Proceedings of the 2015

IEEE Conference on Robotics and Biomimetics

Zhuhai, China, December 6-9, 2015

Robotic Additive Manufacturing along Curved Surface – a Step

towards Free-form Fabrication
George Q. Zhang*, Wiener Mondesir**, Carlos Martinez*, Xiongzi Li*,
Thomas A. Fuhlbrigge* and Hemant Bheda **
Abstract— This paper presents a method and process to evaluation of the AM industry. It predicts a double digit
fabricate parts or structures by printing along curved surfaces increase in the next several years.
with use of industrial robots. Software is developed to generate Figure 1 gives the growth history and prediction for AM
the fabrication path program that can take advantage of the 6-
machine building and service alone. Compound Annual
DOF articulated industrial robotics. Unlike the traditional
vertically layer-by-layer 3D printing scheme, with the angled Growth Rate (CAGR) of the AM industry is 26% in the
printing capability, a robotic fabrication path can be designed industry’s 24-year history. Since AM initiates and expands,
along the direction of curved surfaces or the direction of new and existing industries such as material, software and
physical property requirements such as higher mechanical R&D activities, the actual impact to the manufacturing
strength, specific thermal, electrical and even biomedical industry exceeded the scope of AM industry growth itself.
characteristics. The fabricating path program such as G-code
Next generation manufacturing could combine the traditional
will be then converted to a robot program such as RAPID. The
part building process will be simulated and optimized for given and additive manufacturing technology and spark another
robot specification, mounting configuration and building plate industry revolution.
location. Robot dispensing software and hardware interface are Table 1 shows an AM application timeline in the past and
employed to coordinate the motion speed and material prediction into the future. As can be seen in Table 1, we are
extrusion rate. Robotic Additive Manufacturing (RAM) testing in the time period of expending the AM applications into
cells have been built to carry out and verify the printing scheme
smaller than normal size – nanomanufacturing and larger
along curved surface, which is one step further towards free-
form fabrication. size – architecture (building) applications. For the
architecture AM, normally in open-form 3D printing,
I. INTRODUCTION equipment needs to be larger in size correspondingly, able to
Recent development of additive manufacturing (3D be operated in outdoor environment, and more cost effective.
printing) technology show its bright future for being used in Industrial robots by nature meet the requirements. For some
product prototyping and production. Tremendous effort has of existing applications such as foundry mold and sand core
been carrying out on the technology development and AM, industrial robots have also advantages on large size,
business acquisitions/reorganization in the related fields high flexibility and low-cost. In fact there are many trial
worldwide, especially in the past a few years. applications being done, which show the advantages of using
Cited reference [1] gives an analysis and discussions on industrial robots.
the Additive Manufacturing (AM) technology and market for
Figure 1. Primary AM market sale – products and service (Source:
various industry sectors, such as aerospace, automotive,
Whohlers Report 2012).
biomedical, and electronics. The study shows that the market
penetration by AM is still limited at present time but with a
promising future. The most promising application fields for
success are aerospace, electronics and biomedical industries.
The future of the additive manufacturing depends the
technology advancement and innovation in related fields.
Cited reference [2] further investigated the worldwide trends
in additive manufacturing. It predicts that the applications are
going from modelling and prototyping to direct
manufacturing. In fact, although AM has not been used in
mainstream production yet, the prediction on the revenue
growth of the AM machine production and 3D printing
service market, R&D investment and technology
advancement have been met and exceeded. Furthermore, the
cited reference [3] Wohler Report on Additive
Manufacturing gave a comprehensive assessment and

*ABB US Corporate Research Center, Mechatronic and Sensor ** Arevo Labs Inc., 2960 Scott Blvd., Santa Clara, California 95054,

Department, 5 Waterside Crossing, Windsor. CT 06095, USA (email: USA (email: [email protected]).
[email protected]).

978-1-4673-9675-2/15/$31.00 © 2015 IEEE 721

 Software has been developed to generate unique
Table 1. The AM application timeline (Source: Reference [3]).
fabrication path program that can take advantage of the 6-
DOF articulated industrial robots. Unlike the traditional
vertically layer-by-layer 3D printing scheme, with the angled
printing capability, the part-building path can be designed
along the direction of curved surfaces or along the direction
of other property requirements such as higher mechanical
strength. The path program such as G-code will be then
converted to robot program such as RAPID. The part
building process will be simulated and optimized for given
robot specification, mounting configuration and building
plate location. Robot dispensing software and hardware are
employed to coordinate the motion speed and extruding rate
of the extrusion head. Robotic additive manufacturing testing
cells have been built to realize and verify the angled printing,
which is one step further towards free-form part fabrication
as a new way of additive manufacturing with industrial
robotics.
Cited reference [4] gives a survey and feasibility study on Chapter 2 of this paper describes the method and process
use of industrial robots in additive manufacturing. The to fabricate a part or component with angled printing by use
advantages of the articulated industrial robotics were further of industrial robots. A unique “slicing” software is developed
described and discussed. Figure 2 shows a proposal of use of to generate and verify the fabrication path for robotic
articulated robots in bridge 3D printing. additive manufacturing. Besides a standard industrial robot,
RAM simulation and optimization software and extrusion
Figure 2. Use industrial robot in bridge printing (source: flowrate control hardware interface are introduced and
//http:Businessinsider.com). discussed; Chapter 3 deals with the robotic additive
manufacturing execution testing and test results; and, finally,
Chapter 4 presents the result discussion and conclusions.
Future work in related area is also proposed.

