New developments in fused deposition

modeling of ceramics
Anna Bellini
Department of Materials and Process Technology, Technical University of Denmark, Lyngby, Denmark, and
Lauren Shor and Selcuk I. Guceri
Department of Mechanical Engineering and Mechanics, Drexel University, Philadelphia, Pennsylvania, USA

Abstract
Purpose – To shift from rapid prototyping (RP) to agile fabrication by broadening the material selection, e.g. using ceramics, hence improving the
properties (e.g. mechanical properties) of fused deposition modeling (FDM) products.
Design/methodology/approach – This paper presents the development of a novel extrusion system, based on the FDM technology. The new set-up,
consisting of a mini-extruder mounted on a high-precision positioning system, is fed with bulk material in granulated form, instead that with the more
common filament.
Findings – Previous research showed that the applications of new materials with specific characteristics in a commercial FDM system are limited by
the use of intermediate precursors, i.e. a filament. The new design described in this paper overcomes the problem thanks to the new feeding system.
Research limitations/implications – The work presented in this paper is only the starting point for further development. The new system design was
tested and encouraging improvements of the final product were achieved. However, several parameters, e.g. size of the feeding granules, still need to
be optimized.
Practical implications – This configuration opens up opportunities for the use of wider range of materials, making the FDM to become a viable
alternative manufacturing process for specialty products.
Originality/value – The mini-extruder deposition system developed in this study exploits the advantages of the RP technologies: ability to
shorten the product design and development time; suitability for automation; and ability to build many geometrically complex shapes. Hence,
applying the described technology, it will be possible to manufacture customer-driven product with important cost and time (from design to
final product) savings.

Keywords Advanced manufacturing technologies, Electrodeposition, Rapid prototypes, Ceramics

Paper type Research paper

Introduction structure and shape (as long as the smallest features will
not be less than few millimeters) with theoretically any
Layered manufacturing (LM) is an evolution of rapid material.
prototyping (RP) techniques where the part is built in In this approach, parts are fabricated by stacking layers,
layers. While most of the previous technologies focused on each of which is produced using “roads” in a configuration
building “prototypes” many recent approaches, including the rastering cross sectional surfaces that are generated by slicing
current project, try to achieve an agile fabrication technology three-dimensional solid models (Figure 1). “Roads” are
to produce the final product directly. formed by extruding a mixture of the ceramic particles and a
The objective of the current research, under Office of thermoplastic carrier/binder.
Naval Research funding, is to improve a LM process to Once a part is fabricated using the particle-binder mixture,
fabricate ceramic and multi-component parts. Among all the subsequent binder burnout and sintering steps lead to the
available LM techniques fused deposition modeling (FDM) finished product. Since parts are fabricated on an
process was chosen for this research. The main reason that “evolutionary” fashion road by road and layer-by-layer, the
led to this decision is the capability of this process to build part integrity of the finished product greatly depends on the
objects of relatively contained sizes with any complex void formation and inter road – inter layer – bonding during
fabrication (Bellini and Güçeri, 2003). Exposure to high
The Emerald Research Register for this journal is available at temperature thermal cycles results in undesirable mechanical
www.emeraldinsight.com/researchregister performance characteristics and cracks in the finished part
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1355-2546.htm
This project was sponsored, in part, by the Office of Naval Research
(ONR) in MURI project, #N0014-96-1-1175. Many constructive
discussions with other MURI team members, notable with Drs
Rapid Prototyping Journal Danforth, Safari and Jafari of Rutgers University, are also acknowledged.
11/4 (2005) 214– 220 Received: 3 March 2004
q Emerald Group Publishing Limited [ISSN 1355-2546] Revised: 5 April 2005
[DOI 10.1108/13552540510612901] Accepted: 29 April 2005

214
 New developments in fused deposition modeling of ceramics Rapid Prototyping Journal
Anna Bellini, Lauren Shor and Selcuk I. Guceri Volume 11 · Number 4 · 2005 · 214 –220

