Professional Documents
Culture Documents
Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma
A R T I C LE I N FO A B S T R A C T
Keywords: Design for additive manufacturing (DFAM) guidelines are important for helping designers avoid iterations and
Design guidelines leverage the design freedoms afforded by additive manufacturing (AM). Comprehensive design guidelines should
Powder bed fusion incorporate a variety of features of interest to designers, and given the wide variety of AM processes and their
Selective laser sintering associated capabilities and limitations, those guidelines may need to be process- or even machine-specific. One
Metrology
way to generate detailed DFAM guidelines is to implement a metrology study focused on a strategically designed
test part. This paper describes how quantitative design guidelines are compiled for a polymer selective laser
sintering (SLS) process via a metrology study. As part of the metrology study, a test part is designed to focus
specifically on geometric resolution and accuracy of the polymer SLS process. The test part is compact, allowing
it to be easily inserted into existing SLS builds and therefore eliminating the need for dedicated metrology builds.
To build a statistical foundation upon which design guidelines can be compiled, multiple copies of the test part
are fabricated within existing commercial builds in a factorial study with materials, build orientations, and
locations within the build chamber as control factors. Design guidelines are established by measuring and
analyzing the as-built test parts. The guidelines are summarized in this paper and documented in a publicly
accessible, online web tool.
1. Introduction In recent work, the authors designed a test part specifically for the
polymer SLS process [13]. The test part, shown in Fig. 1, was comprised
Guidelines are important tools for designers selecting the appro- of five panels of features attached to a common base. The features re-
priate manufacturing process for a specific part and tailoring the part presented those of particular interest to designers, and the panels could
for functionality and ease of manufacturing. Although Design for be removed from the base for measurement. The panels were arranged
Manufacturing and Assembly guidelines are prevalent for well-estab- in a cube that measured two inches along each side in an effort to re-
lished processes such as injection molding [1], they are recently duce the total volume required to build the test part. Unlike metal laser
emerging for AM processes. Notable recent work focuses on developing sintering processes, the SLS powder bed is self-supporting, which means
an ontology and knowledge base for representing Design for AM that parts do not require support structures. The self-supporting powder
knowledge [2] and representing design guidelines as modular compo- bed allowed tight grouping of the test part features and panels and easy
nents that can be implemented and interpreted efficiently across AM separation of them during post-processing. The resulting arrangement
processes and machines [3]. Several benchmarking studies focus on gave rise to a high density of features in a compact build envelope.
developing design guidelines that are independent of a specific AM By combining features found in previous test parts (e.g., [4,5,8,11])
process or on comparing AM processes to one another utilizing a but arranging them in a nested form, this design included a compre-
standard test part (e.g., [4–6]). However, process-specific investigations hensive variety of features in a compact cube. The part was built as a
may be required to establish a more complete understanding of a par- single unit, similar to the parts proposed by Mahesh et al. [4], Castillo
ticular AM process. For example, test parts have been designed for di- [8], and Moylan et al. [5], but the detachable panels made it easy to
rect metal laser sintering [7,8], material jetting [9], fused deposition access and measure each feature. The detachable panels were analogous
modeling [10], and SLS [11,12]. These test parts can capture the to a suite of test parts used in the Seepersad et al. study [11], which
nuances of specific AM processes more accurately, and the guidelines utilized a separate test part for each type of feature, but in a much more
created from them provide additional information to designers that compact form. Unlike some previous efforts that focused on char-
process-independent test parts may overlook. acterizing the mechanical properties of SLS parts (e.g., [14]), this test
⁎
Corresponding author.
E-mail address: ccseepersad@mail.utexas.edu (C.C. Seepersad).
https://doi.org/10.1016/j.addma.2018.10.035
Received 19 June 2018; Received in revised form 21 October 2018; Accepted 21 October 2018
Available online 28 October 2018
2214-8604/ © 2018 Elsevier B.V. All rights reserved.
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Fig. 1. Polymer test cube as-built (left) and disassembled (right) [13].
part focused exclusively on the geometric properties and surface char- Table 1
acteristics of SLS parts. Polymer test cube feature ranges [13].
