Professional Documents
Culture Documents
crossmark
Thomas Stichela,b, , Thomas Fricka, Tobias Laumera,b, Felix Tennerc, Tino Hausotteb,d,
Marion Merkleinb,e, Michael Schmidta,b,c
a
A R T I C L E I N F O
A BS T RAC T
Keywords:
Additive manufacturing
Polymers
Polyamide-12
Selective Laser Sintering
Round Robin
The mechanical and microstructural investigation of polymer parts (polyamide 12) fabricated by Selective Laser
Sintering as part of a Round Robin initiative is presented. The paper focuses on the microstructural analysis of
the Round Robin samples and their evaluation regarding their eect on mechanical properties with respect to
each other. Therefore optical microscopy on microtomed samples, X-ray computed tomography and Dierential
Scanning Calorimetry is used to determine the morphology of residual particle cores and of internal pores.
The mechanical tensile testing revealed a high variability of the ductility of the samples among the used
machines and a distinctive anisotropic mechanical response. Especially the quite brittle characteristic along the
building direction has shown to be still a crucial challenge for the process. However, one machine delivered
samples with outstanding ductility with total elongation values of about 21% along the building direction and of
about 32% planar to the layer. This result was back traced to a distinctive pore and residual particle morphology
which is characterized by low pore concentration, the absence of coplanar pore or residual particle
arrangements and the highest degree of particle melting measured. Furthermore, the analysis depicts that
both features, pores and residual particles, contribute to the mechanical properties signicantly and that they
are not necessarily linked since they can vary independently in a certain range depending on the machine
conguration.
1. Introduction
Selective Laser Sintering (SLS) is one of the most widespread
Additive Manufacturing processes that emerged in the last decades
[1,2]. It allows the fabrication of three-dimensional polymer components with complex shapes according to CAD data directly out of
powdery material without the need of tooling and setup [3]. Therefore
parts are built up layer by layer. Three steps are repeated for each layer:
First, the platform is lowered by the thickness of one layer. Second,
powder is deposited on the building platform using a moving roller or
blade device and preheated to a temperature close to the material
melting point. And third, a laser beam scans the powder layer with
arbitrary trajectories in order to melt specic powder layer areas. These
steps are applied alternately until the part is completed. Before the
parts can be extracted from the machine, homogeneous cooling has to
take place whereas the viscose melted polymer crystallizes with
minimal distortions.
The main advantage of SLS over other Rapid Prototyping or
Additive Manufacturing technologies using plastics is the ability to
produce parts which feature material properties that are very close to
the material properties obtained by injection molding [4]. However,
this is only achieved when a robust process and material routine is
established leading to high quality and repeatable results. In many
cases, SLS machine users observe minor reproducibility regarding to
mechanical and geometrical properties [5], limiting the industrial
application and cost-eciency of SLS.
Especially the mechanical properties along the build direction have
shown to be a challenge for the process. Ductility and strength are
generally lower along the build direction than perpendicular to it and
they are also very sensitive to varying process conditions and parameters. This can be explained by the stacking-layer-process which
leads easily to insucient interlayer connection, e. g., if energy input or
http://dx.doi.org/10.1016/j.optlastec.2016.09.042
Received 12 July 2016; Received in revised form 23 August 2016; Accepted 26 September 2016
0030-3992/ 2016 Elsevier Ltd. All rights reserved.
T. Stichel et al.
the local process temperatures are too low for the complete melting of
the layer section. Insucient interlayer connection is related to
distinctive microstructural features such as pore formation or residual
particles (or cores) [6,7] which can occur with varying frequency and
morphology (size, shape, orientation).
The occurrence of dierent morphologies of porosity inside polymer parts built by SLS is well known [68]. While it is aected by many
material and process parameters [8], an evidentiary understanding of
its morphology and impact on the mechanical properties is still
missing. Very often the interlayer porosity is blamed for inferior
mechanical properties along the building direction [6]. Besides pores,
residual particle and particle core arrangements due to partial melting
were also identied to aect the mechanical response of SLS processed
samples. It was stated that an increase of the degree of particle melt
(DPM) leads to an improvement of the mechanical properties and
coplanar aligned residual particles and cores promote mechanical
weakness along the building direction [6,9]. This is explained by the
many short binding necks creating high stiness which result in short
elongation [10].