Figure 1 The FDM approach Figure 2 The MED system

due to the presence of voids and/or poor inter-layer and inter-

road bonding characteristics (Bellini et al., 2001).
The product performance often demands use of different
materials for specific applications. Therefore, it is necessary to
have a technology that is adaptable for a wide selection of
materials.
Observations show that the main obstacle for application of
new materials with specific characteristics often comes from
the use of intermediate precursors such as a filament. For
example, in an FDM process, the advantage of using a
precursor filament is thus offset by problems encountered
during its preparation and fabrication (Venkataraman, 2000).
Furthermore, during the extrusion phase in road formation,
the frequent buckling failures cause interruption of the
process (with the necessary cooling down and warming up of
the liquefier) and necessitate frequent operator intervention
(Venkataraman, 2000). Consequently, this problem prevents
an automatic and continuous process diminishing the main
advantage of a filament-based system. In addition, the
backpressure encountered during deposition limits the
powder volume fraction in the filament (McNulty et al.,
1999) reducing the possibility of successful sintering of the
built part.
A novel system has been proposed and developed to avoid
most of the material preparation steps in a filament-based
system. The new set-up, called mini extruder deposition
(MED) and consisting of a mini-extruder mounted on a high-
.
high efficiency ball screw drive, which offer high
precision positioning system (Figure 2), operates using bulk throughput, efficiency, accuracy ^10mm 4 ^18mm and
material in granulated form. This configuration opens up repeatability (^ 1.3 mm);
.
encoders that offer direct positional feedback of the
opportunities for the use of wider range of materials, making
carriages location; and
the FDM to become a viable alternative manufacturing .
electromagnetic shaft brake on the vertical axis that halts
process for specialty products.
the carriage motion during a power loss.
The three positioning tables are connected to three
System specifications Compumotor Gemini digital servo drives, which allow
torque, velocity and encoder tracking modes. They are also
In order to generally improve the quality of the entire process, equipped with digital notch filters that provide tools to
the new extruder was mounted on three high precision linear eliminate mechanical resonance.
motor tables, the 400XR series from Daedal division of The three drives are thus connected to one 6K4
Parker Inc. Compumotor controller, which communicates to the
The tables are characterized (Daedal, 2000) by: computer through the serial port SR-232. With the use of
.
high strength aluminum body; the 6K4 Compumotor it is possible to write only one program
.
square rails linear bearing, which provide high load to control position, speed, acceleration, deceleration, etc. of
carrying capability, and smooth precise motion; the three axes at the same time (Bellini, 2002).

215
 New developments in fused deposition modeling of ceramics Rapid Prototyping Journal
Anna Bellini, Lauren Shor and Selcuk I. Guceri Volume 11 · Number 4 · 2005 · 214 –220

A separate unit (Figure 2(a)), costumer-built by Shinko sieve. Subsequently, the obtained powder was introduced into
Sellbicn Co in Tokyo, controls the heaters and the motor for the liquefier via a pipe and a funnel to get a constant feeding
the extruding screw (Figure 2(b)). The temperature is rate. However, when such small powder size was used, the
checked by two thermocouples at the entrance of the formation of bigger aggregates frequently occurred in the
liquefier and approximately at 1.5 cm from the exit. The feeding pipe (Figure 2(a)), preventing the incoming of new
thermocouples are connected to an online control system that material. Because of the high ratio surface/volume of the small
regulates the current through the heaters according to the pellets (circa 1 mm in diameter), the heat at the entrance of
required temperatures. Due to the two small copper heaters the liquefier is indeed high enough to melt the particles
and the two thermocouples, the thermal inertia of the system
together, generating bigger aggregates.
is reduced and the temperature of the melt is more uniform
On the other hand, when the extruder was fed with too big
than in the FDM Stratasys machine 1650.
pellets of material (bigger than 5 mm), a non-uniform flow
The rotational speed of the screw can be regulated either
rate was observed. A possible explanation to this phenomenon
manually or through the interface of a digital device. In this
first step of the testing phase of the system, the manual can be the following: because of their size, big granules of
regulation was preferred. However, in the future the controller material can get trapped between the screw and the chamber
will be connected to the same computer used for the of the liquefier preventing the deposition of further material
positioning system. In this way, only one program will be on the substrate. However, due to the movement of the screw,
written for the motion of the platforms and the rotation of the the granule is continuously forced towards the bottom of the
extruding screw, allowing a better correlation between liquefier. When it finally reaches the exit of the nozzle a very
positioning and delivering systems (Bellini and Güçeri, 2004). large flow can be experienced. Moreover, when air is trapped
between two consecutive big aggregates, deposition can
System performance suddenly stop.
It was also observed that the size of granules influences the
In order to test the performance of the new system, different flow behavior by affecting the temperature of the melt at the
samples were built using ECG9/PZT bulk material (McNulty entrance of the liquefier. Since the upper thermocouple is
et al., 1999), developed at Rutgers University for fabrication placed on the external surface of the liquefier, the monitored
of piezoelectric actuators (Allahverdi et al., 2001; value is only an approximation of the temperature of the
Mohammadi et al., 1999).
melt. Consequently when air is trapped between aggregates,
Influence of the size of the granules the temperature of the melt can increase a couple of degrees
During the experimental phase, it was observed that the size even if the outside value is maintained constant. Because of
of granules of material fed into the system has a strong the strong dependency of the viscosity h on the temperature
influence on the melt flow, even if all other variables, such as (see the next section and Figure 3), it can be concluded that
temperature and speed of the motors, are maintained the granules size, affecting the temperature of the melt,
constant. In order to obtain more consistent results, the influences indirectly also the viscosity, hence the flow during
material was granulated and sifted using a 1 mm reticulated deposition.

