Sie sind auf Seite 1von 13

Additive Manufacturing 25 (2019) 532–544

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Full Length Article

Characterization of process–deformation/damage property relationship of T


fused deposition modeling (FDM) 3D-printed specimens
Tomas Webbe Kerekesa, Hyoungjun Lima, Woong Yeol Joeb, Gun Jin Yuna,

a
Department of Mechanical & Aerospace Engineering, Seoul National University, Seoul, Republic of Korea
b
Department of Mechanical & Manufacturing Engineering, Tennessee State University, TN, USA

ARTICLE INFO ABSTRACT

Keywords: In this paper, we investigated the process variable effects on the damage and deformational behavior of fused
Fused deposition modeling deposition modeling (FDM) three-dimensional (3D)-printed specimens by performing tensile tests and inverse
3D printing identification analyses. A characterization of the effects of different parametric variations of 3D-printed speci-
Failure mechanisms mens on fracture properties are a matter of considerable significance that are often overlooked. By combining
Damage model
the infill density and the layer thickness options that are available in the 3D printer machine, six groups with
Process-property relationship
different structural configurations can be obtained. The data and images obtained from experiments are em-
ployed to investigate the failure mechanism of 3D-printed specimens and demonstrate the relationship that exists
between structural variations and fracture mechanical properties. On the basis of experimental results, a Gurson-
type porous plasticity model was used within a 3D continuum finite element model to characterize the pro-
cess–damage parameter relationship through an inverse identification process.

1. Introduction and three-point flexural strength [20]; the printing temperature, LT,
and layer design of PLA on tensile strength and delamination energy
Additive manufacturing (AM) is the process of building three-di- [11]; melting temperature, LT, and raster pattern orientation on tensile
mensional (3D) parts by adding layers of material [1], which offers a strength [10]; LT and orientation on tensile modulus, strength, and
more cost-efficient manufacturing method that enables flexibility in failure modes [21]; temperature, speed, infill direction, relative infill
geometrical complexity, reduces material waste, and has a shorter density, LT, and perimeter on tensile strength and fracture properties
concept-to-market time than the traditional subtractive manufacturing [22]; and part orientation, raster angles, raster width, and air gap on
method [2]. Aerospace manufacturing is the second largest market tensile strength [23]. Interestingly, most studies focused on the uniaxial
where the AM technology is cost-effective because of its low volume, tensile behavior by different manufacturing and experimental test
customization, and high-value production chains [3,4]. One of the methods. It is worth noting that some of the salient features of FDM-
promising applications of AM is a lattice structure with lightweight printed parts are low quality and structural defects (e.g., porosity),
morphing wing skins [5]. which result due to the nature of fused filament-based manufacturing
Considerable research has been conducted on fused deposition [24,25]. This feature indeed hinders the usage of FDM-printed speci-
modeling (FDM) process changes and their effects on mechanical mens for structural applications. Therefore, more research on elim-
properties. Most studies considered FDM process variables such as inating the surface and internal defects is needed to use them as
width, distance between filaments, path of the filament, layer thickness structural materials [26]. However, similar to previous research, we
(LT) [6–9], nozzle temperature, chamber temperature, first layer ad- focus on FDM materials printed with defects and their consequences on
hesion to the printing table, layer design, and material feed rate mechanical properties and failure mechanisms to understand the root
[10–17]. In addition, build direction and orientation are important causes for the failures. Understanding failure mechanisms is critical for
parameters that also influence the part's quality [18]. Because of lim- engineering the design of FDM materials and their applications. The
ited space, we cited some representative works that investigated effects fracture properties of FDM-fabricated parts were also investigated in
of the infill density of acrylonitrile butadiene styrene (ABS) on tensile available literature [21,22,27]. For example, LT and orientation were
strength [19]; build orientation, LT, and feed rate of PLA on the tensile considered for their effects on tensile strength and fractographic


Corresponding author at: Department of Mechanical & Aerospace Engineering, Seoul National University, Gwanak-gu Gwanak-ro 1, Building 301, Room 1308,
Seoul, 08826, Republic of Korea.
E-mail address: gunjin.yun@snu.ac.kr (G.J. Yun).

https://doi.org/10.1016/j.addma.2018.11.008
Received 15 August 2018; Received in revised form 22 October 2018; Accepted 5 November 2018
Available online 15 December 2018
2214-8604/ © 2018 Elsevier B.V. All rights reserved.
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 1. uPrint SE Plus default infill options for (a) LOW, (b) HIGH, and (c) SOLID.

Fig. 2. (a) Dimensions of the dog-bone specimen (all values are in millimeters) and (b) isometric view of the building orientation.

