studies indicating that meter changes cause melt pool physics to transition away from the
surface roughness is a major driver in fatigue behavior. However, conduction mode. Therefore, in this paper the individual laser powers
without consideration to the processing conditions linking the entire and speeds are reported.
PSPP relationship, attempts at optimization of processing conditions to The bulk processing parameter set that is used to fill in the internal
improve the fatigue performance remains a difficult task. material is held at the default setting, along with all other parameters.
Many of the fatigue studies in the literature are considering only the The bulk parameter would control other material aspects such as por-
default processing conditions and often not reporting the exact para- osity and microstructure, therefore it is assumed that material char-
meters used for fabrication. This makes it difficult to replicate results acteristics other than the surface roughness are consistent through all
and form the entire PSPP relationship. The thermal history of a part will specimens.
influence the melt pool behavior and affects many aspects of the ma- The fatigue bars created in this build were tall pillars; therefore,
terial such as porosity, microstructure and surface roughness. The support was required to help avoid build issues from component de-
thermal history depends on the processing parameters used as well as formation, which can occur because of contact with the recoater as it
the local deposition geometry [17]. A thick section [18] versus a thin spreads each new layer of powder. Cylindrical components were also
section [19] results in different heat conduction conditions changing fabricated in the same build between each as-built bar to help support
the thermal history. Additionally, changing of processing parameters the as-built geometries and are being used for a separate study. The as-
[20] and scanning strategies [21] are being considered to modify the built and cylindrical bars were placed closely together at the grip sec-
material structure in specific sections of a part. Therefore, it is im- tions with rows of bars perpendicular to the recoating direction. The
portant to understand the entire PSPP relationship’s influence and rows were staggered such that the recoater only contacted the edge of
whether it is the result of a purposeful modification or an inherent one component at any given time. Triangular supports were added to
geometrical effect [22]. Sheridan et al. has shown the varied para- support the end of each row. In total, 15 as-built bars were included in
meters produce different porosity type defects resulting in different the assembly. One of the bars with parameter set A was deleted mid-
PSPP relationships [23]. Therefore, in this work, a method is proposed build, because of a recoater jam, so 14 as-built fatigue bars were fully
to determine the relationship of surface roughness to processing para- fabricated. The build layout can be seen in Fig. 1 and the fatigue bar
meters and fatigue performance using measurements from structured geometry is shown in Fig. 2.
light (SL) surface scanning and computed tomography (CT). The fatigue bars were heat treated post-process in order to relieve
the residual stress and homogenize the microstructure. This will allow
2. Material and methods consistent assessment of the as-built surface roughness. The bars were
stress relieved at 1065 °C for 90 min and furnace cooled.
2.1. Additive manufacturing processing of Alloy 718 The bars were removed from the build plate using wire EDM and
post-processed through the standard solution and aging heat treatment
Round fatigue bars were fabricated using an EOS M290 laser for AM Alloy 718 [24,25]. No additional heat treatment was used to
powder bed fusion additive manufacturing machine using a layer reduce the porosity because it has been shown that the surface rough-
thickness of 40 µm. The contour parameter set controls the laser scan ness will dominate the fatigue failure [12].
settings that trace around the outside of the component geometry layer
by layer as the build progresses and defines the surface roughness of the 2.2. Surface measurements
as-built component. In some optimized parameter sets, multiple con-
tour passes are used. In this work, a single pass was used to reduce The entire surface of the gage section of every fatigue bar was
experimental variables and discover the fundamental PSPP relation- measured using two nondestructive methods. SL scanning and CT were
ships. Thus, the surfaces formed in this study are rougher than typical used to characterize the surface and calculate the surface roughness
default parameter settings. However, the results gained can be used to metrics. These methods were chosen because they can be performed on
guide further process optimization. Three replicate fatigue bars were the material before fatigue testing and can capture the entire surface,
built at each parameter set. which ensures that all surface features are captured. The chosen metrics
The laser powers and speeds are varied as shown in Table 1. The used to quantify areal surface roughness measurements, designated
laser power and speeds are varied individually to limit required speci- with S, are defined and parameterized in ISO 25178 [26]. The surface is
mens and so the conclusions can relate to AM melt pool characteristics. measured in the x-y plane with z representing the surface height. The

381
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

Fig. 3. SEM images showing the as-built surfaces. (a) Protrusions and powder
particles on the surface obscuring other features and (b) powder particles
dominating viewing of the surface.

