Professional Documents
Culture Documents
Application of Mechanical Surface Finishing Processes For Roughness Reduction and Fatigue Improvement of Additively Manufactured Ti-6Al-4V Parts
Application of Mechanical Surface Finishing Processes For Roughness Reduction and Fatigue Improvement of Additively Manufactured Ti-6Al-4V Parts
Application of Mechanical Surface Finishing Processes For Roughness Reduction and Fatigue Improvement of Additively Manufactured Ti-6Al-4V Parts
a r t i c l e i n f o a b s t r a c t
Article history: Currently, additively manufactured (AM) parts have a high initial surface roughness after the manufac-
Received 25 January 2017 turing process, which can be a limitation for application in highly stressed and cyclically loaded areas.
Received in revised form 8 May 2017 In the present study, Ti-6Al-4V samples were manufactured by laser beam melting, annealed and hot iso-
Accepted 9 May 2017
statically pressed afterwards. They showed an initial surface roughness of Ra = 17.9 lm depending on
Available online 11 May 2017
their build direction (45°). Subsequently, four different mechanical surface finishing processes were
applied separately on plates and fatigue coupons in order to reduce the surface roughness: Milling, blast-
Keywords:
ing, vibratory grinding and a micro machining process. The effectiveness of each treatment is evaluated
High-cycle fatigue
Additive manufacturing
with respect to the surface topography, as well as the fatigue properties based on axial fatigue tests per-
Titanium formed in accordance to DIN EN 6072. The initial roughness could often be reduced to values Ra < 1 lm.
Ti-6Al-4V The roughness decrease led to a substantial increase in the fatigue performance from initially 300 MPa to
Laser beam a maximum of 775 MPa (after 3 107 cycles).
Surface treatments Ó 2017 Elsevier Ltd. All rights reserved.
Milling
Micro machining process
Vibratory grinding
Blasting
Surface roughness
http://dx.doi.org/10.1016/j.ijfatigue.2017.05.008
0142-1123/Ó 2017 Elsevier Ltd. All rights reserved.
136 S. Bagehorn et al. / International Journal of Fatigue 102 (2017) 135–142
the modification of the surface morphology. Furthermore, fatigue each SEM analysis, the samples were sputtered with a platinum
samples were built and smoothened afterwards in order to shed layer with a thickness of 4 nm. SEM micrographs were also used
light on the influence of each surface finishing process on the fati- for a detailed investigation on fractured surfaces of broken fatigue
gue performance. samples.
The roughness of each plate was determined by means of a
Hommel Etamic Waveline profilometer T8000. The roughness
2. Experimental was measured three times on both sides of each plate. The arith-
metic average roughness Ra, the ten-point height roughness Rz
2.1. Material and the maximum height of the profile Rt were measured accord-
ing to DIN EN 4287 [21–23].
For the study two sample geometries were chosen. In order to In order to characterize the microstructure of the bulk material,
characterize the surface morphology prior and after each surface the samples were sectioned normal to the build direction. After-
finishing process, plates with the dimensions of wards they were embedded into an epoxy resin, polished and
105 mm 75 mm 3 mm were built up. They were manufactured etched with Kroll’s reagent [24]. The etched surface were then
at an angle of 45° with respect to the base platform. This angle was analysed using a Keyence VHX-5000 digital microscope.
chosen to obtain the worst case surface roughness, which can be
achieved without the use of support structures. The angular depen- 2.3. Fatigue testing
dence of surface roughness can be attributed to the so-called stair-
case effect, which is described elsewhere [5,18,19]. For the The high-cycle fatigue (HCF) tests were performed on two dif-
investigations of the fatigue behaviour, samples of type FCE 25A ferent resonance machines. The fatigue samples with as built,
according to DIN EN 6072 [20] were manufactured. The samples milled and blasted surfaces were tested on a Rumul Fractronic test
depicted in Fig. 1 had a length of 80 mm and a thread size of rig. The vibratory ground and micro machined samples were tested
M16x1. For each surface condition, eleven samples were manufac- on a Rumul Mikrotron test rig. Both machines were equipped with
tured with Kt = 1. All fatigue samples were built up at an angle of a 24 kN load cell. Prior to testing both machines were calibrated
90° with respect to the base platform, as initial trials had shown and the stress amplitudes were found to be within ±1% of the nom-
that the final contour of the samples tend to lose symmetry, if they inal values for both machines. All samples were tested with an
were built at the same angle as the plates without support struc- axial fatigue stress ratio of R = 0.1 at constant maximum stress
tures. In this case, the threads were more difficult to manufacture. levels varying between 300 and 900 MPa. The tests were per-
Moreover, this would have resulted in a bending moment upon formed at ambient temperature and a frequency of approximately
mounting the samples into the fatigue test machine, as the threads 110 Hz. The fatigue limit was set to 3x107 cycles subsequently
were not accurately centred anymore. referred to as ‘run-out’.
