...

Barclift-Properties-3DPintedParts.pdf

by user

on
Category: Documents
10

views

Report

Comments

Transcript

Barclift-Properties-3DPintedParts.pdf
EXAMINING VARIABILITY IN THE MECHANICAL PROPERTIES OF PARTS
MANUFACTURED VIA POLYJET DIRECT 3D PRINTING
Michael W. Barclift and Christopher B. Williams
Design, Research, and Education for Additive Manufacturing Systems Laboratory
Department of Mechanical Engineering, Virginia Tech
REVIEWED, Accepted August 15, 2012
ABSTRACT
In Objet’s PolyJet process, part layers are created by selectively inkjetting photopolymers
onto a build substrate and then cured with ultraviolet lamps. With an eye towards using PolyJet
as a manufacturing process to fabricate end-use products, the authors examine the sensitivity of
part material properties to variation in process parameters. Specifically, a design of experiments
is conducted using a full-factorial design to analyze the effects of three parameters on the
specimens’ tensile strength and tensile modulus: the in-build plane part orientation (X-Y), the
out-of-build plane part orientation (Z), and the distance between specimens. Results show that
part spacing has the largest effect on the tensile strength, but the three parameters produced no
statistically significant effects on the tensile modulus. Orienting specimens in XZ orientation
with minimal part spacing resulted in the highest tensile strength and modulus. Whereas,
orienting specimens in the YZ orientation at the farthest part spacing led to the lowest
mechanical properties.
Keywords: Objet, PolyJet, Design of Experiments, 3D Printing
1. INTRODUCTION
PolyJet Direct 3D Printing (PJD-3DP) is
an Additive Manufacturing (AM) process that
was developed by Objet Geometries, Ltd
(Israel). In the process, layers of an acrylicbased photopolymer are selectively jetted onto
a build-tray via inkjet printing (Figure 1). The
jetted photopolymer droplets are immediately
cured with ultraviolet lamps that are mounted
onto the print carriage. The differentiating
functionality of the PJD-3DP process (for the
Connex machine line) is its ability to
simultaneously jet multiple materials with
Figure 1. The PolyJet Direct 3DP Process [1]
different mechanical (and optical) properties.
For example, by simultaneously jetting "TangoBlack" (an elastomeric material) and
"VeroWhite" (a rigid polymer), a designer is able to manufacture products with functionally
graded material properties. As such, PJD-3DP offers design freedom in specifying both the
placement and the properties of materials on a point-by-point basis.
876
While primarily used as a process for fabricating prototypes, the design freedoms offered by
the process offer the opportunity to produce end-use products with unique, tailored geometries
and materials. One example is Objet’s marketing of FullCure 630 material, which is specifically
formulated for the production of end-use custom-fit hearing aids. To be suitable for use as a
manufacturing platform, the technology’s structure/process/property relationships must be well
established. However, as it is a relatively new AM process, research on this technology is still
emerging.
1.1. AM Structure/Process/Property Relationships
Identifying structure/process/property relationships is critical for every manufacturing
process; especially in AM where anisotropic properties can emerge due to material/energy
patterning techniques, the build orientation of the part, and the interfaces between layers. Such
research has been a key component of the advancement of all AM processes. For example,
Agarwala and co-authors investigated how build strategies affect part porosity, and thus
structural quality, in Freeform Filament Fabrication (FFF) [2]. Ahn and coauthors discovered
that poor interlaying bonding in FFF contributed to significantly reduced tensile strength when
parts were pulled perpendicular to the direction of layer construction [3]. Researchers have
demonstrated that parts created by the Laser Sintering (LS) process are also anisotropic, with
parts created as XY or XZ (part orientation and coordinate system used in ASTM F2921) were
the strongest and parts oriented as ZX or ZY were the weakest [4, 5].
Existing research of the PJD-3DP process has primarily focused in characterizing the
resultant parts. Pilipovic and coauthors note that their measured values for tensile strength are
lower than that which is reported by Objet [6]. Udroui and Mihail experimentally determined
that the “glossy surface” finishing option provided a higher quality surface roughness than the
“matte surface” option [7]. Existing structure/process/property research is limited to Brajlih and
coauthors’ investigation of algorithms for scaling parts to improve process accuracy [8] and
Kesy and Kotlinksi’s exploration of the effects of part orientation on material properties [9].
