Download as pdf or txt
Download as pdf or txt
You are on page 1of 35

Wire and Arc Additive Manufacturing

Pre printing strategy for torque arm

Mattias Karlsson
Axel Magnusson

Mechanical Engineering, master's level


2020

Luleå University of Technology


Department of Engineering Sciences and Mathematics
1 Introduction

Additive Manufacturing (AM) is a manufacturing method that is based on layer by layer material
addition [1]. AM methods has been developed during the past 30 years and for each year, it’s
potential is more realized [2]. Initially, the AM process was applied to create plastic prototypes
in a short time directly from CAD models which simplifies the product development process.
Recently, the focus of AM research has shifted towards the possibility to fabricate complex metal
components [3]. In this area of use, there are several benefits compared to traditional subtracting
approaches such as CNC machining. First, it enables manufacturing of complex components that
would be hard or impossible to manufacture with subtracting methods. The method also means
a reduction in material waste due to the geometrical accuracy [1, 3]. Generally one can say that
the benefit of the AM method is higher for lower production quantities, since the cost per unit
becomes lower compared to conventional manufacturing techniques [4]. See Figure 1.

Figure 1: A graph showing the cost per unit vs quantity for conventional vs additive manufacturing
techniques [4].

Though there are several advantages using AM, it also implies new challenges and constraints that
needs to be solved [5]. In addition to that, there are several different methods in AM which are
significantly diverse where the advantages and challenges differs. Some can achieve high material
deposition rates with lower tolerances while some can achieve good tolerances but leads to a higher
manufacturing cost [3]. BAE Systems Hägglunds (BAE) have found an interest in an AM method
called Wire and Arc Additive manufacturing (WAAM). The company have done some projects to
investigate the area, and have invested in a new test-lab for further investigation.

This project is divided in two parts. The first is a continuation of previous projects at the company
dealing with the working procedure of progressing from a solid model to a WAAM manufactured
product, the so called ”Part Programming process” [6]. The second part of the project is to apply
the product development process described by Ulrich and Eppinger [7], to redesign a torque arm
that is suitable for manufacturing with WAAM (Design for Additive Manufacturing, DfAM). This
needs to be done with the knowledge gained from the first part of the project. There are already
an existing torque arm in the vehicle, but it needs to be designed in a different way to increase the
performance of the vehicle. Both parts of the project will be going on parallel to each other since
they are associated.

1.1 Design for Additive Manufacturing

In all manufacturing methods, there are different constraints and limitations which needs to
be taken into account when designing a part. To deal with such constraints and challenges

1
regarding the manufacturing process, guidelines is important for designers to tailor the part for
functionality and ease of manufacturing. Although such guidelines are prevalent for well-established
manufacturing processes, they are currently emerging for AM processes [8].

Several studies are focused on developing guidelines that is general for all AM methods, (e.g.
[9]). But there are several different AM processes that all have their different constraints and
opportunities. Design for WAAM needs its own approach from others due to the differing manufacturing
constrains and unique manufacturing capabilities. An example for that is when building overhanging
structures, no need of support is required. Instead, the process could use coordinated motion to
change the build orientation during deposition by using a manipulator. Another difference from
most AM technologies, is that the WAAM process is not constrained to work in a closed chamber.
The only constrain is the reach capability of the robotic arm [10].

1.2 Purpose

The goal of this project is to gather more information about the WAAM process, investigate how
the previously stated challenges [6] can be solved and evaluate the already stated solutions. With
the knowledge gathered during the project, it should be established how a component should be
designed to be suited for WAAM. The final goal is to design and manufacture a new torque arm
that can be implemented in one of the company’s vehicles. This is summarized in four questions:

• Which of the stated challenges is most relevant to solve to manufacture a torque arm with
WAAM?

• How can these challenges be solved?


• Are there any other challenges to solve?
• What aspects should be considered when designing a torque arm for WAAM?

1.3 Delimitations

To narrow the project, some delimitations have been made based on previous work and the
possibilities on the company:

• The project will be focusing on the equipment that is available in the new AM-lab that BAE
has invested in.
• BAE has chosen what software they want to proceed with, hence the different software
possibilities proposed by Vannucci [6] will not be further investigated.
• University West is researching welding parameters and material properties achieved with
WAAM. Their results will be used. Hence this project will use the same specific weld wire
as them, and will not investigate weld parameters.
• The market needs is stated for the vehicle, hence will not be investigated for the torque arm.
The requirements of the vehicle will be the foundation for the requirement specification. The
torque arm needs to be designed in a way that the current fastening points on the vehicle
can be used, to avoid the need to redesign other components.

2
2 Theoretical framework

This section describes the theoretical framework which has been relevant through the project.

2.1 Wire and Arc Additive Manufacturing

As mentioned, in this project the Wire and Arc Additive Manufacturing (WAAM) method is used.
WAAM works by melting a metal wire with an electric arc. When the wire melts, it extrudes in
form of beads as a robot or CNC machine feeds the extruder in a particular path. The beads sticks
together and creates a layer. The process repeats layer by layer until the component is finished.
The material on which the material are built on is called the substrate. The substrate can either
be included in the part or removed when the WAAM process is finished. Depending on the part
size and building orientations, it is important to have a suitable substrate [3].

WAAM is suitable for producing medium to large sized components with good structural integrity
in a range of materials. The process could lead to a lower manufacturing cost, with shorter lead
time and reduced material waste compared to conventional subtracting manufacturing. Although
WAAM has a limited design freedom compared to other AM methods. Therefore, it is important
to consider build orientation, build sequence and design constrains of the process. Due to the
high heat input and rough surface finish, post-process is also needed sometimes and has to be
considered. However, WAAM opens up a wide range of design opportunities [10, 11].

2.1.1 Cold metal transfer

Cold metal transfer (CMT) is a modified gas metal arc welding (GMAW) process that is well suited
for WAAM due to its high deposition rate and low heat input [3, 12]. In CMT, the welding current
and voltage is controlled together with the wire feed. Figure 2 describes the steps of the CMT
cycle [6].

Figure 2: Describes the CMT cycle. a) The wire approaches the material and an arc occurs. b)
The wire is pushed in to the molten pool and the arc is put out. c) The wire is retracted, and a
droplet is formed and released on the weld pool. d) The cycle repeats [6].

3
2.1.2 Challenges and possible solutions

Previous work [6] at BAE resulted in some stated challenges and solutions with the WAAM process.
These are presented below.