For the manufacturing process Ti-6Al-4V powder with an aver-
aged particle size of 44 lm (Dp50) was provided by TLS Technik 2.4. Surface finishing processes for roughness reduction
GmbH. The chemical composition is shown in Table 1. With
parameters resulting from previous studies, all samples were built For roughness reduction, four mechanical surface finishing pro-
with an energy density of 54.3 J/mm3 on a laser beam machine EOS cesses were separately applied on both sides of Ti64 as-built plates
M270 at the Fraunhofer Institute ILT. A scanning velocity of 1.2 m/s and fatigue samples. In order to have a reference sample series,
was used and the thickness of each layer to be built was set to both sample geometries were milled in accordance to the standard
30 lm. The averaged laser spot size was 80 lm. Prior to the sepa- DIN EN 6072. The plates were machined on a Deckel FP3NC milling
ration from the build plate, all samples were stress annealed at machine. The fatigue samples were machined on a contour-
540 °C for four hours under vacuum and cooled within an Argon controlled lathe GDW 3000CS.
gas atmosphere. In order to minimize the risk of fatigue failure In addition, plates and fatigue samples were manually blasted
caused by inner pores or imperfections, all samples were addition- with premium corundum (Al2O3) in accordance to aerospace stan-
ally hot isostatically pressed for 2 h at 810 °C and 2000 bar within dards. The blasting media had an average particle size distribution
an inert gas environment. of 0.5–1.0 mm (Dp50) and the particle shape was sharp-edged. All
samples were blasted manually within a closed blasting chamber
2.2. Surface characterization and a blasting pressure of 5 bar. The blasting lance was orientated
approximately 45° towards the surface to be treated and within a
In order to characterize the surface morphology before and after distance of 50 mm. The treatment time was approximately 2 min
each surface finishing process, scanning electron microscope (SEM) per plate side and less than 5 min per fatigue sample.
micrographs were recorded using a JSM Jeol 6320F SEM. Prior to During vibratory grinding, all samples were treated together
within a vibrating bowl from Roessler filled with a fluid compound
and small abrasive particles, which were adjusted for Ti64. In this
study, the samples were treated in two steps: Within the first 39 h
the main roughness was decreased by coarser particles. Within the
following 9 h, finer particles were used for the surface optimization
up to very low roughness values.
The fourth surface finishing process applied was the so-called
micro machining process developed by the company BestInClass.
In a first step, the roughness profile was analysed in order to
choose the right media for the roughness reduction. Next, so-
called micro tools were chosen, which were adapted to the corre-
sponding roughness. In comparison to the vibratory grinding pro-
cess, micro machining can also be a batch wise process, in which
a fine polishing step can be applied in the end. It took approxi-
Fig. 1. Round fatigue sample according to DIN EN 6072 [20]. mately 50 h to decrease the roughness of the current samples to
S. Bagehorn et al. / International Journal of Fatigue 102 (2017) 135–142 137
Table 1
Chemical composition of the Ti-6Al-4V powder used for the laser beam melting process.
Al V Fe O N Ti
Content, wt.% 6.4 4.2 0.2 0.1 0.007 Balance
the final values. The plates were treated separately from the fati- of partially molten particles, as well as of all surface furrows on a
gue samples. milled part. Only slight milling stripes remain on the surface.
Fig. 4b) demonstrates an SEM micrograph of the plate surface
after blasting. All particles sticking to the surface could be
3. Results
removed. However, the measured roughness values are still fairly
high (Table 2) and the surface shows an irregular shape with sharp
3.1. Surface morphology
edges and furrows.
Fig. 5 demonstrates an SEM image recorded in backscatter
The surface morphology of the as-built metallic AM compo-
mode along with an EDX analysis of the blasted surface. All darker
nents showed a high initial surface roughness, which can directly
areas in Fig. 5a) have chemical compositions that deviate substan-
be seen and felt after the manufacturing process and Table 2
tially from the bulk material Ti64. The element distribution spec-
demonstrates very high roughness values for the as-built material.
trum shown in Fig. 5b) refers to such a darker spot, which is
The standard guidelines for the surface qualities for aerospace
marked with a white circle in Fig. 5a). In addition to the alloying
applications typically call for a surface roughness of Ra = 3.2 lm
components, a huge amount of oxygen, carbon and an unusual
or even lower for cast or wrought parts. Clearly, the roughness pro-
high amount of aluminium was measured. Therefore, the darker
vided in Table 2 by far exceeds this value. With regard to the plate
spots were identified to be aluminium oxide particles, which corre-
surface formerly facing the build plate and the opposite side, dif-
late with the blasting media sticking to the surface.
