JOM, Vol. 67, No.

5, 2015
DOI: 10.1007/s11837-015-1367-y
Ó 2015 The Minerals, Metals & Materials Society

Mechanical Property Optimization of FDM PLA in Shear

with Multiple Objectives

JONATHAN TORRES,1,2 JOSÉ COTELO,1 JUSTIN KARL,1 and

ALI P. GORDON1

1.—Department of Mechanical and Aerospace Engineering, University of Central Florida,

Orlando, FL 32816, USA. 2.—e-mail: jonathantorres@knights.ucf.edu

This study presents the influences of key processing parameters on the

resulting material properties of fused-deposition-modeled (FDM) polylactic
acid (PLA) components tested in torsion. A reduced experimental matrix was
produced through the use of a Taguchi L9 orthogonal array with three pa-
rameters at three levels each. The processing parameters included the layer
thickness, infill density, and postprocessing heat-treatment time at 100°C.
Testing of components at varying times is conducted to facilitate heat-treat-
ment time testing range and show the effects of prolonged heating. The layer
thickness and infill are tested across the entire useful range available for the
FDM machine used. Shear stress–strain response curves are acquired and
average ultimate shear strength, 0.2% yield strength, proportional limit,
shear modulus, and fracture strain are calculated for each run. An analysis of
results via regression analysis is used to determine influences levels of pa-
rameters of the mechanical properties. The layer thickness and infill density
are shown to be of high importance when optimizing for strength, with heat-
treatment implementation slightly improving the resulting properties. Duc-
tility is mainly affected by infill and heat treatment, with layer thickness
having only a slight effect on the fracture strain achieved. Recommendations
are made based on results of a method to optimize for either strength or
ductility and how to compromise between recommended settings when a
balance between the two is desired. The ability to produce parts with me-
chanical properties at or near those of bulk PLA is shown.

expiration of key patents in 2009 and economies of

INTRODUCTION
scale in component supply chains have driven
Proliferation of rapid prototyping (RP) and rapid widescale adoption by hobbyists, students, and re-
manufacturing technologies has steadily progressed in search and development laboratories in both aca-
the past three decades, moving from the first com- demia and commercial settings.
mercially-available stereolithography (SLA) hardware With high levels of utilization in a number of
in the late 1980s to the desktop-sized fused-deposition industries, need arises for proper mechanical
modeling (FDM) systems of today. The aforementioned evaluation of the resulting printed parts. The
types of systems along with several others are all most common source materials for commercially-
variances on the concept of additive manufacturing. In available FDM printers are acrylonitrile butadiene
this family of methods, three-dimensional objects are styrene (ABS) and polylactic acid (PLA). While
created by building layer upon layer via material de- each of these thermoplastics has well-character-
livery along two-dimensional tool paths. ized behavior, the ultimate behavior of a macro-
Most recently, a number of factors have led to scale printed component is highly dependent on
exponential growth in the use of such systems, with the complex patterned layer structure established
the largest gains coming from the lower-cost ther- during the printing process. The characteristics of
moplastic extrusion FDM consumer market. An bulk PLA1–3 are available in Table I.