Furthermore, polymer SLS enabled fabrication of all of the features. Feature Feature ranges Increment
Even complex features such as the thin rods and hinges, as well as the
nested configuration of panels, could be built successfully regardless of Holes 0.8–2.60 mm diameter 0.2 mm
build orientation. The part could not be built in various orientations 1.0–10.0 mm wall thickness 1.5 mm
Thin rods 0.3–0.9 mm diameter 0.1 mm
with a material extrusion process (e.g., FDM) or a metallic powder bed Thin walls 0.2–0.8 mm 0.1 mm
fusion process (e.g., DMLS) without support structures. Gaps 1.4–2.0 mm gaps 0.2 mm
The test cube features and feature sizes are documented in Table 1. 1.0–10.0 mm wall thickness 1.5 mm
Based on the results of the Seepersad et al. study, some feature sizes Cylinders 2.0–8.0 mm diameter 3.0 mm
Hollow cylinders 5.0–25 mm diameter 10.0 mm
were dimensioned such that they range from sizes that are too small to
Domes 6.0 mm diameter -
be resolved to sizes that are large enough to be resolved [11]. This Cones 6.0 mm diameter -
range allowed the resolution of the process to be characterized. 5.2 mm height
The details of the test part's features and procedures for measuring Linear accuracy 5.0–12.5 mm 2.5 mm
the features are described in detail in a previous publication from the Surface roughness 0–90° 15°
Hinges 0.6 mm and 1.0 mm shaft clearances –
authors [13]. As documented in [13], one of the unique characteristics
Lettering 10–18 pt font 2 pt
of the test part was its high feature density, defined as the number of 0.5–2.5 mm embedded/raised depth 0.5 mm
unique features per unit volume, combined with its minimal bounding Snap fits −0.5 to 0.25 mm offsets 0.05 mm
volume. These characteristics allowed it to be placed into unoccupied
volumes in existing SLS builds, enabling characterization of these builds
with respect to a very large number of features without displacing other interior (I) where the cube was built in the center of the chamber and
parts or requiring dedicated builds. In this way, feature density and exterior (E) where the cube was built along the edges of the build
build volume were prioritized over potential heat transfer issues that chamber. Here, exterior is defined as the region bounded by the peri-
may arise from the closely spaced panels. meter of the build area and a second perimeter that is offset from the
In this paper, the focus is on describing how the geometric cap- first by two inches toward the center of the build chamber. Table 2
abilities and surface characteristics of polymer SLS can be characterized shows a summary of the factors and levels that are used in the ex-
using this compact, comprehensive, process-specific test part. A me- perimental design, where each cell represents a specific combination of
trology study was conducted by fabricating the test part repeatedly in material, orientation, and location within the build chamber.
more than 100 commercial polymer SLS builds to investigate and sta- For a statistical characterization of polymer powder bed fusion
tistically analyze the effect of material choice, orientation, and location (PBF), several copies of the test part were built for each combination of
in the build chamber on the accuracy and resolution of a variety of experimental factors. For each unique material, orientation, and loca-
features. Actionable design guidelines were extracted from the results. tion combination, five replicates were built on one machine, and three
replicates on a second machine in order to assess the variability be-
tween PBF machines. In total, 144 test parts were required for the
2. Experimental procedure characterization.
All test parts were built by Stratasys Direct Manufacturing on 3D
A factorial-style study was conducted with the test part to char- Systems Sinterstation 2500 Plus and 3D Systems Sinterstation
acterize the resolution and accuracy of SLS parts under a variety of HiQ + HS machines featuring Integra multizone heater upgrades. The
conditions. Although a wide variety of process parameters can affect the machines were maintained by Stratasys Direct Manufacturing and
resolution and accuracy of features, four specific factors were selected tuned to production level specifications. Each of the test parts was in-
because they are often specified by the designer. The four factors were serted one-at-a-time into a production build filled with other commer-
material choice, orientation of the test part within the build chamber, cial parts. The build parameters were based on the 3D Systems default
location of the test part within the build chamber, and machine iden- settings, but modified by Stratasys Direct Manufacturing in an effort to
tity. Previous studies showed these factors to significantly affect the produce high quality, production parts. Compared to the default set-
mechanical performance of SLS parts [15–21]. However, corresponding tings, the modified parameters sought to balance feature detail with
studies on the geometric effect of these parameters were limited in mechanical strength.1 Nominally, the layer thickness and fill laser
scope [4,11]. By varying each of the four factors, a factorial experiment power were 100 μm and 42 W, respectively. After the build process, the
was conducted. Three different materials were considered: nylon 12 parts were cleaned by Stratasys technicians using a combination of
(PA 12), nylon 12 reinforced with glass beads (GF PA 12), and fire compressed air bead blasting and brush-like tools to remove any re-
retardant nylon 11 (FR PA 11). The test cube was built in three different sidual powder. The cleaning methods were consistent with industry
orientations, denoted by the plane of the build chamber in which the
base was aligned: XY, XZ, and YZ with the z-axis orthogonal to the build
platform and the x-axis parallel to the front window of the machine. 1
The specific values of the process parameters are proprietary to Stratasys
Location within the build chamber was reduced to two specifications: Direct Manufacturing.
