454 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 16, NO.

1, JANUARY 2020

Application of Multidirectional Robotic Wire Arc

Additive Manufacturing Process for the
Fabrication of Complex Metallic Parts
Lei Yuan , Donghong Ding, Zengxi Pan , Member, IEEE, Ziping Yu, Bintao Wu, Stephen van Duin,
Huijun Li, and Weihua Li , Member, IEEE

Abstract—Robotic wire arc additive manufacturing

(WAAM) for the fabrication of metallic parts is garnering
interest due to its advantages of low capital investment, high
deposition rates, and good material properties. Although
many achievements have been made, the build direction
of WAAM remains confined to the vertical-up direction,
requiring additional supporting structure to be deposited
while fabricating metallic parts with overhanging features.
In this article, a novel multidirectional WAAM process, using
robotic gas metal arc welding, to additively manufacture
metal components in multiple directions is presented. Sev-
eral novel modules, including positional bead modeling,
multidirection slicing, and deposition process optimization
of the deposition process, are highlights of this article. In
addition, the performance of the proposed multidirectional
WAAM strategy is evaluated by the successful deposition
of a sample workpiece with complex geometrical features.
The proposed multidirectional WAAM process would signif-
icantly reduce the manufacturing time and cost.
Index Terms—Additive manufacturing (AM), gas metal
arc welding (GMAW), multidirection, overhang, wire arc AM
(WAAM).

I. INTRODUCTION
Fig. 1. (a) (i) Build part with support structure; (ii) table was rotated;
ECENTLY, robotic wire arc additive manufacturing
R (WAAM) process has drawn great interests from
aerospace [1] and maritime industry [2] where such components
and (iii) continue to build the component along another direction [7].
(b) Demonstration of fabricating parts with an inclined angle in flat
position [10].
often suffer extremely high buy-to-fly ratios. The WAAM
is a rapid manufacturing process that first slices an input
three-dimensional (3-D) model into a set of two-and-a-half- dimensional (2.5-D) layer models along the vertical direction
and then builds up the model by depositing materials in a
layer-by-layer process [3]. However, this vertical-up fabrication
Manuscript received February 9, 2019; revised June 25, 2019 and
July 23, 2019; accepted August 7, 2019. Date of publication August paradigm makes it impossible to deposit parts with overhanging
14, 2019; date of current version January 4, 2020. This work was features beyond the overhang angle thresholds [4]. Additional
supported in part by the National Natural Science Foundation of China supports or scaffolds are normally required to print parts with
under Grant 51805085 and in part by China Scholarship Council under
Grant 201708200016. Paper no. TII-19-0423. (Corresponding author: such overhanging features, resulting in increased costs due
Donghong Ding.) to low material utilization rate and additional postmachining
L. Yuan, Z. Pan, Z. Yu, B. Wu, S. van Duin, H. Li, and W. Li are process requirements [5], [6].
with the School of Mechanical, Materials, Mechatronic and Biomedical
Engineering, Faculty of Engineering and Information Sciences, Uni- To reduce the reliance on supporting structures, various
versity of Wollongong, Wollongong, NSW 2522, Australia (e-mail:, strategies have been developed for additive manufacturing (AM)
[email protected]; [email protected]; [email protected]; processes. Some researchers explored multidirection deposition
[email protected]; [email protected]; [email protected];
[email protected]). by integrating part positioners with additional degree of free-
D. Ding is with the School of Mechatronics Engineering, Foshan Uni- doms (DOFs) into the AM system, as shown in Fig. 1(a) [7]. This
versity, Guangdong 528225, China (e-mail:, [email protected]). configuration allows overhanging features to be deposited by
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org. reorienting the part accordingly during the deposition process,
Digital Object Identifier 10.1109/TII.2019.2935233 removing the requirement for supporting structures. Ding et al.
1551-3203 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
YUAN et al.: APPLICATION OF MULTIDIRECTIONAL ROBOTIC WAAM PROCESS FOR THE FABRICATION OF COMPLEX METALLIC PARTS 455