ferent roughness values were measured. The surface facing the
After vibratory grinding low surface roughness values were
build plate showed 20 % higher roughness values.
measured. This is also reflected by the micrograph in Fig. 6. Again
Fig. 2 shows SEM micrographs in two magnifications of the sur-
a significant removal of the partially molten particles and a very
face of an as-built plate. The micrograph in Fig. 2a) reveals a huge
smooth surface could be observed. At high magnification, however,
amount of partially molten powder particles, which still stick to
lots of roughness valleys can still be seen on the surface (Fig. 6).
the surface. Preliminary tests also showed these could not be
This is also reflected by the relatively high roughness values Rz
removed by blasting with pressurized air or by a treatment in an
and Rt, which generally indicate remaining roughness peaks or
ultrasonic bath. Due to these particles, the surface demonstrates
valleys.
an irregular, rough and irreproducible morphology.
Fig. 7a) shows an SEM micrograph of a plate surface after the
Fig. 2b) shows a higher magnification detail of the as-built sur-
micro machining process. The roughness was also significantly
face. The partially molten particles are clearly visible. Additionally,
decreased. The micrograph, which was taken from the upper side
a high waviness can be recognized between the particles on the as-
of the plate, demonstrates a very smooth surface without any par-
built surface, which contributes to the surface roughness as well.
tially molten particles on it. By contrast, the SEM image recorded
This waviness also led to deep furrows between the particles.
from the bottom side of the plate (Fig. 7b)), i.e. the one that was
Fig. 3 shows an etched cross section of an as-built LBM plate
facing the build plate during the manufacturing process still fea-
normal the build direction of the plate. As this was built at an angle
tured remaining roughness valleys.
of 45°, the microstructure as shown in Fig. 3a) also follows this
angle and the different layers of the build-up are visible. As previ-
ous studies have shown, hot isostatically pressed AM Ti64 parts
3.2. Fatigue performance
feature a lamellar a + b microstructure [6,7]. The higher magnifica-
tion in Fig. 3b) reveals both the partially molten particles and the
The fatigue results of the as-built samples are shown in Fig. 8a).
waviness of the surface contour. In some of the surface near
The samples failed within a range of 2 104 and 106 cycles. The
regions and particles, slight oxygen enrichment can be recognized
sample tested at 300 MPa was manually stopped after 3 107
as brighter areas.
cycles without failure (‘run-out’). Fig. 8b) shows two magnifica-
The averaged surface roughness values after the roughness opti-
tions of the fracture surface of the fatigue sample, which failed
mization by milling, blasting, vibratory grinding and micro
after 9.4 104 cycles at 400 MPa. The white rectangle shows the
machining are summarized in Table 2. In comparison to the other
area of crack initiation, and the high magnification detail demon-
treatments, blasting demonstrated the smallest roughness reduc-
strates that the crack nucleated from a roughness valley with a
tion of approximately 44%.
depth of approximately 100 lm. The furrow seems to lead even
By contrast, milling, vibratory grinding and micro machining
deeper into the bulk material. Crack initiation at the outer surface
reduced approximately 95% of the initial surface roughness. Within
contour could also be observed for all of the other fractured
these three processes, milling showed the lowest roughness values
samples.
with respect to Rz and Rt. In fact, Fig. 4a) shows an efficient removal
In order to investigate the influence of a surface roughness
improvement on the fatigue performance, different surface finish-
Table 2 ing processes were applied in the subsequent fatigue tests. The
Averaged surface roughness of LBM Ti-6Al-4V plates in as-built condition and after results of the milled, blasted, vibratory ground and micro
the corresponding surface finishing processes. machined samples are displayed in Fig. 9. Comparing Fig. 8a) and
Surface condition Ra, lm Rz, lm Rt, lm Fig. 9, it becomes obvious that all samples with improved surface
As-built 17.9 ± 2.0 121.9 ± 12.6 141.7 ± 11.7
conditions showed a substantial increase of the fatigue perfor-
Milled 0.3 ± 0.1 1.9 ± 0.8 2.9 ± 1.5 mance in comparison to the as-built samples.