(Published online March 17, 2015) 1183

1184 Torres, Cotelo, Karl, and Gordon

mise between properties which can require con-

Table I. Material properties of bulk polylactic acid flicting process settings to suggest an overall
as found via literature review1–3 optimized solution to promote all properties. In the
case of this work, DoE was applied to determine an
Material property Units Value experimental matrix of factors that include printer
Density (q) g/cm 3
1.24 settings and postprocessing levels. Analysis of re-
Elastic modulus (E) MPa 3500 sults via the ANOVA method is used to determine
Shear modulus (G) MPa 1287 the factor impact on 0.2% yield strength, ultimate
Poisson’s ratio (m) – 0.36 shear strength, proportional limit, fracture strain,
Yield strength (ry) MPa 70 and shear modulus.
Ultimate tensile strength (Sut) MPa 73
Elongation % 7 EXPERIMENTAL
Background Topics and Method Justification
An examination of the dependencies on control-
lable printer parameters can lead to optimized The motivation for this study is twofold. First, as
components from each of the source materials. Print additive manufacturing with PLA continues to ma-
orientation and infill density alone have been shown ture and grow in popularity, increasing the knowl-
to alter compressive, tensile, flexural, and impact edge base of these topics both independently and in
strengths by an order of magnitude in some cas- tangent becomes ever more important. This re-
es.4–6 While significant work of this nature has been search strives to aid in advancing both the
performed with ABS, only recently has PLA been manufacturing processes and the materials in-
the subject of such a study. Industrial and general volved by providing data and methodologies for RP
use of PLA is increasingly more popular due to its users and manufacturers with specific objectives.
environmental friendliness and biocompatibility, Second, this effort reinforces and expands the va-
and many desktop consumer printer models now lidity of the processes described in our previous
exclusively use PLA. As a consequence, scrutiny study, which provided a method for the optimization
now falls on the user-controllable aspects of the of mechanical properties of FDM PLA in tension
FDM process and the resultant effect on PLA parts. and fracture. An expanded experimental array with
Although some mechanical properties of PLA an increased factor level and more thorough analy-
have been studied more thoroughly, there is little to sis explores the effects of interactions between pa-
no information on property evaluation under shear rameters on material properties and the possibility
loading. Therefore, a direct aim of this specific re- of using these methods for multiobjective optimiza-
search is to provide characterization and printer tion.
parameter optimization information for PLA parts Few testing standards exists for additive
intended to primarily be subjected to torsion loads. manufactured components largely due to variability
Additionally, the effect of a postmanufacturing amongst methods and processes; most standards
temperature treatment is considered. After print- are based on terminology and reporting guidelines.
ing, PLA can be annealed to promote microstruc- The ASTM E143 torsion testing standard is the
tural changes that improve uniformity, relieve most relevant, although this standard is not nearly
stresses, and strengthen the PLA part. By extend- as rigorous in its requirements for samples and
ing the time at elevated temperatures below the testing procedures than those for tensile and frac-
glass transition point, PLA has been shown to in- ture. The relevant standards for fused deposition
crease crystallinity, which leads to superior tensile modeling include ASTM F2792 for general termi-
and flexural strength and stiffness.7,8 Consequently, nology, ASTM F2791 for data reporting purposes,
it may be inferred that the effects of both annealing and ISO/ASTM 52921 for generalized guidelines
temperature and time will have noteworthy effects related to test methodologies and coordinate sys-
on resulting shear properties. tems in component manufacturing. These ISO/
A similar study was previously carried out in Part ASTM guidelines were referenced for both testing
I9 to characterize and optimize the PLA parts loaded and reporting during experimental methodology,
in the tensile sense. As shown by the results of the data analysis, and report preparation. Heat deflec-
previous study, a significant number of parameters tion temperatures were measured as outlined in
can be evaluated without the need for a full factorial ASTM D648.
experiment (FFE) program. A reduced number of
specimens can be tested in conjunction with the FDM Printer Setting Selection
application of the Taguchi design of experiments The samples were manufactured in an XY orien-
(DoE) method and analysis of variance (ANOVA). tation, with the X direction being the long direction
The use of a Taguchi matrix can minimize the parallel to the front of the build plate. The first
number of experiments necessary while creating a sample for each subset of three was centered on the
correlation between process parameters to optimize build plate with the other two samples manufac-
select material properties and attempt to compro- tured for that run placed directly in front and be-
Mechanical Property Optimization of FDM PLA in Shear with Multiple Objectives 1185

strength values. The speed setting of 90 mm/s had

little effect on material properties in the previous
study. The infill direction was shown to have little
effect on the results, although additional complexity
was introduced when producing parts. To alter the
default infill pattern, an external slicing program
and custom script must be used. A square-gridded
pattern that can be angularly reoriented is helpful
in drawing out directional dependencies. The hex-
agonal infill pattern, which is standard to many
extrusion printers including the Replicator 2, does
not lend itself to property reliance on infill orienta-
tion. Thus, this factor was also eliminated. Finally,
the dependency contribution of the perimeter layer
is well established, and the prior work showed the
importance of component strength. Of the typically
Fig. 1. Part location and orientation during build: three samples altered parameters, this leaves only the layer height
printed in the center of the build platform in the XY orientation. and percentage infill to be explored, both of which
have been shown to be of high importance.9,10 Ad-
ditionally, heat treatment as a finishing step for
components produced from PLA is a growing topic of
discussion among desktop FDM users. Preliminary
bench tests using a directed air heat source on
samples for varying periods of time revealed sig-
nificant differences in the tensile and torsional be-
haviors of otherwise identical components.
Intrinsically, postmanufacturing heat treatment of
FDM PLA components becomes as important as
some fundamental printer settings. As such, the
Fig. 2. Torsion specimen design and dimensions. effect of altering the duration of such a process will
be examined.