analysis [21]. Riddick et al. employed scanning electron microscopy testing standards for AM technologies [35], applications of diverse
images to use fractographic methods. This helped to investigate the methodologies and model approaches [36], characterization of me-
effect of build direction and orientation on the failure mechanism and chanical properties for FDM [37], experimentation and measurement
mechanical response of ABS FDM-printed specimens [27]. However, technologies [38], and characterization and modeling of FDM parts
Torres et al. did not consider effects of infill density on failure me- [39]. As surveyed, there has been a lack of studies on the mechanical
chanisms but considered various process variables to yield optimized properties, progressive failure mechanism, and predictive material
mechanical properties under tensile- and fracture-type loadings. Their modeling of FDM-printed specimens.
results focused on observations of fractography at the final failure state This paper investigated the relationship of FDM process variables
[22] but not progressive failures. with mechanical and fracture properties of the printed specimens by
Numerical material modeling and tools are required to engineer the uniaxial tensile tests and a predictive modeling method. A pore damage
design of FDM-printed specimens. Research on finite element (FE) plasticity model was applied, with parameters identified by the self-
modeling of 3D FDM-fabricated materials can be found in literature optimizing inverse analysis method (Self-optim) [40] and FE model
[28–31]. Consequently, Bhandari et al. created a space frame lattice updating (FEMU) method. In situ failure mechanism and progressions
and shell FE model to predict the linear elastic response of test coupons both in external and in internal microstructures were captured by three
made up of ULTEM 9095 to provide an efficient design and optimiza- different methods, namely, digital image correlation (DIC), X-ray CT,
tion procedure for 3D-printed parts [28]. The material model they used and microscopy. By analyzing the experimental observations and
was the linear elastic isotropic model. Guessasma et al. presented ani- identifications of damage model parameters, insights into the damage
sotropic damage of FDM-printed polymers under compressive loading mechanisms in relation to FDM process variables are possible. There-
[29] using a 2D FE model with a simple damage kinetics model to fore, the present study helps to expedite the adoption of FDM because of
explain damage mechanisms [29]. X-ray micro computed tomography the valuable knowledge and increased understanding for the FDM parts.
(CT) and FE computations with a linear elastic isotropic material model
were also used to reveal the microstructure and pore connectivity-re-
lated stress heterogeneity [30]. Bellini et al. determined the stiffness 2. Experimental tests
matrix of an ABS part built with the FDM technique combining data
from tensile test and FE analysis [31]. However, in available literature, A single type of dog-bone specimen was used for performing the
there is a lack of investigation on predicting the damage behavior of uniaxial tensile test. For this test, 30 3D-printed specimens were divided
FDM 3D-printed specimens. According to literature reviews, there is into six groups (five samples per group). Each group had a different
considerable research related to FDM-printed specimens. Moreover, configuration of structural characteristics. However, geometrical di-
there are numerous studies related to FDM technologies such as FE si- mensions and material compositions remained the same for all 30
mulations of the FDM process [32,33], improvements in FDM materials specimens.
by using property-altering additives [34], specification of material-

533
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Table 1
Six specimen types with different infill density and layer thickness options.
Infill density Layer thickness (LT)

Configuration 1 LOW 0.254 mm


Configuration 2 LOW 0.330 mm
Configuration 3 HIGH 0.254 mm
Configuration 4 HIGH 0.330 mm
Configuration 5 SOLID 0.254 mm
Configuration 6 SOLID 0.330 mm

Fig. 5. Top view of DIC cameras, load frame, and specimen.

shown in Fig. 2(a). The specimen thickness is 7 mm, and the geometric
dimensions are the same for all 30 specimens.
The other processing parameter is LT, which can be 0.330 mm or
0.254 mm. Therefore, six combinations were considered, where all
Fig. 3. Cross-sectional view of a specimen after the test and schematic drawing specimens were printed using ABS-M30 in the same horizontal building
of sectional arrangement. orientation on the XY plane as depicted in Fig. 2(b). This orientation
allows the tension load application to be parallel to the layers. Thus,
2.1. Processing variables and geometrical features of FDM 3D-printed delamination and fracture in the bonding of the layers are reduced.
specimens Some 3D-printed parts have inherently anisotropic properties, meaning
they have higher strength and stiffness in the XY direction than in the Z
The 3D printing machine used was Stratasys® uPrint SE Plus. This direction, as is the case for specimens used in this investigation.
3D printer used ABS-M30 as the model material and Soluble Release Six groups of specimens with five samples per group were prepared.
Support-30 (SR-30) as the support material. The maximum build size Therefore, 30 specimens were tested. Because of negligible scattering of
was 203 mm x 203 mm x 152 mm (8 in. × 8 in. × 6 in.), and LT was test data, we decided to include five samples per case. Each group had
0.254 mm (0.10 in.) or 0.330 mm (0.13 in.). The ABS filament diameter one combination of processing variables. The printing variable combi-
was 1.75 mm, and the nozzle diameter was 0.4 mm. The default tem- nations are detailed in Table 1.
perature settings were 300 °C for the nozzle and 80 °C for the en- The printing pattern is a default parameter that cannot be changed
vironment (i.e., inside the printer chamber). in the uPrint SE Plus. This default setting results in an infill mesh-like
We considered only two processing variables: infill density and LT. structure made by layer interleaves. The first layer is printed at a 45°
The infill density was LOW, HIGH, or SOLID. Images captured with a angle, and then, the second layer is printed on top of the first one at a
10X augmentation of the three infill options are shown in Fig. 1. A dog- -45° angle. This process is repeated until all the infill layers required to
bone specimen used for the uniaxial tensile test has dimensions as complete the part are printed.

Fig. 4. Schematic representation of printing order, pattern, and nozzle path for the (a) side shell, (b) bottom shell (SOLID), and (c) infill.

534
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 6. (a) Geometry division of the outer shell and infill for the FE model; (b) FE model.

Table 2
Material parameters to be determined.
E Young’s modulus y0 Initial yield stress
C, Kinematic hardening parameters ult Ultimate plastic strain (until hardening)
p
f0 Initial relative porous density q1, q3 Fitting parameters
N Mean value of the normal distribution of the nucleation strain sN Standard deviation of the normal distribution of the nucleation strain
fN Volume fraction of nucleating voids

Fig. 8. DIC matching with the FE model.


Fig. 7. Combined inverse characterization method for the experimental re-
sponse of the 3D-printed material.
The side shell printing process is depicted in Fig. 4(a), where (1),
(2), and (3) single filaments are printed as a continuous line following
The shells are printed parallel to the printing table. Side shells are
the specimen perimeter. In (4), the side shell is not complete, while (5)
composed of two walls and printed according to the model perimeter
shows the first side shell layer completed by printing the second fila-
direction. The top and bottom shells are printed in the same mesh
ment inner and alongside the previous one. These steps are repeated for
pattern as that in the infill. However, the top and bottom shells are
the entire process. There is a side shell layer for every layer of infill and
printed in SOLID density despite the choice of infill density as shown in
top/bottom shells.
Fig. 3.
The bottom shell printing process is depicted in Fig. 4(b), where (1),
A schematic representation of the complete specimen printing pro-
(2), and (3) single filaments are printed as a continuous line following
cess is shown in Fig. 4. In this schematic representation, the printing
the default pattern at a 45° angle. In (4), the bottom shell is not com-
order of the outer shells, as well as the infill, is numbered. The printing
plete, whereas in (5), it is complete. The second filament is printed on
pattern is drawn in the same way as the real printing pattern. However,
top of the previous one with the same pattern but in the opposite di-
the clearance between filaments is drawn much larger for schematic
rection at a -45° angle.
understanding purposes. The printing track direction, i.e., the path in
The infill printing process is also depicted in Fig. 4(c), where (1),
which the nozzle follows while depositing the extruded and melted
(2), and (3) filaments are printed as a continuous line following a de-
ABS, is marked with an arrow.
fault pattern at a 45° angle. In (4), a layer of infill is not complete,

