Академический Документы
Профессиональный Документы
Культура Документы
By
Ing. R. Kuitert
Copyright R. Kuitert
All rights reserved
II | P a g e Remko Kuitert
Delft University of Technology
Author: ________________________________
Ing. R. Kuitert
IV | P a g e Remko Kuitert
Preface
Preface
In the last year of the Master Aerospace Engineering at Delft University of Technology in Delft, the student
will get the opportunity to perform a thesis project. During this graduation period a thesis has to be written.
This report represents my thesis for the Master of Aerospace Structures and Materials at the faculty of
Aerospace Engineering. I had the opportunity to perform my graduation project at the Delft Aerospace
Structures and Material Laboratory of Delft University of Technology in Delft. The main focus of this project
was to perform a feasibility study on the subject of incremental forming for double curved aluminum sheets
used in GLARE laminates.
This thesis would not have been possible without the guidance of several people who supported me
throughout the graduation period. Their support, feedback and critical eye contributed to the quality of this
thesis report.
In particular I would like to thank my girlfriend, Wouke van Veldhuizen, for standing by my side despite all the
setbacks during this thesis. I would also like to give a special thanks to my supervisor, Jos Sinke, for his
support during the entire project and valuable feedback. Without him the experimental setup would have
never been possible. I would also like to thank Gertjan Mulder and Micha Huizinga for all their help during
the design and construction phase of the experimental setup.
Finally, I would like to say thank you to my mother, my brother, my friends and fellow students for their
support during this Master thesis.
Remko Kuitert
Aalsmeer, The Netherlands
October 28, 2016
Summary
Although GLARE is now applied on the A380 and considered as a material for future aircraft, several cost and
production issues still have to be solved for future applications of this material. GLARE panels are currently
being produced by a hand lay-up and self-forming method, which is a suitable production method for the low
volume production line of e.g. the Airbus A380. But even with this method the production of double curved
GLARE panels remained a problem as wrinkles occur if the aluminum sheets become too large.
In the aircraft industry double curved metal sheets with large radius, for e.g. fuselage application, are
currently made by stretch forming. This process includes the use of a die which is product specific. But for the
production of GLARE the use of dedicated dies becomes too expensive. At the moment several processes
exist, but none of these processes meet the standards for accuracy and/or cycle time yet. Therefore this
research focused on a new type of process, the incremental sheet forming process, in order to determine its
feasibility for the application of preforming metal sheets.
As a start an analysis was made what kind of production rates and accuracy are to be expected. It turned out
2
that the production rate is in the range of 72-129 m /hour and the required accuracy should be at least 90%
of the forming depth. Thus for the process to be feasible a combination of a feed rate of 600 m/min and a
step size of 0.5 mm/pass is needed at a single machine to meet these requirements.
In order to verify if this is feasible an experimental approach was chosen, due to a lack of accurate computer
models. This started by designing and constructing an incremental forming machine, which was required for
the experiments. This machine was capable of operating at feed rates of 0-3000 mm/min and step sizes of at
least 0.1 mm/pass or larger. The workable area was limited to 100x100 mm by both the machine and the jig,
which was used to clamp the specimens. The specimens used in the Thesis were made of aluminum 2024-T3
and were 0.4 mm thick, which is the same as the aluminum layers used in GLARE.
The experiments were split up in four categories. The first category was the geometrical experiments. These
experiments showed that despite the change of geometry, the repeatability of the process was in the range
of 0.25 mm. The second category of experiments was related to the feed rate. These experiments provided
information of the effects of higher feed rates. It was shown that if the feed rates increase the dimensional
accuracy decreased. This implies that the feed rate could not be increased as easily as thought. The third set
of experiments was related to the step size experiments. These experiments showed that a small step size
gives a higher accuracy and vice versa. The fourth and last set of experiments was the geometric correction
experiments, which were used to increase the accuracy of the process by iteration. It was shown that after
only a single iteration the specimens achieved 97.2% of the input model forming depth, while only 65% was
achieved without iteration.
Even though relative high accuracies were obtained the conclusion of this thesis is that it is not feasible for
single point incremental forming to be applied for large scale industrial applications, like the Airbus A320
production line, in this form. It is therefore recommended that future research should focus on different
strategies for incremental forming, such as an origami approach or the use of multiple forming tools
simultaneously. Also it is suggested that more research should be done for a better prediction model in order
to avoid the iteration step. Last, but not least, it is also recommended that the current experiment setup
should be replaced by a more professional setup in order to improve the control of the spindle speed for
example.
VI | P a g e Remko Kuitert
Table of Contents
Table of Contents
REFERENCES ................................................................................................................. 62
APPENDICES ................................................................................................................. 65
List of Figures
Figure 2-1 GLARE Layup [8] ................................................................................................................. 12
Figure 2-2 Different GLARE Grades with their fiber orientation and their specific constituents [16] 14
Figure 2-3 Post stretching effect with Stress Strain curve [20] .......................................................... 15
Figure 2-4 SPIF setup (left) and TPIF setup (right) .............................................................................. 18
Figure 2-5 Back pressure SPIF (left), partial die SPIF (center) and full die SPIF (right) ....................... 19
Figure 2-6 Incremental Stretch Forming with a die ............................................................................ 19
Figure 2-7 Single Point Incremental Forming Inaccuracy [5] .............................................................. 20
Figure 4-1 A 3-axis router machine (left) vs. 3-axis knee milling machine (right) [37] [38] [39] ........ 28
Figure 4-2 HBM BF16 Drilling/Milling machine [39] ........................................................................... 30
Figure 4-3 Conventional Y-axis close up [38] ...................................................................................... 30
Figure 4-4 Catia sketch of the Y-axis stepper motor fitting ................................................................ 31
Figure 4-5 Complete CNC Y-axis of the HBM BF16 ............................................................................. 31
Figure 4-6 Conventional X-axis Close up ............................................................................................. 32
Figure 4-7 Additional Coupling Piece X-axis ........................................................................................ 32
Figure 4-8 Complete CNC X-axis of the HBM BF16 ............................................................................. 32
Figure 4-9 Conventional Z-axis close up [38] ...................................................................................... 33
Figure 4-10 Complete CNC Z-axis of the HBM BF16 ........................................................................... 33
Figure 4-11 Complete Conversion of the HBM BF16 .......................................................................... 33
Figure 4-12 Stepper motor Nema 23 3 Nm [43] .............................................................................. 34
Figure 4-13 Leadshine DM556 - 2 Phase Digital Stepper Drive [42] ................................................... 35
Figure 4-14 PSU 48VDC 6,7A [41] ....................................................................................................... 35
Figure 4-15 Arduino Mega 2560 [38] .................................................................................................. 36
Figure 4-16 Ramps 1.4 Shield [39] ...................................................................................................... 36
Figure 4-17 1SL02 IP66 Wall Box [40] ................................................................................................. 36
Figure 4-18 3D printer wiring scheme [46] ......................................................................................... 37
Figure 4-19 Custom design wiring scheme ......................................................................................... 37
Figure 4-20 Complete Electrical Control Box ...................................................................................... 37
Figure 4-21 Repetier Host interface.................................................................................................... 38
Figure 4-22 The jig............................................................................................................................... 39
Figure 4-23 Example of an aluminum 2024-T3 0.4 specimen............................................................. 39
Abbreviations
Term Description
2024-T3 Aluminum 2024 with T3 temper
3D Three Dimensional
ARALL Aramid Reinforced Aluminum Laminate
CAD Computer Aided Design
CARALL Carbon Reinforced Aluminum Laminate
CNC Computer Numerical Controlled
DASML Delft Aerospace Structures and Material Laboratory
FML Fiber Metal Laminate
G-Code General Code
GLARE Glass Laminated Reinforced Epoxy
KU Katholieke Universiteit (Catholic University)
LVDT Linear Voltage Displacement Transducer
M6/M8/M12 Metric bolt with 6/8/12 mm diameter
PSU Power Supply Unit
R Radius
Rev Revolution
SLC Structure Laminate Company
SPIF Single Point Incremental Forming
Sym Symbol
TPIF Two Point Incremental Forming
TU Technical University
UV Ultra Violet
Chapter 1 Introduction
1.1 Background
Since the application of aluminum in aircraft structures around 1930, sheet metals were used extensively
within civil aircraft. The advantages of having high specific properties provided the material to be
economically very attractive. Even today the amount of aluminum used in modern aircraft is still significant
compared to other materials.
However in the 1950s several accidents occurred due to metal fatigue. This resulted in the development of
new aluminum alloys that provided slightly better properties for corrosion resistance (7000 series) and
fracture toughness (2000 series). But these changes did not lead to any significant improvement of the
fatigue life of the material [1].
The real breakthrough for metal fatigue came from Rob Schliekelmann. After experimenting for a few years
at De Havilland with metal bonding, he switched to Fokker. He introduced the metal bonding technique in
the Fokker F-27 project. During fatigue tests with the F-27 wings it was discovered that the metal bonded
parts had favorable resistance to fatigue crack growth, simply because the crack only grew in a single layer.
This effect considerably slowed down the crack growth in the complete structure [1] [2].
This discovery of the improved fatigue properties provided the basis for Schijve and Vogelesang, at Delft
University of Technology, to perform research in the field of bonded structures. In the 1970s they started
experimenting with adding Aramid fibers between the aluminum sheets, which provided even better
properties in terms of fatigue. But the ARALL material had very poor compressive properties. Therefore an
alternative had to be found to make the Fiber Metal Laminate concept successful in the aircraft industry [1].
As an alternative Delft University of Technology developed a new type of FML called GLARE, which was a
combination between glass fibers and aluminum. This material provided to be very promising for fatigue
sensitive aircraft applications; however costs became an issue on applying the material into aircraft industry.
These were partially solved by the weight savings that were possible and the developed spliced method. This
provided the breakthrough for the application of GLARE on a large aircraft, the A380-800 [1].
Although GLARE is now applied on the A380 and considered as a material for future aircraft several cost and
production issues still have to be solved for future applications of this material. Delft University of
Technology has been involved in the development of GLARE from the start of FML laminates. Due to their
involvement the TU Delft noticed that current production methods and standards are no longer sufficient in
terms of speed and accuracy and thus a new method of producing GLARE has to be found.
Several studies have shown that GLARE can offer significant weight savings in other types of aircraft [1]. As a
result, the demand for GLARE panels is most likely to increase. The current hand lay-up method is not
suitable for dealing with high quantities of production and thus an alternative has to be found.
Other options for forming GLARE focus on deforming post-cured GLARE panels, such as brake forming, roll
forming, stretch forming, laser bending and shot peening. The issue with this method is that only limited out-
of-plane formability is possible as the glass fiber layers significantly limit the post-cured formability of GLARE
panels, especially in thicker laminates.