[8] developed an eight-axis robotized AM system combining a

six-axis robot arm and an external two-axis pan-tilt positioning
system. With the aid of additional DOF from the part posi-
tioner, a multidirectional deposition algorithm provides a sig-
nificant reduction in production time. In the study conducted by
Panchagnula and Simhambhatla [9], an additional rotation table
was employed to tilt the substrate at an angle equals to the
growth direction in thin-walled parts. The results showed better
geometry qualities of the deposited material, and a more stable
deposition process.
Instead of using a positioner to adjust the orientation of the
workpiece, other researchers have focused on the fabrication of
overhangs with inclined features directly in a flat position, as
shown in Fig. 1(b). With proper control, gas metal arc welding
(GMAW) based welding processes can be carried out in any
direction/orientation, as the surface tension force of the molten
weld pool is able to bear the weight of the deposited material.
This characteristic was exploited in [9], where small overhangs
were easily deposited in a WAAM process.
Xiong et al. [10] studied the influence of process parameters
on the formation of an inclined wall in a GWAM-based WAAM Fig. 2. Architecture of the multidirectional WAAM process.
process. In these experiments, the influences of offset distance,
wire feed speed (WFS), and welding torch travel speed (TS) on
the inclination angle were examined. The experimental results II. ARCHITECTURE OF THE MULTIDIRECTIONAL
showed that higher TS and lower WFS allowed for a larger WAAM SYSTEM
inclination angle. The maximum inclination angle achieved was Fig. 2 presents the architecture of the proposed multidirec-
over 45° with proper weld settings. In the research of Li et al. tional WAAM system, from initial CAD model to the final near-
[11], a mathematical model of the inclined multilayer multibead net product. It consists of several essential modules including
parts was initially established for a GWAM-based WAAM positional bead modeling, multidirection slicing, and deposition
process. Additionally, the authors mentioned that the material process optimization, which are distinct in multidirectional
shortage areas were found at the edges of the parts with inclined WAAM process and will later be introduced in detail.
features and extra material should be deposited in these areas Weld bead modeling is a necessary first step in the WAAM
to improve local geometry. It was determined that the amount process [1]. This modeling is used to establish a relationship
of compensated filler material depends on the inclination angle, between the weld process parameters and the deposited bead
the profile of the initial bead, and the step-over distance between geometry. It also provides a database used during the slicing
adjacent beads. However, although a number of advancements procedure to select a reasonable layer thickness, for the path
have been made in this area, the range of possible inclination planning procedure to control the offset distance between adja-
angle can be fabricated using a flat deposition process remains cent beads, and for the deposition process to adopt the optimum
limited. welding parameters. Several methods for bead modeling, such
Most recently, Kazanas et al. [12] reported a feasibility study as linear regression, second-order regression, artificial neural
on the fabrication of overhanging structure directly, without network, and Taguchi method, have been reported in the lit-
printing the supporting structures. This was done by taking erature. Other studies focus on finding the optimal model for
advantage of positional welding techniques. In their experi- the bead profile in WAAM process [1], [5], [13]. It has been
ments, a few horizontal and inclined walls were deposited and found that the weld bead geometry is largely dependent on the
several process control strategies were presented. Through this selection of WFS and torch TS. Although weld bead modeling
article, multidirectional WAAM capability was demonstrated, a methods are widely reported in the literature [3], [14], [15], the
comprehensive WAAM-based positional deposition system has positional bead modeling targeting the deposition in an arbitrary
yet to be developed for practical applications. direction has not been developed.
This article aims to develop a process planning algorithm for In the slicing module, the 3-D CAD model (in STL format)
positional welding based multidirectional WAAM. Following is first imported into our in-house MATLAB based WAAM
this section, Section II presents a general overview of the programming software. Multidirection slicing of a CAD model
proposed multidirectional WAAM process. Section III presents is an another challenge associated with the proposed mul-
positional bead modeling strategies. Section IV introduces a tidirection deposition process. Ding et al. [16] developed a
multidirection slicing algorithm. A deposition process optimiza- new decomposition-regrouping method for multidirection slic-
tion strategy is provided in Section V. A case study is presented ing. After decomposing the CAD model into subvolumes, a
in Section VI, and Section VII concludes this article. depth-tree structure based on available topology information is

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
456 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 16, NO. 1, JANUARY 2020

Fig. 3. Weld bead geometry.

introduced to regroup them into new orders for slicing. The

proposed method proved simple and efficient on various test
parts. More multidirection slicing methods can be found in the
work of Ruan et al. [17] (centroid axis extraction method), Fig. 4. Utilized CMT and robot-based multidirectional WAAM system.
Yang et al. [18] (transition wall method), and Dwivedi and
Kovacevic [19] (skeleton method), etc. The proposed algorithm
for obtaining a set of 2.5-D layers through a multidirection an infrared pyrometer, and controlled by pausing the deposition
slicing method, as shown in Fig. 2, consists of four steps, process between layers until temperature readings drop below
including the following. predetermined limits. A structured laser scanning system was
1) Overhanging part identification, in which overhangs are also integrated to measure the geometries of the deposited
identified along the user-defined initial build direction. beads.
2) Part decomposition, in which the input part is divided
into buildable and unbuildable subvolumes according to B. Modeling of the Weld Bead
the identified overhangs.
3) New build orientation, in which new build orientations In this article, the parabola model reported in [1] was used
for unbuildable subvolumes are determined and the to model the geometry of the overhanging weld bead. Twelve
overhangs are consequently eliminated. groups of experiments with different settings were chosen to
4) Slicing, which slices all subvolumes according to their evaluate the performance of the parabola model. As listed in
calculated build directions. Table I, the WFS ranges from 1 to 4 m/min and the TS ranges
In the process optimization module, a process control strategy from 0.1 to 0.3 m/min. Mild steel weld wire (ER70s-6) with a
is developed for the specific welding setup used. The welding diameter of 0.9 mm was used as the welding consumable. Note
setup is the most important step in multidirectional WAAM, as that the ranges of the WFS and TS were selected according
we are trying to form a regular weld bead under an unfavorable to the earlier studies conducted by Kazanas et al. [12]. In this
downward gravitational force. This setup includes the selection article, deposition ratio is defined as WFS/TS, which provides
of proper welding mode and welding process parameters. In an indication of the cross-sectional area of deposited material at
addition, the robot code generation module is provided to a given point.
optimize the deposition sequence and collision avoidance during In Table I, it can be found that small deposition ratios (group
the deposition process. 5, 9, and 10) cannot guarantee the complete formation of the
weld bead due to insufficient materials. For other groups, the
weld bead height (h) and width (w) were recorded. The bead
III. POSITIONAL BEAD MODELING height (h) and width (w) calculated from the parabola models
A. Experimental Setup for Bead Geometry Measurement are also provided. For simplicity, the computing process was
omitted here, as the detailed information can be found in [1].
Fig. 3 shows the cross section of the horizontal wall, the
The relative errors of bead width and height, ew and eh , defined
geometry of the bead, the bead width (w), and the bead height
as the percentage of the deviation between the predicted and the
(h). As illustrated in Fig. 4, a robotic welding system was
measured bead geometry can be calculated as
developed. Fronius TPS 4000 CMT advanced welding machine
is used as the power source for the system. Relevant process h − h
variables, such as WFS, can be adjusted through either the eh = ∗ 100% (1)
h
robot teach pendant or the welding machine controller (which
w − w
is connected to the robot controller). The welding torch was ew = ∗ 100%. (2)
mounted on an ABB IRB 1400 robot. The RobotStudio (ABB’s w
proprietary robotic simulation and programming software) was The maximum error regarding deposition ratio WFS/TS was
used to program robot/torch motion and coordinate weld set- plotted in Fig. 5. It can be seen that the parabola model gives
tings. The temperature of the deposited layer is monitored by the maximum error value as 2.13% and 2.46% of the bead width