Nevertheless, a quantication of the inuence of both eects
(porosity and residual particles) on the mechanical properties in
comparison has not been reached yet. Typically applied research
methodology base on the variation of few parameters which leads to
complementary changes in both features and thus do not allow the
estimation of their respective signicance for the mechanical properties. This can be exemplarily seen in [8] where the variation of the laser
beam energy leads to congruent change of the crystal weight fraction
and the porosity with strong impact on the elongation at break but with
no clue of which eect is more important. Thus experimental methodologies should be applied which include the broad variation of
parameters and dierent machines resulting in more versatile microstructural morphologies.
In order to measure the inuence of pore formation and residual
particle (cores) on mechanical properties with respect to each other, a
comprehensive microstructural analysis is performed upon tensile test
samples built during a Round Robin initiative based on six dierent
machines and parameter sets. The internal pore morphology of the
samples is measured by X-ray computed tomography while the residual
particles (and particle cores) are investigated using microscopy on thin
sections and Dierential Scanning Calorimetry (DSC). The results are
used to trace back conspicuous features of the tensile testing outcome
to the source in order to value the signicance of pore and residual
particle morphology.
Fig. 1. CAD design of a testing sample set consisting of three dierently orientated
tensile bars: Horizontal orientation corresponds to in-plane direction, vertical orientation to build direction.
medium.
3. Characterization methods
In Fig. 3 an overview of the performed characterization methods is
given. The mechanical properties which resemble the functionality of
the samples are determined via tensile testing according to ISO 3167.
Microstructural features which are represented by the residual particle
(core) arrangements and the porosity can help to explain the appearance of functionality (mechanical) issues. While arrangement of
residual particles and cores are veried by optical analysis of thin
sections, the general amount of residual particle is resembled by the
degree of particle melt which is determined using Dierential Scanning
Calorimetry. In order to analysis the pore morphology of the samples
comprehensively, X-ray computed tomography is used which delivers
information about porosity, pore density, pore arrangement, pore
orientation and pore size distribution. All respective determination
methods are commented in detail in the following sections.
The Round Robin includes six dierent SLS machines and users
(M1-M6) which produced a range of tensile samples (ISO 3167, Type
A) with dierent build orientations (see Fig. 1). For each machine
dierent parameter sets (laser power, scanning speed or layer thickness, etc.) were used, which were specied by the applicants to be the
optimum for the respective machine. The samples were made from the
standard material polyamide 12 Duraform PA (3D Systems) or PA2200
(EOS) which base on the same basic powder VESTOSINT produced by
Evonik. An overview of the production parameters are displayed in
Table 1.
Six sample sets were produced in machine M1, M3, M4 and M6,
three sets in machine M2 and one set in machine M5. The nominal and
measured linear dimensions of the samples as well as the classication
details according to ISO 2768-1 are displayed in Fig. 2. The samples of
M6 could not be measured due to strong deformations and displacements and are beyond of any fabrication tolerances. The samples
fabricated with machine M5 cannot be classied by the norm since the
sample length is exceeding the dened deviation of 2.5 mm for the
very coarse category by far. Instead, the samples of the machines M1,
M3 and M4 can be categorized as coarse, the samples of M2 as
Tensile tests are carried out using the Zwick/Roell 100 AllroundLine machine with a test speed of 5 mm/min. The values that are
evaluated and compared are the ultimate tensile strength (UTS) and
the total elongation (TE) which is also known as elongation at break.
The yield point is dened for plastics as the rst point on the stressstrain curve at which an increase in strain occurs without an increase in
stress [11].