While hardness was unaffected by orientation, both tensile strength and elongation at break were
found to be lowest for parts created with XZ orientation.
Although structure/process/property relationships for PJD-3DP have not yet been fully
defined, there perhaps is an opportunity to draw from established research in the
stereolithography (SL) process, due to their common photopolymerization approach. Paul
Jacob’s extensive characterization of the SL process connects laser-based UV exposure of
photopolymer resins to the cure depth of single vector scan to the mechanical properties of the
finished part [10]. Specifically, elastic modulus and tensile strength increase (to a certain point)
with increased exposure. This fundamental process characterization, which identifies the residual
stresses caused by curing shrinkage [11, 12] and the “print-through” error caused by accumulated
exposure across several layers, is linked to, and has driven the design of, SL laser scanning
patterns [13]. SL process understanding has also provided a means of interpreting part anisotropy
caused by varying layer thickness [14] and part orientation [15-17]. In [17] it was found that
tensile samples oriented as XZ or YZ were statistically stronger than those printed flat on build
tray as XY or YX due to differences in inter-layer bonding.
877
1.2. Research Scope
Although PJD-3DP and SL are functionally linked by their use of photopolymerization [18],
the manner in which the UV irradiation is patterned is quite different. The vector scan-based
approach of SL provides the ability to precisely pattern UV energy across the resin bath. The
Objet process, however, selectively patterns resin while indiscriminately distributing UV energy
via lamps mounted on the printing-block. Given that mechanical properties of photopolymer
parts are so closely linked to the UV exposure, gaining an understanding of how PJD-3DP part
properties relate to process parameters is paramount to ensuring its relevance and reliability as a
manufacturing platform.
In this paper, the authors explore the process/property relationships of PJD-3DP via a design
of experiments. Specifically, a full-factorial design is used to analyze the effects of three
parameters on the specimens’ tensile strength and tensile modulus: (i) the in-build plane part
orientation (X-Y), (ii) the out-of-build plane part orientation (Z), (iii) and the distance between
specimens. This work is guided by the research question, “What controllable factors cause
variability in the tensile strength and tensile modulus of VeroWhite parts manufactured through
PolyJet Direct 3D Printing?” The PolyJet process is detailed in Section 2. The experimental
design is outlined in Section 3 and the experimental methods are presented in Section 4. Results
and discussion are presented in Sections 5 and 6, respectively.
2. POLYJET DIRECT 3D PRINTING
As briefly described in Section 1, the PJD-3DP process features direct inkjet printing of
photopolymer resin and subsequent curing by UV lamps. The Connex machine features eight
print heads that can simultaneously print three materials: four heads are dedicated for jetting
support material; the other four heads jet two distinct build materials. A small roller in the
printing block flattens jetted droplets to provide an even surface for the deposition of subsequent
layers. The process offers a relatively high resolution in the X-Y build plane (600 dpi printing
resolution [19]) at small layer thicknesses (32 m, or 16 m in “High Quality” mode [19]). To
provide scaffolding and stability to jetted droplets, printed parts are completely encased by the
support material. The top-most surfaces are also typically coated in support material to provide
an even surface finish (known as “matte finish”); however, the operator can choose to leave the
top surface uncoated (known as “glossy finish”). Once the part has finished printing, the support
material is removed from the part via a high-pressure water jet.
During printing, the print-block moves along the X-axis (Figure 2a) and deposits material in
two back-and-forth translations (which constitutes a full “pass”). Jetting occurs only in the first
forward translation; the remaining translations are solely to complete polymer curing. After
completing a full pass, printing continues in the next printing “path” along the Y-axis of the
build tray (Figure 2b). The extents of each printing path are defined by the width of the
printheads. For example, the tensile specimens arranged on the build tray in Figure 2a each span
three printing paths. The print-block offsets itself within each path between printing layers to fill
in any voids that may have occurred due to clogged print nozzles.
878
(a)
(b)
Figure 2. Objet PJD-3DP Terminology
As translation between printing paths is a process bottleneck, the process’s CAM interface
(Objet Studio) automatically places parts in the build tray such that the longest dimensions are
aligned along the X-axis, near the print origin. Additional “build rules” for orienting and placing
parts within the build tray are presented in Table 1. These rules are derived as methods for
minimizing total build time, minimizing support material consumption, and improving part
surface quality. These rules are used for developing and selecting test parameters for
experimental analysis, described in Section 3.