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
YUAN et al.: APPLICATION OF MULTIDIRECTIONAL ROBOTIC WAAM PROCESS FOR THE FABRICATION OF COMPLEX METALLIC PARTS 457

TABLE I
PROCESS PARAMETERS AND GEOMETRY INFORMATION OF WELD BEAD WITH MILD STEEL ER70S-6

of the input 3-D model. The unbuildable part is

identified as the overhanging part and therefore
decomposed from the input model, as shown in
Fig. 7(c).
Step 3. Determination of build direction for overhangs: The
Gauss map method is adopted to identify the optimal
build directions for subvolumes (overhangs). For the
unbuildable subvolumes, the optimal build orienta-
tion can be determined using the Gauss map. The
Gauss map method provides normal vectors from
all points on a curve or a surface to a corresponding
point on a unit sphere [21], [22]. A spherical crown
with a minimum radius bottom surface that contains
all the points can be found on the sphere surface,
Fig. 5. Relative errors of bead height and width. the unit vector from the center of the sphere to the
center of the bottom circle of the spherical crown
and height, respectively. Note that the bead modeling error is C is the desired optimal build direction. A detailed
irrelevant to the quality of bead appearance. description of the Gauss mapping procedure can be
found in [21] and [22]. Fig. 7(c) demonstrates the
ideal build orientation for the overhanging parts.
IV. MULTIDIRECTION SLICING STRATEGY Step 4. Iterating Steps 1–3: Here, steps 1–3 are iterated,
A critical step in the multidirectional WAAM process is a however, the overhangs previously identified are
multidirection slicing algorithm. Primarily, the slicing method now used as the input model. Step 4 continues until
used should be able to decompose a 3-D model into a set of all decomposed subvolumes are deemed buildable.
optimal subvolumes, and then generate sliced layers along with Step 5. Slicing all subvolumes according to their calculated
the most suitable deposition directions of these subvolumes. build directions: In the slicing procedure, our in-
Based on our previous study [16], a modified multidirection house MATLAB based is used to slice each subvol-
slicing strategy is developed for fabricating complex structures. ume along its identified orientation using facet-plane
The key steps of the slicing strategy are presented in Fig. 6, and slicing method [23]. Note that the distance between
are introduced as follows. adjacent slicing planes is also the effective layer
Step 1. Creation of the build direction: The initial build thickness. The layer thickness data obtained from
direction (or orientation) of the model is determined previous bead modeling process are considered as
manually by the operator. The build direction B [0 an input for 3-D model slicing. Fig. 8(a) shows the
0 1] is adopted for the input model, as shown in sliced layers of the substrate subvolumes along the
Fig. 7(a). vertical-up direction with the input layer thickness.
Step 2. Overhangs identification: Overhangs of the input To slice other subvolumes in their optimal build
model along the build direction are identified us- directions, the normal vectors of the cutting planes
ing the silhouette edges projection method [20]. need to be rotated parallel to the build directions.
Through the silhouette edges projection method, Rotation matrices are used to transfer the normal
the input model can be divided into buildable and vector data of the cutting plane P, as expressed in
unbuildable parts along a certain build direction. (3)–(5), three 3-by-3 matrices Mx , My , and Mz are
Fig. 7(b) shows the unbuildable subvolume (in red) used to represent plane rotations by angles α, θ, and

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
458 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 16, NO. 1, JANUARY 2020

Fig. 8. (a) Buildable subvolume sliced in vertical-up direction. (b) Entire

3-D model sliced in multidirection.

γ about the x-, y-, and z-axis, respectively. Each rota-

tion angle is defined to be positive for a rotation that
is counterclockwise when viewed by an observer
looking along the rotation axis toward the origin,
the detailed information of using rotation matrix can
be found in [24] and [25]. With the help of rotation
matrices, the cutting plane P is transformed into a
new cutting plane Pnew = Mx My Mz P , which is
used to slice the volume along the corresponding
new direction. Finally, the remaining subvolumes
are sliced according to their identified build orienta-
tions, as presented in Fig. 8(b)
⎛ ⎞
1 0 0
Mx = ⎝ 0 cos (α) −sin (α) ⎠ (3)
0 sin (α) cos (α)
⎛ ⎞
cos (θ) 0 sin (θ)
My = ⎝ 0 1 0 ⎠ (4)
−sin (θ) 0 cos (θ)
⎛ ⎞
cos (γ) −sin (γ) 0
Fig. 6. Flowchart of the multidirection slicing method. Mz = ⎝ sin (γ) cos (γ) 0 ⎠ . (5)
0 0 1

V. DEPOSITION PROCESS OPTIMIZATION

In multidirectional WAAM, deposition process optimization
includes the selection of welding mode for positional accuracy
of the weld bead, a practical welding process control strategy for
preventing weld bead sagging, and practical robot programming
algorithms for a collision-free robot motion strategy.