240
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Table 2 significant, and the multiple comparisons test quantifies the effect of
Summary of factors and levels for experimental design. the factor on the feature of interest. The result of a multiple compar-
isons test provides an indication of the differences in the means be-
tween different levels of a statistically significant factor. The results are
summarized in Table 3. As with the significant effects results, “Average
Mean Difference” is calculated by averaging each of the mean differ-
ences generated from the multiple comparisons test in the ANOVA
study. The results can be interpreted as the average variation associated
with changing between levels for a given process parameter. The results
of the ANOVA study are described in greater depth in the next section in
the context of each of the test cube features.
standard practices so that the test part measurements would reflect
production–quality SLS capabilities. A variety of measurement techni- 3. Results
ques were used to analyze each of the different test cube features. The
measurement techniques are described in detail in [13] and are sum- The measurement results are compiled into a set of guidelines that
marized here alongside the results for each feature. can be used by designers creating parts for polymer SLS. The guidelines
To identify which input parameters have a statistically significant are available publicly through an online web tool that can be accessed
effect on part accuracy, an analysis of variance (ANOVA) test was at http://DesignForAM.me.utexas.edu. A screenshot from the tool is
conducted. The output of the test specifies which factors are significant shown in Fig. 2. The web tool describes the test part, the measurement
at a five percent significance level (α = 0.05). As a designer, these procedure, and the general design guidelines derived from the mea-
factors should be considered when creating a part to be manufactured surements. It also includes graphs and tables summarizing the mea-
using selective laser sintering. A summary of the significant effects for surements of each feature and allows the user to customize those results
all features is provided in Table 3. In the table, “Percentage Significant” for a specific material, orientation, and location in the build chamber.
refers to the fraction of settings for a particular feature for which the In the rest of this section, the general design guidelines are reported,
factor is statistically significant. For example, material choice is statis- along with average measurements across all material choices, orienta-
tically significant for all gap sizes, but location is a significant factor for tions, and build chamber locations for all 144 test cubes in the ex-
only 57% of gap sizes. perimental design. The average values for specific combinations of
After performing the ANOVA test, a multiple comparisons test was material, orientation, and location within the build chamber are tabu-
conducted to quantify the effect of each significant factor on the feature lated in Appendix B. Important insights regarding the significance of
of interest. The ANOVA test tells the designer which factors are the effect of material choice, orientation, or build chamber location on
those measurements are also included in Table 3. Section 3.11 describes
Table 3 the additional information available in the web tool, namely customi-
Significant effects and mean differences for test cube features. “Percentage zation of each of the tables and graphs in this section for a specific
Significant” refers to the fraction of levels within a particular feature in which combination of material choice, orientation, and build chamber loca-
the given factor is statistically significant at a p-value of 0.05. For example, tion.
material choice is statistically significant for all gap sizes, but location is a
significant factor for only 57% of gap sizes. Similarly, “Average Mean 3.1. Surface roughness
Difference” is calculated by averaging the mean differences generated from the
multiple comparisons test in the ANOVA study. For linear accuracy, the average
The test part features for evaluating the surface roughness at various
effect of changing materials causes a 0.23 mm difference in the reported linear
build angles are shown in Fig. 3. Surface roughness was measured using
dimensions.
a Zeta-20 3D optical surface profiler. The Zeta software used internal
Significant effects algorithms to measure the profile of the scan area, and average
roughness (Sa) and root mean square roughness (Sq) values were re-
M O L M*O M*L O*L Machine
ported. The measurement procedure for surface roughness is described
Surface roughness further in Appendix A.