Heat Treatment of PLA

hind it with a minimal gap in between, as shown in
Fig. 1. Due to the sample geometry, the need to test Previous literature emphasizes the influence of
in the YX direction is nullified, as the samples would postprocess annealing on certain mechanical prop-
be identical in this orientation. Testing of samples erties in tensile loading. In simplistic tests, in-
manufactured in the ZX (or ZY) direction was for- creases in tensile modulus and heat deflection
gone as preliminary tests indicated that torsion temperature (HDT) versus neat PLA are noted after
samples will delaminate before deforming when annealing at 100°C.11 At lower temperatures, ten-
manufactured in this orientation. The samples were sile modulus also increases with both temperature
produced in a climate-controlled ambient environ- and anneal times in the 65–80°C and 15–60 min
ment, at 20°C and minimal humidity with a range, respectively. From Ref. 12, there is a clear
Makerbot Replicator 2 (MakerBot Industries LLC, indication that in temperature ranges from 65°C to
Brooklyn, NY). The extruder temperature was set to 80°C rigidity, fracture toughness and HDT in PLA
230°C, the number of perimeter layers set at 2, and samples also increase proportionally with tem-
the linear travel speed set to 90 mm/s for all sam- perature. Toughness and rigidity acquired postan-
ples. These are the manufacturer default levels for neal are attributed to a crystallinity increase with
these settings, as well as the recommended settings an associated increase in secondary bonds, thereby
for general use as shown in our previous paper. A forming a less fluid material with increased stability
schematic of the samples under examination can be and heat resistance at higher temperatures.13 An
seen in Fig. 2. increase in useful ductility range coupled with an
Numerous processing parameters can be explored increase in crystallinity due to recrystallization ef-
in a study such as this, yet previous experimenta- fects is also correlated with the annealing process.
tion9 supports that only some key parameters war- Flexure-related properties are augmented by post-
rant scrutiny when a new study is based on the processing annealing, with trials between 30 and
same combination of production technology and 60 min at annealing temperatures of 80°C, the
material. The extruder nozzle temperature of 230°C flexural modulus, strength, and HDT increase in
corresponds to a material temperature of 180°C; direct proportion to annealing time.14
this setting is optimized for the formation of com- Near or beyond 60 min, no further notable in-
ponents with the least warping and the highest crease in strength or modulus occurs. After this
1186 Torres, Cotelo, Karl, and Gordon

Fig. 3. Stress–strain response of 0.1-mm layer thickness, 20% infill components heated for various times of 0–60 min at 100°C.