Figure 3 Temperature dependence of the viscosity of PZT/ECG9, hðT; g_ Þ ¼ h ðg_ Þ · H ðT Þ

216
 New developments in fused deposition modeling of ceramics Rapid Prototyping Journal
Anna Bellini, Lauren Shor and Selcuk I. Guceri Volume 11 · Number 4 · 2005 · 214 –220

Influence of the temperature at the entrance Figure 4 Cross section of free extruded strands
In order to determine the influence of the temperature in the
upper part of the liquefier, experiments were performed
varying the values from 145 to 1608C. The lower boundary
value was decided according to the melting temperature,
which is 1158C, and the temperature dependence of the
viscosity. Since according to Figure 3, at 1158C the viscosity
of ECG9/PZT is approximately 20 times the one at 1358C,
the latest value was considered more acceptable at first.
However, experiments showed that the viscosity of the
material was still too high for the extrusion process when the
temperature was set to any value below 1458C. In these cases
no flow was seen at the exit of the nozzle (0.3 mm in
diameter). For this reason, the lowest temperature used for
successive experiments was set to 1458C.
The upper boundary value of the temperature range was
decided according to the material properties of the binder:
temperature greater than 1658C were discarded because they
could lead to chemical degradation of the polymer. However,
since the thermocouple, which is connected to a close-loop
temperature control system, checks the temperature outside
the cylinder of the liquefier, five extra degrees were added to the stop of the flow. Hence in order to use small nozzle sizes,
the desired temperature of the melt; hence a temperature of the heat convection to the air needs to be reduced. For this
1708C was considered as the upper boundary value. With this reason, in the future, the working area could be enclosed in a
expedient it was possible to take into account the heat controlled environment (see section Influence of the Room
dissipation through the wall. Temperature).
As mentioned above, a T16 nozzle was also mounted on the
Influence of the temperature at the exit mini-extruder deposition system. Because of the observed
The influence of the temperature in the lower part of the
improvements in the continuity of the flow, only this nozzle
liquefier was also tested. Experiments were performed by
was considered for the successive experiments. It was
varying the values from 135 to 1478C. Results showed that
observed that, using this nozzle, samples can be successfully
for temperatures higher than 1458C the material is too
built when consecutive roads are deposited with a positive
“fluidic” after deposition, thus the strand degenerates into a
“gap”, i.e. distance between two consecutive roads (Figure 8
flatter road. For temperatures below 1408C the strength of
and Plates 1 and 2). However, when the gap between the
the interface between two adjacent roads is not sufficient to
roads is reduced, the interference of the heated tip with
ensure satisfactory bonding conditions (see the paragraph
about the influence of the room temperature). In these cases
the delivered material does not adhere to the substrate, but
remains attached to the external surface of the nozzle. Due
Plate 1 Deposition of a single-road wall box
to the temperature of the die, this quantity of material
melts and, either clogs the orifice of the nozzle or glues on
to previously deposited roads, damaging the already built
layers.

Influence of the size and design of the nozzle

In the commercial Stratasys machine four nozzle sizes are
available: T10 (7 mills, or 0.1778 mm, in diameter), T12 (12
mills, or 0.3048 mm, in diameter), T16 (16 mills, or
0.4064 mm, in diameter), T20 (20 mills, or 0.508 mm, in
diameter). In order to be consistent with previous studies, two
nozzles from Stratasys Inc., i.e. the T12 and T16, were tested
with the new system. With the purpose of giving an idea of the
size of the deposited road, a typical cross-section of a free-
extruded strand obtained using a T12 nozzle is shown in
Figure 4.
It was observed that with the T12 the continuity of the flow
is sometime interrupted because of clogging problems. This
phenomenon is probably due to the very high heat dispersion
through the walls of the nozzle and of the adapter (the part
that connect the nozzle to the liquefier chamber). Because of
the reduction of temperature, due to the heat dissipation, the
small volume of the melt present in the tip (the final
cylindrical part at the end of the nozzle) solidifies, causing

217
 New developments in fused deposition modeling of ceramics Rapid Prototyping Journal
Anna Bellini, Lauren Shor and Selcuk I. Guceri Volume 11 · Number 4 · 2005 · 214 –220

Plate 2 Single-road spiral sample can be avoided by an online controlling of the flow during the
building of the sample.