3.2. Thin section analysis
Optical analysis was performed on thin sections with a thickness of
50 m prepared by a microtome. Polarized light was used in order to
distinguish between non-molten, partly molten and fully molten
particles in the sample inside. The preparation of the section was
realized in that way that they display the layer stack.
Already mechanically tested tensile test samples were picked and
prepared by the microtome. The thin sections were extracted from the
furthest clamping areas of the tensile bars where no plastic deformation aecting the samples microstructure is introduced upon tensile
and clamping force. This was veried by controlling the geometry of the
32
T. Stichel et al.
Table 1
Parameter protocol for the different machines M1M6.
Machine
M1
M2
M3
M4
M5
M6
Machine manufacturer
Sample count (H, D, V)
Laser power (W)
Beam diameter (FWHM) (m)
Hatch distance (m)
Hatch strategy
Layer thickness (m)
Scanning speed (mm/s)
Recoating speed (mm/s)
Powder material
Powder mixing rate (used-new)
Processing temperature (C)
Total build time (hh: mm)
Cooling time in machine (h)
Finishing routine
3D Systems
6, 6, 6
42
464
279.4
X and Y alternately
101.6
10000
254
Duraform PA
5050
182
12:35
1+4 (at room temp.)
bead blast
EOS GmbH
3, 3, 3
21
430
250
X and Y alternately
100
2500
n. a.
PA 2200
5050
173
12:30
~10
bead blast+water wash
EOS GmbH
6, 6, 6
n. a.
600
n. a.
n. a.
120
n. a.
120
PA 2200
5050
160
10:20; 27:37
10
sand blast+water wash
EOS GmbH
6, 6, 6
38
500
310
X and Y alternately
120
6000
200
Duraform PA
5050
165.5
16:31
48
sand blast
DTM Corporation
1, 1, 1
7
400
150
X and Y alternately
100
1256
120
PA 2200
5050
172
n. a.
n. a.
sand blast
EOS GmbH
6, 6, 6
30
600
600
X and Y alternately
150
300
50
PA 2200
5050
178
n. a.
2
compressed air
.
While the total crystallinity of the sample can be extracted from the
rst heating cycle of the DSC measurement of the sample, the crystallinity of melted and recrystallized material is derived from the second
heating cycle of the same sample. Instead the core (or powder)
crystallinity is determined by means of the DSC measurement using
pure powder probes of the powders used with the respective machines.
DPM (%)
Total crystallinity of sample Core (or powder) crystallinity
=
Crystallinity of melted and crystallized material Core crystallinity
(1)
Fig. 2. Boxplot of the sample thickness (a), width (b) and length (c) measured using tactile method and compared to nominal value (dotted line) and general tolerance values according
to ISO 2768-1. The measured samples were built with dierent orientation (D: diagonal; H: horizontal; V: vertical) with dierent machines (M1-M6) and parameters.
33
T. Stichel et al.
Fig. 4. Thermogram of a DSC measurement of a sintered sample built with machine M3.
Fig. 6 displays the UTS and the TE of the tensile testing results of
the samples of the machines M1 to M6. Generally, the measured UTS
and TE values show that they depend on build orientation. So the
values decrease with the build angle increasing from 0 (H) and 45 (D)
to 90 (V). Moreover, scatter is also quite large for the D and V builds.
Only the H builds are close to the specication of the powder supplier
which are 45 MPa for PA2200 (EOS) and 43 MPa for DuraForm PA
(3D Systems) for the UTS and 20% for PA2200 (EOS) and 14% for
DuraForm PA (3D Systems) for the TE. Instead, the V builds show
mostly brittle fracture behavior resulting in TE values below 10%.
These observed anisotropic characteristics which are a consequence of
the stacking-layer-process are well known in literature [6,9].
Exceptions are the samples of machine M2 which achieved best
values by far for the UTS and the TE. Despite to the other machines,
higher UTS and higher TE for the V builds than for the D builds were
measured. The V builds feature a TE of around 20% which is as high as
the TE values of the H builds of the other machines. Thus, they are the
only V builds which yield the desired ductile fracture mode which
indicates a fracture that occurs after the onset of strain hardening [6].