535
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 9. Algorithmic procedures of (a) Self-optim and (b) FEMU.

whereas in (5), it is complete. The second filament is also printed on top complete, the printing process is also complete, and the specimen is
of the previous one with the same pattern but in the opposite direction completely formed.
at a -45° angle. These steps are repeated until the printing reaches the
top shell step. 2.2. Tests and measurements
The top shell is printed in the same fashion as the bottom shell. The
top shell is printed on top of the last infill layer. Once the top shell is To perform the tensile test, a Psylotech® μTS – Meso Scale Universal

536
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 10. Tensile stress–strain curves of each configuration.

Table 3
Average values of mechanical properties from uniaxial tensile tests.
Infill Density Layer Thickness (mm) Ultimate Strength (MPa) Modulus of Toughness (Pa) Young's Modulus (MPa) Initial Yield Stress (MPa) Elongation at Break
(cov) (cov) (cov) (cov) (mm) (cov)

LOW 0.254 23.634 (0.008) 1.561 (0.148) 562.381 (0.018) 24.638 (0.013) 3.030 (0.119)
LOW 0.330 26.602 (0.004) 1.513 (0.104) 618.372 (0.041) 23.102 (0.015) 2.730 (0.075)
HIGH 0.254 26.333 (0.010) 1.509 (0.103) 774.062 (0.030) 24.920 (0.018) 2.600 (0.082)
HIGH 0.330 31.780 (0.008) 1.745 (0.044) 881.637 (0.031) 29.827 (0.006) 2.560 (0.032)
SOLID 0.254 34.740 (0.026) 1.974 (0.144) 1015.602 (0.017) 33.514 (0.013) 2.660 (0.120)
SOLID 0.330 34.718 (0.068) 1.982 (0.121) 1030.968 (0.102) 33.680 (0.061) 2.520 (0.103)

Load Frame – was used, which had 10 kN load capacity and wedge was applied in displacement-controlled steps of 100 μm and reduced to
tension-type grips. Psylotest™ software provided the time, loading, and 50 μm as the fractures progressed. This considerably low displacement
displacement history data. To perform the DIC test on the specimens, a per step allowed the observer to scrutinize the progress of each fracture
Dantec Dynamics® Q-400 System was used with two Allied Vision® MG- meticulously. All fractures were recorded as the tensile loading was
201B ASG cameras and Ricoh® FL-CC2514-2 M lenses to capture the applied because even one fracture could develop into the fracture point
images. All 30 specimens were coated on the face looking up with of the specimen.
Nabakem® LCP-77 Matt Black Lacquer Spray paint to create speckle To observe how tensile loading affected the infill structure damage,
patterns. the X-ray CT test was conducted with pristine and damaged specimens
The equipment was set up in such a way that the cameras have the using Bruker® SkyScan 1278 with a flat panel CMOS sensor X-ray de-
optimal angle and direction of view toward the specimen from the top. tector and a nominal resolution of 52 μm.
The load frame and specimens were positioned horizontally. A picture
of the test setup with DIC is shown in Fig. 5.
In situ tensile tests were performed by applying a uniaxial tension 3. Finite element model for 3D-Printed specimens
loading in the X direction. Simultaneously, the cameras from the DIC
system recorded a video of the test. Loading was applied at a speed of To predict the macroscale tensile force and displacement response,
1.0 mm/s until the specimen was fractured. Istra4D® software was used we constructed a 3D FE model. The Gurson–Tvergaard (GT) damage
to adjust the video recording settings and record the video at a sample model was adopted for the model plastic and damage behavior of the
rate of 38 frames per second for DIC calculations. Psylotest™ was used 3D-printed specimen. Details of the FE model are described in this
to record the displacement and force at a sampling frequency of 20 Hz. section. Furthermore, material properties and model parameters were
linked to process variables through inverse analyses.
2.3. In situ damage progression tests

To observe the continuous damage formation on the surface, a se- 3.1. 3D continuum finite element model
parate in situ tensile test was conducted under an Olympus® BXFM
microscope with a LMPLFLN 5X objective lens. The tensile loading was The 3D cubic element C3D8 (8-node linear element) in ABAQUS
applied at a speed of 0.5 mm/s until the first sign of fracture appeared. [41] was used. The tensile load was applied on the right end, and the
From this point onward, tensile loading was applied in a moderate and left end was fully fixed. Two different materials were defined for the
intermittent fashion according to the observer who was constantly outer shell and infill, respectively, as shown in Fig. 6. The side shell
sweep searching the gage area with the microscope looking for new thickness was measured as 0.8 mm. The infill part had a lattice struc-
signs of fractures. All potential fracture points were recorded as the ture, but we assumed a solid homogeneous isotropic model because it
tensile loading was increased because some of these points would de- had a periodic lattice pattern. Furthermore, the outer shell material was
velop into the final fracture. From that point onward, tensile loading assumed to be the same as the SOLID infill material.

