Biaxial Testing of Sheet Metal: An Experimental-Numerical Analysis

Biaxial Testing of Sheet Metal:
An Experimental-Numerical
Analysis
Gerard Quaak
MT 08.10
TU/e Master Thesis

May, 2008
Engineering thesis committee

prof.dr.ir. M.G.D. Geers (Chairman)
dr.ir. J.P.M. Hoefnagels (Coach)
ir. C. Tasan (Coach)
dr.ir. P.J.G. Schreurs
dr.ir. H. Vegter
Eindhoven University of Technology

Department of Mechanical Engineering
Computational and Experimental Mechanics
Abstract
With the recent increase in the popularity of advanced high strength steels
(e.g. dual phase, TRIP) in automotive industry, new challenges have arisen.
Conventional continuum models are found not to capture the reported pre-
mature ductile failures in such steels, which are governed by damage evolution.
Another challenge is understanding and predicting metal behaviour under com-
plex strain paths. The ability to precisely capture these eects in continuum
models is important for the sheet metal forming industry, in order to carry out
these processes as ecient as possible. However, improvements of the numerical
tools highly depend on the development of accurate and practical experimental
techniques. A testing device for biaxial deformation of sheet metal is such an
experimental tool that has been studied by many researchers before. Although
several experimental set-ups have been proposed in the literature, most of these
designs are not capable of providing information up to the point of fracture.
Additionally, these set-ups are not usable for real-time, in-situ examination of
the deforming structure with advanced microscopic techniques such as SEM,
AFM, surface prolometry, or digital image correlation, because a miniaturized
form is not available or was never investigated.
The main goal of this project therefore is to nd a practical and accurate set-
up that can deform sheet metal specimens under varying complex strain paths,
while allowing for real-time, in-situ microscopic examination. For this pur-
pose, several experimental set-ups (e.g. bulge, punch, Marciniak and cruciform
tests) have been studied and compared both experimentally and numerically.
For testing cruciform samples a simplied in-plane biaxial-loading set-up was
designed and build, while Marciniak tests were carried out at Corus RD&T.
The corresponding experimental results were used to verify and compare these
tests in terms of practical aspects (e.g. specimen preparation). The computa-
tional results are used to analyze stress and strain distributions and for better
understanding of the eects of miniaturization.
Combining literature, numerical and experimental test results, it was concluded
that the cruciform and Marciniak test are the most promising set-ups for minia-
turized biaxial testing of sheet metal with in-situ microscopic examination.
Both test have their limits, but when taken these into account can provide
valuable data.
i
Samenvatting
Met de recente populariteit van 'advanced high strength' staalsoorten (dual

phase, TRIP) in de automotive industrie, zijn nieuwe vraagstukken ontstaan.
Conventionele continuum modellen blijken niet in staat om de aanwezige taaie
breuk te beschrijven, waardoor het voorspellen van het materiaal gedrag tijdens
omvormen van deze metalen niet mogelijk blijkt. Wanner niet lineaire rek paden
een rol spelen blijken de huidige continuum modellen zelfs nog minder accuraat.
Voor het verbeteren van de numerieke gereedschappen zijn echter betrouwbare
experimentele technieken nodig, zodat het materiaal gedrag voor deze groep
metalen is te bepalen. Een voorbeeld van zo een opstelling, is er een om biaxiale
deformatie van plaat staal te bestuderen. Hiernaar is reeds veel onderzoek aan
besteed, maar hoewel verschillende gereedschappen onderzocht zijn, bestaat
er tot op heden geen opstelling om tot breuk te deformeren zonder externe
invloeden. Daar komt nog bij dat om de micro structuur van het materiaal
te onderzoeken geavanceerde technieken als SEM en AFM of digital image
correlation gebruikt moeten worden, wat inhoudt dat een miniatuur opstelling
nodig is. Mogelijkheden voor een dergelijke opstelling zijn nog niet eerder
onderzocht.
Het belangrijkste doel van dit project is dan ook het vinden van een praktische
manier om plaat staal onder veranderende rek paden, met de mogelijkheid real-
time metingen te doen. Hiervoor zijn verschillende experimentele opstellingen
(bulge, punch, Marciniak en cruciform test) numeriek en experimenteel onder-
zocht en vergeleken. De numerieke resultaten geven o.a. inzicht in spannings-
en rek velden en de eecten van miniaturisatie, de experimentele resultaten
worden gebruikt voor vericatie en vergelijk op praktische gebied, zoals het
maken van test samples. Voor het testem van kruisvormige trekstaven is een
opstelling ontworpen en gebouwd, als onderdeel van dit onderzoek. Voor de
Marciniak testen is gebruik gemaakt van een opstelling beschikbaar gesteld
door Corus RD&T.
De literatuur studie, numerieke en experimentele resultaten laten zien dat zowel
de kruisvormige trekstaven als de Marciniak test bruikbaar zijn in een geminia-
turiseerde vorm. Beide tests hebben beperkingen, maar wanneer hier rekening
mee wordt gehouden kunnen zij waardevolle informatie opleveren.
ii
Acknowledgement
This Master project was mainly carried out at Eindhoven University of Tech-
nology, in the group of professor Marc Geers. Experimental work has been
done at Corus RD&T in IJmuiden and in the University Multi-scale Labora-
tory. Special thanks goes to Sjef Garenfeld for his time and patience in the
design process of the biaxial testing set-up and specimens. I also want to thank
Marc van Maris, supervisor of the Multi-scale Laboratory, for his guidance and
advice during experimental work and Tom Engels for his help with the tensile
testing machine. From I want to thank Corus RD&T Menno de Bruine, oper-
ator of the Marciniak test set-up, Carel ten Horn and Louisa Carless for their
time and eort with the experimental work on the Marciniak test. At Philips
Drachten a lot of insight was provided in possible material removal techniques,
for which I want to express my gratitude to Gerrit Klaseboer, Harmen Altena
and Willem Hoogsteen for receiving us and a good discussion of the material
removal problem. I also want to thank Johan Hoefnagels and Cem Tasan for
their input in the project, the many evenings discussing and the help with the
nal report.
Gerard Quaak
Eindhoven, May 2008
iii
Contents
Abstract i
Samenvatting ii
1 Introduction 1
1.1 Formability and strain path dependency . . . . . . . . . . . . . 2
1.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature Survey 5
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Bulge test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Punch test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Marciniak test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 In-plane loading with cruciform geometry . . . . . . . . . . . . 17
2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3 Numerical methodology 23
3.1 Material model . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Bulge test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Punch test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4 Marciniak test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4 Experimental methodology 29
4.1 Marciniak test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.1.1 In-plane testing with the Marciniak set-up . . . . . . . . 29
4.1.2 Specimen manufacturing . . . . . . . . . . . . . . . . . . 30
4.2.1 Design of a test set-up . . . . . . . . . . . . . . . . . . . 30
4.2.2 Tests with in-plane cruciform geometry . . . . . . . . . 34
4.2.3 Specimen manufacturing and characterization . . . . . . 36
5 Results 39
5.1 Bulge test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.1.1 Miniaturization . . . . . . . . . . . . . . . . . . . . . . . 40
5.1.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.2 Punch test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
iv
CONTENTS
5.2.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3 Marciniak test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.3.1 Working principle of the Marciniak test . . . . . . . . . 47
5.3.2 Numerical - experimental study of the Marciniak test . . 49
5.3.3 Minaturization . . . . . . . . . . . . . . . . . . . . . . . 53
5.3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.4.1 Optimization of the cruciform design . . . . . . . . . . . 56
5.4.2 Proof of principle . . . . . . . . . . . . . . . . . . . . . . 57
5.4.3 Specimen manufacturing and characterization . . . . . . 60
5.4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.5 Comparative evaluation of the set-ups . . . . . . . . . . . . . . 68
6 Conclusions and recommendations 72

