Content uploaded by Parinya Punpongsanon
Author content
All content in this area was uploaded by Parinya Punpongsanon on Dec 17, 2020
Content may be subject to copyright.
Programmable Filament:
Printed Filaments for Multi-material 3D Printing
Haruki Takahashi
Meiji University
Tokyo, Japan
haruki@meiji.ac.jp
Parinya Punpongsanon
Osaka University
Osaka, Japan
parinya@sys.es.osaka-u.ac.jp
Jeeeun Kim
Texas A&M University
College Station, Texas
jeeeun.kim@tamu.edu
Figure 1. Programmable Filament is a novel 3D printing technique that enables users to 3D print an object with multiple materials using an FDM
printer without any hardware modification. (From left to right) First, users generate a filament that contains multiple materials, to feed into the
extruder, then 3D print an object in full color.
ABSTRACT
From full-color objects to functional capacitive artifacts, 3D
printing multi-materials became essential to broaden the appli-
cation areas of digital fabrication. We present Programmable
Filament, a novel technique that enables multi-material print-
ing using a commodity FDM 3D printer, requiring no hardware
upgrades. Our technique builds upon an existing printing tech-
nique in which multiple filament segments are printed and
spliced into a single threaded filament. We propose an end-to-
end pipeline for 3D printing an object in multi-materials, with
an introduction of the design systems for end-users. Optimized
for low-cost, single-nozzle FDM 3D printers, the system is
built upon our computational analysis and experiments to en-
hance its validity over various printers and materials to design
and produce a programmable filament. Finally, we discuss ap-
plication examples and speculate the future with its potential,
such as custom filament manufacturing on-demand.
Author Keywords
3D printing; fused deposition modeling; programmable
matters; multiple materials.
CCS Concepts
•Human-centered computing →Interactive systems and
tools;
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for components of this work owned by others than ACM
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specific permission and/or a
fee. Request permissions from Permissions@acm.org.
UIST ’20, October 20–23, 2020, Virtual Event, USA
© 2020 Association for Computing Machinery.
ACM ISBN 978-1-4503-7514-6/20/10 ...$15.00.
https://doi.org/10.1145/3379337.3415863
INTRODUCTION
With recent advances in consumer grade 3D printers, end users
without professional skills are able to fabricate various objects
from a smartphone case to custom mechanical tools. Along
with the progress of machinery, there has been a growing
desire for multi-color and multi-material 3D printing using
fused deposition modeling (FDM), not only to get an artifact
aesthetically appealing but also to enable the production of an
object with properties not found in nature, i.e., metamaterials,
even when faced with a limited choice of materials (e.g.,[1, 2,
9]). However, the materials used in low-cost 3D printers are
fairly limited due to the form factor and affordability. Com-
modity FDM machines are therefore often equipped with only
a single, or dual extruder at most. Furthermore, multi-material
printing with a dual extruder is not simple, and incurs many
practical issues such as oozing, stringing, extruder-object col-
lisions, first layer shifting, and an under-extrusion caused by
drizzling, occurring during the switching between extruders.
Despite advanced techniques to tackle these challenges, such
as FDM 3D printers with a built-in ink cartridge [41], inkjet-
based 3D printing [35, 40] and a 3D printer that can seamlessly
switch different materials within a single nozzle [33], these
techniques are generally inaccessible to average usera because
they often require professional skills to manipulate the materi-
als, machines, and processes.
In this work, we introduce Programmable Filament, a novel
3D printing technique that enables users to print an object
with multiple colors and materials, using an FDM without
any hardware updates. Inspired by the idea of 3D printing a
thin wire resembling a filament by Instructables user DasMia
[3], we expand on the concept to fabricate a programmable
filament connecting several segments of various materials into
a single filament based on the user’s specifications (referred to
as printed filament). We demonstrate that the printed filament
can be used in the same way as a conventional filament, i.e.,
extrude through a standard nozzle, requiring no hardware mod-
ification. We further explore a calibration technique used to
minimize the shifting or mixture of the colors and/or materials
when switching between segments. Our approach addresses
challenges in dual printing listed above, as we only use a single
nozzle to print numerous types and colors of filament.
We first review relevant work in multi-material printing and
advanced techniques to extend the capabilities of desktop
printers. We start an introduction of programmable filament
technique with a step-by-step process of printing filaments
and analyze obtained characteristics. We follow up with a
description of experiments conducted to enhance the quality of
the printed filaments. Integrating findings from these empirical
experiments, we present a filament design system for end-
users, inviting them to program a filament based on 3D objects
or images as input to create a ready-to-print filament. We
conclude with a range of application examples demonstrating
our approach, and discussion on the remaining challenges and
possibilities for future research direction.
RELATED WORK
Our work is related to the 3D printing in multiple materials,
application areas, and various techniques used to overcome the
limitations found in FDM. Our technique takes direct G-code
manipulation approach to yield non-standard effects.
Increasing Needs in Multi-material Printing
Since the commodity FDM 3D printer has been introduced to
the consumer market, significant advances have been made
in materials suitable for these machines. As plastic melts at
relatively low temperatures, filament manufacturers started to
vary its properties by simply infusing special pigments into
standard plastic (e.g., PLA), keeping its nature of easy storing
and safety. Single threaded full-color filament (an example
is available at [31]) seemed promising, despite its inability to
locate each color at the selected parts. The ability to make fila-
ments with special pigments (e.g., a graphite, ferromagnetic,
or carbon fiber) inspired researchers in the HCI community
to create functional 3D objects, such as capacitive objects
[4, 29], magnetically encoded wireless devices [10, 11], or
identifiable artifacts with barcode [21]. Exploiting flexible
thermoplastic such as TPU, through the different buckling
behaviors of heterogeneous materials when attached together,
the transformation of a flat sheet into a sophisticated 3D shape
has become available for desktop 3D printers [2]. Use of
water-soluble material with regular plastic further expanded
application scenarios of 3D printed artifacts, such as optically
identifiable object [18], object with wash-away assembly keys
[23] and more. Nonetheless, these approaches have mostly
been possible with advanced hardware, which still presents
numerous practical issues, listed above. Moreover, additional
potential has not been revealed any further, largely due to only
two materials available at a time.
Techniques for Fabricating Multi-color/material Objects
Many hardware-based attempts have been made to tackle the
3D printing of multiple colors and materials. To make full use
of FDM in multi-color 3D printing, practitioners have been
investigating various hardware-oriented approaches, including
four extruder headers [15] and the use of external hardware
to merge multiple materials before fusing them into a single
nozzle for full-color object printing [14]. Song et al. proposed
a modified print header to allow multiple filaments to be ex-
truded through a single nozzle [34]. These multiple colors are
mixed and provide almost full color gamut, according to the
programmed melting ratio. Jubilee is an open-source toolkit
to enable automated tool-changing for material explorations
[39]. For typical home users, researchers also developed a
low-cost technique, by manually changing a filament with
pause-and-resume [20]. However, multi-material printing re-
mains challenging, even with hardware upgrades, the number
of materials are still limited, and users are limited with the
number of extruders they can apply. Although CleanColor
[8] helped resolve the issue of mixed colors appearing in
dual-nozzle printing, users are typically occupied with tedious
calibrations and required maintenance tasks. Alternatively, an
advanced additive-manufacturing technique has been investi-
gated, such as inkjet 3D printing with special materials [24, 30,
38, 32]. Post-processing is another alternative, painting in full
colors, spraying electro-magnetic [44] or color-changeable
paint [12]. However, hardware upgrades prevent non-experts
from being able to afford these solutions, and additional steps
for post-processing makes it hard to apply them in practice.