537
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

behavior. The increment of the pore volume fraction is expressed as the


sum of the growth rate (dfgrowth ) of the existing pores and the nucleation
(dfnucleation ) of new pores.
df = dfgrowth + dfnucleation (2)
For existing pores, the Gurson model can account for the softening
effect of the material behavior by pores and at the same time predict the
evolution of pore growth rate during plastic deformation. Following the
law of conservation of mass, the existing pore growth rate is expressed
as
dfgrowth = (1 f )d pl: I (3)
where d and I represent the plastic strain increment and the identity
pl

tensor, respectively. The new pore nucleation is expressed in terms of


the equivalent plastic strain ( ¯mpl ) of the matrix as

dfnucleation = Ad¯mpl (4)


The normal distribution of the nucleation strain required is given as
2
fN 1 ¯mpl N
A= exp
sN 2 2 sN (5)
where N and sN are the mean and standard deviation of the normal
distribution of the nucleation strain and fN is the volume fraction of the
nucleated voids.
The hardening behavior is defined in a tabular form for the GT
model with ABAQUS. Because the goal in this section is to identify the
material parameters by using experimental data, minimizing the
number of parameters is efficient in the optimization process. As a so-
lution, the kinematic hardening parameters are assumed to replace the
tabular form under monotonic loading. The following
Armstrong–Frederick type model [43] was used for the kinematic
hardening.
2 p
dx = Cd x dp
3 (6)
where x is the back stress tensor, C and are the kinematic hardening
model parameters, d p is the increment of the plastic strain, and dp is
the increment of effective plastic strain. In its uniaxial form, for
monotonically increasing plastic strain, the back stress can be written in
Fig. 11. Digital image correlation engineering principal strain. (stress in MPa
terms of the magnitude x as
and strain): (a) (2.282, 0.001); (b) (28.500, 0.028); (c) (33.435, 0.043); (d)
(32.545, 0.057); (e) (31.751, 0.071); (f) (31.226, 0.080); (g) fracture. dx = Cd p x dp (7)
Integrating Eq. (7) with x zero at p = 0 , the uniaxial back stress is
3.2. Porous plasticity model for 3D-printed specimens
expressed as a function of the plastic strain.
According to 3D-printed specimen inspections, the materials seemed x =
C
(1 e
p
)
to have many continuous pores between filaments and discontinuous (8)
pores at the filament intersections. Therefore, we adopted a porous
By adding the back stress to the initial yield stress ( y0 ), yield
plasticity model in ABAQUS, which can empirically model the yield-
stresses corresponding to the plastic strains are calculated. Through this
hardening-softening behavior of 3D-printed models. This Gurson-type
process, the hardening curve with the kinematic hardening parameters
model can deal with nucleation, growth, and coalescence of the mi-
(C , , and pult ) in Eq. (8) is converted to a tabular form (i.e., yield
crovoids. The ductile failure behavior of the FDM-printed material stresses versus plastic strains) to be defined in the ABAQUS input file.
seemed to be very similar to what was observed in porous metallic Therefore, 11 parameters that determine elastoplastic and damage-in-
materials. A yield function of the GT model for porous plastic materials duced softening behavior are selected to predict the monotonic re-
[42] has the following form sponse. We assume that the Poisson ratio is 0.3 because it has a minimal
2 effect on the uniaxial response, which is summarized in Table 2.
eq 3q2 m
( eq, y, f, m) = + 2q1 f cosh (1 + q3 f 2 ) = 0
y 2 y (1) 3.3. Inverse identification of material-constitutive model parameters

where
3 dev dev
eq = 2 ij ij is the conventional von Mises equivalent stress, For material parameter estimation, we applied two different nu-
y = y is the yield stress of the matrix material without pores and
(¯ pl ) merical-experimental methodologies: 1) Self-optim [44] for linear
is a function of the equivalent plastic strain ( ¯ pl ), m = 3 ii is the hy-
1
elastic range and 2) FEMU [45] for damage and plastic range. The
drostatic mean stress, and q1, q2 , and q3 are the model constants. As the hybrid approach was used because Self-optim needs to conduct FE si-
ratio of the pores inside the material increases, the pore volume fraction mulations under both force and displacement boundary conditions. The
increases and the yield function decreases. This effect enables softening experiment in our study is the uniaxial tensile test; hence, it is difficult

538
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 12. Progressive failure images from in situ tensile test under the microscope. (loading [N], displacement [mm]) (a) (0.0, 0.0); (b) (414.173, 0.450); (c) (877.122,
0.9); (d) (1244.087, 1.45); (e) (1192.363, 2.15); (f) (1130.997, 2.8).

material properties and are used to construct an objective function to


determine the true material properties. The objective function of Self-
optim is defined as
n
RMSEi (F , D) + MAEi (F , D)
fSO (pk : k = 1, …, m) =
i=1 R i 2 (F , D ) + 1
n
RMSEi (F , D ) + MAEi (F , D )
+
i=1 R i 2 (F , D ) + 1 (9)

where RMSE , MAE , and R are the root mean square error, the mean
average error, and correlation for full-field stresses and strains ex-
tracted from two FE simulations, respectively. Furthermore, pk is the
material property to be identified, and n is the number of load or time
Fig. 13. Progression of the fracture.
steps used to increase sensitivity for the optimization problem. The
notation used in the objective function can be expressed as (F , D) ,
to obtain converged solutions under force boundary conditions. Details which indicates statistical differences (i.e., errors) between force-driven
on Self-optim can be referred to in literature [44,46]. Therefore, after and displacement-driven simulations, respectively. Superscripts and
the yielding point, we applied the FEMU approach where only the in the statistical indices denote the differences in terms of stresses and
displacement boundary is imposed. This combined approach is depicted strains, respectively.
in Fig. 7, where Self-optim identifies Young's modulus and FEMU In FEMU, the reaction force response is obtained through the dis-
identifies other plastic and damage parameters. In both inverse ana- placement-driven analysis. The obtained reaction force data are com-
lyses, force and displacement data, as well as DIC data at five material pared with those from experimental tests. In this paper, an objective
points, are shown in Fig. 8. function form used for FEMU was defined as
Fig. 9 (a) shows the overall algorithmic procedures of Self-optim,
whereas the algorithmic procedures of the FEMU approach are shown fFEMU (pk : k = 1, …, m)
in Fig. 9(b). In Self-optim, the force and displacement profiles have to n
RMSEi (RF sim RF exp) + MAEi (RF sim RF exp)
be obtained from the test specimen before running the Self-optim si- =
Ri (RF sim RF exp)
mulation. Although the experiment in this paper is a uniaxial tensile i=1

test, a complex tested structure under combined loadings is more de- (10)
sirable for Self-optim as long as both boundary force and displacements
The Chaotic Firefly Algorithm (CFA) [47,48] was selected to search
can be measured from a single test. As shown in Fig. 9 (a), force and
for the optimal material properties because it is efficient in searching
displacement data from the single experimental test are separately
for a global optimizer. The CFA is one of the metaheuristic optimization
imposed to two parallel FE simulation models, whereby two sets of full-
algorithms that mimics the social behavior of fireflies [48].
field stresses and strains are extracted. The differences in full-field
stresses and strains between two FE simulations are due to incorrect

539
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 14. Computed tomography scan images.