A. Welding Mode Selection

The first step is the selection of a suitable GMAW weld-
ing mode. Three metal transfer modes, spray, globular, and
short-circuit transfer, are available in typical GMAW processes.
Spray or globular mode is usually used when welding in the flat
position, in which droplets melt on the electrode tip and travel
through magnetic fields between the electrode tip and the parent
layers. The position accuracy is high because the droplets are
pulled toward the center of the weld pool under the influence
Fig. 7. (a) Overall build direction. (b) Unbuildable subvolume of the
model. (c) Decomposed unbuildable subvolumes and their new build of both Lorentz and gravitational forces, as shown in Fig. 9(a).
directions. However, in the positional deposition, the initial velocity of the

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
YUAN et al.: APPLICATION OF MULTIDIRECTIONAL ROBOTIC WAAM PROCESS FOR THE FABRICATION OF COMPLEX METALLIC PARTS 459

Fig. 9. GMAW–WAAM-based deposition process. (a) Using spray mode in flat position. (b) Using spray mode in horizontal position. (c) Using
short-circuit transfer mode in horizontal position.

droplets is no longer parallel to the gravitational force, which

complicates the modeling of the droplets final trajectory. As
shown in Fig. 9(b), the molten droplets appear in a parabolic
trajectory while depositing a horizontal wall, which makes it
difficult to accurately control the locations of the final droplets.
On the contrary, in short-circuit transfer mode, the molten
droplet transfers when the electrode tip touches the previously
deposited layers. The metal transfers directly to the parent layer
in a more accurate and controllable manner, as illustrated in
Fig. 9(c). Taking advantages of this feature, Cold metal transfer
(CMT), as a variant of short arc transfer, is selected for the
proposed multidirectional WAAM process.

B. Welding Parameters Selection

In addition to positional accuracy, a further issue associated
with positional deposition is a "sagging" effect on the deposited
weld bead. Due to the impacts of gravitational force, the molten
pool may sag excessively when welding, resulting in a poor bead
profile. As this effect is exacerbated by poor welding parameter
selection, it becomes necessary to identify the proper range of
welding parameters for multidirectional WAAM.
In GMAW processes, the WFS and torch TS are two key
welding parameters that control the resultant weld bead geom-
etry. To investigate the impact of WFS and TS on weld bead
formation, two groups of tests were conducted using the same Fig. 10. Experimental results for groups A and B. (a) WFS ranges from
setup, as described in Section III. 1 to 5 m/min, TS = 0.1 m/min. (b) WFS = 4 m/min, TS ranges from
1) Impact of WFS on Weld Bead Geometry: To investigate 0.1 to 0.5 m/min.
the effect of the WFS, five thin-walled structures were deposited
in the horizontal position, as shown in Fig. 10. In each test,
WFS was varied between 1.0 and 5.0 m/min and the TS was TABLE II
WELDING PARAMETERS FOR EACH TEST GROUP
kept constant at 0.1 m/min, as listed in Table II. It can be
seen from the results that as WFS increases toward 5 m/min,
irregular bead geometry begins to form. In GMAW welding,
an increase in WFS leads to an increase in welding current,
resulting in higher heat input and weld-pool temperature. As
explained in [10], the surface tension force is inversely related
to the temperature. Thus, the use of high WFS decreases the
surface tension of the weld pool, making it more difficult for
the weld pool to maintain its shape in the horizontal position.
In addition, by increasing the WFS, the size of the molten

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
460 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 16, NO. 1, JANUARY 2020

Fig. 11. Images of molten pool for group B. (a) Trial 1 (side view). (b) Trial 5 (side view). (c) Trial 5 (top view).