• Start and stops


• Voids in solid layer
• Cross of paths

• Surface finish and resolution


• ”Staircase effect”
• Overhang

• Deformation of part due to thermal/residual- stress

Continuous deposition. To avoid start and stops, the path can be planned to be continuous.
This requires that the path is consistent and corners/turns have suitable radii.

Optimized thin wall structures. An alternative to “Continuous deposition” is thin wall


structures. Wall features can be optimized to the same thickness as the deposition of the weld
(after post-processing). This will lead to low numbers of path ends.

Symmetric building. Can be applied when having a moving worktable. By building symmetrically
on both sides of the initial substrate the stress is balanced and deformations is avoided in the
substrate.

Short deposition paths. When long continuous walls is built, it generates stress to the substrate.
To avoid this, the path can be planned to have shorter deposition paths. One alternative is to
orient the part in a certain way to get shorter paths.

Selective deposition. By planning the paths to generate smaller segments of the part and later
binding them together gives them time to cool down. In this way the stress that is built up reduces.

Interlayer cooling. Between the layer depositions, cooling time can be added. This will allow
the metal to completely solidify and improves structural integrity to lower deformations.

Advanced slicing. Multi-direction slicing is possible if you have an articulated robot and can
build in multiple directions. By carefully planning the slicing direction it can be the solution to
solve problems like overhangs and stair-case effect.

Advanced path-planning. Pockets can be avoided by the path pattern and by having a
strategical approach, start and stops can be reduced and surface finish can be improved.

Clamping. When clamping the part it can detain deformations in the substrate. In combination
with heat treatment it can relax the stress before unclamping the part.

Post-processing. Post-processing is often needed for the WAAM process because of the surface
finish and stress that is built up. To achieve a better result of the post-process, material should
be added to the part.

4
2.1.3 3D Slicing

One key factor for good results in additive manufacturing is slicing which includes two issues, the
un-match effect and the stair-case effect. The un-match profile occurs when the constant layer
thickness doesn’t match the original model, as can be seen in Figure 3a. Normally the approach of
Uniform slicing is used, which slices a STL model into 2.5D layers parallel to the building direction
with constant layer height, as can be seen in Figure 3b. The accuracy of the printed part can be
controlled by the layer thickness. Thinner layer height gives better accuracy. Adaptive slicing slices
the model with various layer height based on the surface geometry, see Figure 3c. This strategy is
an effective method to solve the un-match effect [3].

Figure 3: Layer approaches: a) Original model. b) Sliced model with constant layer thickness,
results in a lower height. c) Sliced model with adjusted layer thickness [3].

The stair-case effect occurs when a surface has a high gradient or radii and the deposition layer
has a certain layer thickness. The error is normal to the building (vertical) direction, see Figure
4. For thicker layers, the effect becomes lager [3].

Figure 4: Illustrates the stair-case effect [3, 13].

2.1.4 Path methods

Path planning is important and crucial in the WAAM process due to the large sized deposit, that
is influenced by the geometric complexity. The path guides the toolhead that fills the layers, which
itself builds the geometry of the object. Different methods of tool-paths have been evaluated and
developed for AM processes. In this section different path methods is presented.

Raster path deposes material along one direction by parallel offset lines as can be seen in Figure
5a. It is a common pattern due to its simplicity. Although, the raster pattern builds the whole
component in the same direction so the extruder have to turn frequently. This lead to warpage
and anisotropic problems and this affects outline building quality [14].

5
Zigzag path generation is the most popular used in AM process. As it follows the line by line
approach it combines the parallel lines into one continuous path, see Figure 5b. This decreases
the number of tool-path operations and improves the productivity of the process. The outline
accuracy is poor for the Zigzag path due to small errors on edges that not is parallel to the tool
motion direction [11].

Contour path follows the geometrical boundary as can be seen in Figure 5c. This is to prefer
to solve the poor outline accuracy achieved with the raster pattern. The contour path method
overcomes the warpage and the anisotropic problems by changing the path direction frequently
along the boundary curves of the 2D region. The contour path method has some quality problems
leaving gaps. This is because the path follows the outer boundary deposing lines towards the
middle and gaps can occur [14].

Spiral path in Figure 5d is a commonly used tool-path in numerical controlled machining and
can be applied in AM processes. This path generation is only suitable for certain geometries. It
can solve the problems of the zigzag-path [11].

Continuous path planning is another method for the tool-path, see Figure 5e. There are several
methods to generate this type of path planning. ”Hilbert filing” is one example where a region
is covered without intersecting itself. This is helpful to reduce shrinking during the AM process.
However, for the WAAM process this method is not suitable due to the number of path direction
turning motions [15]. R.Dwivedi, et al. [16] has developed a continuous path pattern which first
combines the zigzag pattern into a continuous path of the 2D layer. The path planning reduces
the number of welding passes. However, the method has issues with surface accuracy [11].

Hybrid path is a combination of the contour- and zigzag path as can be seen in Figure 5f. This
will meet the geometrical accuracy and building efficiency requirement. The zigzag path which is
faster and more universal is applied for the inner regions. For the outline, contour path which is
cooperative for maintaining the surface quality, is used [11].

Figure 5: A brief summary of for the tool-path patterns described above [14]. a) Raster. b) Zigzag.
c) Contour. d) Spiral. e) Continuous. f ) Hybrid.

Modular path planning (MPP) is a new method for the normally used layer by layer approach
from the research by F.Michel, et al. [17]. MPP integrates the feature-based design into an efficient
layer by layer path planning solution. The MPP uses segmentation to integrate modularity in the
path generation. Instead of the traditional approach where a single path planning is used for a

6
layer, the MPP divides the given layer into segments called sections. The sections is shaped into
basic geometries to give easier path planning. Figure 6a shows the traditional single path for one
layer and Figure 6b shows the layer divided into two sections where easier path planning can be
applied.

Figure 6: a) Traditional path planning approach. b) Applied segmentation with two sections to
generate easier path planning [17].

Each section can be also be divided into zones where the path can be different, see Figure 7. The
benefit with zones is the adaptive choice of individual set of deposition parameters. When the
first layer is planned, this approach can be applied on each upcoming layer. This means advanced
geometries can be achieved with simple path generation, as can be seen in Figure 7. However, the
method can be a challenging and time consuming process.

Figure 7: Section with the path divided into zones [17].

2.1.5 Design for WAAM

As mentioned, design guidelines is an useful tool to deal with the different constraints and challenges
regarding manufacturing. Locket et al. have summarized an initial set of guidelines for Design for
WAAM (DfWAAM) [10] based on a few studies [18–22]. These are presented below.