Influence of the room temperature

Self-adhesion of thermoplastic materials, in presence of
prolonged exposure to high temperatures is the basics
mechanism of part strength in FDM. Conformal movement
of polymer chains through interfaces was proposed as the
main mechanism in thermoplastic interface healing by Wool
et al. (1989). According to this study, the degree of interface
healing, defined as the ratio of interface mechanical strength
to bulk material strength, depends on 1/4 power of residence
time t spent above a critical bonding temperature:

si 1
a¼ ¼ K ðT i Þt 4 ð1Þ
sb

The healing function K, dependent on the interface

temperature Ti, has the following form:
8



> Ea 1 1
< AðT c Þ exp 2 ; Ti . Tc
K ðT i Þ ¼ R Tc Ti ð2Þ
>
: 0; Ti , Tc

a previously built road is such to destroy the sample. In order where A and Ea are material constants indicating rate of
to avoid this problem, three different approaches might be healing at critical bond temperature and activation energy for
taken in consideration for future work. healing process, R is the universal gas constant and Tc is the
1 Reducing the width of the tip. However, it should be kept material dependent critical bonding temperature (it can be
in mind that, since the tip function also as “regulator” of consider as the melting temperature).
the fountain flow (Figure 5), it cannot be completely In order to improve bonding between roads (Figure 6) and
removed. layers (Figure 7), the temperature Ti of the interface and/or
2 Increasing the room temperature, in order to improve the the time t (equation (1)) needs to be increased. These two
bonding between the deposited layers. effects can be achieved by only increasing the environmental
3 Controlling the flow during the building process. temperature, thus enclosing the work-volume in a
temperature controlled chamber. In this case, because of the
In this study the extruding screw was manually regulated in reduced heat transfer between the deposited material and the
an ON/OFF fashion. The flow rate was thus kept constant environment, the interface temperature Ti between the roads
during the deposition, even if the speed of the nozzle changed will be maintained higher, and consequently the time t spent
according to the tool-path (Bellini, 2002; Yardimci, 1999). above a critical bonding temperature will be longer.
Using this configuration it was possible to observe an over
deposition of material at the corners of the raster (where the
head accelerates/decelerates). The material deposited in Figure 6 Micrograph of contact between two roads
excess interacted with the nozzle and with previously built
layers, damaging the sample. As mentioned, this problem

Figure 5 Hybrid cross-sectional flow-field

218
 New developments in fused deposition modeling of ceramics Rapid Prototyping Journal
Anna Bellini, Lauren Shor and Selcuk I. Guceri Volume 11 · Number 4 · 2005 · 214 –220