Percentage significant 100% 0% 0% 0% 0% 0% 33% The average surface roughness results from all test cubes are shown
Average mean difference 9.36 – – – – – 8.76 in Fig. 4. The surface roughness is smallest at an angle of 0° relative to
(μm) the build plane. The maximum roughness occurs at 15° then decreases
Linear accuracy at all subsequent angles. Despite the decreasing trend, the roughness
Percentage significant 100% 100% 0% 0% 0% 0% 0% still has a minimum value at 0°. Additional results can be found in
Average mean difference 0.23 0.06 – – – – – Table 13 of Appendix B.
(mm)
241
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Linear accuracy indicates how well the machine holds linear di-
mensions with respect to the prescribed dimensions within the CAD
model. Linear accuracy in each build axis was determined by measuring
the gap between successive walls. Fig. 5 shows the linear accuracy
features in the test part. The measurements were taken with iGaging
digital calipers (0.01 mm resolution), and each feature was measured
three times and the average is reported. The measurement uncertainty
was quantified through a gage R&R study, described in Appendix A. Fig. 4. Average surface roughness as a function of build angle for all test cubes.
The error bars in the figure correspond to 95% confidence intervals about the
mean value.
3.2.1. Accuracy
The linear accuracy results are shown in Fig. 6 and detailed in
242
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Fig. 5. Linear accuracy features in the test part. Features of interest are tabs Fig. 7. Features for evaluating gap accuracy as a function of wall thickness in
aligned along the outer edge of the part. the test part.
increase with the linear dimension. The degree of under-sizing does not
appear to scale with feature size, suggesting a greater contribution from
oversintering than shrinkage. Embedded features built on the interior of
the build chamber are usually more undersized than those built along
the perimeter. With respect to orientation, distances along the Z axis of
the build chamber tend to be closer to the defined CAD dimension than
along the X or Y axes. PA 12 has the highest level of accuracy, followed
by GF PA 12 and FR PA 11.
Gap accuracy was determined by varying the gap size and the
thickness of the wall through which the gap was created. The reported
value corresponded to the gap measured along the axis of interest. Four
gaps of varying dimensions were placed within each of the seven walls,
as shown in Fig. 7. Gap accuracy was measured using iGaging digital
calipers (0.01 mm resolution). The measurement uncertainty for gaps
Fig. 6. Linear accuracy trend for all test cube measurements. The red line in- was quantified as described in Appendix A.
dicates the nominal as-designed value, and the error bars correspond to 95%
confidence intervals about the mean. 3.3.1. Accuracy
Similarly to linear accuracy, gaps tend to be undersized by ap-
Table 14 of Appendix B. SLS tends to undersize dimensions by ap- proximately 0.2 mm. As shown in Fig. 8, there does not appear to be a
proximately 0.2 mm. When built, each of the measured distances in significant dependence on wall thickness. Additional results, separated
Fig. 5 are generally 0.2 mm smaller than their designed value. This by input factor, are provided in Table 15 of Appendix B.
trend appears to be independent of size and is consistent across the
entire range of measurements. 3.3.2. ANOVA results
The results generated from an ANOVA study and subsequent mul-
3.2.2. ANOVA results tiple comparisons test are shown in Table 3. Unlike linear accuracy, gap
The complete analysis of variance results are tabulated in Table 3.
The percentages in the table represent the proportion of feature levels
for which the particular parameter is considered statistically significant.
Linear accuracy is assessed at four distinct levels. The ANOVA results
suggest material and orientation are statistically significant for each of
the four levels. Table 3 also displays the differences in mean that can be
attributed to each of the parameters based on the ANOVA study. On
average, material and orientation can contribute to 0.23 mm and
0.06 mm of variation in linear accuracy, respectively. For designers, a
consideration must be made regarding the accuracy associated with
certain materials and orientations. PA 12 produces the highest degree of
linear accuracy, and changing materials can cause linear dimensions to
be undersized by as much as 0.34 mm.
243
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
244
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
thicknesses of 4.0 mm and smaller. Both material choice and hole or-
ientation contribute to large observed differences in means. Orientation
has the greatest impact, with mean differences of 0.32 mm on average.
PA 12 holes tend to be larger than both GF PA 12 and FR PA 11 holes.