Table 1. Objet Build Rules [19]
Rule
X-Y-Z
TallLeft
Recess
-Up
FineSurface
Description
 Place longest dimension along X-axis
 Place intermediate dimension along Y-axis
 Place smallest dimension in Z-axis
Place tallest part on the left of build tray
Orient indented surface features facing-up
Orient finely detailed features facing-up
Effect
Rationale
Reduced overall
build time
 Reduce number of printing paths
 Reduce total number of layers
Reduced overall
build time
Reduced total part
cost
Improved surface
quality and lower
cost
Reduce translations needed by printblock to create part
Reduce consumption of support
material
 Improved resolution along Z-axis
 Reduce support material
consumption
3. EXPERIMENTAL DESIGN AND METHODS
3.1. Parameter Selection
Given that mechanical properties of photopolymer parts are closely linked to UV exposure
received (Section 1.1) and the manner in which the PJD-3DP process broadly patterns UV
irradiation (Section 2), the authors hypothesize that a part’s mechanical properties might be
affected by build orientation and/or the spacing between parts across printing paths. While
orientation in the X-Y and X-Z planes have been shown to cause variability in mechanical
properties, the samples in that existing study were all printed simultaneously on the same tray,
and thus did not account to variations in exposure levels during the print [9].
879
To identify process variability, the authors chose to employ a design of experiments for the
PJD-3DP process. To properly analyze and select pertinent parameters, [19] and [20] were
consulted based on their "Six Sigma" and quality control methodologies. An Ishikawa Diagram
(aka Cause-and-Effect or Fishbone Diagram) was created based on the authors’ knowledge of the
PJD-3DP process (Section 2), and included potential factors that could produce deviation in the
mechanical properties of VeroWhite material, as seen in Figure 3. Though this diagram does not
guarantee nor reveal the true root cause, it was believed that through its graphical and organized
structure, it would highlight the family where the root cause existed.
Figure 3. Ishikawa Diagram for PJD-3DP Process
Though many branches of the Ishikawa Diagram presented factors that could affect the
material properties of the specimen, the “Machine” branch was further analyzed due to its
uniqueness to PJD-3DP. Three parameters, believed to be the most pertinent and fundamental to
creating PolyJet parts, were chosen for analysis in this study. A description for the selected
parameters (X-Y Orientation, Z Orientation, and Part Spacing), and the rationale for their
selection (i.e., hypothesized effect), is provided in Figure 4.
880
Parameter: X-Y Orientation
XY
Description: The in-plane build orientation of the part
such that its length is parallel or perpendicular with
respect to the front of the build tray.
YX
Hypothesis: Orienting parts across print-head paths, may
lead to lower mechanical properties due to banding from
discretized jetting nozzles.
Parameter: Z-Orientation
YX or XY (flat)
YZ or XZ (angled)
Description: The build direction orientation of the part
such that the width, not the length, lays flat or angled
with respect to the X-Y plane.
Hypothesis: An increase in the number of layers causes
curing “print-through” and thus increased mechanical
properties for parts with widths aligned in the Z plane
(as noted for SL in [17]).
Parameter: Part Spacing
Tight Spacing
Description: The overall spacing between parts in X-Y
plane of the build-tray.
Far Spacing
Hypothesis: Smaller spacing may lead to increased
mechanical properties due to potential UV over-cure
effect.
Figure 4. Experimental Parameters
3.2. Design of Experiments
The goal of the experiments is to identify the variability in the mechanical properties
(specifically, the tensile strength and modulus) of parts created using the VeroWhite resin. Objet
specifies that Fullcure 830 Verowhite has a tensile Strength of 49.8 MPa and tensile modulus of
2495 MPa [22]. Following theory presented in [23], a three parameter and two-level full factorial
design of experiments (DOE) was selected to analyze the effects of the three selected parameters.
The two-level full factorial design was selected because the parameters could be easily varied at
two discrete levels and statistically analyzed using only eight total experiments, making the DOE
easy to regulate and execute due to low complexity.
Each parameter was varied at a "high" (+ 1) and "low" (-1) value and their effect on the
tensile strength and modulus was measured. The full-factorial experiment is shown in Table 2.
Each experiment was given a codified name to quickly identify the experiment and the levels of
881
the parameters. For example, “YXT” corresponds to an experiment with specimens oriented with
lengths along the y-axis, widths along z-axis, and tight part spacing (Figure 5).