pool will increase, and then sag under the effect of gravity. torch pose adjustment. To help understand the workflow, the
Thus, a relatively low WFS is helpful to prevent pending molten major steps of the algorithm are as follows.
sagging. 1) Collision Check Module: Collision mainly happens be-
2) Impact of TS on Weld Bead Geometry: In group B, tests tween the robot, the welding torch, and the workpiece. The
were conducted with TS being varied from 0. to 0.5 m/min, 3-D model of the robot and torch are readily available and
whilst the WFS was kept to a constant 4 m/min, as given in imported to robot motion planning program. The trajectory of
Table II. Although both mass transfer and heat input are reduced the welding torch is the same as the welding path with a distance
with the increasing TS, the weld bead geometry still deteriorates, offset of contact tip to work distance (CTWD), and the initial
as shown in Fig. 10(b). This is a typical welding defect called orientation is the same as the build direction. At each torch
humping, which can be described as the periodic undulation of position, the robot joint angles can be obtained through inverse
the weld bead in high speed welding [26]. In arc welding, a kinematics. After the deposition of each layer is complete, the
strong metal stream is produced, which then flows backward 3-D model of the current workpiece is updated with newly
through the bridge region toward the tail of the molten pool. deposited layer information through 3-D reconstruction from
Captured by a CCD camera, Fig. 11(a) and (b) present side-view the set of two-dimensional (2-D) sliced layer. If no collision is
images illustrating the molten pool behaviors for the group B detected, the next layer will be tested. Otherwise, the collision
trial 1 and trial 5, respectively. When welding at relatively low avoidance module will be called, which includes layer sequence
speeds, the molten metal flow is able to backfill the valley area as sorting and torch pose adjustment module.
it is driven by the surface tension force, as shown in Fig. 11(a). 2) Layer Sequence Sorting Module: When a collision is
Fig. 11(b) presents an image of humped weld bead formation, in detected, the layer sequence will be reorganized and send back
which a swollen region forms at the tail of the molten pool and to the collision check module. As the layers are deposited along
the elongated bridge region solidifies earlier than the humping the build direction and the welding torch approaches from the
area as the heat source keeps moving forward. More detailed opposite direction, collision will only occur at the top few layers
explanations of the humping effect can be found in [26] and [27]. during the collision check. This means there can only a few
In addition, it needs to be mentioned that this humping occurs possible combinations of layer/torch pairings to be tested by the
in the directions of both wall growth and gravitational force collision checking module.
simultaneously, as shown in Fig. 11(c). In this case, it is difficult 3) Torch Pose Adjustment: If collision still exists after
for the molten metal stream to refill the gouged area. Thus, when checking all possible combinations of layers in a collision state,
compared to welding in the flat position, the operational window torch poses can be adjusted to avoid the collision. The collision
of the TS is more limited. A relatively low TS is recommended problem in multidirectional WAAM is formulated as a T-space
for a stable positional deposition. motion planning problem as the path has workspace constraints
such as the actual desired weld path and the weld gun angles
specified in the weld procedure. The details of the T- space
C. Robot Code Generation motion planning problem can be found in [28]. As introduced in
The robot code generation module generates collision-free [29], the torch positions X and Y determine the bead positional
robot motions from 3-D models of the robotic system, torch, and accuracy, while Z (CTWD) and torch orientation, Rx , Ry ,
workpiece. Fig. 12 presents the overall flowchart of the robot and Rz , are redundant DOFs that can be adjusted without
motion planning component of our proposed multidirectional affecting the welding quality and bead geometry. As the collision
WAAM process. The input to the system is the position infor- between the torch and workpiece becomes a priority during path
mation and build direction for a given component (obtained from planning, the strategy is to adjust CTWD, Rx , Ry , or Rz in the
the multidirection slicing process). Layers of each subvolume ranges provided in Table III to avoid the collision. A set of
are grouped and ordered according to their z coordinates. The new torch poses parameters can be generated by our software,
collision between the welding torch and the previously deposited and the optimal solution with the minimum adjustment will be
layers is then resolved through layer sequence sorting and/or recorded for the deposition process.

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
YUAN et al.: APPLICATION OF MULTIDIRECTIONAL ROBOTIC WAAM PROCESS FOR THE FABRICATION OF COMPLEX METALLIC PARTS 461

Fig. 12. Flowchart of the robot motion planning.

TABLE III
RANGE OF TORCH POSES PARAMETERS

4) Robot Code Translation: Finally, the path positions and

torch poses for each layer are translated to ABB robot language
as subroutines. The main routine calls subroutines according to
the updated layer sequence to deposit the input model without Fig. 13. (a) 3-D model of workpiece and its overhang part and required
collision. supports in vertical-up build direction. (b) Optimal build direction of each
subvolume. (c) Final near-net product.

VI. CASES STUDIES AND DISCUSSION

The performance of the multidirectional WAAM strategy the vertical-up direction along z-axis [0 0 1] for subvolume
outlined in this article is demonstrated through the fabrication of one, two, and four, while the horizontal direction along x-axis
a thin-walled workpiece with overhangs. The 3-D model of the [1 0 0] is identified for the subvolume three.
workpiece and the necessary supports for conventional vertical- The key step for the fabrication of this particular part is the
up deposition are illustrated in Fig. 13(a). First, the component bridging of the gap between wall one and wall two, as an
is identified as buildable and unbuildable subvolumes, and the overhanging horizontal wall is required. To reduce the man-
unbuildable subvolume is further decomposed into two parts, ufacturing time, a high TS is selected for depositing subvolume
as shown in Fig. 13(b). Based on the proposed multidirec- one, two, and four, while a relatively low TS is used for a more
tion slicing strategy, the build orientations are identified as stable horizontal deposition on wall three. According to the

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
462 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 16, NO. 1, JANUARY 2020