II. DEVELOPMENT OF A ROBOTIC ADDITIVE

MANUFACTURING SYSTEM
The robotic additive manufacturing system includes a
standard industrial robot, a commercial available extruder
(printing head), heating pad, and extrusion flowrate control
hardware and software. This chapter describes the RAM
system in detail.
A CAD file with building parameters is loaded into an AM
path generation software; APT, G-code or other specific code
format will be the output of the path generator; and then the
Figure 3. Use of industrial robotics in biomedical additive manufacturing code is passed to robotic AM builder module in which the
(Source: Cornel University Medical Center). robot program will be generated and optimized, including the
required embedded process parameters. Furthermore, the
robot program can be also generated directly from the CAD
file with building parameters so that the intermediate step of
APT or G-code generation could be skipped. After the RAM
program generation and optimization, a simulation module
will be used to create the virtual part with the geometrical
and building structure information. The virtual part then can
be used by software such as FEA for analyzing and/or
verification against the mechanical, thermal, electrical as
well as biomedical function requirement. For example, if it
is required to have higher tensional strength in a particular
direction, the building “path” can be designed to print along
that direction, therefore the structural strength will be higher
in that direction. After the analysis and verification, the robot

722
program is loaded into a real robot controller. Robot cell CAD file is loaded into Arevo Kepler Engine software
calibration module may be used to match up the virtual RAM which generates AM path based on the manufacturing
cell with the real one. During the execution, a fabrication process parameters such as the diameter of the extruder and
quality monitoring sensor can be used to send feedback to the the thickness of the building layer. The path program is
robot controller real-time so that the building path can be currently in G-code format. The path could be visualize and
adjusted during the execution to ensure the quality of the verified before being exported to the downstream AM
built object. software for further verification and optimization based on
the RAM system setup. Figure 4 and 5 show two examples
A. Path Generation and Verification of the curved paths which are used in the RAM testing. The
A software has been developed to generate building path hollow wing structure shown in Figure 5 demonstrates the
based the structural geometry, the property (starting with “free-form” fabrication concept. A base layer is firstly
mechanical and thermal, electrical and biomedical can be created; then a frame structure with curved surfaces is built
also further considered), and surface quality requirements. on top of the base at both ends of the wing-shaped object;
Supporting structure path could be generated with coarse then, material is built up to form a curvature starting from
printing path and/or different material. Different from open one of the corners between the base and the frame. Proper
source vertically layer-by-layer path generation, this path cooling device is needed to solidarize the material quickly to
generator can create paths along curved surface (angled form the hollow structure; finally, the surface is finished by
printing), which takes the advantage of additive weaving paths along the curved upper surface of the hollow
manufacturing – combining the fabrication process with the wing. A close look of the surface in Figure 5 will show the
part property to reduce the need of a designer during path pattern.
manufacturing process to optimize the overall design. Figure
4 shows an example of the path generated for a portion of B. Path Conversion, Simulation and Optimization
aerospace part with curved surfaces.
The generated AM path program in G-code is converted
to robot path program such as RAPID. Along with the
Figure4. A sample of curvesd surface disk AM with support material. positions, the printing (extrusion) related specific G-code is
also converted into specifics robot commands. Robot
dispensing software and its hardware interface are employed
in this development. More details in hardware interface will
be described in later this chapter. During the conversion, the
robot reachability is checked. If the positions on the path are
not reachable, the building plate location (attached to the
work object frame of the part to be built) will be relocated
with respect to the robot frame. Finally, the optimal work
object frame will be defined. In case the built object is too
complicated to be fabricated with current RAM cell, a new
cell configuration, including robot positioners track rail or
multiple robots, could be proposed and verified in the virtual
world before an actual robotic cell is built.
After the conversion, the robotic additive fabrication
process can be simulated in a software platform such as
Figure 5. A sample of building path for hollow wing structure. RobotStudio. Path then can be optimized for TCP speed,
extrusion rate, and path termination type (determining the
size of the blending zone between two consecutive path
segments). Finally a virtual part can be generated. The virtual
part contains the geometry and building process information
such as the actual shape of the object built and the direction
of the building path at any particular point. The virtual part
then can be exported to an analyzing software package for
further study. With the software tools described above, the
geometry and properties of the built object could be
designed, analyzed and redesigned virtually. In this way,
unlike the traditional manufacturing, RAM combines the
part’s geometry design, manufacturing process and its
physical properties with a free-form fabrication.