6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.2 Recommendations for future work . . . . . . . . . . . . . . . . . 73
Bibliography 75
A Electrical Discharge Machining 79
B Electrical Chemical Machining (ECM) 83
C Specimen Preparation: TegraPol or Target System 86
v
Chapter 1
Introduction
Metals, and in particular sheet metals, are used in a wide variety of appli-
cations in industry, with the main elds of application being packaging (food
containers, beverage cans), automotive and aerospace industry. As material
costs are a substantial part of the costs of manufactured products and most
of the products are produced in large numbers, large cost reductions can be
achieved by lowering the amount of material used. Moreover, several other
reasons that can be thought of for lowering the amount of used material, e.g.
lowering weight and lowering impact on the environment by polluting. All of
these reasons result in eorts to achieve material use reduction without quality
loss in the product.
Figure 1.1: Advantages and disadvantages of using aluminium instead of steel for a
standard medium size car [44]
In the automobile industry a rst attempt to achieve weight reduction was

done by using low density materials like aluminium, magnesium and plastics,
but recent studies show a promising future for steels instead. The International
Iron and Steel Institute (IISI) computed how the use of aluminium makes the
car body lighter, but does not have a signicant eect on the total weight of
the car and causing more environmental impact because of higher equivalent
CO2 -emissions (see gure 1.1). The development of Advanced High Strength
Steels that can replace the existing steels is therefore closely followed by the
automotive industry. [16, 44]
1
CHAPTER 1. INTRODUCTION
Figure 1.2: Overview of steel grades as used in the automotive industry [44]
Figure 1.2 shows an overview of regularly used types of steel for automotive
applications. In this gure the materials furthest to the left are most suitable
for forming virtually any desired shape, while consuming relative low amounts
of energy. The steels on the right, however, can withstand much larger forces,
which makes manufacturing more dicult. An ideal material would have prop-
erties of both steel types, being highly formable, yet very strong. Under certain
circumstances, a good combination of these properties can be achieved e.g. by
suitable phase transformations.[44]
Advanced High Strength Steels (AHSS) have such properties, e.g. a relatively
high yield strength and high hardening rate compared to conventional steels. In
the past decades signicant amounts of eorts are put in setting up strategies for
improving FE models to capture these failures. However, to make optimal use of
FE-modelling, a good description of the materials behaviour is necessary, which
relies on the accuracy of the used constitutive laws for describing the material
behaviour. The deformation-induced evolution of metal micro structure for
which, as will be explained in more detailed later, new experimental tools are
necessary. [44]
1.1 Formability and strain path dependency
In the previous section the problems with AHSS were shown, which will be
explained in more detail with the help of the Forming Limit Diagram (FLD)
that will be introduced now. The FLD is based on the assumption that for
forming purposes, the maximum deformation is limited by the initiation of
unstable deformation, e.g. necking. When forming metal sheets the material is
subjected to dierent strains and strain paths, which have been found to have
dierent maximum allowable deformations. Therefore in industry the FLD, as
shown in gure 1.3 (a), is used to show these limiting deformations.
On the axes are the strains in the two principle directions in the plane, with
the line giving the point of necking for the combination of strains at that point.
The numbers in the gure show the strain paths pure shear (1), simple tension
2
Figure 1.3: a) Schematic forming limit diagram; b) Stress limit diagram (From Ba-
nabic [3])
(2), plane strain tension (3) and biaxial tension (4). These maxima are only
true for linear strain paths to that point, and are therefore not a simple to use
as it seems.
The limits of the FLD are becoming clear when testing a sheet metal under a
changing strain path. The rst gure, 1.4(a), shows the strain paths up to neck-
ing for linear strain paths on an undeformed sheet of metal. The second gure,
1.4(b), shows where necking starts for a sheet metal that is rst deformed under
uniaxial tension, and then by biaxial tension. A large increase in formability is
found, that was not predicted by the original FLD. When starting with biaxial
tension followed by uniaxial tension, a large decrease in formability is found, as
shown in gure 1.4(c). This eect is stronger for AHSS then for conventional
steels, which makes the need for understanding what happens necessary to be
able to use the new steels up to full potential. [3, 24]
Figure 1.4: a) Linear strain paths on an undeformed sheet; b) Forming with uniaxial
tensile state followed by biaxial; c) Forming with biaxial tensile stage
followed by uniaxial (From Banabic [3])
A quite similar concept, but not less sensitive to strain path changes and thus
the strain history of the material, is the stress forming limit diagram, as shown
in gure 1.3(b). A disadvantage of the stress based forming limit is uncer-
tainty of the computed stresses, which in practice can only be determined from
measured or computed strain elds. A FE model could be used to determine
these forming stresses, but therefore the used material model should accurately
describe the material behaviour.[24]
3
1.2 Objective
The challenges that have grown due to the increased use of AHSS in new de-
signs, lead to the goal of this project. This is the development of an experimen-
tal methodology to deform a sheet metal specimen under biaxial tension. The
need for such a methodology to analyze microstructural changes under biaxial
loading is obvious, as no such set-up exists. The data obtained with such a test
method can then be used to predict damage evolution and thus in the future
make better FE-modelling possible for (re)designing products.
An experimental set-up that can be used to study the microstructural changes
will have to t in or under standard microscope systems and has to be usable
with digital image correlation systems for in-situ examination of the deforming
material. Such a set-up can then be used in future to study and characterize
new materials, as developed by the industry.
The most important properties that will be considered are:
The existence of a homogeneous stress- and strain distribution in the

studied area of the specimen, which is not inuenced by contact, friction
or other eects introduced by the test equipment.
The possibility to deform under complex strain paths, preferably with the
option to change the strain path during a test.
The initial point of fracture and the crack itself should be free to form.
Inuences from specimen or set-up must be minimized so the obtained
material data is as undistorted as possible.
1.3 Strategy
Working towards a suitable test method, various known methods for testing
under biaxial loading are studied. The literature survey in chapter 2 is meant
to provide better understanding of the problem and dierent set-ups, so a choice
can be made which testing methods will be used.
The validation and further studying of the most promising set-ups was done
both numerically and experimentally. The models and the assumptions made
to simplify the computations are being discussed in the rst part of chapter 3.
The second part of chapter 3 contains the experimental set-ups used to validate
the numerical work.
The results of both numerical and experimental work are discussed in chapter
4. The numerical and experimental results are compared in order to nd lim-
itations and possible future improvements for both, resulting in an extensive
overview of all the studied set-ups.
4
Chapter 2
Literature Survey
In the last decades several scientists have studied methods to deform sheet metal
under complex strain paths, including punch tests [3, 36, 39], bulge pressure
tests [49, 50], viscous pressure forming tests [37], biaxial compression tests [30]
or cruciform tests [11, 19, 32].
Currently the most used method for determining FLCs in the industry is by
using punch tests, which are known to overestimate the maximum allowable
strains [3]. As the exact amount of the overestimation of can not be exactly
determined, an unknown error in the resulting FLD makes a relatively large
safety margin (up to 10 %) is applied to the maximum allowable strains when
using the material in a forming process. This means more material will be used
to make a safe structure or product, leading to higher costs. [3, 45, 47]
This chapter will rst describe several properties of the biaxial testing set-ups,
that will be used in the following sections to compare the dierent set-ups.
A denition of biaxial stress is giving, followed by denitions for the working
plane, geometrical constraining and properties to measure. The set-ups to be
discussed are the punch test, the bulge test, the cruciform tensile test and
the Marciniak test, with a study of the workings of each set-up and recent
developments. The last section gives an overview of the studied set-ups, for
easy comparing of each set-up.
2.1 Introduction
Biaxial loading
In the biaxial stress state forces are working in two directions on an innitesimal
small volume, the third direction is the out of plane direction that is related to
the two in plane directions, just like an uniaxial stress state as shown in gure
2.1 on the left. The stresses working on the volume under biaxial stress can
be visualized, as shown in gure 2.1 on the right: forces are acting on the four
areas perpendicular on the plane, from which the stresses can be computed
5
CHAPTER 2. LITERATURE SURVEY
dividing the force by the area it is acting on.