Table 2. Full Factorial Experiment
Parameters
Levels
YXT
YXF
YZT
YZF
Experiment
Code Name
XYT
XYF
XZT
XZF
X-Y Orientation
YX (-1)
XY (+1)
-1
-1
-1
-1
1
1
1
1
Z-Orientation
YX or XY (-1)
YZ or XZ (+1)
-1
-1
1
1
-1
-1
1
1
Part Spacing
Tight (-1)
Far (+1)
-1
1
-1
1
-1
1
-1
1
Figure 5. Build platform layout for “YXT” (specimens oriented in the y-axis, flat z-orientation,
and tight part spacing) experiment.
3.3. Experimental Procedure
ASTM D638-10 was followed to measure specimens’ tensile strength and modulus. This
standard has relevancy in engineering design and enables comparison across processes, including
existing AM process analysis [17]. Figure 6 provides dimensional data for the selected "Type 1"
tensile test specimen. Specimens’ thicknesses and widths were measured using a Marathon
Electronic Digital Caliper (accuracy of 0.01 mm) to monitor any variable resolution and
discrepancy. The specimens’ masses were measured on a digital scale (accuracy of 0.01g) were
monitored to determine variation.
Figure 6. ASTM D638 Type 1 specimen dimensions
882
Each experiment (Table 2) was randomly chosen and printed using an Objet Connex 350.
Parts were printed in “Digital Material Mode” with the “Matte” surface finish. Random selection
prevents experimental bias and propagation of a consistent error [23]. Eight total build trays were
printed in this study. Five specimens were printed per build tray, with each build tray
corresponding to a specific experiment with uniform orientation among all printed specimen (no
mixed orientations). Following the completed print job, specimens were cleaned using a high
pressure waterjet. Upon drying, the specimens were conditioned at 72˚F in a closed opaque
plastic container to prevent excess light exposure until ready for testing. Since [13] and [21]
identify “time” as a possible degradation effect (due to oxygen levels and ambient light), the
authors attempted to maintain a constant time between printing and testing for all experiments
with a minimum conditioning period of 12 hrs. The authors also attempted to control the
humidity of the conditioning specimens by storing the plastic container in metal cabinet located
away from potential sources of moisture.
Tensile tests were conducted in the Virginia Tech Material's Testing Laboratory using an
Instron 4468 Tensile Testing Machine in accordance to D638 standard. Using MTS Systems
TestWorks for data acquisition with the Instron machine, tests were conducted at a stable
environment of 72˚F and tested at a speed of testing of 50 mm/min (constant strain rate) until
specimens ruptured. TestWorks was used to measure the applied force and strain of the
specimen, which was then used to calculate the tensile strength and tensile modulus. The Tensile
modulus was calculated using the Tangent Modulus method. Specimen elongation was measured
with an axial clip-on extensometer (50% max strain). Test specimens were not tempered.
4. RESULTS
4.1. Experimental Data
The mean values for the metrics of each experiment are presented in Table 3. Standard
deviation of tensile strength and modulus measurements are provided in parentheses.
Table 3. Mean Experimental Data
Experiment
YXT
YXF
YZT
YZF
XYT
XYF
XZT
XZF
Mass
(g)
10.1
10.1
10.5
10.1
10.0
10.0
10.8
10.4
Thickness
(mm)
3.23
3.23
3.59
3.43
3.25
3.22
3.66
3.56
Width
(mm)
13.4
13.4
13.0
13.0
13.4
13.3
13.0
13.0
Tensile Strength
(MPa)
32.3 (0.56)
26.0 (1.06)
31.2 (0.44)
24.2 (2.62)
35.3 (1.43)
29.0 (1.50)
37.8 (0.47)
22.9 (0.43)
Tensile Modulus
(MPa)
1501 (32)
1577 (251)
1696 (240)
1176 (232)
1665 (73)
1719 (217)
1874 (89)
1284 (189)
Experiments fabricated with XZ and YZ orientations had thicknesses consistently in the
3.42-3.66 mm range compared to the D638 specification of 3.23 mm. A similar effect is also
shown in that experiments fabricated with XY and YX orientations had widths consistently in the
13.3-13.4 mm range compared to the D638 specification of 13.0 mm. This suggests that Zorientation can affect the accuracy of a part's thicknesses and widths.