723
 C. Extrusion Rate Control Table 2. An example of calculation of extrusion speed for certain layer
thickness, nozzle and filament diameter, and feed speed and extrusion speed
In this RAM development the extrusion flowrate control ratio.
is done by use of robotic dispensing software -
DispenseWare and its analog I/O interface. Figure 6
illustrates the hardware scheme. DispenseWare provides an
analog output signal which is proportional to the speed of the
robot tool center point (TCP). The analog signal is connected
to a pulse generation card which generates certain frequency
pulses based on the input voltage value. The TCP-speed
proportional pulses are then input to a stepper motor drive to
control the motor on the extrusion head. For other types of
printing head to be used, as long as the extrusion controller
takes an analog or digital (using an analog digital converter)
input, the extrusion speed can be controlled based on the TCP
speed of the robot.

Figure 6. RAM extrusion flowrate control hardware.

D. RAM Execution
Because of the nature of additive manufacturing (3D
printing), the RAM program size will be rather large. It could
be challenging to load a whole program into RAM for
execution. The robot program should be broken down into
modules with reasonable size. There are two proposed
methods to execute a RAM program: a) load the RAPID
robot program module directly using dynamic
load/unloading function. The advantage of this method is that
the robot program lines will be explicitly displayed; tracking
Figure 7 gives the measured data for the robot TCP speed and debugging will be relatively easy and convenient. The
versus the dispensing analog I/O output. It has a very well disadvantage of this method is that when loading/unloading
linear relation in the common used TCP speed range, which the robot program, the robot may be held in position for a
set up a base to calculate the programed extrusion speed (in period of time. If the loading/unloading occurs during
RPM) for the stepper motor on the extrusion head. Table 2 continues printing, the surface quality could suffer. This
gives an example of calculation of the motor steps for TCP could be avoid if the loading/unloading of modules happens
motion distance in millimeters. These formulas and at a point between the building layers, so the robot can retreat
calculation can be used by a RAM system builder (integrator) the extrusion head before the loading/unloading. Another
to setup the initial process parameters. The actual parameters method to ensure continuously RAM execution is using of
can be tuned for the specific hardware in an actual RAM
background data loading. Position and process data modules
system.
are loaded by a parallel background task while the robot is
Figure 7. The relationship between the robot TCP speed (in mm/s) and executing the building program module in “front” task. If
DispenseWare analog output voltage (in volts) designed properly, this method can perform continuous
building process without interruption. The disadvantage of
this method is the robot position and fabrication process
information are buried in the background loaded data
modules so it is harder to tack and debug the program.
Ideally, the dynamic loading/unloading method is used in
commissioning and prototyping stage of RAM, where most
debug and building parameter adjustment are performed.
When a RAM program has been debugged and tested, the
background load method can be used in production.
However, further research is required to provide the same
execution model while debugging or printing a part.