Another major challenge with GLARE panels is the production of double curved panels. These days double
curved GLARE panels are made of flat sheets with the earlier mentioned hand lay-up method. These flat
sheets are then forced into a double curved mold to their final shape. Yet this method is limited to certain
curvatures as the sheet edges will start to wrinkle due to the enclosed residual stresses. Some of these
wrinkles can disappear and can be translated into internal sheet stresses during vacuum bagging and
autoclave cycles. However, if the curvature is too large or the sheet is too wide, wrinkles will remain within
the product.
In practice the supplier of these sheets can deliver up to 1500 mm in width. But to avoid wrinkles the applied
width is usually smaller. This leads to a large number of splices in large panels, which is unfavorable for both
weight and simplicity. One way of avoiding wrinkles is by preforming the aluminum sheets to obtain a certain
curvature. In theory this preforming should allow the production of larger panels with larger curvatures and
less splices.
In the aircraft industry double curved metal sheets with large radius, for e.g. fuselage application, are
currently made by stretch forming. This process includes the use of a die which is product specific. But for the
production of GLARE the use of dedicated dies becomes too expensive. At the moment several processes
exist, but none of these processes meet the standards for accuracy and/or cycle time yet. Therefore this
research will focus on a new type of process, the incremental sheet forming process, in order to determine its
feasibility for the application of preforming metal sheets.
To perform a study on the combination of GLARE material and incremental forming to define
and perform an experimental setup and test plan to determine to what extend incremental
forming is feasible as a method to preform the aluminum sheets used in GLARE.
Can die-less incremental forming be a feasible process for preforming the aluminum layers
used within GLARE laminates?
In order to be able to answer the main research question, several sub-questions are formulated. These are
listed below:
The material selected for the specimens is aluminum 2024-T3 with a thickness of 0.4 mm. The specimen
sheet size is 120 x 120 mm.
The forming tool being used is a straight steel tool with a 16 mm diameter. This forming tool is originally from
an impact tower setup, but is sufficient for this research. This forming tool has sufficient wear resistance to
neglect any effects of tool wear during the experiments.
The small size milling machine is capable of moving along the specimen. The movement in X and Y axis are at
least 100 mm in order to cover the area to be processed of the specimens. The movement in Z axis can be
limited as the parts only require small curvature, which limits the depth of the part.
In order to clamp the material, a jig is used to hold the material in place during the process. This jig is already
available in the Delft Aerospace Structures and Material Laboratory at Delft University of Technology as it is
also part of the impact tower setup.
In order to work with this setup, a thorough knowledge of this setup is required. In the first place such a
machine is not available in the Delft Aerospace Structures and Material Laboratory (DASML), thus a custom
designed machine has to be designed and constructed. After which familiarization with the software is
required to provide the input of the process. Therefore an extensive study is done for this special purpose
setup.
1.5.3 Experiments
Geometry experiments
Each of the radius experiments will contain a specific pre-determined shape with a 100 mm or 200 mm
longitudinal diameter and a 100 mm or 200 mm diameter in transverse direction. These diameters are much
smaller compared to the diameters encountered in aircraft industry. This downscaling is required in order to
obtain measurable dimensions during the experiments on small scale. During these experiments the other
process parameters are kept constant at 1000 mm/min feed rate and a vertical step size of 0.2 mm per pass.
After these specimens are formed they are measured by means of linear voltage displacement transducer
before being released from the jig. This provides data about the local spring back during the process.
These radius experiments are performed several times for each combination to provide information about
the scatter of the results. In the first option, which consists of a 100 mm diameter in both directions, the
experiments are conducted five times. If the results obtained provide no significant deviations the following
experiments are conducted three times instead of five times.
The reason to perform these experiments three times is to see if any large deviations occur during these
tests. If these three experiments have a significant deviation to one another, the specific experiment is
conducted again to obtain statistical data of the scatter.
The feed rate is varied from 1000 mm/min to 3000 mm/min in steps of 1000 mm/min, where the step size is
fixed at 0.2 mm/pass. The step size is varied afterwards from 0.1 mm to 0.3 mm in steps of 0.1 mm, where
the feed rate is fixed at 1000 mm/min. This results in a total of two additional experiments for the feed rate
and two additional experiments for the step size.
At last there will be a geometrical correction experiment according to Micaris method of over bending /
optimized trajectories [5]. In this experiment two iterations will be made to verify to what extend this
method improves the geometrical accuracy. During these test it may be that the vertical step size is variable
depending on the trajectory, while the feed rate remains at 1000 mm/min.
Number Step Size / Feed Rate Number Step Size / Feed Rate
1 0.2 mm / 1000 mm 5 0.2 mm / 3000 mm
2 0.2 mm / 1000 mm 6 0.1 mm / 1000 mm
3 0.2 mm / 1000 mm 7 0.3 mm / 1000 mm
4 0.2 mm / 2000 mm 8 Variable / 1000 mm
Table 1 Test matrix of the geometry, feed rate and step size experiments
10 | P a g e Remko Kuitert
Introduction
Data
During the eight different experiments data is generated. This data has to be stored and processed.
Depending on the amount of scatter during the experiments this implies that a minimum of 25 specimens
(1x5, 6x3 and 1x2) is used up to a maximum of 37 specimens (7x5 and 1x2).
In order to collect the geometrical data each specimen is measured four times at first and each point will be
measured twice. If the measurements are proven to be reliable the following specimens will only be
measured once, even though each point is still scanned twice. This is done with the aid of a linear voltage
displacement transducer or LVDT. The data obtained from these measurements can be stored in an excel
sheet containing the following information:
- Process settings
- Intended geometry from the computer model
- Achieved geometry from the actual specimen
- Measured production time by machine
Outcome
The goal of this Master thesis is to determine the feasibility of die-less forming of double curved aluminum
sheets using single point incremental forming. The experiments performed on the different geometries will
show which range of diameters can be achieved and at which accuracy these can be achieved. From these
experiments it is expected that small diameters will provide more accurate results as the ratio between
elastic and plastic deformations becomes smaller.
Regarding the experiments performed with the different settings, different results are expected. It is
expected that lowering the step size provides higher dimensional accuracy with longer production times.
Considering the feed rate, it is expected that increasing the feed rate will decrease the process time, without
affecting the dimensional accuracy.
Relevance
The results obtained from the experiments can be used to link the different settings with the different
process requirements. For example the geometric experiments can be used to obtain a relation between the
intended geometry and the achieved geometry and thus the dimensional accuracy. These experiments also
provide insight in the differences in production times between different forming depths as it requires more
passes to form deeper products. Another important step is that the results of these small scale experiments
have to be translated to large scale production in order to determine the feasibility.
1.5.5 Management
During the research, management is required to keep track of the data, outcomes and the relevance to the
research question. Management is also required for the deliverables required during the research, such as
the interim report, Master thesis and the presentation of the research performed.
Remko Kuitert 11 | P a g e
Literature Review
2.1.1 Patents
Since the introduction of Fiber Metal Laminates in the 1970s many patents have been filed. The original
patent on FMLs was filed by Delft University of Technology with patent NL8100087 and NL8100088 [6] [7].
The first patent contains the information of the constituents, while the second patent contains the
information regarding the manufacturing process. These patents were based on ARALL, but they also covered
the application of carbon (CARALL) and glass fibers (GLARE).
12 | P a g e Remko Kuitert
Literature Review
As can be seen from Figure 2-2, on the next page, the aluminum material is either a 7475- or a 2024-T3
aluminum alloy [8]. The prepreg used is S2 Glass fiber prepreg with an epoxy resin FM94 for aluminum 2024-
T3 or FM906 for aluminum 7475-T761 [1] [14].
Grade 1 has been developed initially as a proof of concept in the research phase of GLARE and was a direct
spinoff of ARALL. This is the only Grade that uses aluminum 7475-T761 and the corresponding FM906 resin.
This material has the highest ultimate tensile stress of the different GLARE Grades; but the material has a
rather low formability.
As a result Grade 2 has been developed with aluminum 2024-T3 and FM94 resin. This material provides a
slight decrease in ultimate tensile stresses, but offers higher strain limits which makes the material easier to
handle in part manufacturing. Both Grade 1 and Grade 2 are mainly used in unidirectional loaded structures
and patching purposes [15].
Nevertheless the unidirectional Grade 1 and 2 werent very useful for the intended applications of fuselage
material and Grade 3 and Grade 4 were developed. Grade 3 is mainly used on top of the fuselage where the
ratio between circumferential stresses and longitudinal stresses is close to one. Grade 4 is used on the sides
of the fuselage, where the loads in one direction are about twice as high compared to the other direction.
Later Grade 5 was developed as an impact resistant material, which proved to be useful for impact sensitive
areas such as lower wing skins, lower fuselage sections and leading edges. Grade 6 is a special Grade
designed for shear areas and is the only Grade that has the fibers in a +/-45 degrees direction. The
applications for Grade 6 are limited to shear webs or door corner doublers.
Remko Kuitert 13 | P a g e
Literature Review
Figure 2-2 Different GLARE Grades with their fiber orientation and their specific constituents [16]
14 | P a g e Remko Kuitert
Literature Review
The aluminum layers are bonded together with the prepreg layers inside an autoclave, at temperatures up to
120 degrees Celsius and a maximum pressure of 6 bars [18]. At these temperatures the materials expand and
cure together. Due to the differences in thermal expansion coefficients between the layers, the aluminum
layers want to shrink more compared to the prepreg layers after cooling down. This implies that the
aluminum layers will have tensile residual stresses, while the prepreg layers have compressive residual
stresses. This has to be taken into account when the material is used. If the material is used at elevated
temperatures the residual stresses will become lower; while at lower temperatures the residual stresses
increase.
Post Stretching
A solution to change the unfavorable stress state of GLARE after curing is post stretching of the material.
During this post stretching process one reverses the stress state within the material by straining the metallic
layers plastically. These layers will be elongated permanently, while the elastic fibers remain elastic. As a
result, the residual stresses within the material will be reversed as can be seen in Figure 2-3, which is
favorable for the fatigue life of the material [19].
Figure 2-3 Post stretching effect with Stress Strain curve [20]
But post stretching isnt used in GLARE skin structures, due to the additional and difficult process step. The
benefit of post stretching is that the crack initiation is extended; but the crack initiation is only a minor part
of the fatigue life. Therefore the costs are much higher than the benefits of this post stretching method.
Remko Kuitert 15 | P a g e
Literature Review
Tensile properties
Tensile properties of GLARE have been measured by standardized tensile tests. In Table 2.1 a summary is
made of some typical values found in literature for the most common GLARE Grades and the aluminum alloys
used in these materials [15] [21] [22].
The tensile properties of GLARE are depending on its constituents: the glass fiber orientation, the aluminum
layer and the adhesive or resin. The yield strength of GLARE is dictated by the aluminum alloy, which is the
constituent with plastic behavior. Despite the other parameters, the ultimate strength is dictated by the
failure of the glass fibers.