Part dimensions. WAAM is not constrained by an enclosed working area, therefore the upper
boundary for a part is only constrained by the reach of the robotic arm. The minimum deposition
thickness for a wall is 2 mm but thinner walls can also be achieved by post processing.

Symmetry. As mentioned before, symmetrical building could be used to deal with the distortions
that could occur due to the thermal stresses [23]. This makes the build-up of the residual stress
more balanced and reduces the distortion during the manufacturing process [22].

7
Enclosed features. In difference from conventional subtracting manufacturing methods, WAAM
parts can have enclosed features. These could be produced by using carefully designed deposition
paths or multi-axis build techniques. However, one should be cautious when designing such enclosed
features since they can be hard to inspect or post-process.

Unfinished faces. Although WAAM results in a rough surface finish, it may not always be
necessary to post-machine some of the deposited faces. Even though unfinished surfaces may be
stress raising and fatigue vulnerable, this option could be considered in some applications.

Machining Considerations. For the areas where post machining is required, conventional
manufacturing guidelines should be followed. Fastening points has to exist to make it possible
for e.g. a CNC-machine to hold the part in place. Suitable measuring points must be provided
on the part to make it possible to measure different features to the correct place [24]. Also, an
offset to the part must be added to give a machining allowance. This is typically 1 mm added to
all machined faces [22].

Corners. It is preferable to add radii at corners to make the deposition more effortless and allow
for continuous deposition. Corners that will be post-processed must be given additional radii to
allow for the tool size [10].

Undesirable features. Some complex features is not suited to manufacture with the WAAM
process. Holes and cavities can be hard to build and should be avoided [6]. These features should
instead be removed and added in the post-process instead. Complex 3D lattices is not desirable for
WAAM and may be difficult to deposit and post-process. Hence, topological optimization should
be undertaken with the WAAM constraints in mind.

Topology optimization. To fully use the advantages of AM, a topology optimization software
can be used. They are able to optimize the material distribution within a given design space for
a given set of load- and boundary conditions. If they are designed such that they represent the
real load cases, the weight of the component can be minimized. Since this can result in complex
component structures, the tool is well suited to combine with AM that allows a great geometrical
freedom [25].

2.2 Product development

The product development process through the project is based on the process described by Ulrich
& Eppinger [7]. The process consists of different steps and activities which is divided into six
phases. The primary steps are described below and can be seen in Figure 8.

Figure 8: Shows the six phases of the product development process [7].

8
2.2.1 Planning

The first step of the process is called Phase 0 which includes the planning of the project. A time
and resources planning is made to determine whether the project can be carried through. The
output from the phase is a plan over the time and resources together with a projects mission
statement that specifies the product market, goals and constrains.

2.2.2 Concept development

In this phase the target needs are identified and translated to requirements for the product. When
the requirements for the product is established, the next step is to generate concepts. This means
generating ideas on how the requirements and demands can be fulfilled. The concepts then needs
to be evaluated and eliminated to make a decision on witch solution that is best suited to solve the
problem. Ulrich & Eppinger [7] describes a method to do this. The first step is called “Screening”,
and is done to make a quick reduction of the concepts. The concepts is ranked against each other
on a number of decided criteria to decide which concepts is good enough to keep to the next step.

The next step is called Scoring, which gives a more accurate evaluation of the remaining concepts.
The team rates the importance for each criteria. Each concept then gets a score for each criteria,
which then is multiplied with is importance. The total score for the concept is determined by the
sum of all criteria. Then the result can be a basis when the team decides which concept to proceed
with.

2.2.3 System-level design

In this phase, the focus is on the product architecture. The system-level design shows how the
subsystems and components are connected, what functions every system has, and which part is
connected to each function.

2.2.4 Detail design

In the detail design phase, the product is defined with exact geometry, material selection and
tolerances for each detail. The output of the detail design is a full documentation on the product.
This documentation can consist of drawings or other digital files that describes the geometry,
simulations, specification of purchased parts, a production plan and assembly descriptions.

2.2.5 Testing and refinement

The next step is to test the actual performance and reliability of the product. Early prototypes
are built with the intended geometry, but not necessarily the intended manufacturing process,
to determine whether the product will perform as intended and satisfy the requirements. Later
prototypes is then manufactured according to the decided manufacturing process. These steps are
made to test the concept and identify necessary engineering changes for the final product [7]. The
last phase called “production ramp-up” won’t be covered in this project.

9
3 Methodology

This section describes the methodology for the WAAM experiments and the torque arm development.
These have been ongoing parallel by guiding each other in the right direction with their results.
The WAAM experiments have been designed to suit the design of the torque arm while the design
of the torque arm have been modified according to the results from the tests.

3.1 The applied hardware

Figure 9 shows the WAAM-lab at BAE used in this project. It consists of an IRB 2600 industrial
robot with an IRC5 control system and a Fronius TPS 500i welding power source. Attached on
the robotic arm is a Fronius WF 60i Robacta Drive CMT weld gun and the worktable consists of
a 2-axis positioner called IRBP A250. The positioner can be used for coordinated movements, but
is mostly used to change the build direction for parts to prevent building with overhangs.

Figure 9: The robot cell with an IRB 2600 robot, an IRBP A-250 positioner, a Fronius TPS 500i
power source and a Fronius WF 60i Robacta Drive CMT weld gun attached to the robot.

3.1.1 Welding wire and shielding gas

Tensile test have been made by S. Dahat [26] for a specific wire, ESAB OK Tubrod 14.03
1, 2mm. The shielding gas used was Mison 18. The results from [26] showed a yield strength
for the wire to be 720 MPa as worst. This value has been used when developing the torque arm
and the same wire have been used through the project. Also the working envelope stated by Dahat
have been used. The span of the bead width (LW) is presented in Figure 10. The width depends
on the current (I) and travel speed (TS).

10
Figure 10: Shows the expected bead width for different input of current (I) and travel speed (TS)
[26].

3.2 Workflow

There is no typical workflow for the WAAM process. In Figure 11 a suggestion for the workflow is
shown. This method have been tested and evaluated from previous work at BAE [6]. The method
starts with a CAD model of the solid component design in the software CATIA V5. The model is
exported as a STL-file into Simplify3D where the model is sliced into a set of layers and the tool
deposition-path is generated. The layers and deposition-path is generated in a G-code, which is
the input to RobotStudio. RobotStudio is the software that controls the printing process and also
simulates the program before printing, to see if there are any problems. In the program the weld
parameters is stored and the path orientation can be edited. The last step in RobotStudio is to
generate a RAPID-code that controls the robotic arm, weld and worktable. When the component
is printed the last step is the post-processing, because of the near net shape output.