2.2.1. Structured light scanning

Measurements of the surfaces of all geometries were performed
using a Keyence VR-3200 wide area 3D measurement macroscope,
Fig. 1. AM build chamber layout for vertical fatigue bars. which uses angled SL scanning to capture the surface heights. For the
fatigue bars, eight measurements were evenly spaced around the cir-
cumference of the gage section of each bar in order to capture the entire
surface as seen in Fig. 4. Stitched line measurements with a length of
7 mm at each radial location were used to capture the surface along the
majority of the length of the gage section while avoiding the filleted
regions on each end. Each measurement had a width of 0.5 mm to avoid
extreme curvature. The secant curvature correction was applied in the
Keyence software to correct for the cylindrical shape of the fatigue bars.
The raw height data was then exported to allow for calculation of the
surface roughness metrics for all radial locations. The arithmetic mean
height, Sa, and the maximum pit height, Sv, were calculated at each
radial location. The surface metrics are used in order to capture a re-
presentation of the entire surface. A representative example of one
height map in micrometers (µm) from each parameter set is shown in
Fig. 5.
Fig. 2. Fatigue test bar dimensions in mm.
2.2.2. Computed tomography
heights are stored in a matrix z(x, y). The mean plane is from all It is known that the surface protrusions on AM surfaces, such as
measured points on the surface, and the surface metrics are calculated powder particles stuck to the surface, make it difficult for non-contact
from the height variations from this plane. For both methods, Sa (ar- methods to capture an accurate representation of the surface roughness
ithmetic mean height) and Sv (maximum pit height) were calculated [27]. Therefore, CT was performed on the gage section of each fatigue
using Eqs. (1) and (2) respectively, where A is the area of the surface bar prior to testing to provide an unobstructed view of the surface by
and z(x, y) is the matrix of surface heights. Both surface roughness
metrics are related to the contour process parameters used in fabrica-
tion. Sa and Sv are chosen for this investigation because these metrics
are typically reported in relation to fatigue life when considering tra-
ditional manufacturing processes and have been shown to relate to the
contour processing parameters [8].
1
Sa =
A
∬A |z (x, y)| dxdy (1)

Sv = |min z (x , y )| (2)

Scanning electron microscope (SEM) images of the as-built surface

are shown in Fig. 3. Fig. 3a is a view down the 90° surface exposing
caverns hidden below powder particles that are stuck to the surface. In
Fig. 3b, the melted surface is evident under protrusions and powder
particles partially melted to the surface. Through a comparison of the
surface roughness data found using CT scans and SL surface scanning,
this work provides an understanding of the limitations and advantages
of each method in capturing a true representation of the surface. The
surface features from each method are then related to the fatigue life to
create a relationship that can be used by process developers to improve
fatigue performance. Fig. 4. Radial measurement locations on fatigue bar gage section.

382
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

Fig. 5. Example surface scans from one radial location for different contour parameter sets.