The limit strain depends on the orientations of the fibers. If there are fibers in the strain direction, the limit
strain will be dictated by the fibers. But if there are no fibers in the strain direction, the strain is dictated by a
combination of the epoxy limit strain, the aluminum limit strain or a consistency failure in the material. This is
only applicable to GLARE Grade 1 or 2.
Another important feature of GLARE is the stress strain relation after the yield point has been reached. The
strain hardening beyond the yield stress is almost linear. This is caused by the elastic deformations of the
fibers; their contribution to the stress is significant after yielding. As a result the stress strain relation
becomes a straight line, resulting in a bilinear stress strain curve.
What can be seen in Table 2.1 is that Grade 1 and 2 have high ultimate tensile stresses compared to the
other GLARE Grades and aluminum. This has to do with the fiber orientation in these Grades, which is
unidirectional. This leads to superior properties in the main direction (L) over the aluminum alloys; in the
secondary direction (LT) however, the material is inferior.
The properties of Grade 3 and Grade 5, in both L and LT direction, are roughly equal to each other. This can
be expected as the fibers are orientated in both directions in equal amounts and the aluminum itself is
isotropic, which should result in a material that is more or less similar in both directions.
Grade 4 on the other hand has twice as many fibers in one direction opposed to the other direction. This
result in higher properties in one direction compared to the other direction. Though the difference between
the two directions is not as large as with Grade 1 and 2. Therefore Grade 4 has in superior properties in both
L and LT directions compared to aluminum.
Table 2.1 Mechanical properties for GLARE laminates and aluminum alloys [15] [21] [22]
Property Sym Dir Dim GLARE 1 GLARE 2 GLARE 3 GLARE 4 GLARE 5 2024-T3 7475-T6
16 | P a g e Remko Kuitert
Literature Review
Fatigue resistance
The most important property of GLARE is its high fatigue resistance. Typically for aluminum alloys the
majority of the fatigue life is required for crack initiation. This implies that once a crack is initiated the crack
grows relatively fast. For GLARE this is the opposite; the minor part is used for crack initiation and the
majority is used for crack growth. And since the amount of cycles to initiate a fatigue crack in both aluminum
and GLARE is almost identical, GLARE has a longer fatigue life at higher stress levels.
The crack growth goes slowly in GLARE, because of the so called fiber bridging. The crack typically initiates in
the metal layers, keeping the fiber layers intact. This means that the load is being bypassed by the fibers,
which reduces the stress concentrations at the crack tips in the metallic layers. Due to the ability to bypass
the load in the fibers instead of the crack tip, the crack grows slower in GLARE compared to monolithic
aluminum, which results in the superior fatigue properties.
Impact resistance
Another good property of GLARE is the impact behavior. Vlot [23] studied the effects of impact on Fiber
Metal Laminates. This research showed that several Fiber Metal Laminates provided smaller post-impact
damage zones compared to fiber reinforced plastic materials due to the metallic layers within FMLs.
An impact damage zone in a monolithic metallic material will leave small local damage in which the region is
plastically deformed absorbing most of the energy. The fibers in a fully composite structure do not have the
ability to deform plastically and absorb the loads. As a result impact on a composite structure will provide a
large impact zone as the energy is spread through the different layers, possibly causing large spread
delamination. In the case of an FML impact will cause the metallic layers to absorb most of the energy as they
deform plastically, which prevents the impact from spreading as much as in a pure composite. This provides a
smaller post-impact damage zone, which results in a higher residual strength after impact damage has
occurred [21].
Fire resistance
FMLs also excel in fire resistance. When a monolithic metal alloy is exposed to a heat source, the heat will
penetrate the entire sheet. In GLARE the heat penetrates the outer metallic layer first. This layer will
distribute the heat and possibly melt away, exposing the glass fiber layer. This glass fiber layer doesnt
conduct heat as well as the metallic layer, resulting in a smaller heat affected area at the next layer.
Therefore this layer will act like insulation for the second metallic layer. Due to this process the second layer
of metal is exposed to less heat and as a result it takes significant longer for heat to penetrate the entire
material [21].
Corrosion resistance
The corrosion resistance is another advantage of FMLs, which is due to the composite layers in between the
metal layers. The composite layers between the metal layers will prevent corrosion to penetrate to the next
metallic layer and thus the corrosion is limited to the outer layer of the FML in general. The composite layers
deteriorate under influence of UV radiation and moisture, but these layers are shielded from these influences
by the metal layers. So the deterioration of the composite layers is limited, which improves the general
corrosion resistance behavior of the entire laminate [21].
Remko Kuitert 17 | P a g e
Literature Review
Classification
Incremental sheet forming has its origins in stretch forming processes and metal spinning processes. Its a
relative new technique for deforming sheet materials by applying a stepwise increment to the deforming
tool. The main process consists of four elements, which are the sheet material, a blank holder, a single point
forming tool and a CNC machine. Within incremental sheet forming there are two main categories; the
conventional incremental sheet forming techniques and the hybrid incremental sheet forming techniques.
18 | P a g e Remko Kuitert
Literature Review
Figure 2-5 Back pressure SPIF (left), partial die SPIF (center) and full die SPIF (right)
In his research he tries to solve the limitations in process speed and geometrical
accuracy with a full die, as can be seen in Figure 2-6. In this process the die is
first pushed into the sheet material and afterwards the incremental forming tool
is used to create the concave parts of the product. This does significantly reduce
the process time as the surface for incremental sheet forming is reduced and the
use of a full die improves the accuracy compared to die-less forming techniques.
The down side of this process is that it still requires the use of full dies, where industry is looking for die-less
solutions. The setup for this process does provide a suitable basis for the die-less incremental forming setup.
The main difference between the setup used by Araghi and the setup used in this research is that the die will
be removed.
Remko Kuitert 19 | P a g e
Literature Review
Till then it was assumed that shear deformation was the only deformation occurring. However, Filice [29]
noticed that if the tool moves straight on a horizontal plane, the deformation is mostly plane-strain
stretching. Except at the beginning and ending of the tool path the deformation is mostly biaxial stretching.
Therefore if the radii of a tool path increase, the deformation mode is more biaxial stretching instead of
plane-strain. Formability in metal is optimal under plane-strain stretching as the minor strain is zero in this
case, providing the largest major strain limit; this implies that small radii of the tool path reduce the
formability of the sheet material. Hussain [30] proved the findings of Filice after conducting several
experiments for formability with different radius of curvatures in which the formability increased as the
curvature decreased.
2.2.4 Accuracy
Cerro [31] did several experiments measuring the geometric accuracy with the process model as compared to
the actual product made in the CNC machine, which showed that the computer model provided very close
resemblances with the actual product.
However, these researches were done using a die or partial die within the process. In the case of die-less
incremental forming there is no tool on the other side of the sheet. This method of single point incremental
forming provided to be less accurate as shown by Micari [5]. The inaccuracy can be split up into three main
effects: the sheet bending effect, pillow effect and sheet spring back effect as can be seen in Figure 2-7.
These inaccuracies are influenced by process parameters, material parameters and design parameters. In the
case of a process parameter the step size, the distance between two successive loops, influences the
accuracy. This implies that a large step size results in less accurate parts. In the case of material parameters
strain hardening and anisotropy influence the accuracy. While in the case of design parameters the blank
thickness and geometry of the part are influencing the accuracy.
Nevertheless the shape and dimensional errors occurring during die-less forming can be reduced using
optimized trajectories as mentioned by Micari [5]. The concept of optimized trajectories is very similar to that
of over bending: the idea is to apply more deformations during the process in order to compensate for the
relaxation. However Micari noticed that the main issue with this method is that there is to-date no
completely accurate model in order to predict the material behavior in single point incremental forming.
20 | P a g e Remko Kuitert
Literature Review
The step size influences both the process time and the surface roughness. A large step size results in a less
repetitive pattern over the surface; this reduced the total length of the toolpath; but it increases the surface
roughness. Considering the surface roughness they noticed that the combination between rotational speed
and feed rates have a major effect. The higher the rotational speed over the feed rate, the smoother the
surface. For the thickness distribution there are no significant changes if the feed rate is increased. Finally,
they investigated the effects on the microstructure of the material; they noticed that the step size is the
dominant factor on the grain size. A smaller step size results in smaller grains. Although they noticed that a
higher spindle speed or feed rate also reduces grain size.
In 2013 Ambrogio continued the study of Hamilton and Jeswiet by expanding the field of research to
different materials and even higher speeds up to 600 m/min [33]. They also considered the microstructure of
the alloys used during their research and stated that the microstructure is very similar to the as received
state. However in the case of 600 m/min feed rate the grain size is slightly larger compared to a feed rate of
60 m/min. This contradicts the findings of Hamilton and Jeswiet earlier as they stated that increased feed
rates result in reduced grain sizes. On the other hand the data from Ambrogio shows that at feed rates up to
60 m/min the statements of Hamilton and Jeswiet are valid.
Nevertheless both researches state that the increased feed rates do not affect the material microstructure,
meaning that the increased mechanical work done on the sheet material is insufficient to cause any changes.
The material roughness is affected by the increased feed rates, but Hamilton and Jeswiet stated that with a
proper choice of step size and spindle speed one can obtain the desired surface roughness. This implies that
the process speed can be increased significantly solving one of the major disadvantage of this process.
Remko Kuitert 21 | P a g e
Requirements
Chapter 3 Requirements
3.1 Applications of GLARE
3.1.1 Current applications
The current main application of GLARE is on the fuselage of the A380. Due to the superior damage tolerance
properties of GLARE most of the upper fuselage is covered with GLARE, resulting in 27 panels with a total
area of 470 square meters per aircraft. In addition GLARE, due to its good impact behavior, also got applied in
impact sensitive areas, such as the leading edges of the empennage. In Figure 3-1 the application of GLARE
on an A380 is visualized [34].
But the A380 isnt the only application of GLARE. The Learjet 45 has a bulkhead made of GLARE and Comtek
Advanced Structures offers a GLARE based cargo liner interior solution for regional jet aircraft. Also Galaxy
Aviation Security developed a GLARE LD3 luggage container, which is capable of containing blasts of twice the
magnitude of the bomb used in the Lockerbie disaster in December 1988 [1] [34].
So even though the weight savings were possible, each of these studies were declined due to economics or
politics. Considering economics GLARE was 5 to 10 times more expensive per kilogram compared to the
aluminum alloys being used in those aircraft, the weight saving benefits were outnumbered by the added
costs in materials and manufacturing. Considering politics in the 90s Boeing decided to make their new
aircraft, the Dreamliner, out of carbon. As a result Airbus decided that their new aircraft, the A350, also
should contain as much carbon as possible and thus GLARE got neglected. The main reason why GLARE was
applied in the A380 was due to required weight savings in order to make it fly and costs were not as
important [1].