883
The highest mean tensile strength was found in samples in the XZT experiment; the lowest
occurred in the XZF experiment. The highest relative precision occurred in XZT with one
standard deviation of 0.47 MPa (1.24 % its mean tensile strength) and the lowest occurred in
YZF at 2.62 MPa (10.8 % of its mean). In comparison to Objet's quoted Verowhite tensile
strength of 49.8 MPa [22], XZT had the lowest discrepancy at 12.0 MPa (24.1% of the quote)
and XZF had the highest discrepancy at 26.9 MPa (54.0% of the quote). Figure 7 graphically
summarizes the tensile strength data with one standard deviation and shows that experiments
with the tight spacing consistently show higher mean tensile strength values than at far spacing.
Figure 7. Mean tensile strength
Figure 8. Mean tensile modulus
The highest tensile modulus occurred in the XZT experiment at a mean of 1874 MPa; the
lowest occurred in YZF at a mean of 1176 MPa. The highest relative precision occurred in YXT
with a sample standard deviation of 32 MPa (2.16 % of its mean tensile strength); the lowest
occurred in YZF at 232 MPa (19.7 % of its mean). In comparison to Objet's quoted Verowhite
Tensile Modulus of 2495 MPa [22], XZT again had the lowest discrepancy at 621 MPa (24.9%
of the quote) and YZF had the highest discrepancy at 1319 MPa (52.9% of the quote). Figure 8
graphically summarizes the tensile modulus data. It is observed that all experiments with XZ or
YZ orientation and tight part spacing showed higher mean tensile modulus values regardless of
the orientation. A similar effect is shown in that the lowest mean tensile modulus occurred with
specimen widths aligned with Z-axis with far part spacing regardless of In-Build Plane
orientation.
4.2. Statistical Analysis
Using the experimental data, a factorial analysis was performed to determine the effects of a
parameter and its interactions on tensile strength and modulus (Tables 4 and 5).
884
Table 5. Tensile Strength Factorial Analysis
Parameter
A
(X-Y Orientation)
Parameter
Level
Average
Response
Effect
2nd ORDER INTERACTIONS
MAIN EFFECTS
B
(Z-Orientation)
C
(Part Spacing)
A*B
A*C
3rd ORDER
B*C
A*B*C
YX
(-1)
XY
(+1)
YX or
XY
(-1)
YZ or
XZ
(+1)
Tight
(-1)
Far
(+1)
-1
1
-1
1
-1
1
-1
1
28.4
31.2
30.6
29.0
34.1
25.5
29.9
29.7
30.8
28.8
31.0
28.7
30.8
28.8
2.9
-1.7
-8.6
-0.2
-2.0
-2.3
-2.0
Table 6. Tensile Modulus Factorial Analysis
Parameter
MAIN EFFECTS
A
B
(X-Y Orientation) (Z-Orientation)
Parameter
Level
Average
Response
Effect
2nd ORDER INTERACTIONS
C
(Part Spacing)
A*B
A*C
3rd ORDER
B*C
A*B*C
YX
(-1)
XY
(+1)
YX or
XY
(-1)
YZ or
XZ
(+1)
Tight
(-1)
Far
(+1)
-1
1
-1
1
-1
1
-1
1
1488
1636
1616
1508
1684
1439
1564
1559
1573
1550
1717
1407
1568
1556
148
-108
--245
-5
-23
-310
-12
The factorial analysis was then used to produce normal probability plots, Figures 9 and 10.
These plots are a statistical technique for graphically estimating which main and interaction
effects have statistical significance. Assuming the validity of the Central Limit Theorem [23],
measured effects from a consistent process will appear to be from a normal distribution
therefore, a potentially statistically significant effect would be an outlier.
Normal Probability Plot
0.95
Normal Probability Plot
0.95
X-Y Orientation
X-Y Orientation
0.90
0.90
0.75
0.75
Probability
Probability
Z Orientation
0.50
0.50
Z Orientation
0.25
0.25
0.10
0.10
Part Spacing
Part Spacing
0.05
0.05
-8
-6
-4
-2
0
Main and Interaction Effects Data
2
Figure 9. Tensile strength normal probability plot
-300
-250
-200
-150
-100
-50
0
Main Effects and Interaction Data
50
100
150
Figure 10. Tensile modulus normal probability plot
For the tensile strength, the factorial analysis revealed that the largest absolute effect was the
part spacing, which could vary the response by 8.6 MPa. Figure 9 shows that a majority of the
effects and interactions appear to be from the normal distribution but, the Part Spacing may be
the only statistically significant effect on the Tensile Strength. For the tensile modulus, the
interaction between the Z-orientation and the part spacing produced the largest absolute effect at
310 MPa. The normal probability plot shows that all effects and interactions appeared to be from
the normal distribution suggesting that none of them may have statistically significant effects on
885
the tensile modulus. While the part spacing was the largest absolute main effect for both
responses, the Z-orientation was the smallest absolute main effect for both.