point, there is no further increase in crystallinity, noted at annealing times up to 20 min, consequently
yielding a lack of improved mechanical properties. creating an upper limit for annealing time for this
In several trials lasting longer than 1 h, the mod- study.
ulus and strength properties degraded to or fell The stress–strain curves of the 5- and 10-min
below that of the neat PLA.15 This decrease may be trials in Fig. 3 reveal an identical increase in ulti-
tied to the material reorganizing weaker bonds in- mate shear stress, while the 5-min trial also indi-
ternally after specific time intervals of thermal cates an increase in shear strain. This implies that a
treatment allow for enough energy input. period of 10 min may be the start of the increased
Increases in tensile modulus have been reported crystallization and subsequent brittle transition.
at annealing temperatures of both 70°C and 100°C, Beyond 10 min, all of the following annealing times
although maximum modulus values remain in the showed very high stress levels and very low strain
domain of 100°C only.14 Additionally, Harris and levels in the specimens at failure, indicating brittle
Lee14 reported a decrease of fracture strain with behavior. With increased scatter but no discernible
increased annealing. Thus, a relationship between trend beyond, the 20-min annealing length was
fracture strain and annealing time will also be ex- chosen as the upper limit to be representative of this
plored to validate current findings in literature. grouping of high heating times.
Furthermore, it is suggested that a 100°C treatment
of 10 min allows PLA to recrystallize with an opti-
Experimental Parameters and Methodology
mal internal filament arrangement, a by-product of
lower porosity due to induced thermal bonding.16 Based on findings from the previous study, pre-
From these converging sources, a heat-treatment liminary heat-treatment testing, and literature re-
temperature of 100°C was chosen as optimal for views including manufacturer recommendations,
further experimentation. the following conditions and related parameters
A preliminary experiment was carried out to de- were chosen for the study:
termine the heat-treatment times of interest for the
 Print layer thickness varied across the full avail-
study. The torsion samples identical to those em-
able range of the printer, from 0.1-mm layers to
ployed within the main test matrix of this study
0.3-mm layers, in increments of 0.1 mm.
were manufactured with 0.1 mm layer thickness
 Infill relative densities from 20% (lower limit of
and 20% infill, as these components were theorized
useful structural parts) to 100% (solid parts) in
as being the most susceptible to any effects likely to
increments of 40%.
be incurred from heat treatment. Along with a
 Postprint heat treatments at 100°C with anneal
control sample that did not undergo any heat
times of 0 min, 5 min, and 20 min to coincide with
treatment, the parts were heated for 5 min, 10 min,
zero, low, and high recrystallization levels.
20 min, 30 min, 40 min, 50 min, and 60 min at
100°C, and then they were tested in torsion at a rate These settings are summarized in Table II. With
of 0.1 rpm. Strong time-dependent effects were three process parameters each carrying three levels
Mechanical Property Optimization of FDM PLA in Shear with Multiple Objectives 1187

Table II. Ranges of process parameters employed in this study and the corresponding low, middle, and high
settings for each

Process parameter Range Low Middle High

Layer thickness (d) 0.1–0.3 mm 0.1 0.2 0.3
Percent infill (q) 0–100% 20 60 100
Heat treatment (HT) 0–20 min 0 5 20

MECHANICAL EXPERIMENTATION
Table III. Run definition via implementation of an RESULTS
L9 Taguchi orthogonal array
A mechanical response was characterized in the
L9 array Process parameters form of shear stress versus shear strain. Shear
stress s is defined as
Run d (mm) Q (%) HT (min)
TD
1 0.1 20 0 s¼ (1)
2 0.1 60 5 2J
3 0.1 100 20 where J, the polar moment of inertia, is calculated
4 0.2 20 5 as
5 0.2 60 20
6 0.2 100 0 pD4
7 0.3 20 20 J¼ (2)
8 0.3 60 0 32
9 0.3 100 5 and shear strain c was defined as
hD
of interest, a Taguchi L9 orthogonal array was c¼ (3)
2L
constructed to test the impact of layer thickness,
infill, and heat treatment on a variety of material where T is the torque as measured by the torsion
properties. Three identical repetitions for each ex- load cell, D is the diameter at the gage section, L is
periment lead to a total of 27 experiments, which is the gage length, and h is the angular displacement.
reduced versus a full factorial set that would require A typical response curve for each of the trials is
33 9 3, or 81, experiments. The Taguchi array de- shown in Fig. 5 for a direct comparison. Although
signed for this experiment is shown in Table III. some grouping is evident, it is apparent that the
Heat treatment of samples was conducted using a various combinations of processing properties asso-
Barnstead Thermolyne F48055 (Hogentogler & Co. ciated with the different tests yield highly diverse
Inc., Columbia, MD) resistance furnace. The fur- responses.
nace was preheated to 100°C and heat soaked for To properly quantify these differences, the me-
30 min prior to specimen insertion. A Eurotherm chanical shear properties of the specimens were
2416 temperature controller (Schneider Electric, calculated and compared. The properties evaluated
Ashburn, VA) provided furnace control and readout include strength indicators such as 0.2% offset
via feedback from a type K thermocouple. The spe- shear yield strength sy and ultimate shear strength
cimens were treated in batches of three with their Sus. The proportional limit spl is evaluated to signify
testing counterparts and removed promptly from the primary transition from elastic to plastic be-
the furnace at the end of the treatment. Cooling in havior, and shear modulus G quantifies the mate-
the ambient air environment occurred before the rials response to shear stress and is defined as ratio
length and diameters of the gage sections of each of shear stress to shear strain during elastic re-
sample were measured and recorded. sponse. Ductility, the ability of the part to deform
After preparation, the samples were secured in before failure, is quantified by the strain at fracture
the test device, an MTS Bionix Electromechanical cf. The average values for each of these properties
Torsion tester (model no. 100-224-094; MTS Sys- were calculated for each trial and quantified in
tems Corporation, Eden Prairie, MN) with a 50-Nm Fig. 6. The graph containing G and cf uses a nor-
capacity. Chuck-style grips were hand tightened to malized average value for these properties, calcu-
avoid compressing sample ends while providing a lated by dividing the averages by the maximum
grip in which no slippage could occur, shown in sampled value, that is, the highest value found
Fig. 4. The samples were fixed at one end and then among the 27 samples tested. The maximum values
torqued at the other at a slow rate of 0.1 rpm to found were 1312.5 MPa and 3.4454 mm/mm for the
insure that no strain-dependent effects were intro- shear modulus and fracture strain, respectively.
duced. Data in the form of a torque-twist curve were The metrics that quantify strength all follow an
obtained at a rate of 100 Hz. identifiable pattern. Comparing the graphs of Fig. 6
1188 Torres, Cotelo, Karl, and Gordon