Strains in a biaxial deformation can than be computed via equations 2.1 to 2.3.
Often it is more convenient to measure strains, equations 2.4 and 2.5 are given
for calculating stresses from known strains. 3 = 0 as there is no force acting
on the plane. These equations are only valid in the elastic regime, whereas in
the plastic regime pure biaxial loading only takes place up to localization. [13]
Figure 2.1: Uniaxial and biaxial stress states [13]
1
1 = (1 2 ) (2.1)
E
1
2 = (2 1 ) (2.2)
E

3 = (1 + 2 ) (2.3)
E
E
1 = (1 + 2 ) (2.4)
(1 2 )
E
2 = (2 + 1 ) (2.5)
(1 2 )
A complicating factor in the biaxial case is to determine the area that the forces
are acting on, which makes determining stresses 1 and 2 more dicult than
for the uniaxial case. Furthermore, during an actual manufacturing process the
biggest problem is determining the plastic response. This cannot be described
with a set of equations as given above.
An important observation is that real biaxial loading only occurs up to lo-
calization. Due to damage, necking and failure in a material, asymmetry is
introduced and the simplied approaches as used in the elasticity regime are
not correct anymore. Still the elastic behaviour is important, as this is where
the nal failure mode might be determined.
6
Working plane
Some experimental set-ups test a material in-plane, others out-of-plane, de-

pending mostly on tooling. As out-of-plane testing gives rise to bending, it is
preferred to test in-plane. This means stresses and strains are constant over the
thickness of the sheet, which makes computing of the stresses and measuring
the strains less complicated.
Some studies also show an inuence to the forming limit while comparing in-
plan and out-of-plane testing. Forming limits up to 6 % higher where found
with out-of-plane testing of the same material. [47]
Geometrical constraints
The geometry of a tested specimen or the set-up itself can inuence the data
measured during an experiment. Possible inuences are areas of contact where
friction plays a role or a geometry that is sensitive to a certain mode of failure.
These so-called geometrical constraints can be introduced by contact or friction
with the used tool set in the region of interest, by asymmetry of the tool
set, by non-isotropic material behaviour or by a geometry that leads to stress
concentrations. A well known example is anisotropy in sheet metals, which
gives rise to the need to test a material in more then one orientation relative
to the rolling direction of the sheet. When deforming biaxially, the anisotropy
will introduce a weaker direction, which is more likely to fail. [47]
Measuring stresses, strains or forces
Not every experiment has the same possibilities for measuring stresses, strains
or forces. As stated before, directly measuring stresses would be the most ideal
solution in most cases, but this is hardly ever possible. Stresses are normally
calculated from either a strain eld or forces and the area they are working on.
Measuring strain elds can be done with a digital image correlation set-up that
for sheet metal can measure strains on top or bottom surface. Stresses can be
calculated with the use of a FE-model, but therefore depends on an accurate
material model.
The other solution is measuring forces and the area they work on, as is done
for uniaxial tensile tests. This is only possible when both the force on an area
and the area itself can be measured. This works ne for a simple tension test,
but for more complex stress states this is often not possible.
7
2.2 Bulge test
The bulge test is a well described experimental set-up for biaxial loading, where
pressure is used to deform a specimen. The set-up consists only of a pressure
chamber and clamping mechanism. The bulge test is mostly used for testing
thin lms, as bending stresses can be neglected for that case. The pressure can
be build up by a gas, a uid or even a owing polymer [21, 37].
Figure 2.2: Pure biaxial bulge test with rounded die [21]
In gure 2.2 a simple bulge test set-up is shown, with the most important
properties visualized being t0 and td , the initial and nal thickness of the sheet,
dsheet the diameter of the sheet, dc the diameter of the die cavity, hd the height
of the dome and RC the radius of the die edge. Rd is the radius of the bulge in
a circular set-up. Rd is divided in two values R1 and R2 for an elliptic bulge,
with the two radii relating to the bulge radius in the principle directions. For
large apertures, the membrane theory can be used to compute stresses, strains
and pressures, as will be discussed in the next section.
Membrane theory
1 2 p
+ = (2.6)
R1 R2 t
where 1 and 2 are the principle stresses on the sheet surface, R1 and R2 the
radii, perpendicular to each other, p the pressure applied to the sheet and t
the thickness of the sheet. In the pure biaxial case where the bulge is a perfect
bowl, R1 = R2 = Rd and = 1 = 2 , so the equations can be simplied to
pRd
= (2.7)
2td
with td the thickness at the top of the dome. The eective stress can be written
as
8

p Rd
= +1 (2.8)
2 td
In the above relations the dome radius Rd is dened as
((dc /2) + Rc )2 + h2d 2Rc hd

Rd = (2.9)
2hd
and the thickness at the top of the dome as
2
1
td = t0 (2.10)
1 + (2hd /dc )2
with dc , Rc and hd dened as shown in gure 2.2. This model includes the
plastic deformations by using a correction for hardening, although in a simpli-
ed way. The hardening component can be adjusted by replacing the exponent
2 in equation 2.10 by (2 m), where m is the hardening power law exponent.
[21]
Slota et al [51] state that a die aperture of at least one hundred times the thick-
ness of the sheet is needed to be able to neglect bending inuences. Especially
for determining reliable strain elds this is important, because the strains vary
with the thickness of the sheet due to bending.
Advantages and Disadvantages
An advantage of the bulge test is the absence of contact (and therefore fric-
tion) in the area of interest, which makes the analytical solution less complex.
There are no geometrical constraints due to the tooling or the geometry of the
specimen.1
Some disadvantages of the bulge test include the large height dierence be-
tween the deformed and undeformed specimen, making it dicult to use lens
systems for online and in-situ measurements (e.g. imaging correlation analy-
sis). Moreover, only strains in the 1 > 0 and 2 > 0 region can be determined,
as the sheet needs to be clamped over the whole outer region to prevent the
pressurizing air or uid from escaping.
The high pressure also leads to uncontrollable neck and crack propagation, be-
cause of the force controlled nature of the experiment. Necking and fracture
might occur in a split second, with no tools available to measure the phe-
nomenon. The high pressure needed in a miniature set-up might even prove
to be a problem to reach in a conventional set-up without the use of a large
hydraulic system or polymer as pressure body. [37]
1
In a bulge test the thickness of the sheet varies, with the thinnest part of the sheet
forming in the centre. This can be considered a geometrical constraint, as it forces fracture
at this point.
9
Further literature
The most important properties described extensively in literature including

out-of-plane bending, especially for small apertures [15, 51], uncertainty of the
exact shape and thickness of the bulge [15, 49] and uncertainty of the moment
of fracture [50]. A big disadvantage is found when changing strain distributions,
as this leads to building new die shapes for every wanted distribution [3, 9, 50].
10
2.3 Punch test
A second, somewhat similar approach, is the use of a punch to deform sheet