Direct G-code Manipulation to Enhance FDM Capability
Recent works have begun to use of G-code to directly con-
trol the mechanisms of machinery so to yield non-traditional
effects from FDM. G-code is a series of commands widely ap-
plied to numerically controlled machines (CNC), commands
to control movement axis (G1), speed (F), extrusion amount
(E), and printing temperature (M109) are generated by inter-
preting the user’s choice of slicing parameters. Although a set
of recommended parameters for specific machine types and
materials have been proposed to guarantee the print quality,
adjusting values or command generation mechanisms yields
special effects that are unlikely to achieve through standard
settings. New ways to overcome the existing limitations in the
current form factors of consumer-grade FDM machines [28]
have been explored, e.g., 3D printed hair proposes extending
a stringing phenomenon of extruded materials for fabricat-
ing fine hair-like textures [17]. WirePrint allows diagonal
z-movements of a delta style printer to construct sparse wire-
frame meshes so reduce the printing time [22]. Controlling
the height of the nozzle as well as the amount of material
extrusion, expressive textures such as a fluffy surface can be
created [37]. 3D printed fabric is an additional technique en-
abling the weaving movements of the header, enabling users to
create a flexible fabric out of rigid plastic [36]. Compositional
3D Printing also enables a pause-to-insert action and droop-
ing effects, resulting in the explorative, intervention-oriented
techniques for design-by-print [13]. Dual-color mixing [25]
and 3D hatching [16] use G-codes at low level to express
the tone of 3D prints by fine-tuning color layering. Our tech-
nique also employs a similar approach to print a programmable
filament, and applies the printed filament to obtain a new mate-
rial/coloring capability in FDM. In the remainder of the paper,
we refer to an FDM 3D printer simply as a 3D printer.
PROGRAMMABLE FILAMENT
We first describe an overview of the printing a single threaded
filament printed by splicing multiple segments.
Printing Filament: Process
Our approach is inspired by a craft technique that splices
segments of multiple resource filaments to fabricate a single
resulting filament. By printing a long spiral consisting of
several segments, the resulting filament can be used as regular
material similar to the off-the-shelf filament for 3D printing [3].
We extend this by directly manipulating G-codes to control
3D printer mechanisms and create a programmable filament.
Even without a fixed .stl file of a spiral introduced in this work
with limited control, we can fine-tune the characteristics such
as the thickness and roundness of the edges, the number of
materials to be spliced, adding a special structure to join each
segment by directly manipulating the G-code.
Figure 2 shows an exemplary procedure of a printing filament—
printing a single threaded, three-colored filament in order of
red, green and blue. First, the 3D printer prints a brim that
helps adhere the first layer to the printing bed, and then starts
printing all red segments onto the brim. The endpoints of each
segment are diagonally cut (Figure 2a), to avoid a collision
with the nozzle when printing an adjacent segment for the
second color. After completing the printing, all red segments
in the spiral, the 3D printer pauses and moves its header to the
XY-home of the build area, to be ready for exchanging filament
(Figure 2b). While pausing, the user manually exchanges the
currently plugged filament with a green filament. We control
the printer mechanisms for user-intervention by inserting G-
code commands in the middle to pause the printing (M0),
which notifies the user with a beep sound (M300). The M0
helps pause the printing without cooling the nozzle or bed
while waiting. If the 3D printer has an LCD, it displays a short
message instruction letting users know the right material to
plug in. Resuming the printing process initiates an extrusion
of the green filament, by purging the remaining materials in
the feeder and nozzle. The green segments are then ready,
allowing the next segments to be printed. If the segment
starts from the endpoint of the prior segment, the printer prints
a green part with an upside-down diagonal cut to match the
contact surface of the prior segment (Figure 2c). After printing
all green segments, the 3D printer pauses again, and iterates
the above process (Figure2d). Finally, the 3D printer adds
stitches that patch the joint of the segments on top (Figure
2e). The printed filament is now ready to be plugged in for 3D
printing in the same fashion as an off-the-shelf filament.
Programming Filament: Computations
Our technique makes use of a programmable filament to create
an on-demand filament to be used as a regular off-the-shelf
filament (Figure 3). We envision a unique usage of a printed
filament that empowers users to control the extrusion of the
printed filament according to the printing path. For example,
the height of each stripe differs when printed by the segments
in the same length. With each color-segment in the same
length in the thread, only approximately 10+ layers of a blue
segment in the body can be printed (Figure 3b), whereas it can
print approximately 30+ layers on its face (Figure 3a) due to
Nozzle
Nozzle
Brim
‘pause’ and filament change
Nozzle Brim
avoid collision
Nozzle
Brim
Nozzle
Brim
joint printed filament
Brim
a
b
c
d
e
Figure 2. Printing procedure of a filament: (a) Printing starts with one
color, (b) it pauses upon completion of printing all segments, allowing the
user to change the material. (c-d) The 3D printer prints the remaining
segments avoiding collision with prior segments, (e) then prints stitches
to join adjacent segments.
Figure 3. An exemplary printed filament and object printed using it:
The layers of (a) and (b) are printed in the same segment length (200
mm), but appear with different layer counts.
the discrepancy of the surface area on each layer. Here, we
describe how to compute the relationship between the amount
of material required to print a series of segments from the
resource filament, as well as the length of each segment in the
resulting spiral path. We then describe the calibration process.
Measuring Source Filament Length for Resulting Filament
The amount of extrusion is determined by the
E
parameter
value in millimeter in the G1 command (we refer this as
e-mm
in unit for E-push length, and how long of a filament is fed
during the movement distance that we refer this as
t-mm
in
unit for travel distance). For example, when a 3D printer
receives the command G1 X10 E1 at the origin
(x,y) = (0,0)
,
the header moves to (10,0)(X10) while extruding 1 e-mm of
material (E1). The E value is calculated from the form factors
of each 3D printer as follows:
E[mm] = Wnozzle ×Hlayer ×Lpr inting
Cf ilament
(1)
Assume a nozzle diameter
Wnozzle
of 0.4
mm
(the most com-
mon size), a layer height
Hlayer
of 0.4
mm
, a printing length
Lprinting
of 1
mm
(unit length), and a cross-sectional area of
a filament
Cf ilament
of 2.405
mm2
using a 1.75
mm
diameter
filament. The computed E value is 0.0665
mm
in this case.
By transforming Equation 1 to compute
Lprinting
from the E
value; for the same case, we can print a 15.033
t-mm
long line
(
Lprinting
) on the bed using 1
e-mm
of the resource filament (E).
Using this equation and an input 3D model, we can compute
the length of each segment in a printed filament. If the 3D
model is already sliced into G-code, we only need to add the E
values in the G1 commands. We can then control the amount
of filament to be extruded under the following two conditions:
The 3D printer (1) can extrude a precise amount of material
and (2) can extrude the materials along a pre-generated path
that enables the 3D printer to correctly switch the materials
(see the implementation section).
Calibration for Actual Use of Printed Filament
We explore a calibration technique to meet the first condition
above by calibrating the stepper motor used to feed a filament
into the extruder. Our calibration is based on a Marlin
1
open-
source system, which calculates the rotation distance from the
number of steps where the stepper motor moves to extrude 1
mm
of filament (E1), i.e., in
ste ps/mm
, called an E-step. In
mechanical terms, this value is known by the characteristics of
the 3D printer’s form factors, such as the motor’s step angle,
belt pitch, and number of teeth in each gear. The E-step needs
to be calibrated, allowing the 3D printer to extrude an accurate
amount of material.
We conducted a manual calibration, following a well-known
calibration technique described by E3D in [26]. In short, we
mark the filament to be inserted into the extruder end at 100
and 120
mm
, and extrude a 100
mm
(E100) filament. After
the extrusion, we measure the actual length of the filament
that has been extruded and the remaining amount from the
mark at 100
mm
. If the 3D printer extruded too much passing
100
mm
mark, we can measure it using the mark at 120
mm
.