The cost of GLARE limited the possibilities for future applications of this material. But several advances in
manufacturing methods of GLARE, such as the splicing concept and integrating stiffeners into the panel
design, have made the application of the material already much cheaper. These ongoing costs reduction
might result in future applications of GLARE in redesigned fuselage panels of A320s, A330s, A340s and A350s
[1].
22 | P a g e Remko Kuitert
Requirements
The critical components with respect to curvature for GLARE production will be the double curved
applications at the empennage or leading edges. These parts may have radii in one or both directions of
about 100-200 mm and therefore during the production higher strains will occur to achieve these radii.
Therefore the process should at least be able to form parts with a radius of 100 mm, which occur in leading
edges, up to 2000 mm, which occur in fuselages.
Production assumptions
In contrast to the production numbers of the Airbus A380, which is currently at around 3 per month, the
Airbus A320, for example, has a production rate of 46 aircraft per month as mentioned by Williams [35]. This
means that a large production volume is required for future applications of GLARE in aircraft as these.
In the current Airbus A380 the fuselage is made of 470 square meters GLARE material, as shown in Figure 3-1.
The cabin length is 49.9 meters with an average diameter of 8 meters, meaning roughly one third of the
entire cabin fuselage is made of GLARE. If kept in mind that the production and manufacturing costs of GLARE
can be lowered successfully the percentage of GLARE used in the fuselage could range between 50% and
90%. If this 50% fuselage material is applied on the smaller Airbus A320, which measures 27.5m cabin length
2 2
and 4m cabin width, the total square meters of GLARE could range from 172 m (50%) to 311 m (90%).
An important notification has to be made that the Airbus A380 is most likely to have 4/3 GLARE lay-up, as the
loads are significant higher compared to the Airbus A320, where a 3/2 GLARE lay-up is most likely. This has
implications for the laminating or preforming to shape method, as fewer layers have to be used. In Table 3.1
an overview is given of the current production numbers of the Airbus A380 vs. the production numbers for
the A320 with either 50% or 90% GLARE Fuselage.
Current (A380) Future (A320 50% GLARE) Future (A320 90% GLARE)
Aircraft Per Year [-] 30 552 552
GLARE Lay-up [-] 4/3 3/2 3/2
2
GLARE / Aircraft [m ] 470 172 311
2
Production / Year [m ] 14100 94994 171662
2
Production / Hour [m ] 7 24 43
2
Aluminum / Hour [m ] 28 72 129
2
Glass Fiber / Hour [m ] 21 48 86
Table 3.1: Production Assumptions Overview
Where hand lay-up is now common practice in the production line of the Airbus A380 for the application of
GLARE, for the Airbus A320 this is no longer viable as the numbers are simply too high. Therefore the process
2 2
has to be capable of producing 24 -43 m of GLARE 3/2 lay-up per hour. This implies processing 72-129 m of
2
aluminum and 48-86 m of glass fiber prepreg per hour.
Remko Kuitert 23 | P a g e
Requirements
Using this method the layers can be preformed before the laminate is cured. Also the required accuracy
doesnt have to match the exact product dimensions during the preforming of the aluminum layers. The
reason is that during the autoclave process the pressure will correct some of the dimensional inaccuracies
before the product achieves its final shape. Because of this the dimensional accuracy, before the autoclave
process, can be in the order of 80 to 90% of the forming depth.
Spring back
The application of a die-less stretch forming process on slightly double curved panels implies that the elastic
energy within the material is significant, as the deformations are minimal. This implies that the spring back
could be a serious issue regarding the accuracy.
Spring back can occur locally, during incremental forming processes, as the deformations are locally applied.
After the forming tool moves on, spring back occurs on these places locally. On the other hand after the part
has been formed and removed from the clamps global spring back occurs, which implies that the residual
stresses left over after deformation are partially unloaded. This might affect the entire dimensional accuracy,
instead of only the local dimensional accuracy.
The easiest way to compensate spring back is to apply slightly larger deformations to the product. This can be
done after a product has been formed and measured in order to determine the amount of over deformation
required to compensate for the spring back.
24 | P a g e Remko Kuitert
Requirements
This doesnt say anything yet of what kind of feed rate or step size must be reached to achieve these
production volumes with this kind of accuracy. In order to determine the minimum required feed rate or step
size an analysis is required of what kind of parts can be expected and what forming depths are to be
expected.
The speed during incremental sheet forming mainly depends on the step size and the feed rate, where the
accuracy mainly depends on the step size. Therefore, one could for example increase the step size to achieve
higher production rates, at the cost of some accuracy. But one could also increase the feed rate instead of
the step size to achieve the same higher production rates, without sacrificing for accuracy. Therefore if
possible increasing the feed rate is favorable over increasing the step size. But increasing the feed rate is not
always possible, as friction might become an issue during the forming process or the feed rate is already at its
maximum. As a result a higher step size is required in order to increase production rates.
A panel with a width of 1500 mm with a 2000 mm radius has a forming depth of 139 mm, as can be seen in
Figure 3-2. This implies that with a step size of 0.1 mm/pass a total of 1390 passes are required to achieve
this depth. These 1390 passes do not include any spring back compensation as well, which results in even
more passes that are actually required to obtain the preferred shape.
Remko Kuitert 25 | P a g e
Requirements
For example we take a random panel, which is required to have a 400 mm radius in one direction and a 2500
mm radius in another direction. This panel has a width of 500 mm and a length of 2000 mm. As a result, the
depth of the panel, as shown in Figure 3-3, is already 273 mm. This is already double the depth of the
fuselage panel and thus requires 2730 passes of 0.1 mm step size to obtain this depth.
2
1500 mm and 5000 mm length, covering a total of 7.5 m of
material. In order to achieve the required production speed,
mentioned earlier in 3.2.2, one has to create 9.6 to 17.5 of
these panels per hour or one every 3.5 to 6.25 minutes;
assuming only one machine is being used.
The feed rate, on the other hand, has no effect on the distance,
but it does have an almost inverse relation with the production
time. Due to some accelerating and deaccelerating effects in the
stepper motors of the machines the relationship between the
feed rate and the process time is not entirely an inverse
relation. Though in practice one may assume that if the feed
rate is doubled, the production time is halved. In Figure 3-5 a Feed rate [mm/min]
plot is made which relates the production time in minutes Figure 3-5 Feed rate vs. Production time for large
panels
versus the feed rate.
26 | P a g e Remko Kuitert
Requirements
In order to get the production time down to 3.5 to 6.25 minutes per panel, the process either requires a high
feed rate or a high step size or a combination of both. In Figure 3-6 one can see that in order to achieve a
production time of 6.25 min / panel (Red line) one could use much lower feed rate compared to the 3.5 min /
panel production rate (Blue line). However one can also see that a combination of high feed rate and high
step size are required in order to actually create these panels at the desired production rate, whether this is
3.5 minutes/panel or 6.25 minutes/panel.
The only possible issue with these large radii panels is that
the amount of elastic deformations is relative large
compared to the plastic deformation, which implies that a
lot of spring back is achieved. But the effects of spring back
in single curved panels are well known and can be Step size [mm]
compensated. Figure 3-6 required step size vs. the feed rate for
minimum and maximum production rates
The small double curved panels might prove more difficult to manufacture. As any small deviation within the
process settings might cause relative large inaccuracies and the complex spring back mechanism within these
double curved shapes also contributes to larger inaccuracies. Therefore it is expected that these small double
curved parts are critical for the feasibility of the incremental sheet forming process.
The above shows that it is important that during the experimental phase of this research the specimens
should resemble this category as much as possible. As the specimen size during the experimental phase is
relative small, the parameters should be compensated for such. This implies that radii of 100 to 300 mm have
to be tested with step sizes up to 0.3 mm and feed rate up to 3000 mm / min.
Remko Kuitert 27 | P a g e
Experimental Setup
Typically milling operations are performed at very high cutting speeds, which imply that each cut only
removes a very small portion of the material. Therefore the cutting force remains typically below 10 N in any
direction, despite the high feed rates.
During incremental forming the tool presses on top of the material and is then moved in the XY plane
afterwards. In order to deform the material the tool is first pressed into the material with a certain
predefined tool path, in which the forces are highly depending on the tool path. Duflou et al. [36] have done
research in the field of forming forces during incremental forming and they suggested the following equation:
In this formula the force in Z direction thus depends on the tensile strength( ), the thickness of the
plate(), the diameter of the tool( ), the step height() and the forming angle between the forming tool
and the material in degrees (). In their experiments they used aluminum 2024 of 0.4 mm thickness, which is
the same type of material being used in this study. The forces they measured during their experiments were
in the order of 200 N in the Z direction for the 0.4 mm thickness and at angles of 20 degrees, which is similar
to what can be expected during this study.
Even though the forces in X and Y directions are significantly lower compared to the Z direction during
incremental forming, these forces are still much higher compared to the milling forces that a typical CNC
milling machine have to endure.
These CNC routers are simple, accurate machines and are perfect for 3D printing, engraving or milling
operations. However as one can see in Figure 4-1 on the left side, if the Z-axis is exposed to the forces during
incremental forming, the Y axis will bend as it has a rather low stiffness.
Due to the low stiffness within the Y-axis the CNC routers in the laboratory were only capable of achieving
forces in the order of 20 N and below to prevent any deformations within the Y-axis.
Figure 4-1 A 3-axis router machine (left) vs. 3-axis knee milling machine (right) [37] [38] [39]
28 | P a g e Remko Kuitert
Experimental Setup
There is also a different type of machine used for CNC milling. These are the so called knee milling machines,
as can be seen in Figure 4-1 on the right side. These machines have their Z-axis separated from their other
axis. The main advantage of these machines is therefore their rigidity.
The X-axis and Y-axis are usually controlled with a solid XY Table, which can hold much higher forces
compared to the rails system of the routers. The Z-axis is capable of delivering more force as the axis is fitted
in a rigid top and fully separated of the other two axes, preventing any deformation to those axes. Therefore
these types of machines are also capable of performing processes which include higher forces, such as
incremental forming.
The DASML has one CNC knee milling machine, the Proxxon Micro Frees MF 70. This is one of the smallest
CNC knee milling machines available and comes with low end stepper motors. As a result this machine was
only able to deliver forces up to 50 N according to the specifications before the stepper engines stop working.
Another problem with this machine was its range, which was limited to 134x46 mm in the XY plane.
After several contacts it turned out that the machine they usually used for incremental forming was damaged
beyond repair. But they had other machines, which were available, but after sending more detailed
information about the research there was no more contact and eventually the option of performing the
experiments was dismissed.