capturing all the previously hidden data as shown in Fig. 6. performed using a 100 kN MTS servo-hydraulic load frame. The applied
The cross-section slices of the CT scans were exported as images load for each test was calculated based on caliper measurements of the
with each bar having a total of one thousand images. Through the cross-sectional area of each specimen. The maximum stress was
utilization of MATLAB [28] all one thousand images for each bar were 900 MPa with a load ratio R = 0.1 at a linear frequency of 5 Hz. Each
read, cropped, centered, and converted to binary images. Next, the bar was tested to failure.
center of each individual adjusted image was determined by finding the
centroid. In some cases, multiple circles were found in each image from 3. Results
particles protruding from other planes on the surface, resulting in
multiple centroids. It was found that the anomalous surface features 3.1. Effect of contour processing parameters on surface roughness
from other places were 12 pixels or fewer. With this information the
threshold value was set above 12 pixels to capture only the circle of The as-built fatigue bars with varied contour processing parameters
interest. were characterized nondestructively using both SL scanning and CT.
After defining the centroid of the circle, the perimeter of the binary The relationship between the contour parameters and the surface
image was determined. Some slices near the top and bottom had in- roughness metrics is discussed.
consistencies occur because of CT limitations. To avoid affecting the
results of the overall bar, and to remain consistent across all measure-
3.1.1. Structured light scanning
ments, the first six and last six slices of each specimen were omitted
Sv and Sa from the SL scanning method are related to the contour
from the study. With the coordinates of the average centroid and
processing parameters in Fig. 7. The measurements from the eight
boundary perimeter for each image found, the mean radius is de-
different radial locations are averaged for each sample and the standard
termined from all surface points. The mean radius is used to define the
deviation is represented with error bars. Replicate samples within each
mean surface. Therefore, the circumference of the bar is flattened and
parameter set are shown separately. Sv shows a clear trend of de-
the surface roughness metrics Sv and Sa are calculated from the height
creasing with increasing power and increasing with increasing speed.
variations from the mean surface using Eqs. (1) and (2). Because the
This is likely because of the change in melt pool size for the contour
entire surface is measured at once, only one measurement is reported
pass. Each surface is comprised of contour passes stacked one upon
for each specimen. Additional research is needed to determine the
another. The melt pool size increases based on the processing para-
variability of the surface and if more than one measurement is required.
meters used [18], and the gap between each layer is decreased, thus
resulting in a smaller Sv. The inverse happens with an increasing speed:
2.3. Fatigue testing the melt pool size decreases and the gap between layers increases.
Additionally, as the energy input and melt pool size decrease, there is
Load controlled axial fatigue tests at room temperature were additional potential for other melt pool instabilities such as balling. A

383
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

3.1.2. Computed tomography

Sv and Sa from the CT scanning methods are related to the contour
processing parameters in Fig. 8. The entire sample is analyzed. A single
value of Sa and Sv are taken for the entire sample. Replicate bars within
each parameter set are shown separately.
It was found that while power was held constant at 80 W, the
maximum valley, Sv, increased as speed increased as shown in Fig. 8(a).
This shows a direct relationship between maximum surface valley and
speed for the CT data, which was the same result shown in the SL
scanning method. A similar trend was seen when speed was held con-
stant at 560 mm/sec: Sv, decreased as power increased as shown in
Fig. 8(b). This shows an indirect relationship between average surface
roughness and power. The same relationship that was observed in the
SL scanning method. No clear trends were found when relating the
average roughness, Sa, to the processing parameters in Fig. 8(c) and (d).
Sa is largely influenced by the powder particles and protrusions on the
surface [4–6]. These superficial features are dominating the measure-
ments and not capturing the solidified surface. Therefore, the average
powder particle size in the raw power would likely highly influence the
measured value of Sa but is expected to have little contribution to the
fatigue performance.

3.1.3. Comparison of SL and CT

When comparing the values of Sv from the SL scanning and the CT
scanning, the values of Sv are larger for the CT scanning data. This is
likely for several reasons. In the SL scanning method, the maximum
notches are obscured by powder particles that are stuck to the surface
resulting in a lower maximum notch measured. In addition, the powder
particles dominating the surface raise the average surface value from
which the maximum notch is measured. The variation is small within
the sampling of the eight largest notches within the component. This
indicates that there are many large notches of a similar size that could
be resulting in multiple crack initiations.
It was shown that the Sa values for the CT data were consistently
lower than those found using the SL surface scanning method. This is
likely because of powder particles that are stuck on the surface. In SL
scanning, underlying surface data is obstructed by powder particles
resulting in less surface to be measured and some lower surface heights
to be missed. With CT a truer representation of the surface is captured
Fig. 6. CT scan of the fatigue bar gage section showing surface roughness.
and it is logical that the Sa value would be lower.