Holes built in the Z-direction are the most undersized because the entire
cross section of the hole lies in the plane of the laser path. The spot size
of the laser is larger than the layer thickness, so holes tend to be more
undersized when they are built in-plane.
3.5.1. Accuracy
As shown in Fig. 13, thin walls are usually oversized by approxi-
mately 0.2 mm. This trend does not depend significantly on the thick-
ness of the wall. The results are further stratified by input factor in are the most accurate. XZ and YZ walls are generally 0.10 to 0.12 mm
Table 17 of Appendix B. thicker than the nominal dimension by comparison.
Fig. 12. Thin wall features in the test part. Fig. 14. Thin rod features in the test part.
245
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Fig. 15. Example of feature resolution for thin rods. The intended geometry from the CAD model is shown on the left. The as-built geometry is on the right, with
arrows to show the pass/fail criteria.
3.6.1. Resolution
Resolution was determined through visual inspection. Features that
did not build successfully were given a “Fail” rating. Unsuccessful Fig. 16. Hinge features in the test part.
features included those that built poorly or failed to build at all. An
example of this convention is shown in Fig. 15, where the bottom four
Table 7
rods received a “Pass” rating and the top three missing rods were given
Hinge resolution at shaft clearances of 0.6 and 1.0 mm for all test cubes. The
a “Fail” rating. values within each cell indicate the proportion of observed instances that cor-
As shown in Table 6, no significant relationship is observed between rectly resolved. (For interpretation of the references to color in this table le-
the support conditions and the ability to resolve thin rods. In order to gend, the reader is referred to the web version of this article.).
ensure reliable resolution of thin rods, they should generally be at least
0.6 mm in diameter. Smaller rods tend to fail or leave an “empty” spot
in their place.
3.8. Lettering
246
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Table 10
Resolution of raised lettering for all test cubes. The cell color follows the pass/
intermediate/fail convention outlined in Table 8 corresponding to green,
yellow, and red, respectively. The color is determined by the highest count of
either pass, intermediate, or fail observations. (For interpretation of the refer-
ences to color in this table legend, the reader is referred to the web version of
this article.).
Fig. 17. Lettering features for embedded (left) and raised (right) letters.
Table 8
Convention used to evaluate lettering resolution. (For interpretation of the re-
ferences to color in this table legend, the reader is referred to the web version of
this article.).
spot size versus layer thickness mentioned in the thin walls section.
Consistent with many of the other features, the best results are achieved
with PA 12, followed by GF PA 12 then FR PA 11. Location within the
build chamber does not produce a discernable effect on lettering re-
solution.
lowercase forms of the letter “A” were used as the test font. Lettering
resolution was determined using the criteria outlined in Table 8.
3.9. Snap fits
3.8.1. Resolution The offset required for mechanical snap fits was determined by
Tables 9 and 10 show the resolution results for embedded and raised testing an extruded square peg against slots of varying size (Fig. 18).
lettering, respectively. For each cell in the tables, the total number of Square pegs of 3.0 mm nominal width were tested against slots of
instances of “pass,” “intermediate,” and “fail” ratings according to the 2.9 mm to 3.5 mm nominal width in increments of 0.1 mm. Snap fit
criteria in Table 8 are tallied. The cell color is determined by the cri- resolution was defined by the offset distance between the peg and slot
terion with the highest number of occurrences. For instance, green cells that provides a snug fit.
indicate more “passing” features than “intermediate” or “failing” ones,
but not necessarily a majority. Lettering resolution does not appear to 3.9.1. Resolution
depend significantly on the depth of the embed/extrude, but depends In general, as documented in Table 11, when incorporating me-
substantially on the size of the font. Embedded lettering resolved for all chanical snap fits into a design, an offset of 0.15 mm to 0.2 mm should
font sizes at all depths. For this reason, a designer should consider be used between the peg and slot. An example design and a depiction of
embedded lettering as opposed to raised lettering when adding text to a the offset distance are shown in Fig. 19.