metal under many strain paths, including biaxial tension. Several standardized
tests are available (e.g Keeler, Nakazima and Hasek tests) as described by
Banabic [3]. Although all these tests are used to determine the same material
properties, there are several dierences. The biggest disadvantage of the punch
test is the presence of contact, as this both gives rise to geometrical constraints
and adds friction to the problem.
An advantage of the punch test is its ability to undergo various strain paths,
all of them up to necking and fracture. Many ideas to achieve dierent strain
paths have been proposed and will be briey discussed. Changing the strain
path during a test is practically impossible for a punch test, as tooling geometry
and specimen are xed in most cases.
Punch set-ups
The Keeler test uses punches of dierent radii to vary the strain path of the
tested sheet metal specimen, introducing dierent strain paths due to geome-
try and friction variations. The specimens are the same for every test, which
makes the test easy to prepare. The dierent punch shapes, as shown in gure
2.3, make the test more time consuming if a larger part of the FLC is to be
determined. The test can only determine the positive part of the FLC, 1 > 0
and 2 > 0.
An alternative where the same specimen, but only one type of punch are used,
is the Hecker test. In this case the amount or type of lubricant is varied, which
gives dierent strain paths. For this test again only the positive part of the
FLC can be found. [3]
Figure 2.3: Punch shapes as used in the Keeler test [3]
In the industry the Nakazima test (or sometimes the similar Hasek test) is
most often used to determine material properties. For both tests a simple
hemispherical punch and a circular die are used, while the shape of the specimen
determines the strain path. Especially for the Nakazima test both tooling and
specimen are relatively simple. The Nakazima specimen, as shown in gure
2.4 on the left side, only dier in width W . Strain paths for 1 > 0 can be
11
found with this test. The main disadvantages apart from those due to friction
are possible wrinkling and measurement errors caused by the curvature of the
punch. Specimens proposed by Hasek (gure 2.4) can be used if wrinkling is a
problem. The advantages and disadvantages are the same as for the Nakazima
test. The advantage of avoiding wrinkling is countered by the extra work needed
to manufacture the specimen.
Figure 2.4: Specimen geometries for Nakazima and Hasek punch tests [3]
Tooling inuence
The test methods as shown up to here have dierent regimes they can be used
for, as shown in gure 2.5. This clearly shows the limits of the uniaxial tension
test, the bulge test and Keelers test. It also shows how dierent tests can lead
to dierent results, mainly because of dierences in tooling and deformations
because of that.
Figure 2.5: FLCs established using dierent testing methods: 1. Hasek; 2. Nakaz-
ima; 3. Uniaxial tension; 4. Keeler; 5. Hydraulic bulge [3]
12
Sheet thickness
Another observation is sheet thickness inuencing the results in all set-ups,

caused by dierences in bending stresses. Both Raghavan and Banabic show a
rising FLC for thicker sheets, showing how a thicker sheet, with more material
to ow, is leading to higher forming limits. This is observed for both in-plane
and out-of-plane testing and can therefore not be described as a pure bending
eect. More likely is the presence of an edge eect, leading to stiening of the
surface of the sheet and leading to earlier fracture in thin sheets as there is
less material in the centre to distribute the stresses introduced by deformation.
[3, 47]
Overestimation of fracture strains
A second disadvantage of the punch test is the overestimation of acceptable

strains, mainly because of tooling introduced geometrical constraints on necking
behaviour. [3, 45, 47]
An eect found in punch tests, mainly due to friction, is localizing of the neck
away from the centre of the specimen. A second problem is that the punch test
does not allow diuse necking of the material, leading to larger formability [38].
This happens as the material on top of the punch sticks to the punch, resulting
in lower strains. Depending on the shape of the punch, the test method and
the lubrication this determines where the material fails and under what strain
path. For most punch tests this behaviour is unwanted, but tests like Keelers
are partly based on this principle. A test with a hemispherical punch and
varying lubrication states can be used to determine failure from pure biaxial
strain paths (in the centre) to almost pure stretching. [3, 8]
Measuring
The punch test can be used with an image correlation system, as the top area
is free from obstacles, but the strain eld can only be measured at the outer
layer of the sheet. As the strain eld will not be homogeneous through the
thickness of the sheet and therefore the measured strains might not represent
the actual strain eld. The eect of friction on the surface of the punch might
also introduce an error that has to be compensated for when wanting to measure
the real material properties instead of the properties under the given set of
restrictions.
13
2.4 Marciniak test
An alternative punch test was proposed by Marciniak and Kuczyski [39], re-
sulting from their theory on loss of material stability under biaxial tension,
which manifests itself by a groove running perpendicular to the largest princi-
ple stress. They showed how their hypothesis about local strains concentrating
in this groove could experimentally be veried with a set-up as shown in gure
2.6.
The idea behind the Marciniak test is it simply converting a vertical force into
a biaxial force in the horizontal plane. This is done by a at punch deforming
a test specimen indirectly via a washer sheet with a central hole. The hole
expands radially as the punch moves in and because of friction the tested sheet
of metal expands with the washer. The radial friction forces in the contact
region between washer and sheet also prevent the sheet from fracturing near
the rounded edge of the punch, with the largest strains found in the at central
part of the specimen. The central part is now uniformly balanced, biaxially
loaded, with no contact in the area, allowing failure to occur anywhere in this
region.
History
In 1977 Tadros and Mellor [53] expanded the theory of Marciniak and Kuczski
by adding dierent tooling shapes. They proposed using elliptical shaped tool-
ing, resulting in various biaxial loads from pure biaxial to aspect ratios of 1:7.
They give results for several materials, for some the test set-up works, for oth-
ers like brass 70/30 it does not. Further research by Mellor showed dierent
damage behaviour up to fracture for brass. [46, 47]
Figure 2.6: Schematic diagram of the Marciniak test tooling set-up for in-plane test-
ing of sheet metal [47]
As dierent punch geometries are a costly method for testing, other options to
achieve dierent strain paths have been investigated. One method in particular
seems to have potential and is described by Raghavan [47] as a simple technique
to generate in-plane forming limits. His proposal, based on earlier work by
Gronostajski and Dolny [20], diers from the others by the use of dierent
14
specimen and washer geometries. With this combination, compared to earlier