Remko Kuitert 29 | P a g e
Experimental Setup
However, it was required that the machine could be converted back to its original state after this research
has been completed. This provided some challenges as any permanent changes to the machine itself were
prohibited, while still making sure that the machine could be fitted with stepper motors on each axis.
As a result, the mechanical design and assembly basically consists of three steps. The first step is the Y-axis
assembly, which allows back and forward movement. This is followed by the X-axis assembly, which allows
sideways movement, and finally the Z-axis assembly, which allows up and downward movement.
As the axis itself was still in good shape, it was decided to keep
the axis as is. The only modification done to the axis itself is the
removal of the black handle to allow a shaft coupler on the axis
in order to close the gap between the different diameters of the
axis itself and the axis of the stepper motor.
Figure 4-3 Conventional Y-axis close up [38]
30 | P a g e Remko Kuitert
Experimental Setup
Also the current axis was secured by two bolts just behind the black handle; one of them can be seen in
Figure 4-3. These bolts will be replaced by longer bolts to secure an additional aluminum backing plate at the
same position. This aluminum backing plate will function as a connection point for the stepper motor in order
to avoid any permanent modifications on the machine.
The design of the stepper motor fitting, as can be seen in Figure 4-4, is kept rather simple. The backing plate
(green) will be fixed with two bolts, as mentioned earlier, on the machine itself. In this drawing, the location
of the holes of these bolts are not shown as they were specified later on during the assembly.
A total of four rods (red) with a diameter of 12 mm are used to go from the backing plate of the machine
towards the stepper motor, while in the center the shaft coupler (yellow) is positioned in place to connect
the two different axes to each other. Both the backing plate of the machine (green) and the stepper motor
backing plate (blue) are fixed to the rods (red) by M4 screws.
The shaft couplers, Oldham D25 L30 D1-8 D2-12, were eventually ordered by Hardware CNC and were rather
easy to install. But due to wear the end of the Y-axis of the machine wasnt exactly 12 mm, this resulted in
some slipping between the shaft coupler and the axis itself. In order to solve this issue two-component metal
glue has been applied, which locks the two parts together.
Eventually, this simple design resulted in a fully functioning Y-axis, as can be seen in Figure 4-5, in which the
stepper motor is directly connected to the axis with the use of the shaft coupler.
Remko Kuitert 31 | P a g e
Experimental Setup
Just like for the Y-axis, the axis itself was still in good shape
and there was no need to replace it with a new axis. Again
the axis consisted of a 15 mm rod with a thread pitch of 5
mm on the inside of the machine and an exterior rod of 12
mm, on which the black handles were attached. Therefore
the black handle, on the left side, was removed in order to
make room for a shaft coupler.
32 | P a g e Remko Kuitert
Experimental Setup
Remko Kuitert 33 | P a g e
Experimental Setup
Stepper motors
There are several types of stepper motors available for CNC purposes, such as servo motors, unipolar stepper
motors and bipolar stepper motors. Servo motors have an advantage over uni- or bipolar stepper motors as
they are capable of delivering a feedback to the system of their actual position. However this comes at a cost,
which makes servo motors much more expensive compared to uni- or bipolar stepper motors.
Even though uni- and bipolar stepper motors do not provide any feedback to the control system of their
current position, they usually do not require it as long as no steps are lost during the process. Loss of steps
occurs when the stepper motor has to deliver more torque than it could. In that case the software will send a
signal to the motor that it should rotate, while physically the motor cant deliver enough torque to actually
rotate. As a result, the software assumed the motor has rotated, while in reality it stood still and only
provided a clicking noise. Loss of steps could thus easily be avoided, if the process itself does not approach
the torque limit of the stepper motor and thus make the feedback option of a servo motor redundant.
This leaves the choice between uni- and bipolar motors. A bipolar motor works basically the same as a
unipolar motor; however it has an additional set of magnets/poles. As a result the bipolar motors are more
efficient and deliver more torque compared to a similar sized unipolar motor. Therefore bipolar motors are
usually preferred for CNC milling operations.
Now the choice has been made to use bipolar stepper motors, the next step is to determine the amount of
torque required for the incremental forming process. As mentioned before Duflou et al. [36] performed
similar experiments on the same type of material to determine a relation between the forming force and
several parameters, with forces of about 200 N in the Z direction.
For the Z-axis a worm drive gear arrangement is being used to transfer the torque of the stepper motor to
the forming tool. Its measured from the mechanical design that for every full rotation of the Z-axis the
forming tool moves 25 mm. This implies that if the forming tool has to press down with 200 N, the torque on
the Z-axis has to be around 0.8 Nm. However this is true for a static situation with 100% efficiency. The
stepper motor will be constantly accelerating and deaccelerating to its correct position and at the same time
the transmission will have some friction. Thus a higher torque is required to achieve this 200 N in reality.
With the above points in mind the choice has been made to
use Nema 23 3 Nm stepper motors from Hardware CNC,
which can be seen in Figure 4-12.
34 | P a g e Remko Kuitert
Experimental Setup
Stepper drivers
In order to control the stepper motors, a stepper drive is usually required. For this research a certain degree
of precision is required. Typically, a stepper motor has 200 steps per revolution, which implies that for the Z-
axis (25 mm movement per revolution) each step accounts for 0.125 mm and for the other axes (5 mm
movement per revolution) each step accounts for 0.025 mm.
To obtain movement accuracies of 0.01 mm or below the stepper driver should be able to use so called
microsteps. Microstepping is a method to control a stepper motor in such a way that each step on its own is
divided into multiple steps. For example 1/8 microstepping allows a stepper motor 1600 steps/revolution
instead of 200, which increases the precision of the stepper motor by a factor of 8.
Remko Kuitert 35 | P a g e
Experimental Setup
Interface card
Even though the stepper motors, drivers and the power
supply unit can work on their own; this system has to be
connected to a computer in order to have a computer
numerical controlled application. Modern stepper drivers
have the option to directly connect the drivers to a
computer. But since this system requires an external power
supply, this would provide 48 Volt 6.7 Ampere towards the
computer. This would ruin the computer and thus some kind
of separation is required between the computer system and
the machine system.
Enclosure box
Another important part of the electrical system is the
enclosure box or control box, in which the electrical system is
mounted. Basically, the enclosure box prevents the user from
touching any of the electrical components in order to avoid
damage to the components or the user from being shocked.
The choice was made to go for the 1SL02 IP66 Wall Box of RS
components, as can be seen in Figure 4-17. This box was fully
made out of thermoplastic material, which does not conduct
electricity. Also the interior dimensions are sufficient to
mount all the electrical components inside, while they can still
be modified and/or accessed by opening the door.
Figure 4-17 1SL02 IP66 Wall Box [40]
36 | P a g e Remko Kuitert
Experimental Setup
Wiring scheme
Now that all components are selected for the electrical design, they
still have to be properly connected in order to get a working system
and also fitted inside the enclosure box. Considering the wiring, a
wiring scheme is setup in order to successfully connect the
individual components to each other for a working system.
The difference comes in the power supply units and the amount of
stepper motors. The CNC machine used in this project only has
three stepper motors and one PSU, instead of four motors and two Figure 4-18 3D printer wiring scheme [46]
PSUs. Also the system will not use any extruders, end stops, PSU for
the Ramps 1.4 or any fans to cool the system. If this wiring scheme
is adjusted with the right components, all that remains can be seen
in Figure 4-19. However this only provides a direct system, without
any emergency stops or any on/off switches.
Remko Kuitert 37 | P a g e
Experimental Setup
4.2.3 Software
Regarding the software being used in this project, the DASML has been working with Repetier in the past for
3D printers. This software is also compatible for CNC milling operations and can thus be converted towards
an incremental forming process.
In order to convert the software for incremental forming it has to be understood that the software package
comes with two separate units; the Repetier firmware for the Arduino board and the Repetier host software
for the computer itself.
Repetier firmware
The firmware is basically the program that is copied onto the Arduino Mega 2560 card. This software
contains all the information to convert the computer signal, received by the USB cable, into an electronical
signal that can be sent towards the correct pins at the Arduino card. This Arduino card can then transfer the
signal to the Ramps 1.4 card and this Ramps 1.4 card can then transfer the signal towards the stepper drivers.
The first change to this firmware is to switch all extruders off, as the machine will not have any. After this is
done the correct motherboard has to be chosen from a list, included within the Repetier firmware. Last but
not least, all the end stops should be turned off in order for the machine to actually work. All these changes
in the firmware have to be made in the Configuration.h file. The full list of all changed lines of code can be
found in Appendix B Repetier Firmware changes.
If all these changes are done correctly the firmware is ready to be uploaded to the Arduino Mega 2560 and
can be used for any incremental forming operation or other 3-axis CNC milling operations without end stops /
limit switches.
Repetier host
On the other side of the USB cable of the Arduino Mega 2560 is a computer with the Repetier host software
installed. This software package is able to communicate with the Repetier firmware located on the Arduino.
However, this software package also requires some different settings compared to the usual settings in order
to perform incremental forming operations, instead of 3D printing operations.
Besides these windows, there are also the control tabs. These are used to load objects, write G-code, slice
objects or move the machine manually to a specific point. During this research the manual movement tab is
used to maneuver the machine into the correct starting position. Also the print preview tab is used to load
the generated G-code into Repetier-Host and to preview and verify the code to see if any flaws are made
within the G-code.
38 | P a g e Remko Kuitert
Experimental Setup
The clamping plates being used for these experiments are 150x150x10 mm and contain nine M12 bolts to
secure them and the specimen in between to the foot. The reason to use two clamping plates is in order to
make sure that the specimen has the same clamping conditions on the lower side as on the upper side, since
the foot only contains nine M12 holes, but not a decent clean clamping surface.
4.3.2 Specimens
The specimens used in this project are made of 2024-
T3 aluminum, which is the aluminum most frequently
used in GLARE laminates. Only the GLARE Grade 1
laminates use a different type of aluminum, but this
type of GLARE is not used in aircraft industry in
general.
Remko Kuitert 39 | P a g e
Experimental Setup
4.4 Measuring
In order to measure the specimens at first a laser scanning setup was preferred. However this system was not
available during the research. Therefore an alternative had to be found to measure the shape of the
produced specimens to compare them with the input model and the other specimens within the same set of
experiments. This resulted in a measurement setup with a linear voltage displacement transducer instead of
the laser.
Also to verify that the LVDT measures correctly, each data point was measured twice. Keeping in mind that
scanning of a specimen shouldnt be too coarse, but also not too slow a compromise had to be made. The
specimens deformed area is 70x70 mm, which means that if the LVDT measures every 5 mm a total of 225
data points are created. The measurement of these 225 data points takes a total of 15 minutes, which is a
reasonable scanning time, and still provides a dense enough grid to extract sufficient detail of the surface of
the specimen. The Matlab script to generate the measurement grid can be found in Appendix A-2.