We then compute a new E-step value based on the extrusion
amount,
Enew =Ecurrent ×100/Lextruded
, and so overwrite the
firmware setting with this value.
Printed Filament: Analysis
Here, we describe the characteristics of the printed filaments
and associated constraints that need to consider when used for
3D printing.
#1 Maximum Length
: The maximum length of a printed
filament is highly dependent on the build area given the type
of 3D printer. With a
300×300 mm
bed size, a printed filament
in a spiral shape can be approximately 20-m long. The inner-
most radius of the spiral should be sufficiently large to prevent
breakage while plugging it into the feeder, upon completion
of the printing filament. We set it to 30
mm
by referring to a
commercial spool size.
#2 Density
: Similar to the existing filament splicing technique
[3], the cross-sectional area of our printed filament is hexago-
nal (Figure 4a). Therefore, when this filament is used in actual
3D printing, the amount of material to be extruded may not
be the optimum value which is measured for a full cylindrical
shape (dotted lines). If an under- or over-extrusion occurs
when using it as a regular filament, the flow rate needs to be
calibrated [27].
#3 Fabrication Cost
: There are three types of fabrication
costs for a printed filament: the printing time, amount of
1Marlin Firmware (https://marlinfw.org/)
1.75mm
abb c
Figure 4. (a) The cross-sectional area of printed filaments is hexagonal
and (b, c) the endpoints of each segment are cut by a wavy-cutting plane.
material, and manual labor. The costs differ according to
the length and number of materials used. As an example, a
1-m long printed filament that consists of 20 segments (50
mm
per segment) in 3 types of materials takes approximately
36
min.
to print (excluding the time required for material re-
placement). Note that we empirically set our printing speed to
1,200
mm/min.
and the movement speed to 9,000
mm/min.
To
change the materials, we estimate that it takes approximately
3–5
min.
from the removal of a material from the nozzle to in-
serting the next material, purging, and resuming to the original
position. Because we fabricate using resource filaments, the
amount of material required to print a filament is almost equal
to the sum of their lengths. The purging requires an extra
amount of material to replace the filament without residue,
and we currently set a 50
mm
filament for purging. Finally,
some manual effort is required for removing and inserting a
material. Because the 3D printer informs the user each time
the material is changed, the number of user interventions for
an exchange is equal to the number of materials minus 1 (i.e.,
a single thread consisting of 5 colors requires 4 exchanges).
#4 Connectivity between Segments
: The endpoints of one
segment in contact with the adjacent segments are diagonally
cut with a wavy shaped surface to increase the contact area
(Figure 4b, c). An additional layer that covers two materials
across is printed on top and on the bottom to enhance the
binding (we refer to this layer as a stitch). The stitch itself
and heat added by having the hot nozzle passes along when
printing the stitch on top contributes to strengthen the adhesion
of two heterogeneous segments. We confirmed that various
colored PLAs can adhere to each other tightly. However, each
material shows different characteristics (e.g., viscosity and
stiffness), and the binding strength between two materials may
differ. We will further extend the experiment to validate the
effects, particularly with various material combinations such
as TPU + PLA in the following section.
PROPERTIES OF PRINTED FILAMENT
We conducted several experiments to explore the new charac-
teristics occurred to the printed filaments, validity across vari-
ous machines and material types, and developed techniques to
mitigate any negative effects.
Validation on Various Machines for Replicability
One main goal of this work is to overcome the challenges in
multi-color/material printing that generally occur when using
a consumer grade FDM 3D printer, particularly with a sin-
gle nozzle. Hence, we conducted several experiments using
various single-nozzle platforms. We also extended our exper-
iments to verify the printability over numerous variations of
FDM. All tests were iterated using four different 3D printers,
namely, a Creality CR-10S, Prusa MK teacup, Creality Ender
3, and NWA3D. These four printers come with similar config-
urations but different gear setups. We also tested a Geeetech
A20M (with a single nozzle and dual extruders) and a Creality
CR-10S Pro (with dual gear extruders).
Validation on Material Adhesion for Generalizability
We examine whether our approach applies to various material
combinations, such as PLA, ABS, TPU (flexible material),
conductive PLA, nylon, and PVA (water-soluble filament),
obtained from several brands for generalizability. We pur-
posefully chose these sets of materials to consider their future
possibilities in fabricating functional objects using produced
programmable filaments. As previously identified by Mosaic
Palette, a hardware attachment manufacturer of splicing mate-
rials [14], various combinations of materials affect the quality
of resulting filament. Here, we describe our results in gen-
erating a short filament consisting of two types of material
segments and present the empirical findings.
We first confirmed that all materials can form a filament shape
in a spiral when printed. It is clearly harder to connect two
heterogeneous materials than the same material in different
colors. To ensure stability, it is necessary to adjust the print-
ing parameters (printing temperature, flow rate, and printing
speed) when switching between two types of materials, e.g.,
ABS from PLA, following the recommended settings from the
manufacturers. For instance, to print TPU segments after com-
pleting PLA segments, the printing temperature of the nozzle
must be increased, and slower printing speed is recommended.
The type of combination and adhesion strength also affect the
feeding of the printed filament when the resulting filament is
used. As the torsional force is applied to the printed filament
when fed, caution must be used to determine the insertion
direction. Furthermore, we found there are various extrusion
conditions in the printed filament. For example, for a printed
filament consisting of TPU and PLA, there are four combina-
tions of which material pushes out which one, and the TPU
is too soft to push a PLA segment toward the nozzle. Even if
we confirm that all materials listed above can be used to cre-
ate a printed filament in any combinations, certain joints (but
not all) will break if bent harshly or when removed from the
bed owing to their original different bending and strengthen-
ing nature and loose binding. The stitch structure sufficiently
compensates this fragility, as detailed in the following section.
Stitches to Enhance Binding Strength of Joints
As mentioned earlier, we added a stitch to connect each seg-
ment and enhance the binding strength between multiple ma-
terials. Creating a strong binding is significant to guarantee a
seamless connection of each resource material segments, (1)
making it easy to remove a printed filament from the print
bed without breaking it, and (2) smoothly feed it to the nozzle
when used for printing an object. We set the stitch length to 7
mm
, to cover the top of both adjacent segments. We consider
two types of stitch: (i) linear and (ii) zigzag. We tested using
three materials in various combinations: colored-PLA, PVA,
and conductive-PLA connected with a colored-PLA stitch.
Figure 5 shows a comparison between two types of stitch
abc
1 mm 1 mm1 mm
Figure 5. Comparison of different stitches to join two materials. The
printed filament (a) without a stitch, and with (b) a line stitch and (c) a
zigzag stitch.
a
b
c
d
printing direction
Figure 6. Printed sheets with gradation. In (a, c), the edges takes more
lines to turn into the pure white (close to d), because the color residue
remains in the nozzle. In comparison, (b) brighter color quickly switches
to darker color.
in the resulting prints. We found that both stitches greatly
enhance the connection, where the zigzag stitch presents a
stronger link (i.e., Linear force presents 2N, 18.3N, 29.6N at
breakage for PLA-PLA links, respectively).
Gradation Behavior
We observed a color gradation when switching between two
segments takes place in the feeder. Technically, it is possible
to mix two materials and obtain properties as a composite
material. For example, at the point at which a material switch
occurs, we can imagine that PLA mixes with TPU and might
present properties similar to those of Nylon. However, spe-
cial material properties such as flexibility, conductivity, and
solvability cannot be easily observed because of external con-
founding factors. Thus, we only focus on the color gradation
behavior from a standard PLA material in this experiment.