11
Figure 11: Shows the used workflow for the WAAM process.

Catia V5

Catia V5 is a multi-platform software developed by Dassault Systems for computer-aided design


(CAD), computer-aided manufacturing (CAM) and computer-aided engineering (CAE). In this
project, the software is used to design models. Offset can also be added here, and the part can be
split into sub-parts.

Simlify3D

Simplify3D is a so called “slicing” software that generates G-code containing toolpaths and instructions,
from a 3D model. The software is mostly developed for plastic 3D printers, but has in this project
been used to generate G-code that is later translated to RAPID-code.

RobotStudio

This is ABB’s robot programming tool, used for offline programming in a virtual 3D environment.
This was used together with the newly released “3DP add in” to create RAPID-code from the
G-code files generated in Simplify3D.

Abaqus

This is a software used for finite element analysis (FEA) and CAE, also developed by Dassault
Systems. This was used for topology optimization in its optimization suite called Tosca. It has
also been used to evaluate concepts using the finite element method (FEM).

3.2.1 Testing

Initially, the first tests were made with welding parameters given from University West to validate
the values and measure the resulting bead geometry and layer heights. These values were then
used for all the tests during the rest of the project. Based on the results from the literature study
and the already stated challenges from [17], a preliminary testing plan were established. Since
the design of the torque arm were developed simultaneously, different possible sequences for the
manufacturing process were discovered accordingly. The testing plan was then modified as the
different sequences became known, together with the results from the previous tests.

12
3.3 Torque arm development

In this section the method to develop and manufacture the torque arm with WAAM is presented.

3.3.1 Requirement specification

To compile a requirement specification, the relevant requirements for the vehicle was used. Additional
requirements was added also, such as that the torque arm needs to be able to manufacture with
WAAM. Some comparisons were also made with the already existing old design to make sure that
the new design will perform better. Using a given operational profile for the vehicle, calculations
were made to decide how many cycles the torque arm need to handle to resist fatigue. In this
report, the requirement specification is not presented due to secrecy.

3.3.2 Concept generation

To generate concepts for the torque arm, it was first divided in to sub parts. The parts defined as
these sub parts was the fastening solutions on both ends of the torque arm and also which design
space to define to the topology optimization. Concepts for these sub parts were then generated
separately using brainstorming. Since these were independent of each other, multiple combinations
of concepts could be composed. The results for the concepts can be seen in Appendix A.

When concepts for each sub part was generated, rough CAD-models was designed to ease the
concept evaluation. Then, to evaluate them, the Screening and Scoring described in Section 2.2.2
was used. This was done for each of the sub concepts individually to finally combine the winning
ones. First a screening was done to get an overview of how the concepts will perform compared
to each other. Here different important criteria were defined which the concepts is compared
according to. Since there was a limited amount of concepts generated, only one was removed in
the Screening.

For the Scoring, different weights for each criteria were defined. In order to achieve as objective an
assessment as possible, criteria for each score were defined as well. Each criteria for each concept
were then scored according to these definitions. The score was then multiplied with the weight of
the criteria. The total score for each concept was summarized by the score of each criteria. The
concept for each sub part with the highest score was then implemented to a final winning concept.
The result from the screening and scoring can be seen in Appendix B.

3.3.3 Detail design using topology optimization

As mentioned before the topology optimization was done in Abaqus. The design which should be
optimized was designed in Catia V5 and, with a STL-file, imported to Abaqus. The first step was
to create partition for the sections that should be optimized and for the ones that should be frozen.
Boundary conditions and loads was applied on the torque arm together with a suitable mesh. For
the optimization a separate task was created for the whole model. Two design responses was
created, one for strain energy and one for volume. An objective function was set to minimize the
maximal strain energy. The last step was to set a constrain value for the design response volume.
This step was iterative with different constrain values to find a suitable design with limited stress.
The topology optimization was used to get an alternative design with reduced weight. When
designing the torque arm, the optimization result was used as a template. Further FEM-analysis
were made for the new design.

13
4 Result

In this section the results from the experiments is shown. Also, the result for each step in the
torque arm development work flow is presented. This describes how all steps presented in Section
3.2 was applied and tuned.

4.1 Experiments

Test 1

The result from testing the parameters from University West as input is shown in Figure 12. Three
beads was places on the substrate and was given a number. The travel speed (TS) of the weldgun
was set to 5 mm/s for all tests. Therefore the width and height of the beads should be the varying
result due to the wire travel speed (WTS). The input parameters and the measured output data
is presented in Table 1. For the purpose to print a quite small component, bead number 1 showed
the most suited result since it gave the smallest width and lowest heat impact. Henceforward, the
parameters used on bead number 1 were used in all the tests.

Figure 12: The result for testing the parameters from University West.

Table 1: The input and results for the test of parameters from University West.

Input Output Result


TS WTS I WTS I Height Width
Number
(mm/s) (m/min) (A) (m/min) (A) (mm) (mm)
1 5 3.6 132 4 127 2.6 7
2 5 5 178 5.4 161 2.9 9
3 5 4.5 160 4.9 152 3 8.3

14
Test 2

Figure 13 shows the result of Test 2, where the start/end-time and start/end-current were varied.
Start time and end time defines as when the weld gun stands still in the start/end-points for a
certain time. When standing still, the current can either be raised or lowered compared to when
the weld gun is moving. This is made to make the bead more even in the start and end. The input
and measured data is presented in Table 2. Bead number 5 gave the best result. From Figure 13
it can be seen that the whole bead has an equal width and height. Henceforward, the start-and
end parameters used in bead number 5 were used in all the tests.

Figure 13: The result from testing different start/end-time and start/end-current.

Table 2: The input and measured data from testing different start/end-time and start/end-current.

Input Result
Start time Start current End time End current Height Width
Number
(s) (%) (s) (%) (mm) (mm)
1 0 0 1 25 2.7 7
2 0.3 135 1 50 2.7 7
3 0.3 110 0.6 75 2.5 7
4 0.2 100 0.8 60 2.8 7
5 0.3 110 1 50 2.7 7

Test 3

The results from Test 3 is shown in Figure 14. The purpose of this test was to see the continuity
for the bead width and layer height. The test resulted in an average first layer height at 2.9 mm
and 2.1mm for the rest of the layers. The width had an average at 7.1 mm. Notable is that is
the width has small dents which will affect the measured result. The input and measured data is
presented in Table 3.

15
Figure 14: The result for five beads with the same input parameters, with five layers where the start
point for each layer was alternated.

Table 3: The input data and the result for the built walls.