TABLE V
COMPARISON BETWEEN CONVENTIONAL WAAM
AND THE PROPOSED STRATEGY

in Fig. 13(a), additional supporting structures were required

to fabricate the part with overhangs by conventional WAAM,
resulting in additional manufacturing time and cost (the cost
is calculated based on the price of BOC 4043 Aluminum MIG
Wire, BOC Ltd., Australia). The production time and cost of
the supporting structure are calculated based on 3-D model
Fig. 14. Building sequence of the third component. (a) Dimension
of welding torch. (b) Diagram of horizontal deposition. (c) Diagram of data, it was found that the additional requirements for the
vertical deposition. (d) Connection between wall two and three. supporting structure end up costing more than the part itself.
It can be seen that reductions of 52% (manufacturing time),
57% (material use), and 57% (material cost) were achieved
TABLE IV
MANUFACTURING INFORMATION OF CASE STUDY when using the proposed multidirectional deposition strategy. In
addition, it should be noted that all data summarized in Table V
are calculated based on the deposition process only, without
considering additional post-machining processes required for
removing the supporting structure required by the conventional
WAAM method.
For practical use of the multidirectional WAAM, some
following suggestions have been summarized.
1) Due to the influence of downward gravitational force, the
process parameters of multidirectional WAAM are lim-
calculated build directions and deposition sequence, the vertical ited compared to WAAM in a flat position. For positional
wall one and two were built first while the torch angle was deposition, it is suggested that relatively low power in
maintained vertically. However, as illustrated in the Fig. 14(b), combination with low WFS and TS are employed for
wall one was initially built to the vertical height of wall three better weld quality, whilst a relative higher WFS and TS
and wall two was built to a slightly lower height than that of can be used for higher productivity.
wall one, in order to give enough space for the welding torch 2) The surface tension force is a major holding force in
when depositing wall three. For the same accessibility reason, multidirectional WAAM process. According to (6), the
wall three was not built to its full size, as shown in Fig. 14(c). As surface tension coefficient can be expressed as [10]
the build progresses, some actions must be made to ensure the γ = γm
0
− As (T − Tm ) (6)
welding arc is kept as far away from the other walls as possible.
This will help avoid situations where the electron stream will where γm 0
is the surface tension coefficient of pure metal at the
flow toward wrong wall structure, and also help maintain a melting point, which should be constant for filler material. As is
stable welding arc. This was done by alternating deposition the negative of ∂γ/∂T for pure metal, which is also considered
between the horizontal and the vertical wall, and the torch angle as a constant when the temperature is much above the melting
kept alternating between vertical and horizontal directions as point. Tm is the melting point of filler material. As temperature
well, as shown in Fig. 14(b) and (c). Subsequently, the last variations can be suppressed by an optimal deposition process
bead was deposited with the torch rotated to 45° to allow for control strategy. γm 0
is another property that affects the surface
simultaneous bonding between walls two and three, as shown tension force, it is an inherent characteristic of the filler material,
in Fig. 14(d). Finally, wall four and the rest of wall one and two so we named it inherent surface tension (IST) in this article.
were fabricated, the successful creation of the enclosed shape is Table VI provides surface tension values of several pure metals
shown in Fig. 13(c). The detailed information of the experiments in the liquid state at the melting temperature [30]. It reveals that
is provided in Table IV. IST is different for various metallic materials. The larger IST
Table V lists the experimental results of multidirectional can provide a more stable deposition process with a relatively
WAAM in comparison with conventional WAAM. As shown high material deposition ratio.

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
YUAN et al.: APPLICATION OF MULTIDIRECTIONAL ROBOTIC WAAM PROCESS FOR THE FABRICATION OF COMPLEX METALLIC PARTS 463