4.4.2 Calibration
Yet a calibration is still required, as the LVDT provides an output in Volts and not in mm. It was found that for
the small range LVDT one Volt was equal to a change of 4.64252 mm, with an accuracy of +/- 0.01 mm in the
range of 0-8 mm. For the large LVDT it was found that one Volt was equal to 10.03886 mm, with an accuracy
of +/- 0.03 mm in the range of 32-100 mm. It had to be kept in mind that if this LVDT was used, the initial
displacement position was already at 32 mm or higher, to ensure that this LVDT was in its linear range.
Measurement Grid
Besides the LVDT the measurement grid also had to be calibrated as the movement of the machine and the
measurement points had to correspond to each other. Therefore, four dents were made, in every specimen,
at the corner points. These dents were large enough for the LVDT to detect and therefore the location of the
LVDT could be matched with that of the machine coordinates.
Although the dents are detectable, it only provides a correspondence with the first point of the measurement
grid and the machine coordinates. If the LVDT is supposed to be measuring a different point every four
seconds, the machine should change position of the specimen every four seconds. However, due to the non-
linear acceleration of the stepper motors, the time it takes to displace a certain distance is not easily
analyzed. It was experimentally determined that it takes the machine 0.54 seconds to displace 5 mm. This
implies that at every point the machine was paused for 3.56 seconds. This provided a match between the
measurement grid and the corresponding machine locations, which could be used for the measurements.
40 | P a g e Remko Kuitert
Experiments
Chapter 5 Experiments
This chapter will cover the experiments performed with the earlier described experimental setup. In this
chapter all the different experiments, as mentioned in the research approach, are described and analyzed.
The results of these experiments are found in the end of this chapter.
5.1 Calibration
Because it is yet unknown what kind of repeatability either the machine or the measurement setup has, a
calibration is required. This set of experiments will be done with 100/100/0.2/1000 specimens. This type of
specimens implies that R1 and R2 equals 100 mm, the step size is 0.2 mm/pass and the feed rate is 1000
mm/min.
Since the specimens are produced with the same shape, the
effect of spring back should ideally be the same and should not
contribute much towards the deviations seen here.
A solution to reduce the deviations is not to calibrate the depth of the forming tool at the specimen itself, but
at a solid point such as the jig. By doing so the deviations are reduced, as will be seen in later experiments.
Remko Kuitert 41 | P a g e
Experiments
42 | P a g e Remko Kuitert
Experiments
Remko Kuitert 43 | P a g e
Experiments
The jig itself is 159 mm long and wide, thus the center of the
jig is located at 79.5 mm from each side. However once a
specimen is clamped, its not practical to determine the
center based on this 79.5 mm. Since the forming tool also
has to be centered at the same position.
Figure 5-9 Starting point determination X and Y
Instead of measuring and indicating the center of the plate, one could also include the forming tool during
this step, as seen in Figure 5-9. It is known that the diameter of the forming tool is 16 mm. Therefore the
distance from the left side of the forming tool towards the right side of the jig becomes 87.5 mm in total, and
the same for any other sides. As a result the forming tool is positioned at the center of the plate and jig. This
has to be repeated for each specimen.
44 | P a g e Remko Kuitert
Experiments
The script first computes the forming depth of the specimen based on the input radii and the size of the
specimen to be created. Afterwards the first X, Y and Z coordinates are generated and the Z coordinate is
checked with the initially compute forming depth. As long as the Z-coordinate hasnt reached the forming
depth, the loop has to continue to the next set of coordinates.
In order to determine the next set of coordinates the distance between these coordinates and the previous
coordinates is set to 1 mm, to keep the points evenly distributed over the length of the tool path. As a result
the script computes the next set of coordinates with very small increments.
In the end, this list of generated coordinates is stored with a G-code G1 representations within a text file, in
order to run the coordinates in Repetier host, and is also shown in a figure for a visual inspection. Therefore,
for every new specimen with different geometry or machine settings this script is used to generate a G-code
list.
Starting position
Create Specimen
Calibrate X
Correct X position No
Specimen Produced No
Yes
Yes Calibrate Y
No
Correct Y position
Secure the jig
Yes
Calibrate Z
Jig Secured No
Correct Z position No
Yes
Yes
Generate G-Code
Clamp Specimen
No
Correct G-code
Yes
Run Experiment
Remko Kuitert 45 | P a g e
Experiments
As this research focusses on the feasibility of incremental sheet forming as a die-less solution for preforming
aluminum layers within GLARE laminates, it should focus on those parts with a double curved profile and
therefore the chosen geometries to be tested are as followed:
1. One set of specimens with 100x100 radii
2. One set of specimens with 100x200 radii
3. One set of specimens with 200x200 radii
All of these specimens will have the same feed rate (1000 mm/min), same step size (0.2 mm/pass), same size
(120x120x0.4 mm) and material (Aluminum 2024-T3). After these tests the specimens are scanned with the
linear voltage displacement transducer in order to obtain a 3D image of the formed part. This 3D image is
than compared to the input model, which was used for the process.
In the next image, Figure 5-14, the deviations Figure 5-13 Geometry Experiment 1: Average obtained geometry
between the different specimens of the geometry
experiments 1 can be seen. What is most important
here is that using the jig as a reference point for the
Z-axis already provides fewer deviations, compared
to Figure 5-3.
46 | P a g e Remko Kuitert
Experiments
In the next image, Figure 5-17, the deviations Figure 5-16 Geometry Experiment 2: Average obtained geometry
between the specimens are plotted. Even though
the deviations are quite different compared to the
previous set of experiments, the scale of these
deviations is still in the order of 0.2 mm or less.
Remko Kuitert 47 | P a g e
Experiments
Next the deviations between the specimens are Figure 5-19 Geometry Experiment 3: Average obtained geometry
determined, as shown in Figure 5-20. Again the
deviations are almost all below 0.2 mm. Although
this specimen has a small peak at 0.2019 mm, which
means that the scale of the axis is slightly different
compared to the previous two experiments. For now
it seems that the deviations between the specimens
remain the same, no matter the geometry.
48 | P a g e Remko Kuitert
Experiments
As can be seen the results didnt provided to be very accurate with respect to the computer model due to
various reasons, such as sheet bending, pillow effect and spring back. However in order to determine the
effect of the feed rate the geometrical accuracy compared to the computer model is not important.
Instead the differences between the obtained geometries with different feed rate provide the required
information. Therefore the specimens of the following feed rate experiments are compared to the data of
the first geometry experiment to determine if the feed rate affects the geometrical accuracy in any way.
Remko Kuitert 49 | P a g e
Experiments
Most of these differences can be found near the center of the specimen. At first it was assumed that local
heating, due to the friction between the tool and the specimen, allowed the material to have a higher spring
back. However the actual reason is that the Z-axis was skipping several steps. It is most likely that the
increase in feed rate increased the force on the Z-axis during the process, which resulted in the loss of steps.
As the process continuous the amount of steps lost accumulates near the center, causing larger deviations. It
was measured that the Z-axis had lost between 0.9 and 1 mm in steps after an experiment at 3000 mm/min.
This explains the majority of the differences seen within Figure 5-23. The remainder of the differences is most
likely caused by either the friction or process deviations.
The step size in these experiments is alternated from 0.1 mm per pass to 0.3 mm per pass in steps of 0.1 mm,
all with a feed rate of 1000 mm/min. This provides three data points, which can be used to see the relation
between the step size and the geometrical accuracy.
In the first geometry experiment section the results of these settings didnt provided to be very accurate with
respect to the computer model due to various reasons, such as sheet bending, pillow effect and spring back.
However to determine the effect of the step size the geometrical accuracy compared to the computer model
is not important.
Instead the differences between the obtained geometries with different step size provide the required
information. Therefore, just like the feed rate experiments, the specimens of the following step size
experiments are compared to the data of the first geometry experiments in order to determine to effect of
the step size on the geometrical accuracy.
50 | P a g e Remko Kuitert
Experiments
This implies that the decrease in step size actually has an Figure 5-24 Step size experiment 2: Absolute differences
effect on the accuracy. However Figure 5-24 doesnt vs. 0.2 mm/pass
provide any information whether the accuracy is
increased or decreased. To determine if the accuracy is
increased, the data obtained from this set (red) is
plotted against the data of the first geometry
experiment (green), as can be seen in Figure 5-25.
Remko Kuitert 51 | P a g e
Experiments
The results have been made visual in Figure 5-28, Figure 5-29
and Figure 5-30. The first thing that has to be noted is that
after only a single correction the forming depth is significantly
increased, compared to the initial specimen. The initial
specimen has about 3-4 mm forming depth, where the first
correction already has 6-8 mm forming depth. For the second
correction hardly any visual changes are directly seen Figure 5-30 Specimen correction 2
compared to the first correction.
In order to provide insight in the overall accuracy improvement Figure 5-32 has been made. This figure clearly
shows in the left image that the first correction specimen is rather accurate compared to the initial input
model. The forming depth is matched up to 97.2% in fact. But this figure also that the differences between
first (middle) and second (right) iteration are very minimal (97.2% vs. 98.3%). Therefore it is probably best to
apply the correction only once, as the results from a second pass correction are minimal.
Figure 5-32 Correction 1 vs. CAD Model (left), difference iteration 1 (middle) and difference iteration 2 (right)
52 | P a g e Remko Kuitert
Experiments
Instead it might be better to compare the relative Figure 5-33 Overview of the contribution of each effect on the
error between the computer model and the actual accuracy mismatch
obtained geometry. This provides a certain
percentage, which can be used to compare the
effect of different radii with respect to the forming
depth. In order to provide such comparison the
forming depth of the specimen is divided by the
forming depth of the actual mode, to see what
kind of percentage of the depth is achieved. The
results can be seen in Figure 5-34. The numbers on
the X-axis represent the number of the
experiments, thus number 1 corresponds with
geometry experiments 1.
Remko Kuitert 53 | P a g e
Experiments
54 | P a g e Remko Kuitert
Experiments
The main reason for this deviation was that the machine itself had problems running at these feed rates,
which caused a deviation of 0.9 to 1 mm in total. But this only partially explains the differences seen between
the different feed rates. Another reason might be that the friction between the forming tool and the material
causes the material to become warm. An increase in temperature usually increases the ductility of the
material, which results in less plastic deformations. If one has the ability to either freely rotate this forming
tool or control the spindle speed, the friction between the forming tool and the material could be reduced. In
order to minimize friction the following formula could be if the spindle speed can be controlled:
=
1
(1 cos(2 )
2
Nevertheless the current results do imply that the feed rate has a negative effect on the dimensional
accuracy and is thus cant be increased over and over to achieve higher production speeds achieving the
same dimensional accuracy as initially assumed.