We first created a G-code to print a multi-colored filament in
special dimensions within a flat sheet, as shown in Figure 6,
to clearly observe a linear gradation. We specified the dimen-
sions: a layer height to be 0.40
mm
in a single-layered sheet
and a width to be 30.07
t-mm
, where each line in the sheet re-
quires approximately 2.00
e-mm
of filament to be printed. We
arranged 150 horizontal lines in parallel to form a square sheet
vertically. The width of each line is 0.48
mm
(i.e., from calibra-
tion,
Wnozzle ×120%
considering the occurrence of swelling
[27]), which is sufficiently sparse to visually examine each line
using a microscope. We prepared a printed filament consisting
of transparent PLA and color-PLA materials to print this sheet.
The transparent PLA is used for the brim and stitch, and the
color-PLA materials form each segment (i.e., a striped pattern
printed on the surface of the sheet).
As shown in Figure 6, the lighter material is easily overridden
by switching to a darker material, while the opposite requires
more steps to remove any residue. This can be easily observed
in the white—black and white—red combinations. Figure 6a
(black to white) and c (red to white) show that the edge of each
line has difficulty turning into a pure white as shown in Figure
6d, owing to the residue color in the nozzle. Figure 6b (white
to black) shows that color changing can happen in just one
horizontal line. Counting the lines consisting of two colors
allows us to find that approximately 5–12 lines are needed to
clearly change two materials, depending on the color mixture.
Because each line is printed with approximately 2.00
t-mm
of
filament, this means approximately 10–24
e-mm
of filament is
needed to change the colors without a gradation.
This length is a significant factor to be considered in imple-
menting an end-to-end design system to facilitate the use of
a printed filament. For example, if a user wants to hide the
gradation effects into the infill region in the printed object, we
can increase the infill density or the wall thickness to spend
the mixed material parts. If the scale of a target 3D model
does not allow this transition, a separate prime tower needs to
be printed to budget the space, similar to the process used in
existing dual color printing.
SYSTEM
Integrating the empirical results found above, we implemented
the design systems to support an end-to-end pipeline. Our
system consists of two functional components, namely, fila-
ment design software to obtain the printed filaments, and three
different modeling software with various input types (i) G-
code, (ii) 3D model designed for dual printing, (iii) 2D image
to obtain the parameters to be used as input in the filament
design software. These software were implemented to pro-
vide various options for end-users, utilizing our programmable
filament technique in different ways to offer unique design
space, using Rhinoceros and Grasshopper, and the GUI using
HumanUI library.
Filament Design Software
The purpose of this tool is to generate G-codes for printing
a filament, taking a list of materials and the lengths of each
segment as user input, as well as a list of material-specific
slicing parameters corresponding to each segment.
User Interface and Workflow
The user interface consists of three sections as shown in Fig-
ure 7: a list of segment lengths representing each material in
sequence, geometric specifications used to construct a spiral,
and material-specific settings (e.g., print temperature, print
speed). First, a user inputs the length of each segment to be
spliced in order, in the first input box. Each time the user
adds a new material in a row, a new drop-down menu speci-
fying the material-specific slicing parameters is added to the
material settings section. These segments constitute a spiral
in a user-specified dimension. A spiral requires the center
position, number of turns, and inner/outer radii. Technically,
it is possible to generate a longer spiral than the total length
of the segments; however, if the bed of the 3D printer is too
small, it is necessary to adjust the specified radii accordingly.
Then the user selects per-material settings from the system-
generated drop-down menus. For example, if the user inputs
material =0,1,2
into the top input box, the software creates
three drop-down lists in the material settings section at the
bottom. Although the default settings show the recommended
Figure 7. Overview of the filament design software with interface win-
dow. Users design a printable filament using the input boxes (right), and
the resultant filament can be previewed (left).
10 mm 10 mm 10 mm length of the bowden tube
plate to split a spiral
0 1 2 0 material
number
joint
joint
brim
printing path
a
b
c
Figure 8. Process of the filament generation, with example input of
length =10,10,10 and mat erial =0,1,2. The software generates a spiral
then (a) labels each segment, (b) splits the spiral into segments using
cutting-planes, and (c) generates a G-code of the filament for each layer.
value of each material for best practice, the user can further
modify the settings if necessary (e.g., increase the printing
temperature). Finally, the system generates a ready-to-print
G-code file by clicking the “Generate” button, which we detail
the internal mechanism in the following section.
G-code generation Mechanism
We walkthrough the G-code generation process with the ex-
ample shown in Figure 8, with lists of
length =10,10,10
and
material =0,1,2
as user input. The system generates a spiral
given the geometric dimensions, and each segment is labeled
according to the list of materials specified (Figure 8a). Note
that the last segment is equal to the length of a Bowden tube
to ease filament feeding, and is labeled to ’0’ matching with
the very first segment. In our setting, the length of the spiral is
50
mm
longer (the length of our Bowden tube) than the sum
of each segment length.
Next, the system generates a wavy slicing-plane to split the
spiral into segments (Figure 8b). The direction of the tilt of
each slicing-plane is determined by each segment’s printing
order, based on the material label. In the example, the first and
final segments (labeled 0) are printed first, followed by the
printing of segment 1. Therefore, a slicing-plane between seg-
ments 0 and 1 tilts toward segment 0, securing the printability.
In the same way, a slicing-plane between segments 1 and 2
tilts toward segment 1, and as a result, segment 1 has two dif-
ferent types of cut at each end towards segment 0 and 2. Using
the complete set of these slicing-planes, the system splits the
spiral into a series of segments. To obtain a printing path of
each segment, the system calculates an intersection between
each segment and XY-planes. The XY-planes lined up in the
z-direction with intervals of the layer height (we empirically
set the layer height to 0.16
mm
), and each intersection makes
the contour of the segments.
Finally, the software generates the G-code from the printing
path. A series of commands that will be used to print each
segment in this G-code are placed in order of the label, and
when swapped to the next label, a pause command is inserted
in between: moving to the XY-home position, unloading the
current material, beeping to alert (M300), and pausing (M0).
Commands to load and purge a new material (about 50-mm)
follow. A command to print a short line at the corner to
prime the nozzle is added, assuming resumed by the user after
manual exchange of the source filament. When generating the
G-code, we add brims around the first layer and stitches on
the top and bottom of every segment across the entire spiral,
to seamlessly join them (Figure 8c).
Material Settings
Each material comes with unique parameters: (i) type, (ii) noz-
zle temperature [
◦
C], (iii) bed temperature [
◦
C], (iv) printing
speed [mm/min], and (v) flow rate [%]. Filament type (i) is
referred to as an argument to call the pause command (M0).
The LCD display of the 3D printer can show a custom message
(e.g., ’Next: PLA’) for the user’s preference. Here, the tem-
peratures are adjusted to fit the optimal conditions based on
the properties of each material switched at a specific moment
using the (ii) M109 command for the nozzle with the obtained
parameters and (iii) M190 for the bed. The printing speed (iv)
and flow rate (v) are accommodated using the G1 commands,
with the F value applied to replace the printing speed, and
using the E value multiplied by the given flow rate.
Modeling software #1: Reinterpreting G-code in Stripes
It is hard for users to know what specifications are required
to print a needed filament. To support users who want to start
with an input 3D model and are already familiar with existing
slicing software, we provide a post-processor that generates
a filament from pre-generated G-gode using Cura (Figure 9).
As an example of post-process, this software allows users
apply a stripe pattern to the G-code. First, a user imports a
G-code file generated using Cura into the software, but any
slicers can be used. We set the extrusion mode to relative in
the slicer to easily calculate the amount of required material
by accumulating the E values. The software interface shows
a slider to add the material information to the input G-code.
The left end of the slider indicates the bottom of the 3D model,
and the right end corresponds to the top. The user can add
anchor points to the slider by clicking on a blank space and
define a color. By sliding the anchor points, a striped pattern
in various lengths can be designed. Using the information
in the slider, the software computes the amount of material
needed to print segments in selected colors, and generates a
list of parameters (segment lengths and material labels). After
designing the striped pattern, users copy and paste the lists of
parameters into the input box in the filament design interface
above to obtain a G-code to print a filament.