4. Results and discussions configuration. Above all, infill density was the most relevant parameter
affecting the specimen properties. As shown in Table 3, Young’s mod-
4.1. Effects of 3D-printing process variables on mechanical properties ulus, initial yield stress, ultimate strength, and toughness increased
with an increase in infill density. Torres et al. obtained the same results
Tensile stress and strain curves were obtained for the six groups and explained that this is attributed simply to the fact that the more the
with different combinations of FDM process variables as shown in materials, the more is the strength and stiffness [22]. The same results
Fig. 10. The stress–strain curves were characterized by five properties: were also observed in available studies [19,49,50]. However, compared
1) Young’s modulus, 2) initial yield stress, 3) ultimate strength, 4) to other properties, the increase in toughness per infill density was
modulus of toughness, and 5) elongation at break (EB). The initial yield relatively small because EB decreased as the infill density increased. It
stress was obtained by the 0.2% offset yield strength method. The is worth noting that infill density affecting EB has not been reported in
modulus of toughness was obtained by calculating the area under the available literature. The coefficient of variation (cov) for properties due
stress–strain curve up to the fracture point. As seen in Fig. 10, there was to infill density changes did not have a clear regular tendency.
considerable variation depending on the 3D printing parameters even LT showed a moderate influence in affecting the specimen’s prop-
when the bulk material properties were the same. This depicts the erties. Young’s modulus was 0.33 mm, which was higher than the value
importance of understanding the effects of printing process variables on of 0.254 mm. In addition, pores were considered premature material
the mechanical properties. flaws. Because smaller thickness contributed to more pores, Young’s
The average values obtained from the results of uniaxial tensile tests modulus decreased. The same result was observed in the study by
are shown in Table 3. Every property is analyzed for each printing Carneiro et al., where Young’s modulus increased with a higher LT
configuration. [50]. The ultimate strength increased with 0.330 mm LT, except for the
The tensile test results showed the performance of each printing case with SOLID density. However, the difference is minimal. Torres

540
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 15. Parameters’ results from inverse identification analyses.

et al. and Carneiro et al. obtained the same results in their studies 0.330 mm specimen are shown in Fig. 12. A fracture location was
[22,50]. The opposite result was obtained when the ultimate strength identified and progressively monitored up to the fracture moment. As
increased with a lower LT [10,13,21,51,52]. The initial yield stress shown in Figs. 12 and 13, the fracture apparently occurred at a pre-
slightly increased as the LT increased for SOLID and HIGH, except for mature pore location. The observed area corresponded to the edge of
LOW. This is because LOW specimen yielding is governed more by the the gage section (i.e., where the side shell meets the top shell).
infill density than by the LT. As LT increased, the toughness slightly Fig. 12(c) shows necking and heavy local strain of the side shell fila-
increased for HIGH and SOLID, except for LOW. EB slightly decreased ment as indicated by the color change to a whitish tone (i.e., stress-
as the LT increased for all infill densities. Chacón et al. and Shubham whitening zone with voids). Fig. 12(d) shows top shell tearing in the
et al. also obtained the same results [20,52]. As shown in Table 3, cov void made by the side shell and the top shell. Fig. 12(e) shows the side
values for ultimate strength, modulus of toughness, and EB when LT shell filament fracture, fracture progression from side shells to top shell
was 0.254 mm were much larger than those when the LT was 0.3 mm. void, and nucleation of new pores at the interface between two adjacent
This implies that a smaller LT provides larger variability at the post- top shell filaments. Fig. 12(f) shows that fractures from the top shell
yielding stage. However, cov differences in Young’s modulus and initial filament bonding and top shell to side shell void are grown and become
yield stress between two thicknesses were irregular because these united. Moreover, the side shell filament is completely fractured.
properties characterized the initial deformation stage. The fracture progression in accordance to the loading vs. displace-
ment graph is shown in Fig. 13. Microscopic fracture images are located
4.2. Progressive failure mechanisms at their corresponding places in the curve. Although this failure me-
chanism was observed on the surface, similar fracture mechanisms are
Failure mechanisms of 3D-printed specimens were investigated by expected through the thickness due to the presence of similar pores
three experimental techniques (i.e., DIC and in situ observations by through the thickness.
optical microscopy and micro CT). A series of images from DIC are Therefore, specimens were scanned using the X-ray CT scan device
shown in Fig. 11, which belong to a specimen with SOLID infill density to observe the behavior of the infill filaments under tensile loading. A
and 0.254 mm LT. (a), (b), (c), (d), and (e) start from 0.5 s and reach up CT scan can show the inside of the specimen without cutting or altering
to 2.5 s with a 0.5 s interval. Furthermore, the contours corresponded to its structure. Six specimens were scanned: two for each infill density
engineering principal strains. Although the specimen is under uniaxial (i.e., LOW, HIGH, and SOLID). Three specimens were previously sub-
constant strain loading, strain distributions are nonuniform along the jected to a tensile loading of 850 N, while the other three remained
longitudinal direction, which is attributed to 3D-printed filament con- unaltered. All specimens were printed with a 0.254 mm LT. A com-
figurations [53]. The strains become concentrated at equally spaced parison between pristine and damaged specimens is shown in Fig. 14,
multiple locations along the side free edge as the loading increases. where the ± 45° printing pattern cannot be seen in the pristine setup
These locations are where premature pores are located at the inter- because they are sectioned images.
sections of side shells and top shells, and the fracture occurred at one of As shown in Fig. 14, premature pores equally spaced between infill
the locations. Others also reported similar heterogeneous strain dis- and side shell were observed. There were premature uneven air gaps
tributions on the uniaxial tensile test specimen [54] and fracture between two side shell filaments even in the pristine state. These air
toughness samples [55]. However, our results showed additional pro- gaps were also partly observed between ± 45° infill filaments. At a
gressive failure patterns. tensile loading of 850 N, necking and delaminations between long-
A series of fractured side edge location images for a SOLID and LT itudinal filaments of the side shell were observed. For SOLID and HIGH