Remko Kuitert 55 | P a g e
Experiments
Total difference
The total difference is once again the maximum
difference between the obtained specimen and the
computer model and is the summation of the pillow
effect, the sheet bending effect and the spring back
effect. In Figure 5-40 this is visualized.
Pillow effect
Most surprisingly what can be seen is that if the
step size is decreased sufficiently the pillow effect
seems to play a rather large role in the inaccuracies,
approaching the spring back effect. However as
soon as the step size is increased, the pillow effect
fades away.
56 | P a g e Remko Kuitert
Discussion
Chapter 6 Discussion
This chapter will cover the discussion regarding the results obtained in Chapter 5 and how these relate to the
large scale application as mentioned in Chapter 3.
After the calibration phase, the deviations seen were still in the same range of 0.20-0.25 mm as can be seen
in Figure 5-14, Figure 5-17 and Figure 5-20. However the average deviation was decreased significantly. The
magnitude of these deviations is similar to the data found by Singh and Agrawi during their experiments at
similar sized specimens with different geometry [40].
For the large scale application it is assumed that the deviations, seen at the repeatability of the experiments,
are linear with respect to the size of the specimens. This implies that the deviations become in the order of 3
to 4 mm for 1500x1500 mm panels. This size of deviations is not uncommon for aircraft industry in fuselage
panels. Therefore the repeatability shouldnt be an issue for large scale products as long as the assumption
that the deviations behave linear holds.
Nevertheless after a specimen has been formed, the repeatability of the process allows a rather accurate
prediction of the next specimen. With the help of the correction method one could come much closer to the
actual aim, which increases the predictability. Therefore the initial predictability is low based on computer
models, but within a series production one could achieve predictable results after a first sample has been
made and corrected.
The problem arises if one wants to produce large panels in small series. As the process lacks a good initial
prediction, the production of the first panel will be very difficult. However after the first panel has been
successfully formed within its tolerances, the second panel can be made the same way. But the predictability
might become an issue in large scale manufacturing, due to its lack of a good initial prediction by an accurate
computer model.
The only possible deviation in the controllability arises from the backlash in the axes of the machine itself.
The backlash occurs when the axes turns in the opposite direction. A small gap between the bearing and the
threaded rod allows some backlash. However the backlash in most of these machines is limited to 0.01 mm
or less. Therefore the controllability should not be an issue in large scale manufacturing.
Remko Kuitert 57 | P a g e
Discussion
Initially it was stated that the step size had an influence on the dimensional accuracy. This statement has
been verified by the results of the step size experiments. There is a clear trend that if the step size is
decreased, the dimensional accuracy is increased and vice versa.
It was also stated that the feed rate has no influence on the dimensional accuracy. This statement has been
proven wrong. The experimental results showed that, with increasing feed rates, the dimensional accuracy
went down. Where the 3000 mm/min feed rate experiment shows this best.
Nevertheless the overall dimensional accuracy at each of these experiments was still low and did not even
approach the input model given. This was solved with the correction method experiment. This experiment
used an iteration to correct the spring back in a specimen. As a result the forming depth approached the
depth of the input model up to 97.2%. But at certain places the difference between the input model and the
obtained specimen was still +/- 0.8 mm.
These differences do provide an issue for large scale implementations. In these experiments the best case
scenario was achieved with a single iteration, but still provided +/- 0.8 mm variation at certain places in the
specimen. For the large scale implementation, larger sheets will be formed at a time and therefore the
variations will most likely increase. As a result the dimensional accuracy of these plates may have such large
variations that the process itself is not accurate enough.
But if the predictability of the process becomes better, one could predict the material behavior of these large
plates during the process. This could provide much higher dimensional accuracies and solve this problem.
This does require more research on the different effects that cause these variations, which could result in an
improved computer model for the process.
During the step size experiments it was shown that increasing the step size affects the dimensional accuracy.
Therefore there are clear limitations towards the step sizes in order to stay within the dimensional accuracy
limitations, which limit the maximum process speed.
Regarding the feed rate experiments it was expected that doubling or even tripling the feed rate should have
a minimum effect or even no effect to the dimensional accuracy and thus the feed rate could be increased up
to the required feed rate. However the feed rate does influence the dimensional accuracy, as seen in the
experiments. A higher feed rate cause more friction, which heats up the material and allows more relaxation
and therefore less dimensional accuracy.
The specimens in these experiments were processed between 1.5 and 5 minutes. For the large scale
implementation a much higher process speed is required, which implies a much higher step size and feed
rate. The experiments have shown that in order to reach these step sizes or feed rates, the dimensional
accuracy gets affected. Therefore its not possible to obtain the required feed rate and step size.
In order to obtain the required process speed a different approach is required. For example the process could
use an origami method to fold or bend the material at certain locations, which requires much less toolpath
length. Or multiple forming tools can be used at once by the same machine, which reduced the toolpath
length per forming tool. Another solution would be to use multiple machines. But with the current limitations
in feed rate and step size the amount of machines required would be tremendous in order to produce
enough panels per hour.
58 | P a g e Remko Kuitert
Discussion
6.2 Limitations
So far there have been several limitations with respect to the large scale implementation of incremental
sheet forming. The typical arguments against applying single point incremental forming in large industrial
applications are that the process is too slow and inaccurate.
2
In order to produce the required 72-129 m aluminum per hour, the feed rate should be in the order of 600
m/min and the step size should be in the order of 0.5 mm per pass. Even though machines exist that are
capable of performing at these feed rates and step sizes, its most unlikely that the results are accurate.
As seen in the experiments a step size of 0.5 mm per pass would be feasible, but would decrease the
dimensional accuracy quite a bit. The main issue is the feed rate of 600 m/min. These feed rates cause a lot
of friction, which decrease the overall accuracy. The friction can be reduced, by allowing the forming tool to
freely rotate or have a controlled rotation. But even if this is done, the feed rate is usually limited to 10 to 30
m/min to avoid friction/heating problems or material surface damage.
These results severely limit the feasibility of incremental forming for large scale applications like the one
proposed in this research. But it might be possible that in the near future a better prediction model is
developed. This prediction model can then be used to predict the material behavior at higher feed rates to
compensate the input model in advance.
6.3 Improvements
There are several improvements or ideas that can be used to make incremental forming more feasible for
large scale applications, like the one suggested in this research.
Origami method
A second improvement can be made, as already mentioned on the previous page, by changing the toolpath
philosophy. Right now the toolpath is a spiraling toolpath that goes from the outside in, with equal steps in
vertical direction. But one might want to investigate the option to use an origami folding method to reduce
the toolpath tremendously. The main issue with this method is that it does require a thorough knowledge of
the process and material behavior. But it could significantly reduce the required feed rate.
Iteration Method
Even though the iteration method is used in this research on an existing formed specimen, one might be able
to use the new toolpath obtained by the iteration method and apply it to a new flat specimen directly to see
if the dimensional accuracy is also increased. If successful, this method could be used, while an improved
prediction model is being developed.
Remko Kuitert 59 | P a g e
Conclusions
Chapter 7 Conclusions
7.1 Consolidation of the work
The main goal of this thesis was to determine the feasibility of incremental forming in terms of dimensional
accuracy and process speed for preforming aluminum sheets used in GLARE laminates. This was done with a
literature study on GLARE and incremental forming. After a theoretical analysis was done in which the
process requirements were determined. This was followed by an experimental phase, which consists of
obtaining an experimental setup and performing several experiments. The results of these experiments were
than combined with the theoretical analysis to determine to which extend the process was feasible or not.
To perform a study on the combination of GLARE material and incremental forming to define
and perform an experimental setup and test plan to determine to what extend incremental
forming is feasible as a method to preform the aluminum sheets used in GLARE.
During the literature study, the definition of GLARE was described with the different Grades and their specific
applications. Also the current production of GLARE, the hand lay-up method, was analyzed and in
combination with the material properties it became clear why this current method was not feasible for
double curved panels as described in this research. It was also shown that the material most likely to be used
for these applications was either GLARE Grade 3 or 4.
In the second part of the literature study the die-less incremental forming sheet is described and the most
recent developments in the field of single point incremental forming were mentioned. It was shown that feed
rates up to 600 m/min were achievable. But it was also shown that the accuracy depends on the sheet
bending effect, the pillow effect and the spring back, and that there are no accurate computer models.
Afterwards the literature study was used as a basis for the process conditions, where it was found that, in
theory, the production speed of large fuselage panels shouldnt be a problem as long as the combination
between feed rate and step size doesnt affect the accuracy. However for the small double curved panels,
with radii between 100-2000 mm, the accuracy might become an issue, therefore this research was focused
at these smaller, more curved panels.
The experimental setup in combination with the experiments has proven that the process itself is repeatable,
predictable and controllable, as seen during the calibration experiments. However, at first the dimensional
accuracy of the process was around 60% with respect to the forming depth. The correction method,
suggested by Micari, was proven to be useful to increase the accuracy of the process with a single iteration to
97.2% of the forming depth.
It was also shown that the feed rate had a negative effect on the accuracy and that increasing the step size
also had a negative effect on the accuracy. This implied that the step size could be increased to a certain
limit, before the accuracy becomes too low. The same applies to the feed rate, where higher feed rates cause
a decrease in dimensional accuracy. Therefore the feed rate could only be increased to a certain limit.
The experimental results were used to analyze the feasibility of incremental sheet for large scale
manufacturing, such as the Airbus A320 production line. According to the results the accuracy drops if either
the feed rate or step size is increased and therefore limited. Therefore it is suggested to stay below a feed
rate of 30 m/min and a step size of 0.3 mm/pass, to stay within the required dimensional accuracy.
For large scale industrial applications, e.g. the production line of the Airbus A320, the above mentioned
combination is not fast enough with respect to the required production rates. Especially if the splices are
supposed to be reduced, the forming depth increases and the production time per panel increases. This
makes the process, at this moment, not feasible to be used for the Airbus A320 production line. But for the
smaller, lower volume, double curved parts the process might be feasible if multiple machines are used.
60 | P a g e Remko Kuitert
Conclusions
7.2 Conclusion
The conclusion of this report is the answer to the main research question stated in Chapter 1, and repeated
below, which is:
Can die-less incremental forming be a feasible process for preforming the aluminum layers
used within GLARE laminates?
In order to be feasible the process should be able to produce parts with an accuracy of 90% or higher and at
rates fast enough to keep up with the production rate of the Airbus A320. The experiments have shown that
the accuracy can be achieved with the correction method suggested by Micari. But the main issue lies within
the process speed, which requires a combination of a high feed rate and step size. As a result the incremental
sheet forming process is not feasible for the fuselage panel production of the Airbus A320. Yet it is feasible to
produce the smaller, highly double curved parts of the leading edges and empennage with this process, due to
the lower required production volumes of these parts.
