Professional Documents
Culture Documents
1 s2.0 S014294181931030X Main
1 s2.0 S014294181931030X Main
Polymer Testing
journal homepage: www.elsevier.com/locate/polytest
A B S T R A C T
Selective laser sintering (SLS) is an additive manufacturing technique that has become increasingly prevalent in today's rapidly advancing production industries. SLS
utilizes a layer-by-layer processing technique that allows for complex geometries to be produced without expensive tooling or molding. Because of the technology's
widespread use in fulfilling functional demands, parts produced by SLS should have a good surface quality. However, parts manufactured via SLS have a com-
paratively higher surface roughness than alternative manufacturing methods. To better understand the various factors that affect surface roughness, a design of
experiments was implemented which investigated the effects of four process parameters using a Sinterstation 2500, a commercial SLS printer: laser power, roller
speed, powder type, and scan spacing. To draw a comparison with other SLS systems, two process parameters were investigated on a Sinterit Lisa, a desktop SLS
printer: power ratio and powder type. These effects were studied across the entire powder bed, a study of which has not been previously conducted. Since many other
factors play a role such as layer height, this study concentrated on the top surface of the SLS printed part. The surface roughness of the produced parts was studied
using Focus Variation (FV) technology. Conditional formatting was used to show the surface roughness and area ratio of the powder bed, and an analysis of variance
(ANOVA) is performed on the data collected from the Sinterstation.
∗
Corresponding author.
E-mail address: spetzold@wisc.edu (S. Petzold).
https://doi.org/10.1016/j.polymertesting.2019.106094
Received 21 June 2019; Received in revised form 7 August 2019; Accepted 31 August 2019
Available online 01 September 2019
0142-9418/ © 2019 Elsevier Ltd. All rights reserved.
S. Petzold, et al. Polymer Testing 80 (2019) 106094
dosage. For roller speed, the default value of 12.7 cm/s (5 in/s) and a
slower rate of 7.62 cm/s (3 in/s), were selected. For laser power, 10 and
13 W were used, bounding the default value of 12 W. Lastly, for fill
spacing, the default value of 152 μm (0.006 in) and a higher density of
127 μm (0.005 in) were investigated.
To draw a comparison with other SLS systems, a Sinterit Lisa
desktop SLS printer was also used to produce builds. In contrast to the
Sinterstation, this system uses a 5 W infrared laser diode with an
emission peak of 808 nm [11]. For the powder to effectively absorb the
energy of the laser and begin the melting process, a darkened pigment
must be employed. Since the scan spacing and roller speed could not be
directly altered on the system, only alterations in powder and power
ratio were investigated. Power ratio values of 0.84 and 1.08 were used
to provide similar power variations to the Sinterstation. Both recycled
and virgin Sintratec PA12 powder were used, using the same thermal
history of 5–10 h of 140 °C exposure for each build. According to the
Fig. 1. Schematic of SLS process.
manufacturer of the Lisa printer, an energy density between 400 and
700 J/cm2 is supplied to the powder during the build process, where the
agent, over the region that is intended to become the final product. variation in this parameter is controlled by several variables which
Since MJF uses infrared heating instead of a laser, the regions sprayed cannot be disclosed.
with the fusing agent are able to absorb the energy supplied by the Post-processing of the samples bypassed the customary media
heating system, forming the sintered part. This technique was devel- blasting with glass beads as this could alter the surface roughness.
oped to improve processing speed and allows for multi-material and Samples blasted for 30 s with glass beads showed an average reduction
multi-color options through the ink-printing process [4]. In contrast to in surface roughness by 72% for recycled powder and 62% for virgin
SLS and MJF, SHS uses a thermal printing head to selectively melt powder, which was conducted as a part of this study. A milder cleaning
portions of powder through contact heating. During a sintering cycle, method was developed, which first removed residual powder on the
the print head deposits a thin layer of powder over the build platform, surface with compressed air and then subjected the surfaces to 10 min
while the printer head selectively melts the required areas during the in an ultrasonic water bath with deionized water. Then the coupons
same pass. By implementing a simple and cheaper heating system, SHS were air dried overnight. To prepare the coupons for analysis, the
offers an economical alternative in comparison with SLS. Both MJF and samples were gold sputtered two times to deposit about 80 nm of gold
SHS still produce parts with high specific surface roughness, which on the sample surface.