smaller Sv is expected to result in a higher fatigue life, therefore ac- 3.2. Effect of processing parameters on fatigue life
cording to these results, a higher power should be used on the contour
pass. However, there will be limitations in this relationship. If the The fatigue life is related to the processing parameters in Fig. 9. As
power continues to increase, a keyhole type melt pool will form re- the contour laser speed increases, the general trend in fatigue life de-
sulting in a change in shape of the melt pool. Additionally, keyholing creases. As the contour laser power increases, the fatigue life increases.
results in the formation of subsurface porosity within the melt pool that There is a significant amount of spread in the fatigue life within each
would be a further detriment to the fatigue performance. parameter set. Additional samples would likely be required to verify the
Sa is shown to decrease with an increasing power, keeping with the observed trends between processing parameters and fatigue life. How-
same trend as Sv. However, when the speed is varied, there is no clear ever, there is also an observed spread in the surface roughness mea-
trend in Sa. With the highest speed, there is a large variation in the surements. Therefore, it is useful to compare the fatigue life to the
measured Sa values for the replicate samples. This indicates that at surface roughness measurement from each individual specimen.
higher speeds, the process may be unstable and create additional var-
iation in the surface results. It is expected that Sa will have some ob- 3.3. Relation of surface roughness to fatigue life
scurity in the results because of the powder particles that are stuck to
the surface and other protrusions superficially raising the roughness It has been shown that fatigue failure is likely to initiate on the
and making it difficult to observe changes in the melted surface. For surface when AM components are left in the as-built state. Therefore, it
both Sa and Sv, there is a significant amount of scatter in the data is useful to understand the relationship of the surface roughness metrics
shown through the plotted standard deviation; indicating a high to the fatigue life in order to predict mechanical performance of an as-
amount of variability in the surface and in the measurement technique. built component non-destructively. However, it is not known what
Because a new mean plane is defined for each radial location, the plane surface roughness measurements relate to the fatigue performance and
can be skewed by a large anomaly on the surface. This variation em- how the magnitudes compare.
phasizes the need for multiple measurements to accurately capture the The measured Sa and Sv for each specimen are related to the fatigue
surface roughness metrics. life for each specimen and each measurement method. For each fatigue
bar, the fatigue life is shown in Fig. 10 relative to Sa and Sv for both the
SL scanning and CT methods. It should be noted that the trend lines are

384
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

Fig. 7. Comparison of surface roughness from structured light scanning to the contour processing parameters (a) Sv to Speed, (b) Sv to Power, (c) Sa to Speed and (d)
Sa to Power.