7.3 Recommendations
As could be read in the conclusions the incremental forming process as used in this thesis is not a feasible
method to be applied for large scale industrial applications, such as the Airbus A320 fuselage. However the
author would like to provide a few recommendations for future research.
Toolpath variation
Various methods exist to decrease the toolpath length, as mentioned in section 6.3. One of these solutions is
to use the art of origami to bend thin aluminum sheets into certain shapes. This could significantly reduce the
required toolpath length. So far, no research exists with incremental forming being applied in such a way and
could be the key solution to preform aluminum sheets.
Another possibility is the use of a backing plate or tool, as seen earlier in Figure 2-4. This doesnt decrease the
process speed, but it could increase the process accuracy. At this moment such as machine is used for
research at Ford, namely the F3T. Even though it seems promising the application has not be used for series
production yet [41].
Machine selection
Last, but certainly not least, the machine used in this research was handmade by the student itself. This
implies that the machine doesnt have the same specifications as professional machines, as seen in Leuven
and Ford. It is therefore recommended that, if future research is done at the DASML, an investment should
be made in a more advanced machine. For example, the machine used in this research was not able to
control the spindle speed, which increased the friction between the forming tool and the specimen. A
dedicated professional machine is usually capable of controlling the spindle speed.
Remko Kuitert 61 | P a g e
References
References
Picture used on the front-page of the report: http://www.lamieranews.it/files/2015/06/fig.2.jpg
[1] A. Vlot and J. W. G. , Fibre Metal Laminates: An Introduction, Dodrecht: Kluwer Academic Publisher,
2001.
[3] P. Broest, Assembling 3d-fml panels of undeformed aluminium sheets. estimation of internal stresses.,
Delft: Delft University of Technology, 1999.
[4] J. Sinke, "The production of 3d glare panels," Delft University of Technology, Delft, 1999.
[5] L. Filice, G. Ambrogio and F. Micari, "Shape and dimensional accuracy in Single Point Incremental,"
Journal of Materials Processing Technology, vol. 191, p. 390395, 2007.
[6] T. H. D. A. L.-. E. RUIMTEVAARTTECHNIEK, "Laminate of aluminium sheet material and aramid fibres.".
Nederland Patent NL8100087, 2 Augustus 1982.
[7] T. H. D. A. L.-. E. RUIMTEVAARTTECHNIEK, "Laminate of metal sheet material and threads bonded
thereto, as well as processes for the manufacture thereof.". Netherlands Patent NL8100088, 2 August
1982.
[8] L. B. Vogelesang and G. H. Roebroeks, "Metal-resin laminate reinforced with S2-glass fibres". United
States of America Patent US5039571 A, 11 Oktober 1988.
[10] G. Roebroeks and A. Mattousch, "Impact resistant laminate". Patent US5547735, 20 August 1996.
[11] G. Roebroeks, "Laminaat uit metaalplaten en elkaar kruisende dradenlagen uit verschillende
materialen in kunststof.". Patent NL1022706, 17 February 2003.
[12] L. Vogelesang, L. Verbruggen and C. Paalvast, "Armour plate composite with ceramic impact layer".
Patent US4836084, 6 June 1989.
[13] J. Gunnink, "Composite laminate of metal sheets and continuous filaments-reinforced synthetic
layers". Patent US4935291, 19 June 1990.
[14] R. Alderliesten, "On the Development of Hybrid Material Concepts for Aircraft Structures," Recent
Patents on Engineering, vol. 3, pp. 25-38, 2009.
[15] W. Guocai and J. Yang, "The Mechancial Behaviour of GLARE laminates for Aircraft Structures," in
Failure in Structural Materials, JOM, 2005.
62 | P a g e Remko Kuitert
References
[17] Cytec Industries Inc., "BR127 Corrosion Inhibiting Primer Datasheet," Cytec Industries Inc., 2003.
[18] E. Kroon, "Influence of general quality on Glare material performance," FMLC, The Netherlands, 2002.
[19] C. Pegels, "A study on the Residual Stress of GLARE 1," Delft University of Technology, Delft, 1995.
[20] R. Alderliesten, "Fatigue Crack Propagation and Delamination Growth in Glare," Delft University Press,
Delft, 2005.
[21] J. Sinke, Forming technology for composite/metal hybrids, Abington: Woodhead Publishing Limited,
2007.
[22] J. Schijve, "Development of fibre-metal laminates, ARALL and GLARE new fatigue resistant materials,"
TU Delft, Delft, 1993.
[23] A. Vlot, "Impact loading of fiber-metal laminates," Journal of Impact Engineering, vol. 18, pp. 291-307,
1996.
[24] E. Leszak, "Aparatus and Process for Incremental Dieless Forming". Patent US3342051A1, 19
September 1967.
[25] J. Jeswiet, F. Micari, G. Hirt, A. Bramely, J. Duflou and J. Allwood, "Asymmetric Single Point Incremental
Forming of Sheet Metal," CIRP Annals - Manufacturing Technology, vol. 54, no. 2, pp. 88-114, 2005.
[26] B. Araghi, G. Manco, M. Bambach and G. Hirt, "Investigation into a new hybrid forming process:
Incremenal sheet forming combined with stretch forming," CIRP Annals - Manufacturing Technology,
vol. 58, pp. 225-228, 2009.
[27] T. Kim and D. Yang, "Improvement of formability for the incremental sheet metal forming," Int J Mech
Sci, vol. 42, pp. 1271-1286, 2000.
[28] Y. Kim and J. Park, "Effect of process parameters on formability in incremental forming of," J Mat
Process Technol, vol. 130, pp. 42-46, 2002.
[29] L. Filice, L. Fratini and F. Micari, "Analysis of material formability in incremental forming.," CIRP Annual
Manufacturing Technology, vol. 51, pp. 199-202, 2002.
[30] G. Hussain, L. Gao, N. Hayat and L. Qijian, "The effect of variation in the curvature of part on the," Int J
Mach Tools Manuf, vol. 47, pp. 2177-2181, 2007.
[31] I. Cerro, E. Maidagan, J. Arana, A. Rivero and P. Rodriguez, "Theoretical and experimental analysis of
the dieless incremental sheet forming process.," Material Process Technology, vol. 177, pp. 404-408,
2006.
[32] K. Hamilton and J. Jeswiet, "Single point incremental forming at high feed rates and rotational speeds:
Surface and structural consequences," CIRP Annals - Manufacturing Technology, vol. 59, pp. 311-314,
2010.
[33] G. Ambrogio, F. Gagliardi, S. Bruschi and L. Filice, "On the high-speed Single Point Incremental Forming
of Titanium Alloys," CIRP Annals - Manufacturing Technology, vol. 62, pp. 243-246, 2013.
Remko Kuitert 63 | P a g e
References
[34] R. Hallet and P. Camanho, Composite Joints and Connections: Principles, Modelling and Testing,
Philadelphia: Woodhead Publishing Limited, 2011.
[35] Airbus, " Airbus to raise A320 Family production to 46 a month by Q2 2016," Airbus, 26 February 2014.
[Online]. Available: http://www.airbus.com/presscentre/pressreleases/press-release-
detail/detail/airbus-to-raise-a320-family-production-to-46-a-month-by-q2-2016/. [Accessed 5
November 2015].
[36] J. Duflou, R. Aerens, P. Eyckens and A. Van Bael, "Force prediction for single point incremental
forming," The International Journal of Advanced Manufacturing Technology, vol. 46, no. 9, pp. 969-
982, 2009.
[37] "Wood Cnc Router Table Machine Kit Plans PDF Plans," [Online]. Available: http://pcria.org/MjI0Yjc1-
cnc-router-table-machine-kit-plans.html. [Accessed 8 June 2016].
[38] Trochilidesign, "Building your own CNC router/milling machine," [Online]. Available:
http://www.instructables.com/id/Building-your-own-CNC-milling-machine/. [Accessed 8 June 2016].
[40] A. Singh and A. Anupam , "Comparison of Dimensional Repeatability and Accuracy for Deformation
Machining Stretching Mode with Sheet Metal Componen," in 5th International & 26th All India
Manufacturing Technology, Rupnagar, 2014.
[41] P. Ponticel, "Ford sheet-metal-forming technology on journey from lab to application," SAE
International, 14 Augusts 2013. [Online]. Available: http://articles.sae.org/12385/. [Accessed 30 June
2016].
[46] "Driver Stappenmotoren: Leadshine DM556 - 2 Phase Digital Stepper Drive; Max 50 VDC / 5.6A,"
[Online]. Available: https://www.hardware-cnc.nl/nl/winkel/drivers-stappenmotor/leadshine-dm556-
2-phase-digital-stepper-drive-max-50-vdc-56a-detail. [Accessed 20 June 2016].
64 | P a g e Remko Kuitert
Appendices
Appendices
Remko Kuitert 65 | P a g e
Appendix A Matlab Scripts
zlim([0 140])
66 | P a g e Remko Kuitert
Appendix A Matlab Scripts
%Scan Area
XDir = 35; %mm half side
YDir = 35; %mm half side
Step = 5; %mm per step
Coords=[];
Round=1;
for XCoords=-XDir:Step:XDir;
Round=Round+1;
if Round/2==round(Round/2);
for YCoords=YDir:-Step:-YDir;
Coords=[Coords;XCoords,YCoords];
end
else
for YCoords=-YDir:Step:YDir;
Coords=[Coords;XCoords,YCoords];
end
end
end
fclose(fid);
Remko Kuitert 67 | P a g e
Appendix A Matlab Scripts
%Diameter definition:
XRadius = 100; %mm (Radius of curvature X 200 300 or 400mm)
YRadius = 100; %mm (Radius of curvature Y 200 300 or 400mm)
%Forming Depth
XDepth = XRadius-(XRadius^2-XStart^2)^0.5;
YDepth = YRadius-(YRadius^2-YStart^2)^0.5;
Depth = min(XDepth,YDepth);
while (Z<Depth);
Distance=0;
%Calculate the first location
X = -((2*XRadius*StepSize*t-StepSize*StepSize*t*t)^0.5)*cos(2*pi()*t);
Y = -((2*YRadius*StepSize*t-StepSize*StepSize*t*t)^0.5)*sin(2*pi()*t);
Z = StepSize*t;
%To prevent any coords being addedon the list above the limited depth
if (Z<Depth);
%If Z is still below the allowed depth
%Add the step, x coord, y coord, z coord and the distance to the
%previous point into the list
Coords=[Coords;Step XNew YNew -(Depth-ZNew) Distance];
end
end
68 | P a g e Remko Kuitert
Appendix A Matlab Scripts
%In order to view the specimen in scale enable the following limits
xlim([-50 50])
ylim([-50 50])
zlim([-10 10])
Remko Kuitert 69 | P a g e
Appendix B Repetier Firmware changes
70 | P a g e Remko Kuitert