methods, a wide range of strain paths can be prescribed.
Strain paths
The Marciniak set-up has been used to study the role of material defects under
balanced biaxial stretching conditions, but the Raghavan proposal makes it
useable for failure under dierent strain paths. Any strain path from uniaxial
to balanced biaxial can be achieved with the right washer and sheet geometry.
The used sheet and washer geometries are shown in gure 2.7 and divided in
several types, depending on strainpaths that can be generated with them. [47]
Figure 2.7: Typical specimen (top) and washer (bottom) congurations used for
drawing and stretching strain states in the in-plane Marciniak test fol-
lowing from the Raghavan proposal [47]
In tests with dierent types of steel and aluminium both the elliptical tool-
ing suggested by Tadros and Mellor, and the the varying washer geometries
suggested by Raghavan were capable of reaching strains of up to 40 %. The
largest dierence is the fact that Raghavans method can go into negative minor
strain paths, i.e. in his tests he spans a range from -25 % to 40 %, while the
elliptical punch only reaches positive strains. The type I Raghavan geometry
spans minor strains from -25 % to -10 % for determining forming limits in
the draw region, the type II geometry can be used for determining the plane
strain region, with minor strains from -10 % to 10 %. Geometries III and IV
give strain paths in the stretching region, with the latter equal to the classic
biaxial balanced Marciniak test. Both Tadros and Mellor, and Raghavan found
positive minor strains from 15% to 40% with these geometries. [47, 53]
15
Tooling geometry constraints
The mentioned articles not only show high strains can be reached, they even
deform up to fracture and succeed in that. In the classic Marciniak biaxial
balanced case, the type IV geometry, the central part of the specimen is under
uniformly biaxial tension and failure can therefore initiate anywhere in the
central region. This results in several nearly biaxially loaded necking areas.
With all other geometries, both for Raghavan and the elliptical tooling, the
strain distribution is not perfectly uniform, resulting in failure near the centre
of the specimen. This again is considered a geometrical constraint, although the
observed fracture paths suggest at least some defect sensitivity, as the fracture
paths vary between similar tests. [47]
Further literature
Some other properties of the Marciniak test discussed in literature include the
ability to see the inuence of anisotropy (r-value) [2, 47] and the inuence of
sheet thickness [47, 53]. Also the simple set-up that can be build on a conven-
tional tensile tester with only the specimen geometry to vary for dierent strain
paths [14, 47]. set-ups in several sizes have been used ranging from diameter of
75mm [53] to large enough plates to cut out tensile specimen for uniaxial test-
ing [14]. Strain measurements can be done with an image correlation system,
which is easy because of the at nature of the area of interest [14].
16
2.5 In-plane loading with cruciform geometry
The set-ups shown earlier mostly test the sheet metal out-of-plane, introducing
bending stresses. As this is undesired when producing "clean" material data to
determine the properties of a material, in-plane alternatives have been studied
for a long time to replace the out-of-plane set-ups.
Figure 2.8: Flat cruciform shape with necking widths for the arms (1) and the centre
(2) shown
A lot of research is focusing on cruciform shaped specimen to overcome the

out-of-plane problem. The basic idea of a cruciform specimen is based on a
standard tensile test, but with a second direction of loading added. The four
arms of a cruciform specimen can be given a displacement, thus introducing
tensile forces in two directions perpendicular to each other in the centre of the
specimen, as shown in gure 2.8.
History
The idea of using at, cruciform shaped specimen has been researched since
the sixties, by Shiratori and Ikegami (1967), Hayhurst (1973), Kelly (1976),
Makinde (1989) and several others [11, 30]. The methods described by them
full the requirements as mentioned in the introduction, by generating an ho-
mogeneous strain distribution in the thickness direction, yielding in the central
part of the specimen and being capable of describing dierent strain paths. Not
all methods are useful for reaching necking or fracture conditions though, for
dierent reasons.
Several authors studied the possibilities of cruciform specimen for determining
yield loci or hardening, which has the advantage of only going into the yield
region and no further. Promising results for determining yield-locus where
found by Mller and Phlandt [42] by using a specimen as shown in gure
2.9(a). For this geometry, high stress localization is found near the notches,
but for deformation up to yield the geometry is useable.
A similar goal, but with a dierent geometry, was achieved by Hoferlin et al.
17
[26], who used a square sheet with multiple small clamps to prevent introducing
an in-plane bending moment. The method was used to experimentally deter-
mine the yield locus of ve dierent materials and comparing it with nite
element simulations.
Several authors used cruciforms related to the geometry as shown in gure
2.9(b). [19, 31, 32, 57] These all have slits in the direction parallel to the tensile
forces in common, with the idea behind it being the avoidance of bending forces
in the plane of interest. Kuwabara et al. [31] claims the slits to make the strain
distribution in the biaxially loaded zone almost uniform. Some of these studies
use curved arms, others use straight arms, some even use both to compare.
Kuwabara [31, 32] tested low-carbon sheet metal and determined experimen-
tally the plastic work for a strain range up to < 0.03 in the biaxially loaded
zone under load ratios of 4:2 and 4:4. In a second work they determine the
yield surface, with the use of an abrupt strain path change. A similar specimen
is used by Wu et al. [57] for testing a biaxial tensile set-up capable of realiz-
ing complex loading paths. No local strain measurements where done for this
set-up though, making it dicult to compare the useability.
An optimization of the Kuwabara specimen was performed by Gozzi et al. [19],
in order to study the mechanical behaviour of extra high strength steel. They
had a problem with reaching the desired amount stress, as failure occurred
before reaching that stress in the biaxially loaded region. A geometry as shown
in gure 2.9 b) was used and optimized, where the notches where changed to
keep the stress in the corners low enough to prevent failure there. Dierent
lengths of slits where found to be preferable in some situations.
A dierent geometry is proposed by Yu et al. [58] in a study on forming limits
for sheets under complex strain paths. Using a nite element model to optimize,
they come up with a cruciform shape as shown in gure 2.9(c). The centre of
the cruciform has been thinned, with a cross-shape thinned area surrounding
a bowl shaped area that is even further thinned. The general idea behind this
shape is obtaining the most uniform stress distribution in the central region.
According to the authors, complex strain paths can be achieved by adjusting
the velocity ratios imposed on the specimen arms.
Demmerle and Boehler [11] describe several cruciform geometries in their arti-
cle, ranging from uniform thickness specimen with trapezoidal arms to plates
clamped by three or more limbs from each side. The geometry they investigate
most is the one proposed by Kelly, as shown in gure 2.9(d). From their they
look into two alternatives based on the original specimen of Kelly, namely 2.9(e)
and 2.9(f) where the centre area is thinned in a square form or round form re-
spectively. They nd a uniform stress distribution in the centre, but with stress
localizations in the corners. A last design proposed by them is shown in gure
2.10 and is found the best as it has the lowest stress localization out of the
centre. The design was found to expensive to realize though and never tried.
A problem described by Demmerle and Boehler [11] is loading of anisotropic
materials, which will result in a distortion of the loading axes. Some solutions
18
Figure 2.9: Several cruciform geometries as used in dierent studies; a) Mller and
Phlandt [42]; b) Kuwabara and Gozzi et al. [19, 32] ; c) Yu et al. [58]
; d) Kelly [27] ; e) Square modied Kelly [11]; f) Round modied Kelly
[11]
19
Figure 2.10: Optimal design of a biaxial specimen by Demmerle [11]
for this problem that are proposed in literature are making slots in the arms
[19, 31, 57] or giving the tensile set-up that is used more degrees of freedom
[7, 11]. No studies where found on using the cruciform geometry up to fracture
under biaxial loading, other then for composites [52].
Alternative approaches
An alternative to weakening the centre part of the cruciform is strengthening

the arms. This method is used in composite testing Smits et al. [52], where
it is relatively easy to shape the test specimen in the desired form. For sheet
metals no literature was found on strengthening the arms, although some al-
ternative solutions to be though of are to glue extra material to the arms or
to change material properties in the arms, for instance by changing the mate-
rials microstructure by case- or surface hardening (e.g. carburizing, nitriding,
boriding or titanium-carbon diusion).
Manufacturing
The main problem mentioned for a thickness reduced cruciform specimen is the
change in material properties due to manufacturing and the change in sheet
properties due the removal of the outer layer. The former problem can only be
minimized, by studying removal methods that introduce only minimal damage
to the original material, including Electro Discharge Machining as is used in
other studies and said to have little impact on the tested material. [56]
Test set-ups
The set-up as proposed by Kuwabara et al. [31] and also by Smits et al. [52]
uses hydraulics to drive the tensile tests. This has the advantage of being
able to directly connect the two opposing hydraulic cylinders via to a common
hydraulic reservoir and thus keeping the pressures (and therefore forces) exactly
the same. With a servo controlled system it is possible to change the strain
path.
20
A second approach found in literature is the use of a mechanism, xed in a