Fig. 4. Torsion specimen during testing in an MTS Bionix Electromechanical Torsion Tester.

Fig. 5. Typical stress–strain response curves characteristic of each run.

versus the test matrix, it is noted that strength rises smaller strands clustered together results in higher
with increasing infill percentages. This finding is resistance to deformation.
logical, as components with higher levels of infill Torsional rigidity response in Fig. 6b has a simi-
physically contain more material to resist deforma- lar pattern to the strength metrics, although the
tion. In a more localized view, a pattern is evident in shear modulus is dependent on the parameter val-
trials 1–3, 4–6, and 7–9 in Fig. 6a. This correlates ues to a lesser degree. The fracture strain does not
inversely with a change in layer thickness, as a follow a likewise trend. The three trials with the
decrease in strength is seen with increasing layer highest average fracture strain in descending order,
thickness. This is hypothesized to be related to runs 9, 6, and 2, also have high infill levels of 100%,
thinner layers containing smaller gaps between the 100%, and 60%, with seemingly little dependency on
extruded strands and thus having overall lower layer thickness. Trial 3, however, which also uses
porosity, or it could be that a greater number of 100% infill and yields the highest strength values,
Mechanical Property Optimization of FDM PLA in Shear with Multiple Objectives 1189

Fig. 6. Comparison of the average shear properties across all runs showing (a) ultimate shear strength Sus, 0.2% shear yield strength sy,
proportional limit spl; and (b) the normalized values of shear modulus G and fracture strain cf relative to their respective maximums.