4.3. Analysis of Variance (ANOVA)
Given that the Z-orientation parameter had the smallest absolute main effect, and appeared to
be from the normal distribution for both the tensile strength and modulus (Figures 9 and 10), it
was assumed to be statistically insignificant. This assumption permits the treatment of the
performed three parameter and two level full-factorial design as a pseudo-replicated two
parameter and two level full-factorial design (i.e., YXT to YZT and YZF to YXF were
considered replicated experiments) [23]. By using this pseudo-replication, Analysis of Variance
(ANOVA) could be performed on the two remaining effects (X-Y orientation and Part Spacing)
and their interactions. ANOVA is a statistical technique in which the sum of squares and error
terms are compared against an F-ratio statistic along with p-values to determine the statistical
significance of the parameters.
Using SAS JMP, a statistical analysis software, experimental data was used in a "Fit Model"
to define a two-level and two parameter full factorial design to perform an ANOVA. The DOE
was modeled using a "Standard Least Squares" behavior. Each effect and interaction had its Fratio calculated and compared against the critical F-ratio calculated at an alpha-value of 0.05.
The ANOVA for Tensile Strength and Modulus is shown in Figure 11 and 12 below.
Figure 11. Tensile Strength ANOVA
Figure 12. Tensile Strength ANOVA
886
4.4. Interpreting ANOVA
The ANOVA analysis of tensile strength provided an R-squared value of 0.88 and a RMSE
of 2.45%, indicating an adequate correlation between the experimental data and the model
suggesting the model is sufficient for prediction, due to minimal residuals and random noise. The
factorial analysis in Figure 11's "Parameter Estimates" shows that the part spacing's was the
largest absolute effect at 8.63 MPa, (which is approximately equal to the value from earlier
analysis (Table 5). Its F-ratio of 24.8 is the only value that exceeds the critical F-ratio of 9.60 and
its 0.0076 p-value is nearly an order of magnitude smaller than the 0.05 alpha value. Based on
these two criteria, part spacing is the only statistically significant effect on the tensile strength.
The tensile modulus ANOVA has an R-squared value of 0.43 but a RMSE of 232% which
indicates that large residuals and a very large amount of noise exists in the data, suggesting that
the model is not very adequate to predict the data. Figure 12's "Parameter Estimates" show that
the part spacing had the largest absolute effect at 122 MPa, which is not approximately equal to
the value found in earlier factorial analysis (Table 6). This discrepancy is due to the interaction
effect between the part spacing and Z-orientation, which is confounded with the part spacing in
this ANOVA. While part spacing's F-ratio of 2.22 does exceed the critical F-ratio of 1.02, further
analysis reveals that its 0.210 p-value is greater than the alpha value. Despite having a large Fratio, part spacing and other effects are not statistically significant due to p-values greater than
the alpha value, indicating a higher risk of Type I error.
5. DISCUSSION
5.1. Analysis of Results
Synthesizing the experimental data presented in Section 4, the following results are
identified:
 Part dimensions aligned in the X-Y plane were fabricated 200-400 m larger than
designed (Table 3). This suggests that Z-orientation can affect the accuracy of a part's
thicknesses and widths. This result is not surprising given that the Z-resolution of the
PJD-3DP process (32-60 mm layer thickness) is better than the X-Y printing resolution
(600 dpi). This is result is unique to the PJD-3DP process as most AM processes have
poor dimensional accuracy for features aligned along the Z-axis due to the discretized
nature of the layer-by-layer approach.
 XY and YX parts did not show any statistically significant effects on material
performance. The authors’ hypothesis that YX oriented parts would be weaker due to
jetting from discretized nozzles was found to be incorrect. XY and XZ parts were, on
average, stronger; however, the results were not statistically significant.