Input Result
Number of First layer Layer height First layer Height Width
Number
layers height (mm) (mm) height (mm) (mm) (mm)
1 5 2.6 1.9 2.7 11.6 7.2
2 5 2.6 1.9 2.8 11 7.1
3 5 2.6 1.9 3 11.6 7
4 5 2.6 1.9 3.1 11.3 7
5 5 2.6 1.9 3 11.2 7.1
Average 5 2.6 1.9 2.9 11.3 7.1

Test 4

For test 4 different patterns and overlaps for solid cubes was evaluated and the result can be seen
in Figure 15. The test showed that the outer boundaries had the best result when an outer bead
is laid before the infill. The best infill of the ones available in Simplify3D showed to be the zigzag
pattern. For the infill parameters, number 4 and 5 gave best result. In Table 4 the input and
measured data of the result is presented. The height was hard to measure when there was too
much overlap in the middle, because then the height did not become constant over the top surface.

Figure 15: The result for testing different patterns. 1) Zigzag pattern. 2) concentric pattern. 3-6)
Zigzag with one outer wall.

16
Table 4: The input data and measured result for different infill and overlap.

Input Result
Number Layer Height Width Bead Beads in Overlap Overlap Height Width
Number
of layers height (mm) (mm) (mm) width (mm) width outer bead (%) (%) (mm) (mm)
1 4 2.9 11.6 65x65 4.9 13 - 35 11 65x67
2 4 2.9 11.6 65x65 4.9 11 0 35 11.1 68x68
3 4 2.9 11.6 65x65 4.9 6 - 35 11.2 67x67
4 5 2.9 14.5 47x47 5.25 9 - 30 13.8 49x49
5 5 2.9 14.5 51x51 6 7 25 20 13.2 52x52
6 5 2.9 14.5 43x43 6.75 5 35 10 13.2 43x43

Test 5

In test 5, walls with overhang was tested. The result shows that it is possible to build a wall under
30 degrees overhang with the parameters used. For the test with 35 degrees overhang, number 2,
the last layers failed. The reason for that was that height deviated to much for each layer. The
weld floated out on the sides as can be seen in Figure 16. The input and measured data for the
test is presented in Table 5.

Figure 16: Test of single bead overhang.

Table 5: Input parameters and measured data for overhang wall.

Input Result
Number Layer Overhang Height Top diameter
Number
of layers height (mm) (deg) (mm) (mm)
1 18 1-3: 2.9 4-18: 1.9 25 39.3 49.4
2 18 1-3: 2.9 4-18: 1.9 35 33.8 66.4
3 18 1-3: 2.9 4-18: 1.9 30 38.1 60.5

17
Test 6

The result of test 6 is shown in Figure 17. The three crosses was printed with different overhang,
see Table 6. The test showed that it is possible to do inner and outer overhang. For number 3,
with 35 degrees overhang, the print failed. For the last layers the weldgun missed the outer corners
and the wire hit the substrate instead. Though, it was possible to continue but the final result had
its defects. The tests shows that the resulting height varies for different overhangs. None of the
test succeeded to get the height that the model had. Best was test number 1 as had the height 64
mm which deviated 0.8 mm from the model height. Worst was number 3 which deviated 2.3 mm
from the model height.

Figure 17: Result of overhang with a solid model.

Table 6: Input parameters and result of overhang with solid model.

Input Result
Number Overhang Layer Height Overlap outer Overlap Height Overhang
Number
of layers (Deg) height (mm) (mm) bead (%) (%) (mm) (deg)
1 24 25 2.7 64.8 25 20 64 30
2 24 30 2.7 64.8 25 20 64.2 35
3 24 35 2.7 64.8 25 20 62.5 40

Test 7

This test was performed to see if it was possible to tilt the worktable to print overhangs. The result
of the three runs is shown in Figure 18. First a solid rectangle (Number 1) was printed. After that,
the worktable was tilted 45 degrees and an angled solid rectangle (number 2) was printed. Finally
the worktable was tilted 90 degrees in the opposite direction and a wall (number 3) was printed.
The input parameters and measured data is presented in Table 7. The visible result is that it built
some extra material in the middle of number 1 and 2. This can be good to know when designing
the part with an offset for near net shape.

18
Figure 18: Result for a part with three different build directions.

Table 7: Input parameters and measured data for the different prints.

Input Result
Number Layer Height Width Bead Beads in Overlap Overlap Height Width
Number
of layers height (mm) (mm) (mm) width (mm) width (%) outer bead (%) (%) (mm) (mm)
1 15 2.7 40.5 36x36 5.5 7 25 20 41.5 36.5x36.5
2 18 2.7 - 36 5.5 7 25 20 11.1 36.5
3 29 2.1 60.9 36 7 1 - - 55 32

Test 8

Test 8 was the first try to print the torque arm. The print was divided into two prints. In Figure
19a the first print and layer direction can be seen in Figure 19b the direction for next print can be
seen where the worktable has rotated 90 degrees. Figure 19 shows the CAD model with dimensions
of the the important heights, widths and overhangs. The numerical values and measured data is
presented in Table 8.

Figure 19: Shows the CAD model of the torque arm for test 8. a) First build direction. b) Second
build direction where the worktable has rotated 90 degrees.

19
Figure 20 shows the printed result for the two prints to the torque arm. All layers were printed
with 20% overlap at a temperature below 50◦ C. The print gave various results. The result for the
all the widths is around 1 mm larger than the CAD model as can be seen in Table 8. The height
for this test varies in different sections but overall it is lower than the cad-model. h2 had the worst
result, that was 13 mm shorter than the cad-model. The reason is that the part is the thinnest
in that section. When slicing it, there was not enough space to do infill. Another parameter that
may have caused the lower height is the overhang of h1 since overhangs has showed a tendency
to collapse. For h3 the height difference became 5 mm. That section had infill until the last two
layers where a single bead was extruded. For the overhang in print two, it also collapsed a bit.
This resulted in that it missed one outer line and weld floated over the edge, see the right picture
in Figure 20.

Figure 20: Shows the printed torque arm for test 8.

Table 8: Input parameter and measured data for the first test of the torque arm. The numbers in
the table refers to the number next to w and h in Figure 19.

Input Result
Height Width Overhang Height Width Overhang
Number
(mm) (mm) (Deg) (mm) (mm) (Deg)
1 30 209 30 27 210 36.7
2 84 12 - 71 13 -
3 106 13 - 101 14 -
4 71.5 73.5 - 74 74.5 -
5 6 39 - 7 41 -
6 - 78 - - 80 -

One problem with the layer height arises when slicing with only one process. Figure 21 shows four
consecutive layers from test 8. The first layer beads fills the whole layer. But for the next three,
the geometry becomes smaller and the number of beads do not match the geometry. Therefore, the
automatic slicing skips the last bead and the layer gets a gap. This can be solved with individual
layer settings for certain layers.