not statistically significant but are plotted here for visual emphasis completed quickly. CT is not easily accessible and adds significant time
when comparing the surface roughness metrics. For Sa measured using and costs for inspection. However, CT may be required if the true
the SL scanning method (SL Sa) shown in Fig. 10(a), a general de- maximum depths of the notches are needed for predictive methods and
creasing trend is shown where an increasing Sa results in a lower fa- modeling. It is promising that the trend from the SL scanning method is
tigue life. This is behavior that is typically seen with traditional man- the same and could be used by process developers when choosing
ufacturing processes; however, the fit gets worse as the roughness contour parameter values as an early indication of better fatigue per-
values go up, in the cases with a higher speed. For Sa measured with CT formance.
(CT Sa) shown in Fig. 10(b), the fit is unsatisfactory with the points are
clustered close together.
3.4. Surface features resulting in failure
For Sv measured with the SL scanning method (SL Sv) shown in
Fig. 10(c), a clear decreasing trend is shown where the fatigue life
The fracture surfaces of the fatigue bars are examined to determine
decreases with an increasing Sv. There is a concern that the SL mea-
the location of failure and gain additional insight into the structure of
surement method is not measuring the true maximum notch, but results
the AM surface. Fig. 11(a) shows the crack initiation location for con-
suggest that consistent measurement procedure can still lead to accu-
tour parameters P = 80 W and S = 820 mm/sec. This parameter set has
rate fatigue life prediction. For Sv measured using CT (CT Sv) shown in
the highest speed and would have the smallest melt pool area. In this
Fig. 10(d), the fit improves further, especially in the cases with higher
case, there are several locations near the surface where there is evi-
notch values, CT provides a better fit than the SL scanning method. This
dence that the layers are not fusing together properly, and the contour
is likely because of the fact that as the roughness of the surface in-
pass is missing. In this experiment, the contour offset was not changed.
creases, more notches are obscured by the protrusions and powder
As the melt pool gets smaller, there is the potential for a gap between
particles on the AM surface.
the bulk melting and the contour pass. If this gap gets too large, the
Both methods provide reasonably accurate trends within an accep-
melt pool may not adhere appropriately. This produces a large notch in
table level of scatter when comparing the fatigue life to Sv. Based on
the surface and a possible crack initiation location. Similar behavior, to
fracture mechanics concerns, it is expected that the maximum defect
a lesser extreme, is also seen in a case with the lower speed, P = 80 W
will be the failure location [29]. The maximum notches in the surface
and S = 560 mm/sec in Fig. 11(c) where the surface of the prior con-
are represented by Sv. In this study, the surface of the entire gage
tour pass is visible. This indicates that the contour passes did not suf-
section was measured ensuring that the maximum overall notch was
ficiently fuse together.
captured. However, this process is time consuming and labor intensive,
When the contour pass appropriately adheres to the bulk material,
particularly for the SL method. Additional research is required to de-
there is still the risk of subsurface porosity from process instabilities.
termine the number of measurements required to capture a re-
This can occur at the boundary of the bulk and contour parameter in-
presentation of the life-limiting notch. SL scanning and other similar
terface. Fig. 11(b), from contour parameters P = 80 W and
non-contact surface measurement methods are readily available in
S = 700 mm/sec, shows a failure location where the failure is likely a
many laboratory and industrial settings and inspection can be
combination of a surface notch and a subsurface pore. The combined

385
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

Fig. 8. Comparison of surface roughness from CT to the contour processing parameters (a) Sv to Speed, (b) Sv to Power, (c) Sa to Speed and (d) Sa to Power.

influence of subsurface porosity and a nearby surface notch could adjacent to a large protrusion. While protrusions do not carry stress in a
dominate over a larger notch and should be considered in process op- loaded scenario, the presence of a protrusion on a prior layer during
timization. Subsurface porosity could also be caused by keyholing processing could contribute to a process instability resulting in a surface
porosity when high energy parameters are used for the contour. In the notch or adhesion issues on the next layer. Fig. 11(e) is at the highest
higher power cases, the layers are appropriately fused together, and the contour laser power, P = 120 W and S = 560 mm/sec. The failure lo-
failure location is occurring at a notch in the surface at the interface cation now exhibits a material inclusion instead of a surface notch. This
when stacking melt pools on the contour laser scan. The exact notch of contour parameter set exhibited the smallest Sv values. As the surface
failure is now influenced by other factors. Fig. 11(d), with contour improves, the interaction of competing failure mechanisms will in-
P = 100 W and S = 560 mm/sec shows a surface notch vertically crease and must be considered during processing parameter

Fig. 9. Comparison of fatigue life to contour processing parameters.