Figure 9. The modeling software #1 generates a list of required param-
eters to input into the filament design software, using an input G-code
with slider to add colors.
Figure 10. Cura slicer with our virtual 3D printer profile. Our settings
define multiple extruders to generate a G-code with various materials
dedicated by each extruder.
Modeling software #2: Slicing with Virtual Print Headers
Often, users want to print a 3D model originally designed for
dual-color printing, easily obtainable from numerous free on-
line repositories (e.g., Thingiverse and Yeggie). To support this
scenario, we provide software enabling users to start with 3D
models to specify the parameters required to attain a dedicated
filament. First, we create a 3D printer profile (i.e., setting the
building volume, extruder count, etc.) in Cura. This enables us
to define a virtual 3D printer with multiple extruders account
for each material spool assigned to them. For example, if a
user has a 3D model that requires three different materials but
owns a single-nozzle printer, the user can begin with the Cura
profile using a virtual 3D printer with three headers. Note that
we confirmed that Cura can add up to ten or more extruders
regardless of the actual physical printer corresponding to this
setting. Here, the user imports a watch model consisting of
three parts; flexible bands, rigid joints, and a rigid display in
three colors (Figure 10), and merged into one for alignment.
Each part is assigned to each header with unique material set-
tings. Slicing the model under this setting spits out a G-code
that assumes it will be used in a physical tri-header printer,
with the materials plugged accordingly.
The modeling software imports this G-code and calculates
the amount of material needed in each extruder. This can be
easily achieved by reading G-code commands T0, T1, ... Tn
(where
n
indicates the extruder number) to switch between
materials. The software captures them while calculating the
amount of each material that will be used to 3D print the model
and generates the lists of length and material settings to be
used in the filament design software. In addition, to reuse this
G-code in actual 3D printing the input 3D model using the
printed filament, the system modifies the original G-code by
removing Tn commands. The system yet leaves other material-
specific commands such as temperature adjustment, as this
abc
Figure 11. Users can use existing CAD tools (a) to select a 3D model
in parts by brushing the surface, (b) automatically offsetting to segment
into discrete volumes to export, (c) then import parts into Cura to obtain
a list of parameters needed to print a filament.
is still required in making use of the printed multi-material
filament in 3D printing. For example, if a single nozzle is
fed with a TPU segment from the end of a PLA segment, the
machine needs to wait until heating the nozzle for flexible
materials (200 to 230+
◦
C). For 3D models needs support in
other materials such as PVA, users can also assign one virtual
header dedicated for supporting material only.
If a user does not own a 3D model with parts segmented
for multi-material printing, several existing CAD tools (e.g.,
MeshMixer) provide a step-by-step guideline to segment one
solid mesh into parts by offsetting [43], as shown in Figure 11.
Modeling software #3: Path-planning using a 2D Image
In addition to software #2 that use existing 3D models, we
further introduce software that takes a delicate 2D image as
input then produce an appropriate printing path for building a
facade. Here, we provide a software that supports converting
2D image to plan 3D-printing paths, by dividing discrete colors
to appear exterior walls, while hiding gradation behavior inside
of them (Figure 12). It does not require a prime tower outward,
as we use this volume in mixed colors to print the so-called
infill. With this structure, the system generates a printing path
having a color change point inside the object such that the
gradient effect does not appear outward. G-code generated by
this software consists of thick walls with several layers in the
cross sections. First, the system generates the walls and divides
the outermost path on the object’s facade according to an input
image as shown in Figure 12. In this example, the outermost
path is divided into three segments according to the number
of colors appeared in the input image. After generating a path
with the first material (red), the system connects the path to an
inner wall to switch the materials (red to green). The length of
the inner path is determined based on the expected gradation
effect of two materials. Empirically, we set the default length
to use a 20
mm
material. When switching is complete to the
next color, the path comes back to the facade (green), and
Figure 12. Our modeling software #3 generates a unique path to utilize
the inner structure of the 3D model and print the transition of the switch-
ing materials (center, right). Therefore, the pattern generated using an
input image can appear in front of the wall (left).
Figure 13. Example prints using our system: (a), (d) are created by
software #1, (b), (e) and (f) are created by software #2 with insets showing
original input 3D models, and (c), (g) are printed by software #3.
the process iterates (green to red). By repeating this process
according to the number of materials in a layer, a clear pattern
appears on the facade without mixed colors.
Table 1 summarizes the unique design space, which enables
users to choose from different input and explore design param-
eters to create multi-color/material.
Software Input Design space
#1 Gcode Height of vertical segment
#2 3D model Number of materials (head count)
#3 2D image Facade dimension
Table 1. Three design software provide unique design benefit for end
users, offering various options to choose from different input to design
parameters in different design space.
VALIDATION WITH APPLICATION EXAMPLES
To validate the system’s general use cases using various mod-
eling methods (software) and materials throughout our end-
to-end pipeline, we created a number of example applications.
Figure 13 showcases these examples with the details summa-
rized in Table 2. We replicated existing 3D models created for
multi-material printing using a dual extruder, such as a popular
3D model used to test dual nozzle printing (Figure 13a), and
an application of prior work (Figure 13b) using our pipeline.
Figures 13d and 13g show that our technique is capable of
printing more than five different materials, which is hard to
achieve even with hardware upgrades— as the commercial,
the latest multi-extruder machine comes with four extruders at
most. Our technique also supports 3D models designed using
common 3D modeling tools, such as CraftML [42] (Figure
13b), which makes it easy to design modular parts and export
them using selective tags similar to jQuery, ThinkerCAD (Fig-
ure 13e), as well as MeshMixer (Figure 13f) that help users
20 min, 2 m
22 min, 2.1 m
shipped
printed filament
42 min
22 min
Figure 14. Upon the user’s order of a 3D model with the desired materials, a custom filament can be manufactured on-demand and shipped to the user.
Name Print time
(filament)
Print time*
(object) Materials
Boat (Figure 1) 5h:4m:32s 1h:50m:17s 5 PLAs
Traffic cone
(Figure 13a) 1h:12m:54s 51m:27s 2 PLAs
Alligator
(Figure 13b) 1h:27m:53s 55m:7s 2 PLAs
UIST2020 logo
(Figure 13c) 5h:5m:43s 2h:44m:55s 2 PLAs
Chessman
(Figure 13d) 1h:33m:3s 58m:56s 4 PLAs
TPU
Watch
(Figure 13e) 3h:49m:1s 1h:48:17s 2 PLAs
TPU
Cat (Figure 13f) 6h:5m:37s 2h:16m:28s 4 PLAs
Cellphone stands
(Figure 13g) 2h:18m:26s 1h:28m:52s 5 PLAs
* We decreased printing speed to between 50-80% from the actual printing
time (i.e., at 100% printing speed) to steady the printing process.
Table 2. Details of printed examples (the printing time was estimated
from Repetier-host software).
align parts and export in place to merge for slicing. Figure
13c demonstrates sample prints made using the software #3
for printing without an external prime tower. Purged materials
during the color transition are used to print inside (replacing
infill) the 3D printed object, reducing the printing time and
material cost.