Blasted 10.1 ± 0.2 73.4 ± 5.7 90.5 ± 6.2 Obviously the milled samples, which had the lowest roughness
Vibratory ground 0.9 ± 0.7 8.1 ± 5.4 17.4 ± 10.5 values, also showed the highest improvement of the fatigue perfor-
Micro machined 0.4 ± 0.3 4.1 ± 3.5 9.5 ± 8.0
mance. A large part of the milled samples failed within the range of
138 S. Bagehorn et al. / International Journal of Fatigue 102 (2017) 135–142
Fig. 2. SEM micrographs of an as-built LBM Ti-6Al-4V plate surface; the two different magnifications (a + b) demonstrate a high amount of partially molten powder particles,
which stick to the surface and cause a high surface roughness [13].
Fig. 3. Optical micrograph of a cross-section of an as-built LBM Ti-6Al-4V plate. Figure a) and b) demonstrates two different magnifications of the etched surface; view
direction is normal to the build direction of the plates.
Fig. 4. SEM micrograph of a milled (a) and a blasted (b) LBM Ti-6Al-4V plate surface. In both cases the partially molten particles could efficiently be removed. The blasted (b)
surface still shows a high roughness with sharp edges resulting from the rough blasting media.
106 and 107 cycles, and two samples tested at 600 MPa and demonstrated. The initiation site is always marked by a white rect-
775 MPa, were stopped after 3 107 cycles. angle and by an arrow within the higher magnification image
Within the blasted test series, two run-outs were obtained at below. Fig. 10a) shows the fractured surface of a milled sample,
500 MPa and 525 MPa. In comparison to the as-built and milled which failed after 4.3 106 cycles at 800 MPa. Similar to the rest
surfaces, the blasted samples showed a slightly higher scatter of of this sample series, the crack was initiated within the bulk mate-
the fatigue lives. The vibratory ground samples demonstrated the rial and no crack initiation at the surface occurred.
highest scatter and three run-outs were obtained at 400 MPa, Fig. 10b) shows a blasted sample, which failed after 2.0x105 -
425 MPa and 575 MPa. The micro machined samples featured a cycles at 575 MPa. In this case, however, the crack was initiated
similar trend with little scattering and one run-out at 500 MPa. at the outer surface. In the higher magnification image, a deep
Fig. 10 exemplary shows the fracture surfaces of one failed fati- roughness valley is visible apart from the rough and wavy surface
gue sample for each surface finishing process. The fracture surface contour around, where the failure initiated. Most of the other
with the most prevalent crack initiation within each test series is blasted samples featured a similar crack nucleation site. In a few
S. Bagehorn et al. / International Journal of Fatigue 102 (2017) 135–142 139
Fig. 5. (a) Backscattered electron image and (b) EDX analysis of the blasted LBM Ti-6Al-4V surface; the EDX data were recorded from the area marked by the ring in a).
4. Discussion
Fig. 7. SEM micrograph of LBM Ti-6Al-4V plate surfaces after micro machining: (a) shows the upper plate side with reference to the build plate, (b) shows the bottom of the
plate.
140 S. Bagehorn et al. / International Journal of Fatigue 102 (2017) 135–142
(a) (b)
Fig. 8. (a) Fatigue results (R = 0.1) of hot isostatically pressed AM Ti-6Al-4V samples with as-built surface morphology (error bars for stress values are smaller than the size of
the symbols used), (b) SEM micrographs of the fracture surface of a sample, which failed at a maximum stress of 400 MPa, see main text for details.
(a) (b)
(c) (d)
(
Fig. 10. Fracture surfaces at two different magnifications of failed Ti-6Al-4V fatigue samples pre-treated by (a) milling, (b) blasting, (c) vibratory grinding and (d) micro
machining.
face and these were less deep according to the roughness data, less ues measured in this study, Rz might be most suitable to character-
scatter in fatigue life was obtained than for the vibratory ground ize the surface roughness as it takes the roughness valleys more
samples. However, the detrimental effect on crack initiation of into account than in case of Ra. If the chosen surface finishing pro-
the remaining roughness valleys was still present. As the basic cess causes an insufficient material removal, the remaining rough-
principles of micro machining and the vibratory grinding process ness valleys play a major role for the fatigue behaviour. Although
are similar, the same disadvantages apply with respect to treat- vibratory grinding and micro machining both achieved roughness
ment time and the accessibility for complex shaped parts. Yet, a results of Ra < 1 lm, the micro machined samples overall obtained
better surface finish could be achieved by micro machining. a better fatigue performance due to less, respectively smaller
All surface finishing processes caused an improvement of the roughness valleys. This effect is reflected within the significantly
fatigue performance. There is no simple correlation between different Rz results and in turn it can be referred to the applied
roughness values and fatigue life. Out of the three roughness val- material removal.
142 S. Bagehorn et al. / International Journal of Fatigue 102 (2017) 135–142