386
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

Fig. 10. Comparison of fatigue life to surface roughness metrics from the SL and CT measurement methods (a) SL Sa, (b) CT Sa, (c) SL Sv and (d) CT Sv.

optimization. shown, using multiple surface roughness measurement techniques, that

the surface roughness is related to the processing parameters. Sa de-
creases with an increasing laser power, but a clear trend is not present
4. Conclusions with increasing laser speed. Sv decreases with an increasing laser power
and increases with an increasing laser speed. The processing parameters
The PSPP relationship is explored for relating the contour process used are influencing the melt pool size. Increasing the laser power
parameters to surface roughness and fatigue for AM alloy 718. It is

Fig. 11. Fracture surfaces from each contour parameter set. (a) P = 80 W, S = 820 mm/sec (b) P = 80 W, S = 700 mm/sec (c) P = 80 W, S = 560 mm/sec (d)
P = 100 W, S = 560 mm/sec, (e) P = 120 W, S = 560 mm/sec.

387
J. Gockel, et al. International Journal of Fatigue 124 (2019) 380–388

results in a larger melt pool size and smaller surface notches. Further [7] Moylan S. Progress toward standardized additive manufacturing test artifacts.
increase of the laser power would result in keyholing that would pro- Proceedings of achieving precision tolerances in additive manufacturing, Raleigh,
NC. 2015.
duce subsurface porosity. Increasing the laser speed produces a smaller [8] Whip B. The effect of processing parameters on surface roughness influencing
melt pool and larger surface notches. This is likely magnified because performance in additively manufactured alloy 718. Wright State University; 2018.
an increase in speed introduces more melt pool instabilities and balling. [9] Kantzos C, Cunningham RW, Tari V, Rollett AD. Characterization of metal additive
manufacturing surfaces using synchrotron X-ray CT and micromechanical modeling.
The CT scanning surface inspection method supports similar trends; Comput Mech 2018:575–80.
however, CT measures larger maximum notches than the SL scanning [10] Arola D, Ramulu M. An examination of the effects from surface texture on the
method because the surface is not obscured by powder particles and strength of fiber reinforced plastics. J Compos Mater 1999;33(2):102–23.
[11] Arola D, Williams C. Estimating the fatigue stress concentration factor of machined
large protrusions on the surface. surfaces. Int J Fatigue 2002;24:923–30.
When comparing the surface roughness metrics to the fatigue life, [12] Greitemeier D, Dalle Donne C, Syassen F, Eufinger J, Melz T. Effect of surface
Sv is inversely correlated to the fatigue life. As Sv increases, the fatigue roughness on fatigue performance of additive manufactured Ti-6Al-4V. Mater Sci
Technol 2016;32(7):629–34.
life decreases. This trend is seen for both measurement methods, with
[13] Fatemi A, Molaei R, Sharifimehr S, Phan N, Shamsaei N. Multiaxial fatigue behavior
the CT method providing a slightly better fit. Sa does not show a cor- of wrought and additive manufactured Ti-6Al-4V including surface finish effect. Int
relation to fatigue life from either measurement method. Fractography J Fatigue 2017:347–66.
shows that in the cases of high roughness and maximum notch values, [14] Pegues J, Roach M, Williamson RS, Shamsaei N. Surface roughness effects on the
fatigue strength of additively manufactured Ti-6Al-4V. Int J Fatigue
the contour pass is not adhering to the layer below, resulting in the 2018;116:543–52.
large notches. When proper fusion is achieved, the failure locations are [15] Beretta S, Romano S. A comparison of fatigue strength sensitivity to defects for
now influenced by other factors such as material inclusions, and sub- materials manufactured by AM or traditional processes. Int J Fatigue 2017:178–91.
[16] Masuo H, Tanaka Y, Morokoshi S, Yagura H, Uchida T, Yamamoto Y, Murakami Y.