20
Figure 21: Shows the problem with slicing with only one process for different layer geometries.

4.2 Design process

In this section the result and strategy for DfAM is presented. The first step is to set the design
that further will be optimized. This is done from the concept generation and selection in Appendix
A and B. Figure 22 shows the design space that will be optimized in grey. The blue areas is frozen
which means that they will not be affected by the optimization. The frozen area for the arm is the
attachment points and the outer boundaries of the design space is limited by other parts in the
vehicle.

Figure 22: Shows the design to topology optimization, the blue area is frozen and the gray is the
design space for the optimization.

21
The next step in the process is topology optimization and the result is presented in Figure 23. The
optimization reduced the weight of the arm by 50 % and the maximum stress becomes 800 MPa
for a certain node. However, this result is more used to see a possible designs.

Figure 23: Shows the result from topology optimization.

The result from the topology optimization and the testing resulted in the final design of the torque
arm that can be seen in Figure 24. The intent is to implement a bushing where the big hole is
designed, to allow some movements in the system. The size of the hole can be changed depending
on the decided bushing.

Figure 24: Shows the final design of the torque arm.

Iterative with the design of the torque arm, finite element analysis was performed. The result
can be seen in Figure 25 with the maximum stress of 660 MPa at maximum load. The maximum
deformation becomes 1 mm.

22
Figure 25: Shows the result of the final FEM analysis.

The next step was to design the torque arm for the slicing, see Figure 26. Features like holes,
chambers, and thin walls were removed to get a better result from the slicing and the print.
Offset was added to the whole part and the minimum offset was 2 mm to give margin to the
post-processing. For constant travel speed as the software currently provides, the layer thickness
becomes constant and it can be hard to match the layers with the geometry. Therefore, extra
vertical offset were applied on the part. However, with more time and tests, the added offset could
be trimmed significantly.

Figure 26: Shows the modified design for slicing.

Figure 27 shows the final design of the torque arm in grey compare to the design for slicing is
shaded. Before slicing, the substrate needs to be selected and designed. In this case, the substrate
for the torque arm is integrated with the design. The thickest part of the design was chosen to
reduce the print time and also the thermal stress. When designing the substrate offset is added
to fit the printed area. This could also be trimmed when more experience is gained. Due to the
thickness of the substrate three attachment points was designed to be able to mount it to the
worktable.

23
Figure 27: Shows the final design of the torque arm compared with the design for slicing design
with offset. Both designs is placed on the substrate.

4.3 3D Slicing

From the testing, overhang showed a tendency to collapse which resulted in a deviating layer
height. To prevent this, the slicing was divided into three parts to be able to rotate the worktable
and always print in vertical direction. Figure 28 shows the three parts. Part 1 in red is the first
print. Part 2 in blue is sliced so that the worktable is rotated 90 degrees and is printed onto Part
1. For Part 3 in yellow, the worktable rotates back and is printed onto Part 2.

Figure 28: Shows the design for slicing spited into three parts, Part 1 in red, Part 2 in blue and
Part 3 in yellow.

24
Before starting to slice, the parts position need be thought out. When the G-code is exported,
the coordinate system from Simplify3D will follow. This will later be aligned with the workobject
in RobotStudio. A suitable reference point on the substrate needs to be chosen to calibrate the
workobject to where the substrate is placed in reality. To calibrate the workobject, the robot can
simply be jogged to the reference point to se if the values match up. A corner is a good point to
place the coordinate system as can be seen in Figure 29.

Figure 29: Shows Part 1 and selected position of the coordinate system.

From the testing, the zigzag pattern with an outer edge (Hybrid path) showed best result with
20-30 % overlap. To get the slicing match the geometry for Part 1, 12 processes was needed for the
50 layers as can seen on the different colours in Figure 30. Each process had different parameters
to fit the geometry. This method gave good result but was time-consuming.

Figure 30: Shows Part 1 divided into 12 processes with different colours.

For Part 2 the coordinate system was selected to match a corner on Part 1, see Figure 31a. This
was done for the same reason as mentioned before, to easier calibrate the workobject when the
worktable is rotated 90 degrees. Part 2 had a simpler geometry than Part 1 so the slinging could

25
be done with one process as can be seen in Figure 31b. The infill overlap was set to 29 % and to
the outer line the overlap was set to 25 %.

Figure 31: a) Part 2 with the position of the coordinate system. b) Shows the slicing with one
process.

For Part 3 the same strategy and input was used as for Part 2, see Figure 32a. The result gave
some staircase effect because the radii, see Figure 32b. This was notable but it should not be a
problem when the area later on is post-processed.

Figure 32: a) Part 3 with the position of the coordinate system. b) Shows the slicing with one
process.

4.4 RobotStudio

To begin with the simulation in RobotStudio and later create the RAPID-code, the substrate is
the first object to import and a position is chosen to mount it on the worktable. Next step is to
create a workobject for Part 1 on the substrate where the coordinate system from the slicing is

26
localized. When the part has the correct position, settings in the ”3DP add in” is selected and
simulated to see if the print is successful. Figure 33a shows the position of the substrate with Part
1. For Part 2 and 3 the process is repeated and a new workobject is created for their positions, see
Figure 33b and c. When the print is simulated, the RAPID-code can be exported to the robot.

Figure 33: a) Position of the substrate with Part 1. b) Position of Part 2 with the worktable tilted
90 degrees. c) Position of Part 3.

4.5 Robot Print

As mentioned, before printing, the workobject needs to be calibrated to match the substrate. This
is done by jogging the robot to the reference point and read if the coordinates of the workobject
is (0, 0, 0). If not, the values can simply be adjusted to make them match. When the first part is
printed, the same procedure goes for the other parts. The result of the print can be seen i Figure
34. The whole print was successful and the arm got correct geometry. The layers became smooth
and the slicing method gave a good result unlike the previous try to print the torque arm. The
thermal stress had a minimal effect on the substrate and because of the offset, it shouldn’t be any
problem to post-process.

Figure 34: Shows the printed result of the torque arm.

An approximate model of the print result have been designed. A comparison of the printed torque
arm with the final design can be seen i Figure 35. The shaded area is the material that needs to
be post-processed. The most material to remove is from the substrate because a simpler substrate
was manufactured due to lack of time. The print also resulted in extra material in height which
was expected. This because the input layer height was chosen to be a bit lower than the real value
to make sure that the print would get high enough.