means that the results of a surface roughness study on SLS parts will be
of concern in these two technologies [4]. Silliani et al. investigated the 3. Focus variation – 3D optical metrology system
area surface roughness of both MJF and SLS and found that the trend of
surface roughness for top and bottom surfaces of a cube were almost A Focus Variation method was used to determine the surface
equal [5]. It was not possible to find published results on surface roughness parameters Sa, and Area Ratio, or the total area of the sample
roughness values for SHS. surface over the theoretical area of a smooth sample. While many
studies have shown significant discrepancies in the results between FV
2. Description of SLS technique methods and traditional tactile methods, this can be attributed to the
fact that sufficient surface variation is required for measurement with
All samples were fabricated using a DTM Sinterstation 2500, Serial FV [12]. Recently, new roughness standards have been developed
#: S9031787, shown in Fig. 2. Prints were completed using 3D System which introduce a certain degree of surface nano-roughness, making
Software, SLS System Version 3.3, including Build Setup v3.3 for in- them suitable for both tactile and FV methods. It has been shown that
putting print parameters. Both virgin powder and recycled powder with these new standards, FV is able to produce repeatable and accurate
were investigated as numerous studies have shown that surface quality measurements of surface roughness that are comparable to tactile sys-
decreases when the powder has been exposed to extended thermal cy- tems [13].
cles [3,6,7]. This is caused in part by an increase in molecular weight of FV methods have been making a strong foothold in industry due to
the polymer chains, increasing the melt viscosity and amount of the benefits of non-contact measuring. Tactile measurement systems, in
shrinkage in the part. Pham et al. showed there is a rapid increase in comparison, suffer from many severe limitations. Firstly, the contact
melt viscosity after 15–20 h of heat exposure, measured in melt flow stylus has been shown to leave scratches, indentations, and visible trace
rate (MFR), and then a gradual increase past this time frame [6]. Re- paths on the surface of specimens [14]. Secondly, high maintenance
cycled powder was regarded as material that had been re-collected after costs are associated with tactile systems, due in part to the large number
at least one print cycle. While this degradation was not controlled, the 8 of parts that must be regularly replaced. And lastly, the radial form of
arrays printed with recycled powder used powder with an almost the stylus tip has a smoothing effect on a surface which can conse-
identical thermal history. Therefore, the powder should be equally quently influence the measurement result. FV methods, on the other
degraded, allowing hypotheses to be drawn if certain parameters hand, efficiently bypass these limitations with the addition of true color
change in importance as the material becomes degraded. images [13].
Three parameters, the roller speed, laser power and slicer scan fill The FV system uses a white light source to project light beams onto
spacing, were selected for this study, since previous research has shown a specimen's surface. Reflected light rays emerge from the surface and
they significantly impact the surface quality of SLS parts [8–10]. The are processed through a precise sensor. The FV system, schematically
speed at which the roller moves across the part cylinder dictates how depicted in Fig. 3, has a very limited depth of focus, which ensures that
the powder particles pack. If the particles are packed with a different only small regions of a surface are in focus at a given time. By vertically
bulk density, the bulk heat capacity will change, which would alter the scanning the surface along the optical axis, the distance between the
melting and coalescing behavior. Altering the laser power and slicer optical lens and specimen surface changes; different sections of the
scan fill spacing, the distance between the parallel lines of the raster, surface are thus brought into and out of focus. The measuring algorithm
changes the melting behavior of the powder by varying the energy identifies the height of each point on a surface by determining at what
2
S. Petzold, et al. Polymer Testing 80 (2019) 106094
3
S. Petzold, et al. Polymer Testing 80 (2019) 106094
Table 1 Table 3
Investigated process parameters and their values. 22 factorial design matrix – sinterit lisa.
Printer Type Parameters Levels Build Power Ratio Powder Type Surface Roughness (μm) Area Ratio
Table 2
24 factorial design matrix – sinterstation 2500.
Build Laser Power (W) Roller Speed (cm/s) Powder Type Scan Spacing (μm) Surface Roughness (μm) Area Ratio
4
S. Petzold, et al. Polymer Testing 80 (2019) 106094
that as the nominal thickness of the part was reduced the less accurate
the actual thickness of the part [21]. This is a valid and important in-
vestigation, the results of which can be implemented in several different
areas across additive manufacturing, not just SLS. However, it was
determined that it is not necessary to include the accuracy of part
thickness in our study, since the thickness of parts was not a variable in
the design of experiments. All parts were printed to 2 mm in thickness
to mitigate the effects of warpage or curling, which would adversely
affect the ability to measure surface roughness.. This is a valid and
important investigation, the results of which can be implemented in
several different areas across additive manufacturing, not just SLS.