DISCUSSION & OUTLOOK
Post-processing vs. Pre-processing
Our technique flips the process to obtain a full color 3D printed
object using a commodity 3D printer. Users often take a post-
processing approach, cautiously brushing and painting the
3D printed result. By contrast, we propose a pre-processing
approach, preparing a filament for multi-material printing in
advance, and leaving the printer to take over. Printing a single
threaded, multi-segmented filament still requires manually
exchanging the resource filaments when switching the color
and/or material. However, this process significantly reduces
the number of manual filament exchanges required during
conventional printing when using a single extruder printer
[20]. The current layer-by-layer printing nature of an FDM 3D
printer requires lots of numbers of exchange of materials even
in one layer. For example, the boat shown in Figure 1 is printed
using five different materials, requires only four exchanges
(equal to the number of materials minus 1); however, if printed
using a regular approach by manually exchanging the material
every time a layer meets the new color segment, 211 exchanges
are required, which is the number of extruder exchanges (Tn)
appear in G-code—not ideal in practice.
On-demand Filament Manufacturing
Utilizing a pre-processing approach may, therefore, empower
manufacturers in the future to produce a filament on-demand,
opening a door for customizable materials. For example, upon
a user’s request to customize filament to print a 3D model
using various materials, this on-demand filament that embeds
the required filament properties, such as colors and location
where these colors appear, can be ordered, manufactured, and
finally shipped, as shown in Figure 14. Currently, materials are
frequently purchased in bulk, from a single color batch, mak-
ing it difficult for individuals to be equipped with full ranges
of colors without purchasing 20+ filament spools for example.
Our technique would unshackle users from budget issues in
testing various colors and materials at low-investment. Our
systems could serve as a first-class interface for the ordering
system. In this future filament supply-chain, the customer and
manufacturer can also closely work together, empowering the
potential for manufacturers to become aware of the emerging
needs in producing new materials for mass manufacturing.
Sharing of Data along with the 3D Model
The above speculative future is based on the possibility that
users can share data describing the material requirements, as
well as the 3D model. Using our pipeline, users can easily
create a G-code to produce a special filament dedicated to
printing a desired 3D model in multiple parts. In addition,
once the user obtains the necessary input into the filament
design system as the parameters (e.g., the type of materials,
sequence, and length) that are determined by utilizing any of
our modeling software (#1-#3), the user can simply share this
model-specific data with others who can input the data into
the filament design system, to replicate the filament with some
variations in color or materials if needed.
Programming of Properties and Alternative Production
As introduced earlier, our technique also has the potential to
produce materials with new characteristics, enabled by its flex-
ibility in blending the materials. Current way of concatenating
materials segments by segments could be laborious. Although
mass-manufacturers in our visionary scenario can do a bet-
ter job at concatenating, an FDM technique presents unique,
wider opportunities in programmability, as presented in Fig-
ure 15. Our experiments demonstrated that a 3D printer can
extrude two or more materials at the same time while mixing
them, allowing the extruded material to become a new mate-
rial. As shown in Figure 15 a-c, two materials are synthesized
into a printed filament in various color ratios at each segment,
exhibiting a gradual color gradation in the printed 3D object.
Applying diverse composition methods of each segment, e.g.,
enclosing with another material (Figure 15c), yields new fil-
ament properties such as strength [19]. Similarly, the same
Figure 15. Various filaments printing techniques producing new prop-
erties. Filaments can be printed with (a) tiny segments, (b) a segment in
which one material is layered on the other in varying portions resulting
in gradient colors, and (c) a white filament with a red material inside.
phenomenon can emerge for different material combinations
such as PLA + TPU [5] or PLA + Polycarbonate [6], if the
material is mixed with a controllable portion of two materi-
als to reproduce a programmable stiffness. The next step of
this work includes exploring composite material production
techniques in filament generation using an FDM technique,
developing a prediction model to obtain programmable me-
chanical structures computed by mixture ratio upon quantified
strength, then can be interpreted into Gcode for production.
LIMITATIONS & FUTURE WORK
Extra Hardware Configuration for Some Printers
During the process of printing a filament, we added z-hop by
adding G1 Z commands to lift, to avoid a collision between the
nozzle and the printed segments along the movement pathway,
then lower the nozzle back to z-0 to print the next segment
on the bed (see Figure 2b). However, some 3D printers have
extra hardware that touch and drag the printed segment in the
next spiral inward/outward even with this collision-proof tech-
nique, such as a magnetic end-stop to detect the proximity of
a print-bed (e.g., in Printrbot and Prusa Teacup) or a cooling
fan around the nozzle. In such cases, a special calibration tech-
nique needs to be applied, i.e., removing a fan or physically
adjusting the end-stop height. When the end-stop is adjusted
to be slightly above the true home (Z0), an offsetting distance
must be manually found to start the printing at lower than
where it homed. This offset needs to be added to the default
setting (e.g., adding G1 Z-4 as z-offset in Repetier-host).
Manual Works and Metrics for Success Measure
Our technique relies on the initial calibration process of each
3D printer. However, it is possible that residue of the previous
segment is not completely purged and thus remains in the
nozzle, particularly when printing with an extraordinarily long
filament. Because E-steps can be manipulated only up to the
first decimal place in the software setting, external factors such
as the stiffness of the material, the temperature of the heater,
and the stepper motor form factor, cannot be perfectly con-
trolled via calibration. Empirically, we found that the amount
of extrusion may vary during each printing process in our ex-
periments. Further, common printing failures might also occur,
such as clogging of the extruder or an under/over-extrusion,
resulting in a shifted location where the actual material was
supposed to appear. Figure 16(left) shows some samples with
errors, such as a pattern shifted out of alignment, where the
color did not correctly appear. Figure 16(right) also shows a
red color mixed into a yellow color owing to the insufficient
material cleaning. To solve these calibration and extrusion
problems, we currently inspect the whole printing process and
adjust the amount of extrusion when needed. Although the
Figure 16. Failure cases of our method. Left: The color is shifted
from a shift in the step motor feeding the material into the nozzle. Right:
The red color is mixed into yellow owing to insufficient material cleaning
during the printing process.
performances of machines are expected to be continuously
improved in the future, a monitoring system can significantly
help confirm the state of extrusion and which material is being
extruded, such as when using an OctoPrint [7]. Along the
way, we will need to develop user-centered metrics to measure
success qualitatively, as it is possible that users feel frustrated
due to the higher chance of clogged nozzle or not satisfied
with quality of final prints.
Validating Mechanical Properties and Material Range
From the experiment, we integrated empirical results into the
system implementation with the best choice (e.g., zigzag stitch
to blend two segments), but we have not gone through a full
structural analysis for material composition and types. The
nearest next step that will be followed by investigating pro-
grammable properties is conducting a holistic experiment to
understand the mechanical behaviors of composite structure.
For example, we will print segments (a) with a more variety
of stitches (b) in various material compositions at each end
and (c) in different lengths and the cross-sectional shapes (e.g.,
wavy cut and linear cut in varying angles), then test strength
with two measures: (i) linear fracture by pulling two ends
from each other in increasing forces, (ii) axial bending and
breakage by adding growing weights at the seam. We expect
to obtain the maximum strength of each material joints, and
optimize structures upon input parameters of above variables
that minimize laborious work needed in producing printed fila-
ments. Further, we observed bigger differences in mechanical
properties on each end result in easier breakage (e.g., TPU-
PLA) in our initial experiment, thus, structural analysis of
printed filaments with various materials will help us define
appropriate range of materials to join, as well as recommend a
better stitching strategy in each case.
CONCLUSION
In this paper, we proposed a programmable filament, a novel
technique to 3D print multi-material objects using a low-cost
FDM machine without any hardware modifications. We con-
ducted a set of experiments to understand characteristics of the
printed filament, then integrated these findings into designing
an end-to-end fabrication pipeline. In this pipeline, we provide
parametric design interfaces for end-users, used to design the
filament directly, or by obtaining the specifications of a print-
able filament from input 2D images, G-code, or 3D models.
We validated our approach is replicable with various types
of machines and materials, then showcased printed examples.
Finally, we discussed remaining technical challenges and a
potential future achievable by the proposed technique.
ACKNOWLEDGMENTS
We thank Instructables user DasMia for the inspiration of this
work, external reviewers for providing constructive feedback.