27
Figure 35: Shows the result of the print compared with the final design.

The test showed one problem when using the method multi-direction slicing when rotating the
worktable. When printing, sharp edges gets rounded which becomes a problem when another part
is to be built on the previous. It is hard to find the correct position of the coordinate system to
build the next part. Figure 36 shows the problem with the multi-direction slicing. An offset of 1-2
mm for the surface would have solved the problem.

Figure 36: Shows the result and problem with the multi-direction slicing.

28
It is hard to see from the test results. Figure 37a illustrates the problem and Figure 37b shows a
solution where offset can be applied in the surface that corresponds to the coordinate system of
Part 2.

Figure 37: Shows two built parts with the white arrows as their respective build direction. a)
Highlighting the problem with the rounded edge. b) Highlighting the offset that needs to be added
on Part 1.

29
5 Discussion and conclusions

To summarize the project, WAAM is a manufacturing method with lots of opportunities even
though it needs to be further developed. The current design of the torque arm was possible to
manufacture, though with a noticeable amount of material offset due to lack of time. With more
tests and refinements, a more near net shape print would be possible to achieve. Some more
advanced features in the design could also be achieved.

However, the printed torque arm still needs to be post processed to evaluate if the print is accurate
enough. When post processed, it also needs to be tested to see if it performs as stated in the
requirement specification. Regarding the questions stated for the project, they can be answered
as follows:

Which of the stated challenges is most relevant to solve to


manufacture a torque arm with WAAM?

The conclusion from this project is that all challenges stated in Section 2.1.2 needs to be considered,
more or less, when building a torque arm with WAAM. However, most of them such as the ”starts
and stops” and ”cross of paths”, were solved with the automatic slicing in Simplify3D. To verify
the ”Voids in solid layer” point, further investigation of the torque arm needs to be done. However,
according to the information found in previous works, this should be solved by the overlapping
beads.

How can these challenges be solved?

As mentioned, some of the challenges is solved with the automatic slicing in Simplify3D. However,
some of them needs to be further considered. To obtain the best possible result, overhang should
be built using an external positioner as mentioned in Section 2.1.2, multi-direction slicing. In that
way overhang can be built while always building in a vertical direction. However, it is possible to
build overhang up to a limited angle.

To reduce the problems regarding the thermal stress built up during the process, several actions
can be taken. A suitable substrate needs to be used. To make sure that it doesn’t deform, it needs
to be thick enough. For the torque arm, the thickest part of the design were chosen to be the
substrate integrated to the design. ”Interlayer cooling” were also used through the project. This
was done manually using compressed air, but it is not a viable method in production. Because of
the design, symmetric building and selective deposition were not used. However, these methods
could also be useful for some designs.

Are there any other challenges to solve?

During the project, some additional problems were discovered. The first problem could be connected
to the ”viods in solid layer” item. When printing with constant weld parameters, the width and
height of the bead will be constant. Since the geometry for each layer on the torque arm is different,
this will cause a gap when the bead geometry doesn’t match the layer geometry as can be seen
in Figure 21. Due to the graphical representation of the toolpath in Simplify3D, this could be
discovered already in the slicing phase. To prevent this, several processes were created. In each
process, the overlap were adjusted to match the layer geometry. This was a time consuming process
but gave a good result.

A problem that was discovered with the welding wire used, was that slag occurred on the surface
between the layers. Sometimes this caused a stop in the program since the robot failed to get

30
contact and create a short circuit. In this project this was solved by manually remove the slag.
This was time consuming and requires an operator to monitor each layer, hence it’s not a viable
solution.

Another problem is the input of the layer height. To receive an accurate result of the print height,
it is crucial to have a correct input of the layer height. If the input is higher than the actual layer
height, the total print height will get lower than the CAD model. Hence, it is better to input a
lower layer height than the actual since this will result in a higher print and gives a safety margin
for the post-processing. However, with more testing and experience this should not be a problem.

What aspects should be considered when designing a torque arm for


WAAM?

Since the time weren’t enough to find a good solution to build overhang, these features were
removed where it was not possible to use the positioner. All the holes were also removed to drill
them afterwards instead.

As mentioned, material needs to be added give room for the post-processing. For the torque arm
this is no exception. For each feature, offset needs to be considered with the post processing in
mind.

Also, the different build directions during the WAAM process needs to be considered. This because
the printed area gets reduced due to the fact that corners gets rounded, which means that the area
is reduced as described in Figure 37. If multi-directions method is used on an already printed area,
it is important that the area is big enough.

5.1 Future work

As mentioned, there are some things that need to be investigated and evaluated more. This section
presents what was discovered in this project, and also some proposals on how it could be solved.

The slag that results in a stop in the program is something that need to be solved. By doing some
quick tests with other weld wires, it showed that it could be solved by using other wires. Hence,
other welding wires should be considered for this reason also. Another problem was the limitations
in the slicing software Simplify3D. Other options should be considered to be able to try if other
possible infills could be even better.

The results from test 5 and 6 showed that it is possible to build overhang. But with the parameters
used in this project, the overlap angle became about 5◦ higher than the input angle. This led to a
shrinkage in the total height of the component, hence overlap were removed from the design of the
torque arm. More tests needs to be done to investigate what parameters is suitable for building
overlap. For example, the tests showed that the overlap could be influenced of the additional
overlap of the outer contour bead. One proposal is also to try other welding wires, since a smaller
layer height should mean a higher possible overlap. Also, in this project, the weld gun has always
been welding in a vertical direction. Tests should be done to investigate if better results could be
received if the weld gun is rotated to be aligned with the overlap angle.

This project have not been focusing on investigating how cold it needs to be to start printing a new
layer. This is something that should be investigated since it could spare a lot of manufacturing
time to trim the waiting time. Some alternatives to speed up the cooling and make it automatically
controlled should be considered. Some proposals on how this could be done is presented below.

31
• A temperature sensor could be mounted on the robotic arm. In this way, a maximum
temperature could be set as a limit to start a new layer.

• Investigate the opportunity to blow compressed air through the nozzle of the weld gun. If it
is possible, it could be a solution to automatically speed up the interlayer cooling.
• The size of the substrate is affecting the cooling time, investigate how this can be optimized.
• Cooling channels could be implemented in the substrate.

• Multiple parts could be built simultaneously. In this way, one part has time to cool down
when another is printed.

Note that if the cooling is speeded, the properties of the material could be affected. Hence, this is
something that needs to be considered if some of these methods should be implemented.