541
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Fig. 16. Comparisons of the force–displacement curves between tests and simulations.

infill densities, single necking of the longitudinal filaments was ob- damage mechanisms, we ensured that the GT porous plasticity model is
served, while multiple necking was observed for the LOW infill density. suitable for FDM 3D-printed parts. To link the 3D printing process
In case of infill filaments, the delamination and necking also occurred at variables to properties and parameters defined in the model, we con-
the loaded state. ducted inverse identification analyses through the approaches ex-
Three different experimental methods were used in this work to plained in Section 3. The identification results are summarized in
inspect the specimen structure and analyze its behavior under tensile Fig. 15, where the mean values and the standard error (= /n2 ) of each
loading. Applying different methods to obtain the failure mechanism identified parameter are indicated.
revealed that the tensile loading unevenly affects each different section Surprisingly, as the infill density and LT decreased, the identified
of the specimen structure. This is because the specimen structure sec- initial relative porous density ( f0 ) increased. These identified results
tions are built in different orientations and patterns. The failure me- matched well with actual material states of FDM 3D-printed specimens.
chanism behaves in a chain reaction among the specimen structure Moreover, the volume fraction of newly nucleating voids ( fN ) also in-
sections and the fracture is initiated at the side shells. Consequently, it creased as the LT and infill density increased, with the exception of
develops to the top and bottom shells until the infill, which is the last SOLID and 0.33 mm LT. It is natural that LOW and 0.254 mm LT
section where the fracture develops. samples had more initial voids than newly nucleating voids. The mean
nucleation strain ( N ) showed an increasing tendency as the LT and
infill density decreased, which matches the actual ductility between
4.3. Process–property relationship by inverse parameter identification samples. The exception of LT and SOLID infill density seemed to be due
to a trade-off between N and fN , as well as the relative brittleness of
As this was the first-time application of the GT model to FDM 3D- behavior. Considering the results of all these analyses, the GT porous
printed specimens, we characterized the damage mechanisms through plasticity model is validated as a predictive model for FDM 3D-printed
careful and multiple experimental techniques. From the observed

542
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

Table 4
Property comparison between the simulation and experiment.
Infill Layer Classification Young s Initial Ultimate Modulus of
density thickness (mm) modulus yield stress strength toughness
(MPa ) (MPa ) (MPa ) (MPa )

SOLID 0.330 Simulation 1018.213 35.936 35.936 2.036


Experiment 1030.968 33.680 34.718 1.982
Error (%) 1.237 6.698 3.508 2.725
SOLID 0.254 Simulation 986.069 33.153 34.894 2.097
Experiment 1015.602 33.514 34.740 1.974
Error (%) 2.908 1.077 0.443 6.231
HIGH 0.330 Simulation 902.27 31.399 31.416 1.723
Experiment 881.637 29.827 31.780 1.745
Error (%) 2.340 5.270 1.145 1.261
HIGH 0.254 Simulation 807.930 25.221 26.010 1.514
Experiment 774.062 24.920 26.333 1.509
Error (%) 4.375 1.208 1.227 0.331
LOW 0.330 Simulation 643.84 22.920 26.112 1.435
Experiment 618.37 23.102 26.602 1.513
Error (%) 4.119 0.788 1.842 5.155
LOW 0.254 Simulation 587.14 23.980 24.224 1.603
Experiment 562.38 24.638 23.634 1.561
Error (%) 4.403 2.671 2.496 2.691