displays very low average fracture strain when with little to no heat treatment show less scatter in
compared with these others. The difference therein the height, shape, and length of the curves, as seen
lies in the associated heat treatments, with trials 2, in Fig. 7a and b. The 5-min annealing trials show
6, and 9 having low heat-treatment times of either 0 the least scattering of results, with all but one curve
or 5 min, whereas trial 3 used the extended 20-min conforming to the same qualitative profile for each
heat-treatment time. Trials 5 and 7 show poor run, as visible in Fig. 7b. Overall, this indicates an
ductility and were also treated for 20 min, leading increase in variance of material properties for parts
to the assumption that extended heat-treatment heat treated for periods longer than 5 or 10 min,
times result in ductility loss. This is consistent with which is further supported by the results acquired
effects of the recrystallization induced in the PLA by in the heat-treatment time-variation experiment.
heating, which results in a stronger, generally more The microstructures pictured in Fig. 8 compare
brittle material.17 the internal structure of a 0.3-mm layer thickness
An additional inspection of the stress–strain re- and 60% infill component treated at 100°C for
sponse also reveals some detrimental behavior in- 20 min to an untreated sample. Based on these
duced when utilizing a heat-treatment procedure on representative views, it can be inferred that the
FDM PLA parts. As was noted in the Experimental increase in heat-treatment time correlates to an
section, although an increase in annealing time increase in the inclusion of flaws within the PLA.
showed a high increase in strength, there was no Most notably, gas bubbles are formed within the
discernable trend in the increases beyond the 10- heated sample of Fig. 8b, which create sites for
min mark. The most notable feature is an increase crack initiation, which eventually propagates
in the variance of results, which is readily observ- throughout the structure and leads to failure. It is
able between the samples of each trial as shown in conjectured that the random sizes and locations of
Fig. 7c, which compares all specimens that under- these bubbles could be a significant contributor to
went a 20-min heat treatment. By contrast, those the varied properties between samples. An inspec-
1190 Torres, Cotelo, Karl, and Gordon

Fig. 7. Shear stress–strain response curves of all samples tested grouped by heat-treatment times of (a) 0 min, (b) 5 min, and (c) 20 min to
emphasize the variance in material properties induced by extended heating times.
Mechanical Property Optimization of FDM PLA in Shear with Multiple Objectives 1191

Table IV. ANOVA ranking table indicating influence of process parameters on specified material properties
and preferred setting for each

Parameter ranking by influence on property

Material property 1st 2nd 3rd

3 1
0.2% shear yield strength (sy) Infill Thickness Heat treatment3
Ultimate shear strength (Sus) Infill3 Thickness1 Heat treatment3
Fracture strain (cf) Infill3 Heat treatment1, 2
Thickness1
Shear modulus (G) Infill3 Thickness1 Heat treatment3
Proportional limit (spl) Infill3 Thickness1 Heat treatment3
Within the table, 1 = low setting, 2 = middle setting, and 3 = high setting.

tion of Fig. 8b also reveals some shrinkage of the closely matched by the layer thickness with only a
individual strands and shows that the outer 7% difference in influence between the two pa-
perimeter of each strand has changed in color and rameters for both properties. Heat treatment rates
opacity—a visual indicator of heating-induced re- at an average of 10% influence for these properties,
crystallization of the PLA. indicating that elasticity is nearly wholly and
equally dependent on both percentage infill and
ANOVA layer thickness. Conversely, the fracture strain, and
consequently the ductility, is largely affected by the
A further examination of experimental mechanics
heat treatment, and the thickness has only a slight
results was conducted via regression analysis and
effect at 8% influence. The nearly equal reversal of
ANOVA, which evaluates the variance between
the influence levels between the thickness and heat
different groups to weigh the overall effects of each
treatment for the elasticity and ductility metrics
process parameter on performance. The regression
suggests a conflict between the parameter settings
analysis was run using the program Quantum XL
necessary to optimize for these characteristics. The
(SigmaZone, Orlando, FL). Pareto plots of these re-
shear modulus is the least disparately affected by
sults in Fig. 9 indicate the level of the effect on each
the parameters, as it is a measure related to both
material property. The values assigned to the pa-
strength and ductility.
rameters for each property on the Pareto plots have
To make a recommendation for the optimization
been normalized by the average of all of the pa-
of specific material properties, the Pareto plots and
rameters’ scores for the property of interest so as to
ANOVA results were aggregated to create an AN-
align them to a common scale. These have been
OVA ranking table9 (Table IV). The ANOVA rank-
catalogued in two distinct groupings: by material
ing table summarizes the influence ranking of each
property and by processing parameter. The pa-
parameter as well as the setting favored to optimize
rameter-based grouping shows that infill
for the property of interest. Each process parameter
unequivocally has the greatest impact on all resul-
is ranked from first to third in descending order of
tant material properties. Contrastingly, heat treat-
importance, and is marked with a value of 1 indi-
ment has only a slight effect on the metrics purely
cating the low setting, 2 the middle setting, and 3
associated with strength when compared with those
the high setting as defined in Table II. The combi-
metrics that give an indication of ductility. The in-
nation of 100% infill in combination with the low
fluence of layer thickness, although significant,
thickness setting of 0.1 mm is favored to achieve
rates lower than that of infill, and it has only a
maximums for all material property values. How-
slight effect on the fracture strain. The groupings of
ever, whereas all of the strength metrics improved
thickness and heat treatment show an almost in-
when using the high heat-treatment setting of
verse effect for all but the shear modulus, which has
20 min, the fracture strain is highest when heat
an intermediate rating in both groupings. These
treatment is forgone or minimized. This means that
associations are corroborated by the material prop-
strength and ductility, as can often be the case, re-
erty grouping, which shows the percentage contri-
sult in conflicting settings when it comes to opti-
bution of each parameter on each property. The
mizing for either one.
ultimate shear strength is heavily dependent on
infill as compared with the other parameters; infill
OPTIMIZATION
makes a 56% contribution, which is nearly twice
that made by the thickness and four times as in- Recommendations can be made for the optimiza-
fluential as the heat treatment. For the proportional tion of mechanical properties based on the ex-
limit and yield strength, the influence of the infill is perimental mechanics results and their subsequent
1192 Torres, Cotelo, Karl, and Gordon