However, it was determined that it is not necessary to include the ac-
Fig. 8. Surface roughness in Build 1.
curacy of part thickness in our study, since the thickness of parts was
not a variable in the design of experiments. All parts were printed to 2
mm in thickness to mitigate the effects of warpage or curling, which
would adversely affect the ability to measure surface roughness.
The surface characteristics of the manufactured specimens were
then analyzed using the Alicona InfiniteFocus machine. Fig. 5 illustrates
how a typical scan looks from this system. One principal objective in
the measuring of the parts’ surfaces was to eliminate the scanning of
“holes”. These holes are portions of the scan that have near vertical
walls, minimizing the scanning of these features ensures accuracy in
retrieving measurements.
In the full factorial design of experiments, a total of 16 unique trials
were executed for the Sinterstation 2500, and a total of four trials were
conducted for the Sinterit Lisa. In each of these trials, the laser power,
Fig. 9. Surface roughness in Build 8. roller speed, powder type, and scan spacing for the Sinterstation, while
power ratio and powder type for the Sinterit, were assigned to one of
two levels according to Table 1.
As can be seen in Table 1, the higher value of scan spacing (152 μm)
is associated with the coded lower level of (−) as this is the standard
value used in the SLS. Table 2 lists all of these variables with each build
number labeled for the Sinterstation. The lower value of scan spacing
(127 μm) is assigned to the coded higher level of (+). The following are
SLS process parameters that were not altered: scan speed (1676 mm/s),
layer thickness (0.100 mm), bed temperature (166 °C), spot size
(0.450 mm), scan count (1) and hatch length (0.15 mm). For the Sin-
terit machine, Table 3 lists the variables with the designated level for
the following parameters: power ratio and powder type. The surface
roughness and area ratio values that were obtained for both printing
systems are also provided.
5. Results
5
S. Petzold, et al. Polymer Testing 80 (2019) 106094
Fig. 11. Top: Surface roughness plots for recycled PA12 (Builds 1–4 – top, Builds 9–12 – bottom); Bottom: Surface roughness plots for virgin PA12 (Builds 5–8 – top,
Builds 13–16 – bottom).
Fig. 12. Left: Suface roughness plots for virgin PA12; Right: Surface roughness plots for recycled PA12.
to detect surface roughness profile for a given sample [18]. Therefore, predicted result, the lowest determined values were from recycled
results for surface roughness should be valid for both the grey Sinterit powder and lower power ratio. It is possible that the material had not
and the more reflective gold coated Sinterstation samples. degraded long enough to have a significant impact on increasing the
From the Sinterit builds, both laser power and recycled material did viscosity to be detrimental to surface quality. In a previous study on
not have a significant impact on surface roughness. Contrary to the correlating the melt flow rate (MFR) with degraded powder, it was
6
S. Petzold, et al. Polymer Testing 80 (2019) 106094
7
S. Petzold, et al. Polymer Testing 80 (2019) 106094
Fig. 18. Top: Area ratio plots for recycled PA12 (Builds 1–4 – top, Builds 9–12 – bottom); Bottom: Area ratio plots for virgin PA12 (Builds 5–8 – top, Builds 13–16 –
bottom).
for the build with the highest amount of roughness can be easily dis- 5.1.4. Area ratio powder bed study
tinguished from the build with the lowest amount of surface roughness,. Random behavior was exhibited across the powder bed for the
For virgin powder, since each build has similar surface roughness va- builds with the lowest area ratio in Build 16 and the highest area ratio
lues, it is more difficult to make this distinction. Fig. 11 shows the in Build 10, shown in Figs. 13 and 14, respectively. As can be seen from
surface roughness for each build using recycled and virgin powder. The the color plots below, no detectable trends were observed.
variance between builds using virgin powder is not as great as in re- However, notable trends were observed in the color plots for Builds
cycled ones. This is due to the fact that the range of surface roughness 1 and 8, as shown in Figs. 15 and 16, respectively. In Build 1, cells with
values is smaller for virgin powder, giving lower roughness values in greater area ratios are concentrated in the back of the SLS machine,
similar regions from build to build. It should be noted that the scale is while cells with smaller area ratios were concentrated in the front. This
different for recycled and virgin powder. If both powder types were behavior can be explained using the same reasoning as presented in the
shown using the same scale, the recycled plots would appear com- surface roughness powder bed distribution for Build 1. In Build 8, area
pletely red, while the virgin plots would be completely green, since ratios of lower value were located in the back-right region of the ma-
recycled powder has inherently rougher surfaces. chine.