specimens. material states in terms of initial relative porous density ( f0 ), the vo-
All the curves from simulations with identified parameters match lume fraction of newly nucleating voids ( fN ), and mean nucleation
the experimental curves as shown in Fig. 16. strain ( N ). Through an inverse identification analysis, the proces-
As shown in Table 4, for each of the 3D printing process config- s–property relationship was established using the GT porous plasticity
urations, the properties were obtained both experimentally and nu- model. However, the macroscale continuum modeling approach could
merically. In all cases, the percentage error between the simulation and not predict nonuniform strains observed in the experiment. To predict
experimental results was sufficiently small. Because of the inverse the nonuniform stress distributions and explicit progressive failure,
identification analyses, the 3D printing process variables were linked to further research on the development of multiscale FE models capable of
numerical models that enabled 3D-printed specimen and structure de- analyzing the preselected structure and patterns of FDM-printed
sign. structures is needed, which involves thermo-induced residual stress
predictions, consideration of nonlinearity, explicit damage modeling,
etc.
5. Conclusions
Acknowledgments
In this paper, we evaluated the effects of FDM 3D printing process
variables on mechanical properties. The progressive fracture me-
This research is based on the work supported by the Air Force Office
chanism was elucidated by three experimental techniques: 1) DIC, 2) X-
of Scientific Research under award number FA2386-17-1-4081. This
ray CT, and 3) in situ tensile testing by optical microscopy. On the basis
work was also supported by the Institute of Advanced Aerospace
of the observed failure mechanisms of the FDM 3D-printed specimens,
Technology and the Institute of Engineering Research at Seoul National
we selected and validated the GT porous plasticity model to predict the
University. Authors are grateful for their support.
experimental elasto-plastic damage behavior. Because of an experi-
mental–numerical inverse analysis, the material properties and model
References
parameters were successfully identified and linked to the 3D printing
process variables to design 3D-printed specimens and structures.
[1] A. Gebhardt, Rapid Prototyping–Rapid Tooling–Rapid Manufacturing, Carl Hanser,
Infill density (LOW, HIGH, and SOLID) and LT (i.e., 0.33 and München, 2007.
0.254 mm), which are 3D printing process variables, were varied to [2] I. Gibson, D.W. Rosen, B. Stucker, Additive Manufacturing Technologies, (2010)
produce six groups of uniaxial tensile test specimens. Five mechanical Google Scholar.
[3] M. Gebler, A.J.S. Uiterkamp, C. Visser, A global sustainability perspective on 3D
properties – Young’s modulus, initial yield stress, ultimate strength, printing technologies, Energy Policy 74 (2014) 158–167.
modulus of toughness, and EB – were evaluated. Young’s modulus, in- [4] Kianian, B., Wohlers Report 2017: 3D Printing and Additive Manufacturing State of
itial yield stress, ultimate strength, and toughness increased with an the Industry, Annual Worldwide Progress Report: Chapters titles: The Middle East,
and other countries. 22 ed ed. 2017, FORT COLLINS, COLORADO, USA: Wohlers
increase in infill density. EB decreased as the infill density increased. LT Associates, Inc.
showed a moderate influence affecting the specimen’s properties, where [5] B. Jenett, et al., Digital morphing wing: active wing shaping concept using com-
an increasing LT apparently increased Young’s modulus, while it de- posite lattice-based cellular structures, Soft Rob. 4 (1) (2017) 33–48.
[6] R. Anitha, S. Arunachalam, P. Radhakrishnan, Critical parameters influencing the
creased EB. We found that smaller LT increased uncertainties (i.e.,
quality of prototypes in fused deposition modelling, J. Mater. Process. Technol. 118
variability) of postyielding properties. (1-3) (2001) 385–388.
With regard to progressive failure, the failure was initiated by strain [7] K. Chin Ang, et al., Investigation of the mechanical properties and porosity re-
concentrations at the premature pores mostly located at the intersection lationships in fused deposition modelling-fabricated porous structures, Rapid
Prototyp. J. 12 (2) (2006) 100–105.
between infill and side shell filaments. The pore size within the infill [8] J. Kotlinski, Mechanical properties of commercial rapid prototyping materials,
filaments increased as the loading increased. The damage was further Rapid Prototyp. J. 20 (6) (2014) 499–510.
progressed with necking and tearing of filaments and nucleation of [9] A.K. Sood, R.K. Ohdar, S.S. Mahapatra, Experimental investigation and empirical
modelling of FDM process for compressive strength improvement, J. Adv. Res. 3 (1)
pores between filaments. The GT porous plasticity model with a 3D (2012) 81–90.
continuum FE model was validated to predict the macroscopic elasto- [10] Ö. Bayraktar, et al., Experimental study on the 3D‐printed plastic parts and pre-
plastic damage behavior of FDM 3D-printed specimens. The identified dicting the mechanical properties using artificial neural networks, Polym. Adv.
Technol. 28 (8) (2017) 1044–1051.
parameters of the GT porous plasticity model matched the actual