 Parts oriented in the Z-plane did not show any statistically significant improvements in
material performance. The authors’ hypothesis that parts with XZ and YZ orientation
would be stronger due to an increase in the number of layers (and thus lead to increased
curing due to “print-through”, as noted for SL in [17]) was incorrect. Specimens with
widths along the Z-axis were, on average, stronger; however the results were not
statistically significant.
887

Part spacing in the X-Y plane showed statistically significant effects on material
performance. As hypothesized by the authors, parts printed closer together in the X-Y
plane were stronger than parts printed further apart. This increase in material properties
is hypothesized to be related to the manner in which the PJD-3DP indiscriminately
patterns UV light during processing. Thus, when printing multiple parts that span
multiple print paths, UV irradiation from the printing-block can over-cure parts in paths
adjacent to the current printing path.
5.2. Sources of Experimental Error
The weakness of DOE is that there are no immediately replicated results to verify and
validate the completed DOE. While ANOVA helps eliminate unnecessary variables, its success
is based on the quality and precision of the collected data, in which replication would provide the
best insurance for determining the consistency of the data. Full-factorial DOE's provide the
highest quality and resolution in an experiment but considering each combination of parameter
level can be time-consuming and lead to excessive use of resources. While a pseudo-replication
can enable ANOVA for two parameters (Section 4.4), this assumption removes a parameter
which provides less precision and insight as to how the original three parameters truly interacted
together. For example, while factorial analysis did not suggest that Part Spacing and Zorientation had statistical significance, they had a practical significance in that their interaction
was the largest absolute effect on the tensile modulus.
In addition to error in statistical techniques, the lifetime of parts (i.e., time between print start
and tensile test) was not consistent between experiments. This is important because time is
suspected to be a potential cause of polymer aging and degradation [21]. In addition, differences
of relative humidity between experiments and testing were not explicitly controlled; humidity
can affect mechanical properties since polymers are hygroscopic and could have softened
surfaces due to ambient moisture [24]. Finally, it is noted that a potential source of
differentiation between the measured values with that specified by Objet could be due to their
use of ASTM D638-03 and D638-04 [25], which might have different test specimens and
conditions for testing and/or the authors’ use of “Digital Material” mode instead of “High
Quality” mode during sample fabrication.
6. CLOSURE
The purpose of this study was to analyze variability in the mechanical properties of parts
created by the PolyJet Direct 3D Printing process due to changes in process parameters. The
authors employed a design of experiments to identify changes in part tensile strength and tensile
modulus due to changes in X-Y orientation, Z-orientation, and part spacing in the X-Y plane.
ANOVA revealed that part spacing had a statistically significant effect on mechanical properties;
specifically, parts printed closer together were stronger than those printed farther apart.
Future studies will further investigate the effect of part spacing on mechanical properties.
The authors will explore the hypothesis that the PJD-3DP process is over-curing parts across
multiple print paths. An AM process wherein part quality is affected by its layout on the build
platform could have significant impact on a designer’s ability to predict part performance – in
888
effect, part properties would be inconsistent across various build platforms (due to part count
and/or size).
7. ACKNOWLEDGEMENTS
The authors acknowledge Daniel Dressner (Virginia Tech ME, ’12) for his preliminary
investigation in effects of build orientation on mechanical properties in the PJD-3DP process.
The authors also acknowledge Mac McCord of Virginia Tech Material's Testing Laboratory for
his assistance with tensile testing.
8. REFERENCES
[1] "Objet Polyjet Process." Objet Geometries Ltd. Www.Objet.com, n.d. Web.
<http://www.objet.com/products/polyjet_technology/>.
[2] Agarwala, M. K., V. R. Jamalabad, N. A. Langrana, A. Safari, P. J. Whalen and S. C.
Danforth, 1996, "Structural Quality of Parts Processed by Fused Deposition," Rapid
Prototyping Journal, Vol. 2, No. 4, p. 4-19.
[3] Ahn, S-H, M. Montero, D. Odell, S. Roundy and P. K. Wright, 2002, “Anisotropic material
properties of fused deposition modeling ABS,” Rapid Prototyping Journal, Vol. 8, No. 4,
p. 248 – 257.
[4] Gibson, I. and D. Shi, “Material properties and fabrication parameters in selective laser
sintering process,” Rapid Prototyping Journal, Vol. 3, No. 4, p. 129–136.
[5] Ajoku, U., N. Salesh, R. J. M. Hague, P. Erasenthiran, 2006, “Investing mechanical
anisotropy and end-of-vector effect in laser-sintered nylon parts,” Journal of Engineering
Manufacture, Vol. 220, No. 7, p. 1077–1086.