conventional tensile testing machine [26, 43]. set-ups like this can be divided in
two main groups - vertical and horizontal loading - as is shown in gure 2.11.
All mechanisms need to be adjusted for diering strain paths, which make a
path change during an experiment impossible.
Figure 2.11: Possible mechanisms in conventional tensile stage. Top pictures: verti-
cal under stage; Bottom pictures: horizontal under compression [43]
21
2.6 Summary
The four methods discussed in this chapter, the bulge test, the hemispherical
punch test, the in-plane cruciform tensile test and the Marcianiak test are the
most practical tests for creating biaxial stress elds. Each of them has its
advantages and disadvantages that will be analyzed in this report with more
details using both numerical and experimental tools.
For the bulge test, the biggest advantage is its simplicity and the existence of
an analytical model. The useability of the analytical model and the typical
dimensions for a miniaturized set-up is analyzed. In the punch test, the most
important property inuencing the results is friction. Therefore the inuence
of friction when miniaturizing is analyzed. The Marciniak test, advertised as
having perfect in-plane, biaxial loading will be both numerically and experi-
mentally veried. The main properties that will be investigated are friction,
tooling geometry and tooling forces, for nding the possibilities for a miniatur-
ized set-up. The last set-up to be analyzed is the in-plane cruciform set-up,
for which the main challenge will be designing and building the set-up and
producing the specimens. Combining the results of the four methods, the most
suitable method for biaxial testing of sheet metal to fracture will be suggested,
with recommendations for possible further investigation.
22
Chapter 3
Numerical methodology
The literature survey describes several experimental set-ups that are used for
biaxial testing of sheet metal in the industry. Each of these set-ups has its
advantages and disadvantages, some of which can be investigated with the help
of a numerical model. The models used in this study are presented here, after
in the rst section the used material model is described, as this is the same for
all numerical models.
3.1 Material model
All numerical and experimental work is done with IF-steel, which is provided
by Corus RD&T. In MSC.Marc the elasto-plastic material model is used to
describe the IF-steel, which is an isotropic model, using Young's Modulus (E ),
Poisson's Ratio ( ) and a plasticity criterium as input parameters.
Table 3.1: Material properties for MSC.Marc isotropic elasto-plastic material model
Property Value
Young's Modulus [GPa] 45
Poisson's Ratio [-] 0.29
Initial Von Mises Yield Stress [MPa] 130
Plasticity is modelled with the piecewise linear method, using a table of equiv-
alent plastic strain and Von Mises stress. The used values for these properties
are given in table 3.1, with the plasticity curve that is given in gure 3.1. The
stress in this gure is determined in a standard tensile test, which per de-
nition is equal to V M (the Von Mises stress) as used by MSC. Marc. From
the obtained stress-strain distribution only the part up to necking is loaded in
MSC. Marc, as the elastic-plastic material model is dened as such.
23
CHAPTER 3. NUMERICAL METHODOLOGY
Figure 3.1: Plasticity curve as used in the MSC.Marc material model
Necking, damage and localization
When interpreting the numerical results, it is important to understand how

plasticity is implemented, as this can otherwise lead to wrong assumptions. The
plasticity curve as used here only describes the stress evolution up to necking
of the material. The reason for this is that local strains become much higher
and result in necking, which is not described by the elasto-plastic material
model. Furthermore the numerical model cannot accurately describe localiza-
tion, which is found when the material starts to neck and fail. To avoid this
problem a failure criterium or local damage model is needed, which is beyond
the scope of this project as such a damage model can only be determined with
the obtained experimental data from the biaxial set-ups.
3.2 Bulge test
The model for the bulge test is based on the analytical model given by Gutscher
[21] as shown in gure 2.2 in the literature survey. The most important prop-
erties to study are the resulting stress- and strain elds and possibilities to
miniaturize.
The FE model
The model was build using straight axisymmetric thick shell elements of type
1, which are suitable for large displacements and large deformations. Shell
elements are chosen for their suitable computational behaviour and bending
incorporation. The degrees of freedom (axial, radial and right hand rotation)
are sucient to describe the problem, and a pressure boundary condition can
be used as well.
24
Figure 3.2: FE Model of the bulge test (expanded view is used for the shell elements)
(MSC.MARC Mentat)
The element only has one integration point for stiness and two integration
points for mass and pressure determination, which means small time steps
are necessary when describing large plastic deformations. Stresses over the
thickness of the element are integrated with Simpson's rule, over 5 layers and
11 integration points (default settings that are suitable for complex plastic
deformation).
To take into account the thickness variation during the bulge test, the updated
lagrange method is used.
Table 3.2: Properties of the bulge test model
Property Value
Sheet thickness t0 [mm] 0.7
Cavity diameter dc [mm] 20 to 150
Clamp rounding Rd [mm] 5.25
To determine the needed pressure and maximum dome height at the point of
necking, it is assumed that failure occurs when the Von Mises strains reach the
maximum of 380M P a as determined in a uniaxial tensile test.
3.3 Punch test
The punch test set-up that will be studied is the hemispherical punch, as this is
the most commonly used punch type in industry, e.g. in Nakazima tests. The
main goal of the analysis will be to study the eect of friction on deformation, as
25
friction is believed to inuence the measured material parameters. The model

is build up from solid elements, as they are more suitable for visualizing stress-
and strain gradients. The set-up is modelled axisymmetric to keep the number
of elements as low as possible.
Figure 3.3: FE Model of the punch test (MSC.MARC Mentat)
FE model
The axisymmetric model has four bodies in it, namely the sheet that is modelled
using solid axisymmetric elements of type 10, a four node quadrilateral element
with bilinear interpolation1 , the punch and the two clamps that are modelled
as rigid body curves, see gure 3.3.
Table 3.3: Properties of the computational punch model
Property Value
Punch diameter [mm] 50
Inner diameter die [mm] 56
Radius die edge [mm] 2
Clamp force [kN] 200
Number of elements 2000
Element Type Axisymmetric solid (quad), bilinear (Type 10)
To study the eect of friction the coulomb model is used, which is used in
MSC.Marc to describe any friction except for shear friction. The coecient of
friction, Mps will be varied between 0 and 1. (For more information on friction
and friction coecients, see Giancoli [18]). To investigate the miniaturization
problem the same model is used with the geometry scaled down to the dierent
punch sizes.
1
Higher order elements can not be eectively used, as in contact MSC.MARC only uses
the bilinear approximation
26
3.4 Marciniak test
Similar to the bulge and punch models, the Marciniak FE model is described
here, followed by the most important properties that will be studied. The
model can be used to gain more insight in the general workings of the set-up
and eects of several design parameters, including punch edge geometry, punch
size and washer geometry.
FE model
The model was build using element type 10, axisymmetric solids, which were
also used in the punch test analysis2 . Tests showed 3D shell element can be used
as an alternative, but for faster computation the axisymmetric solids approach
was chosen. In table 3.4 the used set-up and geometries are described. The
model describes the experimental set-up that is available at Corus RD&T and
will also be used for experimental work.
Figure 3.4: FE Model of the Marciniak test (MSC.MARC Mentat)
The simulations to investigate the inuence of changing parameters are per-