Figure 7 Micrograph of roads in several layers – different layer thickness

Influence of the speed of deposition 2 In order to obtain a smoother and more automatic
Using a T16 nozzle, several experiments were conducted with process, the correct size of the aggregates must be
several speed of deposition, i.e. speed of the head in x and y determined. More experiments thus need to be performed
direction. The cross section of the deposited road can be with granules sizes ranging from 3 to 5 mm in diameter.
indeed reduced increasing the speed of the nozzle. However, This range was decided according to the observations
in this case the acceleration and deceleration needed for the performed on the previous experiments: when pellets of
deposition of a short road are too big. Consequently, without less than 3 mm were used, the formation of agglomerates,
a regulation of the flow rate, the extra deposited material in thus the prevention of new feeding material, could occur;
the corners of the filling raster may cause the bending of the
on the other hand, when pellets bigger than 5 mm were
top surface of the sample (Figure 8(a) and (b)).
introduced in the liquefier, a non-uniform flow rate was
observed.
Conclusions and future directions 3 A closer control of the cross section of the deposited road
The novel system introduced in this paper, consisting of a could be achieved improving the correlation between
mini-extruder mounted on a high-precision positioning velocity of the positioning system and flow of the melt.
system and fed with granulated material, is still under The driving motor of the extruding screw must be
testing, thus significant additional work needs to be done. controlled with the same computer unit and program used
However, the success obtained in building objects of different for the three-axis system.
configurations (Figure 8 and Plates 1 and 2) is encouraging 4 In order to have a better visualization, thus control, of the
for further studies in the use of wider range of materials. deposition process a CCD camera and a boroscope will be
Future directions could be summarized as follows. attached to the depositing head. This will allow a more
1 A better control of the temperature of the melt could be comprehensive understanding of the influence of the
achieved moving the upper thermocouple inside the different parameters.
liquefier. In this configuration, feeding of material will 5 The study of the effect of nozzle size on various material
have less influence on the extruded flow (Section
configurations will give some insights on the road shape
“Influence of the size of the granules”). Furthermore,
that will be possible to obtain.
since the lower thermocouple is acquiring data at 1.5 cm
6 The enclosure of the working platform in a thermally
from the very end of the nozzle, a different mechanism
should be used to determine the temperature at controlled environment will isolate some of the
deposition. A six channels acquisition data system from parameters not related to the process (i.e. moisture) and
Omega will be used for this reason. In order to measure will avoid any contamination of the material.
the temperature and also to understand the cooling 7 The addition of a rotational stage for the extruder will
process in the deposited road, five thermocouples will be allow the use of a square nozzle that will have to be
incorporated on the top of the platform. oriented continuously to match the path of the motion.

219
 New developments in fused deposition modeling of ceramics Rapid Prototyping Journal
Anna Bellini, Lauren Shor and Selcuk I. Guceri Volume 11 · Number 4 · 2005 · 214 –220

Figure 8 Raster sample with zero gap components by fused deposition technique”, Journal of the
European Ceramic Society, Vol. 21, pp. 1485-90.
Bellini, A. (2002), “Fused deposition of ceramics: a
comprehensive experimental, analytical and computational
study of material behavior, fabrication process and equipment
design”, PhD thesis, Department of Mechanical Engineering
and Mechanics, Drexel University, Philadelphia, PA.
Bellini, A. and Güçeri, S.I. (2003), “Mechanical characterization
of parts fabricated using layered manufacturing”, Rapid
Prototyping Journal, Vol. 9 No. 3, pp. 252-64.
Bellini, A. and Güçeri, S.I. (2004), “Liquefier dynamics in
fused deposition”, Journal of Manufacturing Science and
Engineering, Vol. 126, pp. 237-46.
Bellini, A., Güçeri, S.I., Turcu, S., Danforth, S.C. and Safari,
A. (2001), “Nozzle shape, road cross section and space
filling in FDM/SFF techniques”, Conf. Proc. of EUROMAT,
ISBN 88-85298-39-7.
Daedal Catalog (2000), “Electromechanical positioning
system”, Parker Automation, USA.
McNulty, T.F., Shanefield, D.J., Danforth, S.C. and Safari,
A. (1999), “Dispersion of lead zirconate titanate for fused
deposition of ceramics”, Journal of the American Ceramic
Society, Vol. 82 No. 7, pp. 1757-60.
Mohammadi, F., Kholkin, A.L., Jadidian, B. and Safari, A.
(1999), “High-displacement spiral piezoelectric actuators”,
Applied Physics Letters, Vol. 75 No. 16, pp. 2488-90.
Venkataraman, N. (2000), “The process-property-
performance relationships of feedstock material used for
fused deposition of ceramic (FDC)”, PhD thesis,
Department of Ceramic and Materials Engineering,
Rutgers University, New Brunswick, NJ.
Wool, R.P., Yuan, B.L. and McGrarel, O.J. (1989), “Welding
of polymer interfaces”, Polymer Engineering and Science,
Vol. 29 No. 19, pp. 1340-66.
Yardimci, A. (1999), “Process analysis and development for
fused deposition”, PhD thesis, Dept. Mech. Engrg,
University of Illinois, Chicago, IL.

Further reading
Venkataraman, N., Rangarajan, S., Matthewson, M.J.,
Harper, B., Safari, A., Danforth, S.C., Wu, G., Langrana,
N., Güçeri, S. and Yardimci, A. (2000), “Feedstock
References material property – process relationship in fused
Allahverdi, M., Danforth, S.C., Jafari, M. and Safari, A. deposition of ceramics (FDC)”, Rapid Prototyping Journal,
(2001), “Processing of advanced electroceramic Vol. 6 No. 4, pp. 244-52.

220