724
 III. RAM TESTING AND TEST RESULTS The chapters above present the development,
implementation and testing of the proposed robotic additive
With a RAM system setup using the scheme mentioned
manufacturing concept. Free-form (with angled printing and
above, several tests have been carried out on an IRB120 and
building parameters) part fabrication path was generated and
an IRB140 RAM cells. Regular ABS and high strength
verified. A RobotStudio based software converted the G-
carbon fiber materials are used in the sample part building.
code path program into robot program. The RAM program is
Figure 8 shows a carbon fiber “hollow” wing structure built
then simulated and optimized for given robot cell
by the RAM method, in which the frame is built first, than
configuration. Existing DispenseWare and its I/O interface is
the upper surface of the wing is formed by extrude the carbon
used in extrusion flowrate control. Finally, some part
fiber material along the frame in space. After the hollow
building examples are presented and discussed.
structure is formed, the final surface is printing along the
curve surface to obtain the hollow wing structure. This free-
form construction is still at its preliminary stage. Better IV. DISCUSSION AND CONCLUSION
extrusion temperature control and printing configuration
This paper first overviews the AM technology and market
design are needed to further refine the process. However, this
test result is promising and encouraging future research and development in general and briefly introduces the Robotic
development towards free-form RAM fabrication. Additive Manufacturing concept and its development. Then
a method and process to fabricate part or component with
Figure 8. An IRB120 RAM cell printing a hollow wing structure. angled printing by use of industrial robots is proposed,
described and discussed. Software has been developed to
generate unique building path program that can take
advantage of the 6-DOF articulated industrial robot and
robotic system, which could include robot positioners, track
rails and multiple robots. Unlike the traditional vertically
layer-by-layer 3D printing scheme, with the angled printing
capability, part-building path can be designed along the
direction of curved surfaces or with the direction of desired
physical characteristics such as higher mechanical strength,
certain thermal and electrical, and even biological properties.
The G-code path is then converted to robot program. The
RAM fabrication process is simulated and optimized for
given robot specification, mounting method and building
plate location. Robot dispensing software and hardware
interface are employed to coordinate the motion speed and
Figure 9 shows an IRB140, wall-mounted RAM cell material extrusion flowrate. Robotic additive manufacturing
printing a curved disc which is a typical structure that could testing cells have been built to carry out and verify the angled
be used for aerospace structure fabrication. With supporting printing concept, which is one step further towards free-form
material, the curved disc can be built with layers along the
part fabrication as a unique way of additive manufacturing
curved surface therefore the strength in certain direction of
with industrial robotics.
the disc can be reinforced.
From the description and discussion, we can draw some
Figure 9. An IRB140 RAM cell printing a disc with curved surface for conclusions as below:
possible aerospace applications.
· Industrial robots present some advantages to be
used with different printing heads in direct AM
processes – replacing the motion portion of the 3-D
printers for some applications such as open-form
mold and sand core building as well as potentially
in free-form structure fabrication.
· Path generation and verification software has been
developed to create RAM path to take the
advantages of angled printing with 6-DOF
articulated industrial robots and potentially
designed and optimized based on mechanical,
thermal, electrical and even biomedical properties.
· RobotStudio based G-code conversion, path
simulation and optimization software is under
developing. Some of the modules work for the

725
 RAM cell testing and demo. More work is needed
to provide a complete RAM system.
· Robotic AM cell has been developed and tested on
two different models of robot, mounting methods
and building materials for both solid part building
and “hollow” structure construction.

Future research and development topics are:

· Further development of the RAM path generation

software to take the physical properties in
consideration.
· Refining and enhancing the RAM conversion,
simulation and optimization program to handle
various paths sources and RAM cell configurations.
· Explore the robot on track, with positioner and
multiple robots coordinated motion in free-form
structure fabrication.

REFERENCES
[1] [1] Jürgen Gausemeier: “Thinking Ahead the Future of Additive
Manufacturing- Analysis of Promising Industries”, Direct
Manufacturing Research Center. Heinz Nixdorf Institute, University
of Paderborn, Online print, 2011.
[2] [2] Terry Wohler’s: “Worldwide Trends in Additive Manufacturing”,
Wohler’s Associate, Inc., Fort Collins, CO, USA, RapidTech 2009:
US-Turkey Work-shop on Rapid Technologies.
[3] [3] Terry Wohler: “Wohler’s 2012 Report on Additive Manufacturing
and 3D printing”, Wohler Associate, Inc., Fort Collins, CO, USA,
2012.
[4] [4] George Zhang and et al., “Use of Industrial Robots in Additive
Manufacturing – A Survey and Feasibility Study”, ISR/ROBOTIK
2014, Munich, Germany, June 2014.
[5] [4] Neil Mantle and et al., “Additive Manufacturing: Opportunities
and Constrains – A summary of a Roundtable Forum Held on 23 May
2013 Hosted by Royal Academy of Engineering”, Royal Academy of
engineering, 3 Carton House Terrace, London, UK
[6] [5] Edward D. Hedrick: “Additive Manufacturing Consortium and
AM Activities at EWI“, Director of AMC, Material Group of EWI,
Columbus, OH, USA.
[7] [6] Karina Arcuate and et al.: Stereo lithography of Specially
Controlled Multi-material Bioactive Polyethylene glycol) scaffolds“,
University of Texas at EL Paso, W. M. Keck Center for 3-D
Innovation, 500 W University Ave., EL Paso, TX, USA.
[8] [7] Ian Gibson and et al.: “Additive Manufacturing Technologies –
Rapid Prototyping to Direct Digital Manufacturing“, Springer,
Springer Science+Business Media.
[9] [8] Martin Hedges and et al.: “3D Aerosol Jet Printing – Additive
Electronics Functionality to RP/RM“, Presented at DDMC 2012
Conference, Berlin, Germany.
[10] B. Smith, “An approach to graphs of linear forms (Unpublished work
style),” unpublished.
[11] E. H. Miller, “A note on reflector arrays (Periodical style—Accepted
for publication),” IEEE Trans. Antennas Propagate., to be published.
[12] J. Wang, “Fundamentals of erbium-doped fiber amplifiers arrays
(Periodical style—Submitted for publication),” IEEE J. Quantum
Electron., submitted for publication.
[13] C. J. Kaufman, Rocky Mountain Research Lab., Boulder, CO, private
communication, May 1995.

726