This work was partially supported by Adobe Gift Award.
REFERENCES
[1] Davide Jose Nogueira Amorim, Troy Nachtigall, and
Miguel Bruns Alonso. 2019. Exploring Mechanical
Meta-Material Structures through Personalised Shoe
Sole Design. In Proceedings of the ACM Symposium on
Computational Fabrication (SCF ’19). Association for
Computing Machinery, New York, NY, USA, Article
Article 2, 8 pages. DOI:
http://dx.doi.org/10.1145/3328939.3329001
[2] Byoungkwon An, Ye Tao, Jianzhe Gu, Tingyu Cheng,
Xiang “Anthony” Chen, Xiaoxiao Zhang, Wei Zhao,
Youngwook Do, Shigeo Takahashi, Hsiang-Yun Wu,
Teng Zhang, and Lining Yao. 2018. Thermorph:
Democratizing 4D Printing of Self-Folding Materials
and Interfaces. In Proceedings of the 2018 CHI
Conference on Human Factors in Computing Systems
(CHI ’18). Association for Computing Machinery, New
York, NY, USA, Article Paper 260, 12 pages. DOI:
http://dx.doi.org/10.1145/3173574.3173834
[3] DasMia. 2019. DIY Rainbow 3D Printing Filament.
(2019). https://www.instructables.com/id/
DIY-Rainbow- 3D-Printing- Filament/
[4] Daniel Groeger, Martin Feick, Anusha Withana, and
Jürgen Steimle. 2019. Tactlets: Adding Tactile Feedback
to 3D Objects Using Custom Printed Controls. In
Proceedings of the 32nd Annual ACM Symposium on
User Interface Software and Technology (UIST ’19).
Association for Computing Machinery, New York, NY,
USA, 923–936. DOI:
http://dx.doi.org/10.1145/3332165.3347937
[5] Yi Guo, Lei Yan, Zhu Zeng, Lin Chen, Mingxue Ma,
Rui Luo, Jun Bian, Hailan Lin, and Daiqiang Chen.
2019. TPU/PLA nanocomposites with improved
mechanical and shape memory properties fabricated via
phase morphology control and incorporation of
multi-walled carbon nanotubes nanofillers. Polymer
Engineering & Science (2019). DOI:
http://dx.doi.org/10.1002/pen.25365
[6]
Kevin R. Hart, Ryan M. Dunn, and Eric D. Wetzel. 2020.
Tough, Additively Manufactured Structures Fabricated
with Dual-Thermoplastic Filaments. Advanced
Engineering Materials 22, 4 (2020), 2070013. DOI:
http://dx.doi.org/10.1002/adem.202070013
[7] Gina Häußge. 2020. OctoPrint. (2020).
https://octoprint.org/
[8] Jean Hergel and Sylvain Lefebvre. 2014. Clean Color:
Improving Multi-Filament 3D Prints. Comput. Graph.
Forum 33, 2 (May 2014), 469–478. DOI:
http://dx.doi.org/10.1111/cgf.12318
[9] Alexandra Ion, Robert Kovacs, Oliver S. Schneider,
Pedro Lopes, and Patrick Baudisch. 2018. Metamaterial
Textures. In Proceedings of the 2018 CHI Conference on
Human Factors in Computing Systems (CHI ’18).
Association for Computing Machinery, New York, NY,
USA, Article Paper 336, 12 pages. DOI:
http://dx.doi.org/10.1145/3173574.3173910
[10] Vikram Iyer, Justin Chan, Ian Culhane, Jennifer
Mankoff, and Shyamnath Gollakota. 2018. Wireless
Analytics for 3D Printed Objects. In Proceedings of the
31st Annual ACM Symposium on User Interface
Software and Technology (UIST ’18). Association for
Computing Machinery, New York, NY, USA, 141–152.
DOI:http://dx.doi.org/10.1145/3242587.3242639
[11] Vikram Iyer, Justin Chan, and Shyamnath Gollakota.
2017. 3D Printing Wireless Connected Objects. ACM
Trans. Graph. 36, 6, Article Article 242 (Nov. 2017), 13
pages.
DOI:http://dx.doi.org/10.1145/3130800.3130822
[12] Yuhua Jin, Isabel Qamar, Michael Wessely, Aradhana
Adhikari, Katarina Bulovic, Parinya Punpongsanon, and
Stefanie Mueller. 2019. Photo-Chromeleon:
Re-Programmable Multi-Color Textures Using
Photochromic Dyes. In Proceedings of the 32nd Annual
ACM Symposium on User Interface Software and
Technology (UIST ’19). Association for Computing
Machinery, New York, NY, USA, 701–712. DOI:
http://dx.doi.org/10.1145/3332165.3347905
[13]
Jeeeun Kim, Clement Zheng, Haruki Takahashi, Mark D
Gross, Daniel Ashbrook, and Tom Yeh. 2018.
Compositional 3D Printing: Expanding & Supporting
Workflows Towards Continuous Fabrication. In
Proceedings of the 2nd ACM Symposium on
Computational Fabrication (SCF ’18). ACM, New York,
NY, USA, Article 5, 10 pages. DOI:
http://dx.doi.org/10.1145/3213512.3213518
[14] Tyler Koslow. 2016. Pattette2S. (2016).
https://www.mosaicmfg.com/products/palette-2s
[15] Tyler Koslow. 2019. New Multi-Material Upgrade
Released For Prusa i3 MK2 3D Printer. (2019). https:
//3dprint.com/150853/multi-material- prusa-upgrade/
[16] Tim Kuipers, Eugeni Doubrovski, and Jouke Verlinden.
2017. 3D Hatching: Linear Halftoning for Dual
Extrusion Fused Deposition Modeling. In Proceedings
of the 1st Annual ACM Symposium on Computational
Fabrication (SCF ’17). ACM, New York, NY, USA,
Article 2, 7 pages. DOI:
http://dx.doi.org/10.1145/3083157.3083163
[17] Gierad Laput, Xiang ’Anthony’ Chen, and Chris
Harrison. 2015. 3D Printed Hair: Fused Deposition
Modeling of Soft Strands, Fibers, and Bristles. In
Proceedings of the 28th Annual ACM Symposium on
User Interface Software & Technology (UIST ’15).
ACM, New York, NY, USA, 593–597. DOI:
http://dx.doi.org/10.1145/2807442.2807484
[18] Dingzeyu Li, Avinash S. Nair, Shree K. Nayar, and
Changxi Zheng. 2017. AirCode: Unobtrusive Physical
Tags for Digital Fabrication. In Proceedings of the 30th
Annual ACM Symposium on User Interface Software
and Technology (UIST ’17). Association for Computing
Machinery, New York, NY, USA, 449–460. DOI:
http://dx.doi.org/10.1145/3126594.3126635
[19] Tiantian Li, Yanyu Chen, and Lifeng Wang. 2018.
Enhanced fracture toughness in architected
interpenetrating phase composites by 3D printing.
Composites Science and Technology 167 (2018), 251 –
259. DOI:
http://dx.doi.org/10.1016/j.compscitech.2018.08.009
[20] Computer Aided Technology LLC. 2019. Changing
Material Colors for FDM 3D Printers | A Tip from a
Tech. (2019).
https://www.cati.com/blog/2019/11/changing-material-
colors-for-fdm-3d-printers-a-tip-from-a-tech/.
[21] Henrique Teles Maia, Dingzeyu Li, Yuan Yang, and
Changxi Zheng. 2019. LayerCode: Optical Barcodes for
3D Printed Shapes. ACM Trans. Graph. 38, 4, Article
Article 112 (July 2019), 14 pages. DOI:
http://dx.doi.org/10.1145/3306346.3322960
[22] Stefanie Mueller, Sangha Im, Serafima Gurevich,
Alexander Teibrich, Lisa Pfisterer, François
Guimbretière, and Patrick Baudisch. 2014. WirePrint:
3D Printed Previews for Fast Prototyping. In
Proceedings of the 27th Annual ACM Symposium on
User Interface Software and Technology (UIST ’14).