formed with dimensions of the experimental set-up that is available at Corus
RD&T. The inuence of friction is studied by varying friction coecients be-
tween the washer and sheet (Mws ), and between washer and punch (Mwp ) are
2
2D axisymmetric shell elements are available and can be used to describe the model
geometry. These shell elements do not work well in contact though when using more then
one deformable body
27
Table 3.4: Dimensions of the Marciniak test set-up for the FE Model
Property Value
Punch edge radius [mm] 10
Inner diameter die [mm] 53.78
Radius die edge [mm] 3
Clamp force [kN] 200
Number of elements 2000
Element Type Axisymm. solid (quad), bilinear (Type 10)
varied between 0 and 1. The friction is modelled using the Coulomb Friction
model in MSC. Marc.
The cruciform model is computed with 3D elements to describe the thickness

reduced geometry, with only one-eight of the cruciform being modelled when
possible.
Table 3.5: Dimensions and properties of the cruciform model [54]
Property Value
Width (w) [mm] 10.0
Radius of rounding (Rw ) [mm] 3.0
Radius of bowl (Rreduced ) [mm] 6.5
Thickness sheet [mm] 0.7
Number of elements (one-eight) 5000
Element Type 3D solid (hex), trilinear (Type 7)
Boundary conditions in the cruciform test are symmetry planes and displace-
ments of the arms only, as they are there is no contact in the set-up outside of
the clamping area. To compare the numerical results with the experimental re-
sults, the numerical strain elds have been calculated for the same strain types
as can be found experimentally. This is total strain, major strain and minor
strain.
The resulting stress eld is used to nd where the maximum tensile stress is
reached, which is assumed to be the starting point for necking. Parameters that
have been varied to optimize the numerical model are the thickness reduction
and the radius between the arms. All models have a bowl shaped thickness
reduction, where the radius in the plane of the sheet is kept equal at all times.
28
Chapter 4
Experimental methodology
To verify the numerical results from the tests as mentioned in the previous
section, several experiments have been carried out. The experimental work can
be divided in two main parts, the Marciniak test and the cruciform geometry.
4.1 Marciniak test
To compare the numerical results with experimental results, IF steel specimens

were tested on a set-up that was made available by Corus RD&T. The set-up
used is an existing tooling set, normally used for testing deep drawing prop-
erties. This means the used geometry is mostly pre-determined by the tooling
available. The properties that were varied were washer hole size and the amount
of friction between washer and sheet.
4.1.1 In-plane testing with the Marciniak set-up
The test set-up is a hydraulic punch with a 100mm die attached. The punch
measures 50mm in diameter and has a rounding of 10mm. The die used to
clamp the specimen has an outer diameter of 100mm and an inner diameter
of 53.78mm. The edge of the die is rounded with a radius of 4.5mm, which
is considered large enough not to cause fracture on the die. The clamping
force used in the tests is set to 300kN , so no material can ow out of the die.
The speed of the punch can be adjusted for each test, but is normally set to
10mm/min.
To make clear photos for strain measurements, some of the tests are carried out
with stops and the last test is carried out at a slow punch speed without stops
to make photos. Both a simple digital photo camera and an advanced photo
camera set-up as used by Corus RD&T where used. Several specimen have
been painted with a high contrast pattern for use with the Aramis system, using
standard black and white paint. This system works by following the distortion
of a pattern that is sprayed on top of the specimen of interest, comparing each
29
CHAPTER 4. EXPERIMENTAL METHODOLOGY
picture that is taken to the previous. Lightening for the test was provided
by ambient light and a clip-on diuse LED light. Furthermore a pattern of
regular dots was applied to each of the specimens for use with the Argus strain
measurement system.
4.1.2 Specimen manufacturing
Manufacturing of specimen for the Marciniak test is done by stamping and wire
EDM. The 100mm sheets to be tested are stamped from a steel strip at the
Corus workshop. The 100mm washer sheets are cut by wire EDM, after which
the hole in the centre is removed with the same process. As the width of the
specimen is much larger then the EDM or stamping aected zone, no inuence
of the manufacturing process is found in the test specimen.
A set of washer was produced with holes of 5mm, 7.5mm, 10mm, 10.5mm,
11mm, 11.5mm and 12mm, as 11mm was numerically found to be the best
choice.
The cruciform geometry will be experimentally tested, to verify the numerical

results and to study possible diculties that occur in an experimental environ-
ment. The experimental set-up that was used to carry out the biaxial tests was
designed rst, and therefore the design of this set-up is explained rst.
4.2.1 Design of a test set-up
A set-up to test the cruciform specimen at the Eindhoven University of Technol-

ogy was designed keeping several guidelines in mind. First of all the set-up had
to go all the way to fracture, resulting in a total elongation of approximately
4mm. The maximum forces to be expected are under 2kN in two directions.
The set-up has to pull the cruciform with equal force from all sides, with all
the forces acting in one plane. The possible use of the Aramis strain measuring
equipment is preferred, and so is the possibility to change the strain path for a
test.
Table 4.1: Properties for the experimental set-up
Property Size
Displacement [mm] 4-6
Force [kN] 2-5
Other Strain path changeable
Strains measurable
Several designs have been discussed, including the use of a Kammrath und
Weiss tensile stage with additional parts to pull in the perpendicular direction.
30
The compactness of such a set-up makes this impossible to achieve in the plane
of the force acting on the cruciform though. Other designs ended up with
the similar problems, leaving a mechanism to be placed in a standard tensile
machine as the best option.
Several of these mechanisms have been analyzed, with the cruciform either
horizontal or vertical position in relation to the mechanism. The vertical solu-
tion has the obvious advantage that the deformation of the cruciform is easily
followed, as it is clearly visible from the sides, with no obstructions by either
the mechanism itself of the tensile machine. Practical design of a set-up with
the cruciform clamped vertically was found to be a big challenge, as a simple
mechanism cannot elongate the cruciform in four directions at the same time.
Slider mechanisms or double hinges would be necessary, which lead to friction
problems that can make the test rig instable. [17]
Figure 4.1: Principle of the biaxial tester and non-linear movement of the joint
The nal design was therefore made with the cruciform in a horizontal position,
where a compression movement of the Zwick 1474 tensile machine is used to
push four arms outwards and thus stretching the cruciform biaxially. The set-up
was build at the university workshop, with the use of standard parts mostly. As
a result of the use of ball joints in the mechanism, the vertical movement of the
tensile testing equipment is not perfectly transferred in horizontal direction. As
shown in gure 4.1, the joint directly responsible for the horizontal movement
of the test specimen is following an arch-like path. For small displacements the
displacement is approximately linear, which means linear displacements can be
accurately described.
The photo of the set-up (gure 4.2) shows the total set-up. This set-up can be
placed in a conventional tensile testing machine, placing the bottom cylinder
(2) in a centre ring on the tensile testing machine and screwing a standard
cylindrical connector to the top cylinder (3) to attach the set-up to the moving
part of the machinery. When the tensile tester then moves in the direction of
the blue arrows, the joints of the arms move out in the direction of the red
arrows. In the set-up available at the Technical University of Eindhoven the
bottom cylinder is xed, with the top cylinder being pressed down by the tensile
tester.
31
Figure 4.2: Experimental set-up for testing cruciform specimen
32
The eight 125mm arms form four mechanisms, with three joints each to convert
the vertical displacement into a horizontal displacement. To change the load
distribution, the arm length can be changed by using longer or shorter arms,
so dierent strain paths can be tested. Small adjustments can be made by
elongating the arms itself. The arms have ball joints as shown in gure 4.2
(5), that can freely rotate in all directions to prevent a clamped cruciform from
being suppressed in more directions then there are degrees of freedom.
The clamping mechanism has been made by pressing a clamp onto the clamping
area of the cruciform, as shown in gure 4.2 (6) and (7). The four clamps
can move in vertical direction and have a thread that is used to pre-stress
the clamped cruciform. This makes it possible to adjust the position of the
cruciform and get it aligned.
The test set-up is used in a standard tensile machine as stated, which gives two
problems. The rst is the fact that the clamped specimen, although loaded
in-plane, will move in vertical direction as the joints that are connected to the
clamps move in vertical direction. The second problem is the fact that the
cruciform is clamped horizontally, with the tensile machine present above and
beneath, which makes it more dicult to have a clear view of the cruciform.
The rst problem is easily solved, as preliminary tests showed the camera is
able to keep focus over the distance the cruciform moves. The other problem
was solved by adding a cavity in the lower cylinder of the biaxial test set-up
(3) where a mirror is placed under 45 . This mirror gives a clear view (4) on
the deformed specimen, with the possibility to add a camera set-up for online
strain measurements in front of the tensile tester. The camera can be pointed
at the mirror in the bottom cylinder through a hole in the cylinder, as shown
in gure 4.2.
Misalignment
Misalignment of the forces in a cruciform set-up can be introduced by clamping