TABLE VI [5] J. S. Panchagnula and S. Simhambhatla, “Inclined slicing and weld-

SURFACE TENSION VALUES FOR PURE METALS IN THE LIQUID deposition for additive manufacturing of metallic objects with large
STATE AT THE MELTING TEMPERATURE overhangs using higher order kinematics,” Virtual Phys. Prototyping,
vol. 11, pp. 99–108, 2016.
[6] J. Jiang, X. Xu, and J. Stringer, “Support structures for additive manufac-
turing: A review,” J. Manuf. Mater. Process., vol. 2, p. 64, 2018.
[7] F. Liou, K. Slattery, M. Kinsella, J. Newkirk, H.-N. Chou, and R. Landers,
“Applications of a hybrid manufacturing process for fabrication of metallic
structures,” Rapid Prototyping J., vol. 13, pp. 236–244, 2007.
[8] Y. Ding, R. Dwivedi, and R. Kovacevic, “Process planning for 8-axis
VII. CONCLUSION robotized laser-based direct metal deposition system: A case on building
revolved part,” Robot. Comput.-Integr. Manuf., vol. 44, pp. 67–76, 2017.
This article presented an innovative multidirectional WAAM [9] J. S. Panchagnula and S. Simhambhatla, “Manufacture of com-
strategy for fabricating complex metal components without plex thin-walled metallic objects using weld-deposition based additive
manufacturing,” Robot. Comput.-Integr. Manuf., vol. 49, pp. 194–203,
the need for supporting structures. Compared to conventional 2018.
WAAM techniques, the proposed multidirectional WAAM pro- [10] J. Xiong, Y. Lei, H. Chen, and G. Zhang, “Fabrication of inclined
cess can significantly reduce manufacturing time and cost. In thin-walled parts in multi-layer single-pass GMAW-based additive man-
ufacturing with flat position deposition,” J. Mater. Process. Technol.,
this article, the overall process planning algorithm for multi- vol. 240, pp. 397–403, 2017.
directional WAAM system from CAD model inputs to final [11] Y. Li, X. Huang, I. Horváth, and G. Zhang, “GMAW-based additive
near-net shape products is introduced. Three following critical manufacturing of inclined multi-layer multi-bead parts with flat-position
deposition,” J. Mater. Process. Technol., vol. 262, pp. 359–371, 2018.
parts of the process are explained and summarized here. [12] P. Kazanas, P. Deherkar, P. Almeida, H. Lockett, and S. Williams,
1) Positional bead modeling: The parabola model was “Fabrication of geometrical features using wire and arc additive manu-
used in multidirectional WAAM process. The optimum facture,” Proc. Inst. Mech. Eng., Part B, J. Eng. Manufacture, vol. 226,
pp. 1042–1051, 2012.
welding parameters corresponding to the desired bead [13] J. Xiong, G. Zhang, J. Hu, and L. Wu, “Bead geometry prediction for
geometry can be obtained using the developed parabola robotic GMAW-based rapid manufacturing through a neural network and
bead model. a second-order regression analysis,” J. Intell. Manuf., vol. 25, pp. 157–163,
2014.
2) Multidirection slicing strategy: A multidirection slicing [14] R. Urbanic, S. Saqib, and K. Aggarwal, “Using predictive modeling and
strategy was described step by step, with the assistance classification methods for single and overlapping bead laser cladding to
of a 3-D model of a bearing seat. In the slicing process, understand bead geometry to process parameter relationships,” J. Manuf.
Sci. Eng., vol. 138, 2016, Art. no. 051012.
a complex input CAD model was decomposed into a [15] S. Meco, G. Pardal, A. Eder, and L. Quintino, “Software development for
set of subvolumes along the corresponding optimal build prediction of the weld bead in CMT and pulsed-MAG processes,” Int. J.
directions. Each of the subvolumes was then sliced along Adv. Manuf. Technol., vol. 64, pp. 171–178, 2013.
[16] D. Ding, Z. Pan, D. Cuiuri, H. Li, N. Larkin, and S. Van Duin, “Automatic
its optimal build direction. multi-direction slicing algorithms for wire based additive manufacturing,”
3) Deposition process optimization: CMT was selected as Robot. Comput.-Integr. Manuf., vol. 37, pp. 139–150, 2016.
the welding mode for the positional WAAM due to its [17] J. Ruan et al., “Automated slicing for a multiaxis metal deposition system,”
J. Manuf. Sci. Eng., vol. 129, pp. 303–310, 2007.
short arc transfer behavior and low heat input. A relatively [18] Y. Yang, J. Fuh, H. Loh, and Y. Wong, “Multi-orientational deposition to
low WFS and TS were recommended for preventing minimize support in the layered manufacturing process,” J. Manuf. Syst.,
effects in the deposited weld bead. A collision-free robot vol. 22, pp. 116–129, 2003.
[19] R. Dwivedi and R. Kovacevic, “Process planning for multi-directional
motion was achieved through the presented robot code laser-based direct metal deposition,” in Proc. Inst. Mech. Eng., Part C, J.
generation module. Mech. Eng. Sci., vol. 219, pp. 695–707, 2005.
The process planning algorithm was validated through fabri- [20] P. Singh and D. Dutta, “Multi-direction slicing for layered manufacturing,”
J. Comput. Inf. Sci. Eng., vol. 1, pp. 129–142, 2001.
cating a metallic part with complex geometry. This case study [21] P. Berger, H. Hügel, A. Hess, R. Weber, and T. Graf, “Understanding of
demonstrated a large reduction in material usage compared humping based on conservation of volume flow,” Phys. Procedia, vol. 12,
to traditional WAAM processes. The better buy-to-fly ratio pp. 232–240, 2011.
[22] B. Lowekamp, P. Rheingans, and T. S. Yoo, “Exploring surface character-
achieved makes multidirectional WAAM ideal for industrial istics with interactive Gaussian images: A case study,” in Proc. IEEE Vis.,
applications using expensive metallic materials, such as titanium 2002, pp. 553–556.
alloys typically used in aerospace applications. [23] S. Choi and K. Kwok, “A tolerant slicing algorithm for layered
manufacturing,” Rapid Prototyping J., vol. 8, pp. 161–179, 2002.
[24] G. G. Slabaugh, “Computing Euler angles from a rotation matrix,”
REFERENCES Retrieved Aug., vol. 6, pp. 39–63, 1999.
[25] H. Goldstein, C. Poole, and J. Safko, Classical Mechanics. Lino Lakes,
[1] D. Ding, Z. Pan, D. Cuiuri, and H. Li, “A multi-bead overlapping model for MN, USA: Assoc. Asphalt Paving Technologists, 2002.
robotic wire and arc additive manufacturing (WAAM),” Robot. Comput.- [26] G. Xu, Q. Cao, Q. Hu, W. Zhang, P. Liu, and B. Du, “Modelling of bead
Integr. Manuf., vol. 31, pp. 101–110, 2015. hump formation in high speed gas metal arc welding,” Sci. Technol. Weld.
[2] C. Shen et al., “The influence of post-production heat treatment on the Joining, vol. 21, pp. 700–710, 2016.
multi-directional properties of nickel-aluminum bronze alloy fabricated [27] D. Wu, X. Hua, D. Ye, and F. Li, “Understanding of humping formation
using wire-arc additive manufacturing process,” Additive Manuf., vol. 23, and suppression mechanisms using the numerical simulation,” Int. J. Heat
pp. 411–421, 2018. Mass Transfer, vol. 104, pp. 634–643, 2017.
[3] S. W. Williams, F. Martina, A. C. Addison, J. Ding, G. Pardal, and [28] Z. Yao and K. Gupta, “Path planning with general end-effector
P. Colegrove, “Wire+ arc additive manufacturing,” Mater. Sci. Technol., constraints,” Robot. Auton. Syst., vol. 55, pp. 316–327, 2007.
vol. 32, pp. 641–647, 2016. [29] N. Larkin, A. Short, Z. Pan, and S. van Duin, “Automatic program
[4] J. Jiang, J. Stringer, X. Xu, and R. Y. Zhong, “Investigation of printable generation for welding robots from CAD,” in Proc. IEEE Int. Conf. Adv.
threshold overhang angle in extrusion-based additive manufacturing for Intell. Mechatron., 2016, pp. 560–565.
reducing support waste,” Int. J. Comput. Integr. Manuf., vol. 31, pp. 1–9, [30] V. Kumikov, “The measurement of the surface tension of some pure metals
2018. in the solid state,” Mater. Sci. Eng., vol. 60, pp. L23–L24, 1983.