Fig. 8. Microstructure of a component in (a) as-printed condition (0 min HT) and (b) after 20 min of heat treatment. Arrows emphasize the
random growth, placement, and size of inclusions caused by heating.

statistical analysis. As seen in Table IV, regardless

of which mechanical property is being optimized,
the layer thickness d should always be set to the
smallest thickness possible. The infill density q
should always be maximized to maximize strength
and resistance to failure as is shown both here and
in the previous work,9 which focused on tensile and
fracture experiments. Higher infill means more
available material to resist deformation, regardless
of the loading type. The heat treatment, however, is
a conflicting parameter setting when optimizing for
strength or ductility. Although an increase in heat-
treatment time is associated with an increase in
strength, especially for low infill parts, it is also
associated with a decrease in ductility. To optimize
solely for strength, the higher 20-min treatment
time is recommended. Times beyond this point
indicate no discernable differences in performance
and an unfavorable increase in variance. For an
emphasis on ductility, however, little to no heat
treatment is recommended. It is recommended to
favor a heat-treatment time of 5 min for general
use, which results in a favorable balance of strength
and ductility.
With the prediction model created by the regres-
sion analysis, it can be seen that the loss in strength
with a decreased heat-treatment time is sig-
nificantly smaller than the loss in ductility resulting
from increasing heat-treatment time. For example,
the prediction model using 0.1-mm layer height and
100% infill forecasts an 8% increase in Sus when
increasing the heat-treatment time from 5 min to
Fig. 9. Pareto plots depicting relative influence levels and percent-
20 min, but at a 67% reduction in cf. The recom- age contributions of process parameters to resultant material prop-
mendation to employ a short treatment time is erties organized by (a) process parameter and (b) material property.
further substantiated by earlier observations of
Figs. 3, 7, and 8, which showed that an increase in
heat-treatment time leads to greater variance in when changing from no treatment to 20 min of heat
resulting material properties along with an increase treatment.
in the internal flaws of the component microstruc- The highest calculated shear modulus value,
ture. The prediction model further corroborates this which resulted from the average of the run 3 trials
finding, with the expected variance for Sus doubling characterized by 0.1 mm layer thickness, 100% in-
Mechanical Property Optimization of FDM PLA in Shear with Multiple Objectives 1193