surface porosity. This emphasizes the importance of future work in- Effects of defects, surface roughness and HIP on fatigue strength of Ti-6Al-4V
cluding interaction effects. Understanding of the entire PSPP relation- manufactured by additive manufacturing. Procedia Struct Integrity 2017:19–26.
ship for the effect of surface roughness on fatigue can be used to guide [17] Yadollahi A, Shamsaei N. Additive manufacturing of fatigue resistant materials:
challenges and opportunities. Int J Fatigue 2017:14–31.
future process parameter optimization and to inform the qualification [18] Gockel J, Beuth J, Taminger K. Integrated control of solidification microstructure
strategy for relating coupon test data to components with different and melt pool dimensions in electron beam wire feed additive manufacturing of Ti-
thermal histories. 6Al-4V. Addit Manuf 2014;1(1):119–26.
[19] Gockel J, Fox J, Beuth J, Hafley R. Integrated melt pool and microstructure control
for Ti-6Al-4V thin wall additive manufacturing. Mater Sci Technol
Acknowledgements 2015;31(8):912–6.
[20] Narra S, Cunningham R, Beuth J, Rollett AD. Location specific solidification mi-
crostructure control in electron beam melting of Ti-6Al-4V. Addit Manuf
The authors would like to thank the Turbine Engine Fatigue Facility
2018;19:160–6.
of The Air Force Research Laboratory, Universal Technology [21] Dehoff RR, Kirka MM, List FA, Unocic KA, Sames WJ. Crystallographic texture
Corporation and the Air Force Summer Faculty Fellowship Program for engineering through novel melt strategies via electron beam melting: Inconel 718.
funding and support. Mater Sci Technol 2015;31(8):939–44.
[22] Hagen JF, Rasch M, Kohl S, Schmidt M. Geometry dependent microstructures: bug
or feature? Procedia CIRP 2018;74:724–7.
References [23] Sheridan L, Scott-Emuakpor O, George T, Gockel J. Relating porosity to fatigue
failure in additively manufactured alloy 718. Mater Sci Eng, A 2018:170–6.
[24] Society AM. Heat treatment wrought nickel alloy and cobalt alloy parts; 2016.
[1] Grimm T, Wiora G, Witt G. Characterization of typical surface effects in additive [25] ASTM. Standard specification for additive manufacturing nickel alloy (UNS
manufacturing with confocal microscopy. Surf Topogr Metrol Prop 2015;3(1). N07718) with powder bed fusion. ASTM Book of Standards. 2014.
[2] Townsend A. Surface texture metrology for metal additive manufacturing: a review. [26] ISO 25178-2. Geometrical product specifications (GPS) – surface texture: Areal- Part
Precis Eng 2016;46:34–47. 2: Terms, definitions and surface texture parameters; 2012.
[3] Barari A, Kishawy HA, Kaji F, Elbestawi MA. On the surface quality of additive [27] Fox J, Kim F, Reese Z, Evans C, Taylor JS. “Investigation of complementary use of
manufactured parts. Int J Adv Manuf Technol 2016. optical metrology and X-ray computed tomography for surface finish in laser
[4] Fox JC, Moylan SP, Lane BM. Effect of process parameters on the surface roughness powder bed fusion additive manufacturing. European Society for Precision
of overhanging structures in laser powder bed fusion additive manufacturing. 3rd Engineering and Technology Special Interest Group. 2017.
CIRP conference on surface integrity. 2016. [28] M R. Natick (MA): The MathWorks, Inc.; 2017b.
[5] Townsend A, Senin N, Blunt L, Leach RK, Taylor J. Surface texture metrology for [29] Masuo H, Tanaka Y, Morokoshi S, Yagura H, Uchida T, Yamamoto Y, et al. Influence
metal additive manufacturing: a review. Precis Eng 2016;46:34–47. of defects, surface roughness and HIP on the fatigue strength of Ti-6Al-4V manu-
[6] Triantaphyllou A, Giusca CL, Macaulay GD, Hoebel M, Leach RK, Tomita B, Milne factured by additive manufacturing. Int J Fatigue 2018;117:163–79.
KA. Surface texture measurement for additive manufacturing. Surf Topogr Metrol
Prop 2015;3(2).

388