For the Sinterit Lisa builds, Fig. 12 shows the surface roughness
plots for both recycled and virgin PA12. Virgin builds are shown on the
left, while the recycled builds are provided on the right. All builds are 5.1.5. Area ratio levels
performed on the same conditional formatting scale, since there is not As the region of interest moves from outer to inner (1–3), the mean
as significant of a change between the recycled and virgin prints, in area ratio remains approximately constant, while the variance of the
comparison with parts produced on the Sinterstation. area ratio decreases. This can be seen in Fig. 17. The variation of area
ratio reduces closer to the center of the build area, which reflects the
8
S. Petzold, et al. Polymer Testing 80 (2019) 106094
9
S. Petzold, et al. Polymer Testing 80 (2019) 106094
Fig. 20. Pareto chart showing significant effects for surface roughness. Main or interaction effects greater than 1.98 are determined to be significant.
significant. To form a better model, a higher performing modeling surface roughness and area ratio using an Alicona Focus Variation
software should be used in the future, which allows for a higher number measurement technique. Conditional formatting was used to illustrate
of replicates. the variation in these values throughout the build cross section. By
visually showing how the relative surface roughness changes
throughout the powder bed based on variations in processing condi-
6. Conclusion tions, it is possible to effectively determine factors contributing to this
disparity.
An investigation into the surface roughness and area ratio for parts The surface roughness of the top surface of SLS parts showed a
produced across the entire powder bed was conducted. On a significant dependency on the processing conditions with which parts
Sinterstation 2500, the powder type, power, roller speed, and scan were made. Specifically, the greatest average surface roughness was
spacing were varied in accordance to a 24 full factorial design to de- recorded for Build 10. In this Build, the laser power was set to 13 W, the
termine their effect on the surface roughness and area ratio of the roller speed to 7.62 cm/s, the scan spacing to 127 μm, and recycled
produced parts. A novel experimental technique was used which in- powder was used. The smallest average surface roughness was recorded
cluded producing a 11 × 11 grid of 15 × 15 × 2 mm parts to determine
10
S. Petzold, et al. Polymer Testing 80 (2019) 106094
Table 5 piece plots – random behavior was observed. However, for Builds 1 and
ANOVA table for area ratio. 8, it is believed that the absorption of heat to the roller assembly caused
Source DF Adj SS Adj MS F-Value a higher degree of surface roughness towards the back of the SLS ma-
chine, while the overhead heating lamps caused higher surface rough-
Model 15 12.3819 0.82546 203.20 ness towards the front-right portion of the build platform. It was shown
Linear 4 9.4262 2.35656 580.09
that the variation of surface roughness and area ratio decreased closer
Laser Power 1 0.0790 0.07900 19.45
Roller Speed 1 1.3105 1.31051 322.59
to the center of the build area in comparison with the outer edge,
Powder Type 1 7.3520 7.35201 1809.77 probably due to a lower variation in temperature towards the center of
Scan Spacing 1 0.6847 0.68471 168.55 the build.
2-Way Interactions 6 2.4630 0.41050 101.05 An ANOVA was used to determine the main and interaction effects
Laser Power*Roller Speed 1 0.1213 0.12128 29.86
for surface roughness and area ratio. The main effects for surface
Laser Power*Powder Type 1 0.0738 0.07375 18.16
Laser Power*Scan Spacing 1 0.1258 0.12579 30.97 roughness were powder type, scan spacing, and roller speed. For area
Roller Speed*Powder Type 1 1.2693 1.26932 312.46 ratio, these same main effects were determined to be significant, with
Roller Speed*Scan Spacing 1 0.0007 0.00072 0.18 the addition of laser power in the model. Since almost all main or in-
Powder Type*Scan Spacing 1 0.8721 0.87212 214.68
teraction effects of the regression model, the sampled data was found to
3-Way Interactions 4 0.3973 0.09933 24.45
Laser Power*Roller Speed*Powder Type 1 0.1304 0.13044 32.11
be random in behavior. To obtain a better model of surface roughness
Laser Power*Roller Speed*Scan Spacing 1 0.1007 0.10067 24.78 and area ratio, a higher performing statistical modeling software, such
Laser Power*Powder Type*Scan Spacing 1 0.1661 0.16612 40.89 as R, should be implemented. Moreover, the surface roughness and area
Roller Speed*Powder Type*Scan Spacing 1 0.0001 0.00010 0.03 ratios were tested for normality and showed normal tendencies for al-
4-Way Interactions 1 0.0954 0.09535 23.47
most all builds, meaning there is not a significant skew towards higher
Laser Power*Roller Speed*Powder 1 0.0954 0.09535 23.47
Type*Scan Spacing or lower values.