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.
464 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 16, NO. 1, JANUARY 2020

Lei Yuan received the B.E. degree in me- Bintao Wu received the B.E. and M.E. de-
chanical engineering from the North Univer- grees in naval architecture and ocean engineer-
sity of China, Taiyuan, China, in 2013, and ing from Harbin Engineering University, Harbin,
the M.E. degree in mechatronic engineering in China, in 2013 and 2015, respectively, and
2017 from the University of Wollongong, Wol- the Ph.D. degree in mechanical, material, and
longong, NSW, Australia, where he is currently mechatronic engineering from the University of
working toward the Ph.D. degree in mechatronic Wollongong, Wollongong, NSW, Australia, in
engineering. 2019.
His research interests include welding and His research interest includes metallic materi-
joint, vibration control, robotic wire and arc als, in which he mainly works on wire arc additive
additive manufacturing (WAAM), path planning manufacturing for design and manufacturing of
strategies for industrial welding robot, and WAAM process. functional structures with high quality performance.

Stephen van Duin received the B.E. and Ph.D.

degrees in mechanical engineering from the
University of Wollongong, Wollongong, NSW,
Australia, in 1999 and 2006, respectively.
From 1991 to 2000, he was with BHP Steel
Donghong Ding received the B.E. degree in as a Plant Engineer. From 2000 to 2009, he
mechanical engineering from the Nanjing Uni- was a Research Fellow with the University of
versity of Science and Technology, Nanjing, Wollongong. Since 2009, he has been a Mar-
China, in 2009, and the Ph.D. Degree in mecha- itime Program Leader with Defence Materials
tronic engineering from the University of Wollon- Technology Centre (DMTC) Ltd., Hawthorn, VIC,
gong, Wollongong, NSW, Australia, in 2016. Australia, a defence research organization, and
He is currently a Professor with the School concurrently holds a Principal Fellow position with the University of Wol-
of Mechatronic Engineering, Foshan University, longong. He has several patents on automation systems. He has more
Foshan, China. His research interests include than 40 publications in the fields of his research interests which include
Additive Manufacturing, industrial robots, auto- lean industrial automation, welding robotics, friction compensation in
matic welding, and CAD/CAE, in which he has servomechanisms, and intelligent control.
authored or coauthored more than 40 journal articles.

Huijun Li received the Ph.D. degree in materi-

als science and engineering from the University
of Wollongong, Wollongong, NSW, Australia, in
1996.
In 1995, he joined Cooperative Research
Centre Materials Welding and Joining, Univer-
Zengxi Pan (M’13) received the B.E. and M.E. sity of Wollongong, as a Postdoctoral Research
degrees in machine design and mechatronics Fellow. In 2000, he became a Research Scien-
engineering from Tsinghua University, Beijing, tist with Materials Division, Australian Nuclear
China, in 1998 and 2000, respectively, and the Science and Technology Organization. Since
Ph.D. degree in robotics and control engineer- July 2008, he has been with the University of
ing from the Stevens Institute of Technology, Wollongong and is Currently a Professor. He has authored or coauthored
Hoboken, NJ, USA, in 2005. more than 400 technical articles in refereed international journals and
He is currently an Associate Professor with conferences.
the School of Mechanical, Materials and Mecha- Dr. Li was the recipient of the Australian Museum Eureka Prize for
tronic Engineering, University of Wollongong, Outstanding Science in Safeguarding Australia in 2013, Australia En-
Wollongong, NSW, Australia. He has authored deavour Fellowship in 2014, and Defence Materials Technology Centre –
or coauthored more than 100 journal articles in the fields of his research Capability Improvement Award in 2014 and 2016.
interests which include industrial robotics, welding technology, and ad-
vanced manufacturing with innovation.
Weihua Li (M’15) received the B.E. and M.E.
degrees in mechanical engineering from the
University of Science and Technology of China,
Hefei, China, in 1992 and 1995, respectively, and
the Ph.D. degree from Nanyang Technological
University, Singapore, Singapore, in 2001.
From 2001 to 2003, he was a Research
Ziping Yu received two B.E. degrees in mecha- Fellow with the School of Mechanical and Pro-
tronic engineering from the University of Wollon- duction Engineering, Nanyang Technological
gong, Wollongong, NSW, Australia, and Beijing University. Since 2003, he has been an Aca-
Jiaotong University, Beijing, China, in 2018, re- demic Staff Member in Mechanical Engineering
spectively. He is currently working toward the with the School of Mechanical, Materials and Mechatronic Engineering,
Master of Philosophy degree in mechatronics University of Wollongong, Wollongong, NSW, Australia. He has authored
and manufacturing engineering with the Univer- or coauthored more than 350 technical articles in refereed international
sity of Wollongong. journals and conferences.
His research interests include welding and Dr. Li was the recipient of many awards, including Fellow of the
joining for defense applications, development of Institute of Engineers Australia, Fellow of the Institute of Physics, U.K.,
manufacturing strategies for fabricating metallic the Japan Society for the Promotion of Science Invitation Fellowship in
parts with complex geometries using wire and arc additive manufacturing 2014, the Endeavour Research Fellowship in 2011, and the Scientific
(WAAM), and robot path planning strategies for welding and WAAM Visits to China Program Awards. He is currently an Associate Editor or
process using industrial robots. Editorial Board Member for ten international journals.

Authorized licensed use limited to: University of Wollongong. Downloaded on May 12,2022 at 03:36:24 UTC from IEEE Xplore. Restrictions apply.