[6] Pilipovic, A., P. Raos and M. Sercer, 2009, “Experimental analysis of properties of materials
for rapid prototyping,” International Journal of Advanced Manufacturing Technology,
Vol. 40, p. 105-115.
[7] Udroiu, R. and L. A. Mihail, 2009, “Experimental determination of surface roughness of
parts obtained by rapid prototyping,” CSECS'09 Proceedings of the 8th WSEAS
International Conference on Circuits, Systems, Electronics, Control & Signal Processing,
p. 283-286
[8] Brajlih, T., I. Drstvensek, M. Kovacic, J. Balic, 2006, “Optimizing scale factors of the
PolyJet rapid prototyping procedure by genetic programming,” Journal of Achievements
in Materials and Manufacturing Engineering, Vol. 16, No. 1-2, p. 101-106.
[9] Kesy, A. and J. Kotlinksi, 2010, “Mechanical properties of parts produced by using polymer
jetting technology,” Archives of Civil and Mechanical Engineering, Vol. X, No. 3, p. 3750.
[10] P. F. Jacobs, 1996, Stereolithography and other RP&M Technologies: from Rapid
Prototyping to Rapid Tooling, ASME Press, New York, NY.
[11] Lu, L., J. Y. H. Fuh, A. Y. C. Nee, E. T. Kang, T. Miyazawa, and C. M. Cheah, 1995,
“Origin of shrinkage, distortion, and fracture of photopolymerized material,” Materials
Research Bulletin, Vol. 30, No. 12, p. 1561-1569.
889
[12] Karalekas, D. and D. Rapti, 2002, “Investigation of the processing dependence of SL
solidification residual stresses,” Rapid Prototyping Journal, Vol. 8, No. 4, p. 243-247.
[13] Gibson, I., D. W. Rosen, and B. Stucker, 2010, Additive Manufacturing Technologies:
Rapid Prototyping to Direct Digital Manufacturing, Springer, New York, NY.
[14] Chockalingam, K. and N. Jawar, 2006, “Influence of layer thickness on mechanical
properties in stereolithography,” Rapid Prototyping Journal, Vol. 12, No. 2, p. 106-113.
[15] Hague, R., S. Mansour, N. Saleh, and R. Harris, 2004, “Materials analysis of
stereolithography resins for use in rapid manufacturing,” Journal of Materials Science,
Vol. 39, p. 2457-2464.
[16] Dulieu-Barton, J. M. and M. C. Fulton, 2000, “Mechanical properties of a typical
stereolithography resin,” Strain, Vol. 36, No. 2, p. 81-87.
[17] Quintana, R., J-W Choi, K. Puebla, and R. Wicker, 2009, “Effects of build orientation on
tensile strength for stereolithography-manufactured ASTM D-638 type I specimens,”
International Journal of Advanced Manufacturing Technology, Vol. 46, No. 1-4, p. 201215.
[18] Williams, C. B., F. Mistree, and D. W. Rosen, 2011, “A functional classification framework
for the conceptual design of additive manufacturing technologies,” Journal of
Mechanical Design, Vol. 133, p. 121002-1 – 121002-11.
[19] "Connex Family: The Polyjet Matrix System Operator Training Guide," Objet Geometries
Ltd., 2009. Print.
[20] Bhote, Keki R., and Adi K. . Bhote. World Class Quality. New York: Amacom, 2000. Print.
[21] Gibson, I., G. Goenka, R. Narasimhan, and N. Bhat, 2010, “Design Rules for Additive
Manufacture,” International Solid Free Form Fabrication Symposium, Austin, TX.
[22] "Fullcure Materials." www.objet.com. Objet Geometries Ltd., Web. 5 Jan. 2011.
<http://www.objet.com/portals/0/docs2/FullCure_Letter_low.pdf>.
[23] Vining, G. G. and S. M. Kowalski, 2011, Statistical Methods for Engineers, Cengage
Learning, Boston, MA.
[24] Callister, William D. Materials Science and Engineering: An Introduction. Hoboken, NJ:
John Wiley & Sons, 2006. Print.
[25] “Digitial Material Data Sheets," Objet Geometries Ltd., Web. 2012.
<http://www.objet.com/Portals/0/docs2/Digital%20Materials_Datasheet_A4.pdf>
890
Fly UP