part that is additively manufactured using selective laser sintering. If
raised lettering is desired, fonts larger than 16 pt should be used. 3.9.2. Design guidance for snap fits
Snap fits that are built with the primary axis orthogonal to the build
3.8.2. Design guidance for lettering platform require a larger offset for the best performance compared with
Embedded lettering is preferred to raised lettering because raised the other orientations. Attributed to the poorer resolution of the ma-
lettering requires many fine “positive” features. At font sizes below terial, FR PA 11 snap fits require the largest offset compared with GF PA
16 pt the features are too small to be resolved by the laser spot and 12 and PA 12. The results do not suggest a significant dependence on
result in significant defects. Embedded lettering is less susceptible to location within the build chamber.
these defects because only the feature “negative” is scanned by the
laser. Lettering oriented along the Z axis of the build chamber performs 3.10. Machine variation
worse than the other directions. This trend is attributed to the effects of
Five test cube replicates were built on one machine, and three re-
Table 9 plicates were built on a second machine for each combination of ma-
Resolution of embedded lettering for all test cubes. The cell color follows the terial, orientation, and build chamber location in an effort to assess
pass/intermediate/fail convention outlined in Table 8 corresponding to green, variability between machines. However, due to manufacturing
yellow, and red, respectively. The color is determined by the highest count of
either pass, intermediate, or fail observations. (For interpretation of the refer-
ences to color in this table legend, the reader is referred to the web version of
this article.).
247
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Table 11
Snap fit resolution for all test cubes. Green boxes indicate the offsets that cor-
respond to the best fit between the peg and slots. The values within each cell
indicate the proportion of observed instances that correctly resolved. (For in-
terpretation of the references to color in this table legend, the reader is referred
to the web version of this article.).
Fig. 19. Example snap fit design. The image on the right shows the peg and slot dimensions, representing an offset distance of 0.15 mm.
248
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Surface roughness was measured using a Zeta-20 3D optical profiler. The profiler created 3D surface scans generated from a series of images. The
images were taken at 0.3 μm steps in the Z-direction and the reported X–Y resolution was 0.9 μm. Each scan had a field of view of 1646 × 1235 μm,
and surface roughness was measured after stitching two surface scans together with a 10% overlap for a total field of view of 3128 × 1232 μm.
A gage repeatability and reproducibility (R&R) study was performed to quantify the uncertainty for the caliper measurements of linear accuracy,
gap size, and wall thickness. The R&R was conducted using 12 parts, 3 trials, and 2 operators. The R&R percentages listed in Table 12 correspond to
the fraction of total part variation attributed to the measurement system. While the R&R values for linear accuracy and wall thickness were within
the range of what is generally acceptable, the R&R for gap size was slightly higher. The observed part-to-part variation for gaps was smaller than the
other features, so a larger percentage of variation was associated with the measurement system.
Table 12
Gage R&R results for caliper measurements.
Feature R&R (%)
249
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Fig. 22. Hole size as a function of wall thickness with sample flatbed scan image.
Table 13
Average surface roughness for each combination of input parameters. The values in the table represent the average surface roughness across materials, locations, and
inclination angles relative to the build platform in microns.
Inclination angle Surface roughness
PA 12 GF PA 12 FR PA 11
Table 14
Average linear accuracy for each combination of input parameters. Accuracy is defined as the measured dimension minus the nominal dimension. The values in the
table represent the average accuracy for a specific material, location, and orientation.
Linear accuracy
PA 12 GF PA 12 FR PA 11
XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ
−0.02 −0.07 −0.09 −0.02 0.03 −0.02 −0.28 −0.23 −0.17 −0.36 −0.25 −0.18 −0.40 −0.34 −0.35 −0.36 −0.31 −0.30
Table 15
Average gap accuracy for each combination of input parameters. Accuracy is defined as the measured dimension minus the nominal dimension. The values in the
table represent the average accuracy for a specific material, location, orientation, and wall thickness.