ACM, New York, NY, USA, 273–280. DOI:
http://dx.doi.org/10.1145/2642918.2647359
[23] Martin Nisser, Junyi Zhu, Tianye Chen, Katarina
Bulovic, Parinya Punpongsanon, and Stefanie Mueller.
2019. Sequential Support: 3D Printing Dissolvable
Support Material for Time-Dependent Mechanisms. In
Proceedings of the Thirteenth International Conference
on Tangible, Embedded, and Embodied Interaction (TEI
’19). Association for Computing Machinery, New York,
NY, USA, 669–676. DOI:
http://dx.doi.org/10.1145/3294109.3295630
[24] Parinya Punpongsanon, Xin Wen, David S. Kim, and
Stefanie Mueller. 2018. ColorMod: Recoloring 3D
Printed Objects Using Photochromic Inks. In
Proceedings of the 2018 CHI Conference on Human
Factors in Computing Systems (CHI ’18). Association
for Computing Machinery, New York, NY, USA, Article
Paper 213, 12 pages. DOI:
http://dx.doi.org/10.1145/3173574.3173787
[25]
Tim Reiner, Nathan Carr, Radomír Mˇ
ech, Ondˇ
rej Št’ava,
Carsten Dachsbacher, and Gavin Miller. 2014.
Dual-color Mixing for Fused Deposition Modeling
Printers. Comput. Graph. Forum 33, 2 (May 2014),
479–486. DOI:http://dx.doi.org/10.1111/cgf.12319
[26] Dan Rock. 2019a. Extruder steps per mm. (2019).
https://e3d-online.dozuki.com/Guide/Extruder+steps+
per+mm/96
[27] Dan Rock. 2019b. Flow rate (Extrusion multiplier)
calibration guide. (2019).
https://e3d-online.dozuki.com/Guide/Flow+rate+
(Extrusion+multiplier)+calibration+guide./89
[28] H. Tanaka S. Koda. 2017. Direct G-code manipulation
for 3D material weaving. In Proceedings of SPIE Smart
Structures and Materials + Nondestructive Evaluation
and Health Monitoring, Vol. 10167. DOI:
http://dx.doi.org/10.1117/12.2261648
[29] Martin Schmitz, Mohammadreza Khalilbeigi, Matthias
Balwierz, Roman Lissermann, Max Mühlhäuser, and
Jürgen Steimle. 2015. Capricate: A Fabrication Pipeline
to Design and 3D Print Capacitive Touch Sensors for
Interactive Objects. In Proceedings of the 28th Annual
ACM Symposium on User Interface Software &
Technology (UIST ’15). Association for Computing
Machinery, New York, NY, USA, 253–258. DOI:
http://dx.doi.org/10.1145/2807442.2807503
[30]
Liang Shi, Vahid Babaei, Changil Kim, Michael Foshey,
Yuanming Hu, Pitchaya Sitthi-Amorn, Szymon
Rusinkiewicz, and Wojciech Matusik. 2018. Deep
Multispectral Painting Reproduction via Multi-Layer,
Custom-Ink Printing. ACM Trans. Graph. 37, 6, Article
Article 271 (Dec. 2018), 15 pages. DOI:
http://dx.doi.org/10.1145/3272127.3275057
[31] Shopify. 2020. FilaBlend. (2020).
https://www.filablend.com/
[32] Pitchaya Sitthi-Amorn, Javier E. Ramos, Yuwang
Wangy, Joyce Kwan, Justin Lan, Wenshou Wang, and
Wojciech Matusik. 2015. MultiFab: A Machine Vision
Assisted Platform for Multi-Material 3D Printing. ACM
Trans. Graph. 34, 4, Article Article 129 (July 2015), 11
pages. DOI:http://dx.doi.org/10.1145/2766962
[33] Mark A. Skylar-Scott, Jochen Mueller, Claas W. Visser,
and Jennifer A. Lewis. 2019. Voxelated soft matter via
multimaterial multinozzle 3D printing. Nature 575
(2019), 330–335. DOI:
http://dx.doi.org/10.1038/s41586-019- 1736-8
[34] Haichuan Song, Jonàs Martínez, Pierre Bedell, Noémie
Vennin, and Sylvain Lefebvre. 2019. Colored Fused
Filament Fabrication. ACM Trans. Graph. 38, 5, Article
Article 141 (June 2019), 11 pages. DOI:
http://dx.doi.org/10.1145/3183793
[35] Stratasys. 2020. Stratasys Connex3, Objet500 and
Objet350. (2020). https://www.stratasys.com/
3d-printers/objet- 350-500- connex3
[36] Haruki Takahashi and Jeeeun Kim. 2019. 3D Printed
Fabric: Techniques for Design and 3D Weaving
Programmable Textiles. In Proceedings of the 32nd
Annual ACM Symposium on User Interface Software
and Technology (UIST ’19). Association for Computing
Machinery, New York, NY, USA, 43–51. DOI:
http://dx.doi.org/10.1145/3332165.3347896
[37] Haruki Takahashi and Homei Miyashita. 2017.
Expressive Fused Deposition Modeling by Controlling
Extruder Height and Extrusion Amount. In Proceedings
of the 2017 CHI Conference on Human Factors in
Computing Systems (CHI ’17). ACM, New York, NY,
USA, 5065–5074. DOI:
http://dx.doi.org/10.1145/3025453.3025933
[38] Ryan L. Truby and Jennifer A. Lewis. 2016. Printing
soft matter in three dimensions. Nature 540, 7633
(2016), 371–378. DOI:
http://dx.doi.org/10.1038/nature21003
[39] Joshua Vasquez, Hannah Twigg-Smith, Jasper
Tran O’Leary, and Nadya Peek. 2020. Jubilee: An
Extensible Machine for Multi-Tool Fabrication. In
Proceedings of the 2020 CHI Conference on Human
Factors in Computing Systems (CHI ’20). Association
for Computing Machinery, New York, NY, USA, 1–13.
DOI:http://dx.doi.org/10.1145/3313831.3376425
[40] Kiril Vidimˇ
ce, Szu-Po Wang, Jonathan Ragan-Kelley,
and Wojciech Matusik. 2019. OpenFab: A
Programmable Pipeline for Multimaterial Fabrication.
Commun. ACM 62, 9 (Aug. 2019), 97–105. DOI:
http://dx.doi.org/10.1145/3344808
[41] XYZprinting. 2020. XYZprinting da Vinch Color.
(2020). https:
//www.xyzprinting.com/en-US/product/da- vinci-color
[42]
Tom Yeh and Jeeeun Kim. 2018. CraftML: 3D Modeling
is Web Programming. In Proceedings of the 2018 CHI
Conference on Human Factors in Computing Systems
(CHI ’18). Association for Computing Machinery, New
York, NY, USA, Article Paper 527, 12 pages. DOI:
http://dx.doi.org/10.1145/3173574.3174101
[43] Jonny Yeu. 2019. Convert Single-Color STL files into
Multi-Body Models using MeshMixer. (2019). https://
support.mosaicmfg.com/hc/en-us/articles/115002827014
[44] Yang Zhang, Gierad Laput, and Chris Harrison. 2017.
Electrick: Low-Cost Touch Sensing Using Electric Field
Tomography. In Proceedings of the 2017 CHI
Conference on Human Factors in Computing Systems
(CHI ’17). Association for Computing Machinery, New
York, NY, USA, 1–14. DOI:
http://dx.doi.org/10.1145/3025453.3025842