Error 128 0.5200 0.00406
Total 143 12.9019
Future work
Fig. 22. Pareto chart showing significant effects for area ratio.
11
S. Petzold, et al. Polymer Testing 80 (2019) 106094
Fig. 23. Normal probability plot for area ratio. The same method to find a regression equation for surface roughness was used to determine the model for area ratio.
Equation (3) shows the regression equation for area ratio.
be shared at this time as the data also forms part of an ongoing study. [10] M. Vetterli, M. Schmid, W. Knapp, K. Wegener, New horizons in selective laser
sintering surface roughness characterization, Surf. Topogr.: Metrol. Prop. 5 (2017)
045007.
References [11] Sinterit, Lisa SLS Printer – Product's Specification, p. 2.
[12] L Giusca Claudiu, James D. Claverley, Wenjuan Sun, Richard K. Leach,
[1] P.B. Bacchewar, S.K. Singhal, P.M. Pandey, Statistical modelling and optimization Franz Helmli, P. Mathieu, J. Chavigner, Practical estimation of measurement noise
of surface roughness in the selective laser sintering process, Proc. IMechE, Pa. Birds: and flatness deviation on focus variation microscopes, CIRP Ann. - Manuf. Technol.
J. Eng. Manuf. 221 (2006) 35. (2014) 545.
[2] A.B. Varotsis, Injection molding SPI surface finishes. 3D hubs, SPI mold finishes, [13] Reinhard Danzl, Franz Helmli, Stefan Scherer, Focus variation – a robust technology
https://www.3dhubs.com/knowledge-base/injection-molding-spi-surface-finishes. for high resolution optical 3D surface metrology, J. Mech. Eng. (2011) 248–250.
[3] M. Zhao, D. Drummer, K. Wudy, M. Drexler, Sintering study of polyamide 12 [14] R. Danzl, F. Helmi, P. Rubert, M. Prantl, Optical roughness measurements on spe-
particles for selective laser melting, iJES 3 (1) (2015) 34–38. cially designed roughness standards, Opt. Fabr. Test. Metrol. III (2008) 7102M – 3.
[4] T. Osswald, Understanding Polymer Processing: Processes and Governing [15] N Duboust, H Ghadbeigi, C Pinna, S Ayvar-Soberanis, A Collis, R Scaife, K Kerrigan,
Equations, Munich (2017) (Print). An optical method for measuring surface roughness of machine carbon fibre-re-
[5] F. Sillani, R. Kleijnen, M. Vetterli, M. Schmid, K. Wegener, Selective laser sintering inforced plastic composites, J. Compos. Mater., p. 295.
and multi jet fusion: process-induced modification of the raw materials and analyses [16] R. Danzl, F. Helmli, S. Scherer, Focus variation – a robust technology for high re-
of parts performance, Addit. Manuf. 27 (2019) 32–41. solution optical 3D surface metrology, J. Mech. Eng. (2011) 247.
[6] D.T. Pham, K.D. Dotchev, W. Yusoff, Deterioration of polyamide powder properties [17] Alicona Manual, IFM G4 3.5.1 EN, (December 17, 2009), p. 2 (Chapter 1).
in the laser sintering process, Mech. Eng. Sci. 222 (11) (2008). [18] Bruker Alicona, Optical Metrology – Understanding Focus Variation, AZO Materials,
[7] W. Yusoff, D.T. Pham, K.D. Dotchev, Investigation of the thermal properties of 2017.
different grades polyamide 12 (PA12) in improving laser sintering process (SLS), [19] M. Schmid, Laser Sintering with Plastics: Technology, Processes, and Materials,
Appl. Mech. Mater. 548–549 (2014) 294–296. (2018), p. 114 Munich.
[8] M. Launhardt, A. LWorz, A. Loderer, T. Laumer, D. Drummer, T. Hausotte, [20] S. Singhal, P. Jain, P. Pandey, A. Nagpal, Optimum part deposition orientation for
M. Schmidt, Detecting surface roughness on SLS parts with various measuring multiple objectives in SL and SLS prototyping, Int. J. Prod. Res. 47 (22) (2009)
techniques, Polym. Test. (2016) 222. 6375–6396.
[9] A. Wegner, G. Witt, Influencing factors on surface roughness in laser sintering and [21] A. Yahamed, M. Joyce, P. Fleming, A. Pekarovicova, Polymers for 3D Printed
their effect on process speed, Fraunhofer Direct Digital Manufacturing Conference, Structures, Precision, Topography and Roughness, International Journal of
Berlin, 2012. Multidisciplinary Research and Studies 1 (2) (2018) 43–59.
12