Wall thickness Gap Accuracy
(mm)
PA 12 GF PA 12 FR PA 11
XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ
1.0 −0.21 −0.22 −0.01 −0.12 −0.11 0.01 −0.26 −0.25 −0.14 −0.32 −0.22 −0.19 −0.26 −0.27 −0.23 −0.26 −0.27 −0.26
2.5 −0.13 −0.17 −0.08 −0.08 −0.08 −0.05 −0.23 −0.25 −0.23 −0.33 −0.28 −0.26 −0.35 −0.40 −0.34 −0.34 −0.36 −0.52
4.0 −0.11 −0.16 −0.09 −0.09 −0.06 −0.06 −0.24 −0.26 −0.25 −0.33 −0.29 −0.27 −0.36 −0.42 −0.40 −0.34 −0.39 −0.60
5.5 −0.12 −0.18 −0.10 −0.09 −0.09 −0.06 −0.25 −0.27 −0.27 −0.32 −0.29 −0.27 −0.34 −0.44 −0.46 −0.34 −0.40 −0.50
7.0 −0.11 −0.18 −0.11 −0.12 −0.09 −0.06 −0.25 −0.26 −0.27 −0.32 −0.28 −0.26 −0.34 −0.43 −0.45 −0.38 −0.39 −0.52
8.5 −0.14 −0.19 −0.12 −0.09 −0.10 −0.06 −0.23 −0.26 −0.27 −0.31 −0.26 −0.25 −0.35 −0.42 −0.42 −0.35 −0.38 −0.51
10 −0.10 −0.19 −0.11 −0.10 −0.11 −0.05 −0.20 −0.24 −0.24 −0.29 −0.23 −0.22 −0.33 −0.36 −0.35 −0.36 −0.32 −0.45
250
J. Allison et al. Additive Manufacturing 25 (2019) 239–251
Table 16
Average hole accuracy for each combination of input parameters. Accuracy is defined as the measured diameter minus the nominal diameter. The values in the table
represent the average accuracy for a specific material, location, orientation, and wall thickness.
Wall thickness Hole accuracy
(mm)
PA 12 GF PA 12 FR PA 11
XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ
1.0 −0.14 −0.23 −0.34 −0.01 −0.03 −0.21 −0.18 −0.24 −0.48 −0.29 −0.24 −0.55 −0.16 −0.18 −0.44 −0.16 −0.16 −0.62
2.5 −0.14 −0.24 −0.53 −0.02 −0.15 −0.30 −0.22 −0.30 −0.62 −0.41 −0.35 −0.75 −0.34 −0.35 −0.73 −0.36 −0.33 −0.81
4.0 −0.21 −0.29 −0.63 −0.16 −0.23 −0.43 −0.27 −0.38 −0.74 −0.50 −0.46 −0.76 −0.52 −0.45 −0.88 −0.46 −0.42 −0.94
5.5 −0.26 −0.34 −0.64 −0.14 −0.25 −0.52 −0.31 −0.41 −0.88 −0.60 −0.55 −0.83 −0.45 −0.50 −0.72 −0.45 −0.49 −0.92
7.0 −0.34 −0.40 −0.68 −0.23 −0.25 −0.53 −0.32 −0.53 −0.90 −0.54 −0.55 −0.91 −0.52 −0.59 −0.75 −0.52 −0.53 −0.99
8.5 −0.26 −0.50 −0.72 −0.23 −0.19 −0.59 −0.36 −0.55 −0.75 −0.49 −0.52 −0.88 −0.56 −0.60 −1.09 −0.59 −0.58 −1.06
10 −0.27 −0.48 −0.64 −0.30 −0.21 −0.61 −0.34 −0.50 −0.81 −0.54 −0.44 −0.78 −0.63 −0.57 −0.98 −0.62 −0.54 −0.99
Table 17
Average thin wall accuracy for each combination of input parameters. Accuracy is defined as the measured dimension minus the nominal dimension. The values in
the table represent the average accuracy for a specific material, location, and orientation.
Thin wall accuracy
PA 12 GF PA 12 FR PA 11
XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ XY XZ YZ
0.04 0.26 0.26 0.04 0.10 0.10 0.10 0.19 0.20 0.21 0.12 0.20 0.25 0.29 0.36 0.22 0.31 0.42
Hole diameter was measured using an optical flatbed scanner and processed in Matlab. As shown in Fig. 22, the part was scanned at a resolution
of 600 dpi and converted to a black and white image. The area of the white space occupied by each hole in the image was used to calculate the
diameter. It should be noted that eccentricity of the holes was not taken into account, so average diameter was reported. This measurement system
was verified against pin gages in increments of 0.001 in. Furthermore, each image was individually calibrated to the dimensions of a square with a
known size (red square in Fig. 22). Additional information on the measurement methodology is provided in [13] and in the accompanying web tool
(http://DesignForAM.me.utexas.edu).
251