As mentioned, the slicing process was very time consuming due to the need of multiple processes
with different settings. If a database is built up with different welding parameters with its
corresponding layer with and bead height, maybe it would be possible to implement these in
the slicing. To make the slicing more accurate and reduce the need of human modification, the
slicing need to be able to adjust the layer height and bead width to make it match the geometry
of the layer.

32
References

[1] F. Martina and S. Williams, “Wire+arc additive manufacturing vs. traditional machining
from solid: A cost comparison”, 2015.
[2] T. Wohlers and T. Gornet, History of additive manufacturing. Fort Collins: Wohlers Associates,
Inc., 2016.
[3] D. Ding, Z. Pan, C. Dominic, and H. Li, “Wire-feed additive manufacturing of metal components:
Technologies, developments and future interest”, The International Journal of Advanced
Manufacturing Technology, vol. 81, no. 1-4, pp. 465–481, 2015. doi: 10 . 1007 / s00170 -
015-7077-3.
[4] G. Goh, S. Agarwala, G. Goh, V. Dikshit, S. Sing, and W. Yeong, “Additive manufacturing in
unmanned aerial vehicles (uavs): Challenges and potential”, Aerospace Science and Technology,
vol. 63, pp. 140–151, 2017. doi: 10.1016/j.ast.2016.12.019.
[5] Y. Zhang, A. Bernard, R. Kumar Gupta, and R. Harik, “Evaluating the design for additive
manufacturing: A process planning perspective”, Procedia CIRP, vol. 21, pp. 144–150, 2014.
doi: 10.1016/j.procir.2014.03.179.
[6] T. Jonsson Vannucci, “Investigating the part programming process for wire and arc additive
manufacturing”, Master’s thesis, Luleå: Luleå University of Technology, 2019.
[7] K. Ulrich and S. Eppinger, Product design and development, 6th ed. New York: McGraw-Hill
Education, Jul. 2016.
[8] J. Alison, C. Sharpe, and C. Seepersad, “Powder bed fusion metrology for additive manufacturing
design guidance”, Additive Manufacturing, vol. 25, pp. 239–251, 2019. doi: 10 . 1016 / j .
addma.2018.10.035.
[9] I. Gibson, W. David, and B. Stucker, Additive manufacturing technologies : rapid prototyping
to direct digital manufacturing. New York: Springer, 2010.
[10] H. Lockett, D. Jialuo, S. Williams, and F. Martina, “Design for wire + arc additive manufacture:
Design rules and build orentation selection”, Journal of Engineering Design, vol. 28, no. 7-9,
pp. 568–598, Aug. 2017. doi: 10.1080/09544828.2017.1365826.
[11] D. Ding, Process planning for robotic wire ARC additive manufacturing. Wollongong: University
of Wollongong, 2015.
[12] B. Cong, Z. Qu, H. Sun, G. Zhao, and J. Ding, “A Comparative Study of Additively
Manufactured Thin Wall and Block Structure with Al-6.3%Cu Alloy Using Cold Metal
Transfer Process”, Applied Sciences, vol. 3, no. 7, p. 275, 2017. doi: 10.3390/app7030275.
[13] A. Ghazanfari, W. Li, and M. Leu, “Adaptive rastering algorithm for freeform extrusion
fabrication proecesses”, Virtual and Physical Prototyping, vol. 10, no. 3, pp. 163–172, Oct.
2015. doi: 10.1080/17452759.2015.1096798.
[14] D. Ding, Z. Pan, D. Cuiuri, and H. Li, “A practical path planning methodology for wire and
arc additive manufacturing of thin-walled structures”, Robotics and Computer-Integrated
Manufacturing, vol. 34, pp. 8–19, Aug. 2015. doi: 10.1016/j.rcim.2015.01.003.
[15] M. Bertoldi, M. Yardimci, C. Pistor, and S. Güceri, “Domain decomposition and space
filling curves in toolpath planning and generation”, International Solid Freeform Fabrication
Symposium, 1998. doi: 10.26153/tsw/608.
[16] R. Dwivedi and R. Kovacevic, “Automated torch path planning using polygon subdivision
for solid freeform fabrication based on welding”, Journal of Manufacturing Systems, vol. 23,
no. 4, pp. 278–291, Aug. 2008. doi: 10.1016/S0278-6125(04)80040-2.
[17] F. Michel, H. Lockett, J. Ding, F. Martina, G. Marinelli, and S. Williams, “A modular path
planning solution for wire + arc additive manufacturing”, Robotics and Computer-Integrated
Manufacturing, vol. 60, pp. 1–11, May 2019. doi: 10.1016/j.rcim.2019.05.009.
[18] A. Sequeira, “Process control and development in wire and arc additive manufacturing”,
PhD thesis, Cranfield: Cranfield University, 2012.

33
[19] P. Kazanas, P. Deherkar, P. Almeida, H. Locket, and S. Williams, “Fabrication of geometrical
features using wire and arc additive manufacture.”, Proceedings of the Institution of Mechanical
Engineers, Part B: Journal of Engineering Manufacture, vol. 226, no. 6, pp. 1042–1051, 2012.
doi: 10.1177/0954405412437126.
[20] F. Wang, S. Williams, P. Colegrove, and A. Antonysamy, “Microstructure and mechanical
properties of wire and arc additive manufactured ti-6al-4v.”, Metallurgical and Materials
Transactions A, vol. 44, no. 2, pp. 968–977, 2013. doi: 10.1007/s11661-012-1444-6.
[21] L. Nan, “Design for wire and arc additive manufacturing: Development of design guidelines
and an automation design tool”, Master’s thesis, Cranfield: Cranfield University, 2015.
[22] S. Williams and F. Martina, “Wire + arc additive manufacturing”, Materials Science and
Technology, vol. 32, no. 7, pp. 641–647, 2015. doi: 10.1179/1743284715Y.0000000073.
[23] J. Bralla, Design for Manufacturability Handbook. New York: McGraw-Hill, 1999.
[24] J. Campbell, Castings Practice: The 10 Rules of Castings. Oxford: Elsevier/Butterworh-
Heinemann, 2004.
[25] E. Meli, A. Rindi, A. Ridolfi, R. Furferi, F. Buonamici, G. Iurisci, S. Corbò, and F. Cangioli,
“Design and production of innovative turbomachinery components via topology optimization
and additive manufacturing”, International Journal of Rotating Machinery, vol. 2019, 2019.
doi: 10.1155/2019/9546831.
[26] S. Dahat, “A methodology to parametrize wire + arc additive manufacturing applied to a
high strength low alloy steel”, Master’s thesis, Trollhättan: University West, 2019.

34

You might also like