543
T. Webbe Kerekes et al. Additive Manufacturing 25 (2019) 532–544

[11] M. Spoerk, et al., Parametric optimization of intra‐and inter‐layer strengths in parts modelling using three-dimensional finite element analysis, Proc. Inst. Mech. Eng.
produced by extrusion‐based additive manufacturing of poly (lactic acid), J. Appl. Part B-J. Eng. Manuf. 222 (8) (2008) 959–967.
Polym. Sci. 134 (41) (2017) 45401. [33] Y. Zhang, Y.K. Chou, Three-dimensional finite element analysis simulations of the
[12] J.-W. Tseng, et al., Screw extrusion-based additive manufacturing of PEEK, Mater. fused deposition modelling process, Proc. Inst. Mech. Eng. Part B-J. Eng. Manuf.
Design 140 (2018) 209–221. 220 (10) (2006) 1663–1671.
[13] M. Spoerk, et al., Effect of the printing bed temperature on the adhesion of parts [34] A.R. Torrado, et al., Characterizing the effect of additives to ABS on the mechanical
produced by fused filament fabrication, Plastics Rubber Compos. 47 (1) (2018) property anisotropy of specimens fabricated by material extrusion 3D printing,
17–24. Addit. Manuf. 6 (2015) 16–29.
[14] M. Spoerk, et al., Optimisation of the adhesion of polypropylene-based materials [35] D.A. Roberson, et al., Comparison of stress concentrator fabrication for 3D printed
during extrusion-based additive manufacturing, Polymers 10 (5) (2018) 490. polymeric izod impact test specimens, Addit. Manuf. 7 (2015) 1–11.
[15] V. Kishore, et al., Infrared preheating to improve interlayer strength of big area [36] S.-H. Ahn, et al., Anisotropic material properties of fused deposition modeling ABS,
additive manufacturing (BAAM) components, Addit. Manuf. 14 (2017) 7–12. Rapid Prototyp. J. 8 (4) (2002) 248–257.
[16] M. Spoerk, et al., Polypropylene filled with glass spheres in extrusion-based additive [37] S.-I. Park, et al., Effective mechanical properties of lattice material fabricated by
manufacturing: effect of filler size and printing chamber temperature, Macromol. material extrusion additive manufacturing, Addit. Manuf. 1 (2014) 12–23.
Mater. Eng. 303 (7) (2018). [38] Y. Jin, et al., Quantitative analysis of surface profile in fused deposition modelling,
[17] C.B. Sweeney, et al., Welding of 3D-printed carbon nanotube-polymer composites Add Manuf 8 (2015) 142–148.
by locally induced microwave heating, Sci. Adv. 3 (6) (2017). [39] S.-i. Park, D.W. Rosen, Quantifying effects of material extrusion additive manu-
[18] A. Farzadi, et al., Effect of layer thickness and printing orientation on mechanical facturing process on mechanical properties of lattice structures using as-fabricated
properties and dimensional accuracy of 3D printed porous samples for bone tissue voxel modeling, Addit. Manuf. 12 (2016) 265–273.
engineering, PloS one 9 (9) (2014) e108252. [40] G.J. Yun, S. Shang, A self-optimizing inverse analysis method for estimation of
[19] C. Alvarez, et al., Investigating the influence of infill percentage on the mechanical cyclic elasto-plasticity model parameters, Int. J. Plast. 27 (4) (2011) 576–595.
properties of fused deposition modelled ABS parts, Ingeniería e Investigación 36 (3) [41] Karlsson Hibbett, Sorensen, ABAQUS/Standard: User’S Manual vol1, Hibbitt,
(2016) 110–116. Karlsson & Sorensen, 1998.
[20] J. Chacón, et al., Additive manufacturing of PLA structures using fused deposition [42] A.L. Gurson, Continuum theory of ductile rupture by void nucleation and growth:
modelling: effect of process parameters on mechanical properties and their optimal part i—yield criteria and flow rules for porous ductile media, J. Eng. Mater.
selection, Mater. Design 124 (2017) 143–157. Technol. 99 (1) (1977) 2–15.
[21] B. Rankouhi, et al., Failure analysis and mechanical characterization of 3D printed [43] P.J. Armstrong, C.O. Frederick, A Mathematical Representation of the Multiaxial
ABS with respect to layer thickness and orientation, J. Fail. Anal. Prev. 16 (3) Bauschinger Effect, Berkeley Nuclear Laboratories, 1966.
(2016) 467–481. [44] G.J. Yun, S. Shang, A self-optimizing inverse analysis method for estimation of
[22] J. Torres, et al., An approach for mechanical property optimization of fused de- cyclic elasto-plasticity model parameters, Int. J. Plast. 27 (2011) 576–595.
position modeling with polylactic acid via design of experiments, Rapid Prototyp. J. [45] M. Springmann, M. Kuna, Identification of material parameters of the
22 (2) (2016) 387–404. Gurson–Tvergaard–Needleman model by combined experimental and numerical
[23] F. Rayegani, G.C. Onwubolu, Fused deposition modelling (FDM) process parameter techniques, Comput. Mater. Sci. 33 (4) (2005) 501–509.
prediction and optimization using group method for data handling (GMDH) and [46] G.J. Yun, S. Shang, A New inverse method for the uncertainty quantification of
differential evolution (DE), Int. J. Adv. Manuf. Technol. 73 (1-4) (2014) 509–519. spatially varying random material properties, Int. J. Uncertainty Quant. 6 (6)
[24] E.G. Gordeev, A.S. Galushko, V.P. Ananikov, Improvement of quality of 3D printed (2016) 515–531.
objects by elimination of microscopic structural defects in fused deposition mod- [47] X.-S. Yang, Firefly algorithm, stochastic test functions and design optimisation, Int.
eling, PloS One 13 (6) (2018) e0198370. J.Bio-Inspired Comput. 2 (2) (2010) 78–84.
[25] L.M. Galantucci, F. Lavecchia, G. Percoco, Experimental study aiming to enhance [48] A.H. Gandomi, et al., Firefly algorithm with chaos, Commun. Nonlinear Sci. Numer.
the surface finish of fused deposition modeled parts, CIRP Ann. 58 (1) (2009) Simul. 18 (1) (2013) 89–98.
189–192. [49] M. Fernandez-Vicente, et al., Effect of infill parameters on tensile mechanical be-
[26] M.K. Agarwala, et al., Structural quality of parts processed by fused deposition, havior in desktop 3D printing, 3D Printing Additive Manuf. 3 (3) (2016) 183–192.
Rapid Prototyp. J. 2 (4) (1996) 4–19. [50] O.S. Carneiro, A. Silva, R. Gomes, Fused deposition modeling with polypropylene,
[27] J.C. Riddick, et al., Fractographic analysis of tensile failure of acrylonitrile-buta- Mater. Design 83 (2015) 768–776.
diene-styrene fabricated by fused deposition modeling, Addit. Manuf. 11 (2016) [51] W. Wu, et al., Influence of layer thickness and raster angle on the mechanical
49–59. properties of 3D-printed PEEK and a comparative mechanical study between PEEK
[28] S. Bhandari, R. Lopez-Anido, Finite element analysis of thermoplastic polymer ex- and ABS, Materials 8 (9) (2015) 5834–5846.
trusion 3D printed material for mechanical property prediction, Addit. Manuf. 22 [52] P. Shubham, A. Sikidar, T. Chand, The influence of layer thickness on mechanical
(2018) 187–196. properties of the 3D printed ABS polymer by fused deposition modeling, Key Eng.
[29] S. Guessasma, et al., Anisotropic damage inferred to 3D printed polymers using Mater. (2016) 706.
fused deposition modelling and subject to severe compression, Eur. Polym. J. 85 [53] J.T. Cantrell, et al., Experimental characterization of the mechanical properties of
(2016) 324–340. 3D-printed ABS and polycarbonate parts, Rapid Prototyp. J. 23 (4) (2017) 811–824.
[30] S. Guessasma, S. Belhabib, H. Nouri, Significance of pore percolation to drive ani- [54] H. Rezayat, et al., Structure–mechanical property relationship in fused deposition
sotropic effects of 3D printed polymers revealed with X-ray mu-tomography and modelling, Mater. Sci. Technol. 31 (8) (2015) 895–903.
finite element computation, Polymer 81 (2015) 29–36. [55] J. Gardan, A. Makke, N. Recho, Improving the fracture toughness of 3D printed
[31] A. Bellini, S. Güçeri, Mechanical characterization of parts fabricated using fused thermoplastic polymers by fused deposition modeling, Int. J. Fract. 210 (1-2)
deposition modeling, Rapid Prototyp. J. 9 (4) (2003) 252–264. (2018) 1–15.
[32] Y. Zhang, K. Chou, A parametric study of part distortions in fused deposition

544

Das könnte Ihnen auch gefallen