Exceptionally distinctive necking with a high TE value of ca. 32% were
measured for the H builds of M2 proving a very high ductility
compared to the other samples. The TE values (TE for H and V builds)
are also clearly higher than the ones reported by the VDI upon a Round
Robin initiative using PA 12 powder with laser sintering [5]. However,
ductility of injection molded PA12 samples is still higher [6].
The outstanding properties of the samples built by machine M2
show that even along the building direction ductile mechanical properties can be reached. This must be the result of distinctive microstructural dierences compared to the samples built in the other machines.
Thus, in the following sections, a comprehensive microstructural
analysis is performed in order to trace back the mechanical performance to distinctive microstructural features.
4.2. Inuence of the pore morphology on the mechanical properties
A detailed analysis of the morphologies of porosity of the dierent
samples measured by X-ray computed tomography is given and
correlated to the mechanical testing results. The overall degree of
porosity of the samples, the pore concentration as well as the pore size
distributions are determined. The generated volume data sets of the
pores are visually reviewed in order to identify conspicuous porosity
arrangements.
4.2.1. Validity of the porosity measurement
In order to estimate the validity of the measurements, an example
of a pore size distribution is compared with its respective contribution
Fig. 5. Visualized CT data and detected pores (Sample size: ca. 1584 mm3).
34
T. Stichel et al.
Fig. 6. Ultimate tensile strength (UTS) (a) and total elongation (TE) (b) of samples built with dierent orientation (D: diagonal; H: horizontal; V: vertical) in dierent machines (M1
M6). N is the number of samples.
Fig. 7. Exemplary measurement: relative frequency of the dierent pore sizes (pore occurrence relative to pore size) (a) and its contribution to the overall porosity (b). Cross section
image of a tensile test sample showing pores of dierent sizes.
to the overall porosity, see Fig. 7. Fig. 7a shows that there is a high
count of small pores, allowing the conclusion that pores with sizes
below the metrological structural resolution are also frequent.
Fig. 8. Boxplots of overall porosity and pore concentration of samples built with the machines M1-M6, analyzed using x-ray computed tomography.
35
T. Stichel et al.
2
1 K
ac= Ic
Yf
(2)
using a plane-strain fracture toughness KIc for nylon of 2.8 MPa m1/2
[14], and fracture stress of 40 MPa as well as a geometrical shape factor
Y of 1. Thus, a critical crack size ac of about 3.5 mm is estimated which
is clearly larger than the initial pore sizes measured for all samples.
Thus mechanical behavior is predominated by sub-critical crack growth
until the critical crack size is reached, leading nally to fast fracture.
The strong anisotropy of the mechanics (best TE and UTS values for
H built specimen) and the sensitivity of the TE values along the
building direction suggest that pores support directly an anisotropic
mechanical response. Hence, elongated pore shapes or coplanar
alignments have to be the reason which will be discussed in the
following section.
However, the correlation between mechanics and pore concentration is not valid for the H built samples which can be noticed when
comparing the TE and UTS of M1 with M2. They dier strongly despite
similar pore concentration distributions are observed. This shows that
pore morphology is not the only relevant factor inuencing the
mechanics of samples built by SLS (see Section 4.3).
T. Stichel et al.
Fig. 9. Pore concentration in dependency on the pore size of the samples built with dierent machines M1M6.
T. Stichel et al.
Fig. 10. Pore morphology of samples of the dierent machine charges visualized using CT data with a depth of eld of approximately 3 mm. Samples were visualized with vertical layer
orientation perpendicular to the image plane.
Fig. 11. Thin section images of samples built with vertical (V) orientation in the dierent machines M1-M6 allow the identication of non-molten particles.
38
T. Stichel et al.
Fig. 12. Boxplots of percentage crystallinity and degree of particle melt (DPM) of samples of the machines M1-M6 estimated out of DSC charts.
5. Discussion
T. Stichel et al.
40