fill, and 20 min heat-treatment time, was shown that bulk PLA properties can be potentially
1265 MPa, about 98% of the bulk value reported achieved in FDM components using proper pro-
earlier of 1287 MPa. Using the von Mises yield cri- cessing parameters and print orientation, though
terion to equate the shear and tensile strength further research should be conducted for substan-
strengths as tiation.
Sus ¼ 0:57Sut (4) ACKNOWLEDGEMENTS
shows similar results. The average values of Sus The participation of José Cotelo is made possible
achieved for the runs, which used 100% infill, runs via the support of the Career Advancement Men-
3, 6, and 9, were 53.35 MPa, 39.54 MPa, and toring Program for Young Entrepreneur and
39.42 MPa, respectively. These values are analo- Scholars (CAMP-YES), a National Science Founda-
gous to the shear strength of 42.15 MPa calculated tion funded program at the University of Central
using the cited Sut value of 73 MPa with the von Florida. Fabrication of test coupons was made pos-
Mises yield criterion. It can therefore be said that sible due to the cooperation of the Center for Mi-
FDM components can closely match the behavior of crogravity Research and Education at the
those produced from bulk if the method suggested is University of Central Florida.
used to properly optimize for the desired properties.
REFERENCES
CONCLUSION
1. M. Bijarimi, S. Ahmad, and R. Rasid, Int. Conf. Agric. Chem.
FDM PLA components were tested in torsion, and Environ. Sci. 1, 115 (2012).
the effects of processing parameters including layer 2. A.M. Clarinval and J. Halleux, Biodegradable Polymers for
thickness, percent infill, and post-processing via Industrial Applications, ed. R. Smith (Boca Raton: Taylor &
Francis, 2005), pp. 3–31.
heat treatment on the resultant shear properties 3. M. Jamshidian, E.A. Tehrany, M. Imran, M. Jacquot, and S.
were evaluated. A Taguchi L9 orthogonal array was Desobry, Comp. Rev. Food Sci. 9, 552 (2010).
defined and actualized with three repetitions for 4. C.S. Lee, S.G. Kim, H.J. Kim, and S.H. Ahn, J. Mater.
each experiment. Process. Tech. 187–188, 627 (2007).
The effects of heat treatment were explored thor- 5. S.K. Panda, S. Padhee, A.K. Sood, and S.S. Mahapatra, In-
tell. Inf. Manage. 1, 89 (2009).
oughly as they relate to material properties and re- 6. J.W. Zhang and A.H. Peng, Adv. Mater. Res. 538–541, 444
liability. Although heat treatment can cause an (2012).
increase in strength, especially in low-infill compo- 7. G. Perego, G.D. Cella, and C. Bastioli, J. Appl. Polym. Sci.
nents such as those tested in the heat-treatment ex- 59, 37 (1996).
8. M. Kowalczyk and E. Piorkowska, J. Appl. Polym. Sci. 124,
periment trials, a large loss in ductility and an 4579 (2012).
increase in the variance of resultant material proper- 9. J. Torres, M. Cole, A. Owji, Z. DeMastry, and A.P. Gordon,
ties accompanies the increase in strength. As such, low Rapid Prototyping J. (in press).
levels of heat treatment were suggested to improve 10. A.K. Sood, R.K. Ohdar, and S.S. Mahapatra, Mater. Des. 31,
strength while preserving ductility and reliability. 287 (2010).
11. J. Wootthikanokkhan, T. Cheachun, N. Sombatsompop, S.
A previously used method for optimization of the Thumsorn, N. Kaabbuathong, N. Wongta, J. Wong-On,
material properties of FDM PLA components was S.I.N. Ayutthaya, and A. Kositchaiyong, J. Appl. Polym. Sci.
validated and expanded upon. An in-depth analysis 129, 215 (2013).
of the contributions of processing parameters on se- 12. Y. Srithep, P. Nealey, and L.-S. Turng, Polym. Eng. Sci. 53,
580 (2013).
lect mechanical properties was given. A method for 13. S.D. Park, M. Todo, K. Arakawa, and M. Koganemaru,
compromising between conflicting properties was Polymer 47, 1357 (2006).
explored, which could extend these methods to other 14. A.M. Harris and E.C. Lee, J. Appl. Polym. Sci. 107, 2246
rapid prototyping technologies and materials to (2008).
establish processing routines that balance conflicting 15. J.H. Lin, H.Y. Chung, K.D. Wu, S.P. Wen, C.T. Lu, and C.W.
Lou, Adv. Mater. Res. 627, 751 (2012).
resultant properties such as strength and ductility. 16. A.P. Mathew, K. Oksman, and M. Sain, J. Appl. Polym. Sci.
Guidelines were suggested for optimizing various 97, 2014 (2005).
properties of components in shear. It has been 17. A. Galeski, Prog. Polym. Sci. 28, 1643 (2003).