the specimen in under an angle in one of the clamps. In gure 4.3 stress
distributions are shown in cruciform specimen under dierent misalignments,
to analyze the resulting behaviour. The four test are all exactly the same, with
the cruciform being pulled 2mm in each direction, so not reaching the maximal
tensile stress. In all three misaligned cases, a movement of 1mm away from the
principle direction is added.
Each of these possible misalignments results in a highest stress in the centre
of the cruciform, with hardly any distortion of the stress and strain elds.
This observation is an important one, as it shows how the test is insensitive
to misalignment due to faulty or skew clamping. The biggest inuence of
misalignment as predicted by the simulations is the development of a preferred
necking direction, as bands of higher stress develop. These bands develop in
places where the material is likely to fail without this extra high stress as well,
which makes the impact less signicant.
33
Figure 4.3: Stress distribution in a cruciform specimen and the inuence of mis-
alignment in the cruciform set-up; a) Symmetric tensile test; b) One
arm misaligned; c) Two arms with opposite misalignment displacement;
d) Two arms with equal misalignment
4.2.2 Tests with in-plane cruciform geometry
To run a biaxial tensile test, the following steps need to be carried out. First
the tensile set-up is fastened in the specially designed holder, in which the arms
can be positioned in a 45 angle. For the current set-up, the holder is xed in
the exact position to achieve this. The mechanism needs to be position in the
holder in the right direction, which can be checked by aligning the black marks
on the holder and mechanism for both the top and bottom cylinder.
With the mechanism in the holder a cruciform specimen can be fastened in
the set-up. It is important to make sure the clamps are on the device before
attaching the cruciform, as the cylindrical part of the clamps cannot be added
with a cruciform in place.
It is recommended to x the cruciform in two opposite clamps rst, while
making sure the clamps are aligned. With the rst two clamps tightened, the
other two clamps can be fastened, where it is important again to align the
clamps before tightening. When the four clamps are fastened and checked, the
set-up is ready to be placed in the tensile testing machine.
To move the mechanism without deforming the clamped specimen, two steel
bars are screwed to the mechanism. While attaching the bars, it is important
that the bottom cylinder is kept in contact with the holder. The alignment
markers have to stay in touch as well. If the bars are xed, the mechanism can
34
be taken from the holder and placed in the tensile machine.

To attach the mechanism to the Zwick tensile machine, a metal connector is
screwed to the top cylinder. The hole in this connector has to be aligned
with the mirror hole in the bottom cylinder. If the connector is in place, the
mechanism can be placed on the metal centre pad in the tensile machine and the
compression rig can be lowered into the connector. Positioning the compression
rig low enough for the steel xation bar to be put in place requires some caution,
the rig may not exert any force yet.
With the mechanism fastened, the compression rig can be directed to the zero
position by balancing the force to zero. To start the test, the Zwick software
is used and set-up for a standard compression test.
Force displacement measurements
As the experimental set-up as mentioned above does not measure forces or

displacements on the specimen itself, these have to be computed from data
obtained from the Zwick tensile testing equipment. The force-displacement
curves from the tensile testing equipment can be converted are obtained via
the controlling software and give time, displacement and force.
As the angle of the arms of the test rig are approximately 45 , the force and
displacement of the tensile tester are simple to determine. The movement of
the joint to which the clamps are xed is approximately equal for the horizontal
and vertical displacement. The force on the cruciform in one direction is ap-
proximately half the force measured by the tensile machine. This does not take
into account any losses in the testing rig though, which cannot be determined
without adding load cells on the separate arms.
Aramis strain measurements
Using the mirror set-up as described above, strain measurements are made with
the Aramis system. When the biaxial tensile set-up is placed and fastened in
the tensile tester, the Aramis system can be installed. The camera is pointed at
the 45 mirror and adjusted so it shows the cruciform specimen in the Aramis
software. A (diuse) light source is then installed, after which the shutter time
of the camera can be set. If needed the needed shutter time can be changed
by adjusting the diaphragm of the camera1 . After the camera is set-up and
focused, the Aramis software is set-up to take a photo every second, which is
the minimum time step. For slower tests this might be adjusted, a number of
50 to 100 photos are needed for the strain eld calculations.
1
For shutter times close to 50Hz the background lighting inuences the photos, which
makes using the diaphragm needed to be able to increase or decrease the shutter time without
over- or underlightening
35
4.2.3 Specimen manufacturing and characterization
During the manufacturing process of the cruciform specimen several manufac-

turing steps are used, of which some might lead to damage or errors in the
geometry. The steps will therefore be explained here, and several test to inves-
tigate the inuence of the main production step, EDM, are introduced.
Figure 4.4: Shape and size of the cruciform specimen
The rst step in manufacturing the cruciform specimen is cutting the cruciform
shape, with dimensions as given in gure 4.4. The cutting is done by wire
EDM, this will not have a huge inuence on the material, as the thickness
of the aected material is small compared to the thickness and width of the
specimen. The second step for manufacturing the cruciform specimen is die
sink EDM, to thin the centre of the cruciform. EDM is used as it can achieve
the wanted precision and can cut the wanted geometry, without a big distortion
of the original material, according to several studies (see Appendix A).
The inuence of the EDM process is characterized by several experimental
tests. These include height prolometry, SEM, grain size measurements, nano
indentation and tensile tests, which are briey discussed.
Surface prolometry
The EDM process copies the geometry of the electrode into the workpiece
material, removing material from both the workpiece and the electrode. The
nal product should be a perfect copy, which will be veried using the Sensofar
Optical microscope. The microscope is used to make a height prole.
36
The settings used for the height prole are a resolution of 4 with the 5x mag-
nication. A height of 1000m is scanned to capture the cruciform and the
removed bowl shape.
Scanning Electron Microscopy
SEM images of the cross-section of the as received steel sheet and the thickness
reduced specimen are prepared by grinding and polishing with the Struers tar-
get machine. The specimen where cut mechanically from the sheet and grinded
for more then 2 mm to remove any deformed material. The results are com-
pared with the results for the as received material. To make the grains visible,
the cross-section was etched with the same method as the microscopy samples.
Grain size measurements
Using the Zeiss Axioplan 2 grain size measurements of thickness reduced spec-
imen are compared to specimen from as received material. The preparation
of the specimen is the same as for the tests with as received material. The
samples are cut from as received material and then a cross-section is grinded
and polished to make the surface at enough for microscopy. After the nest
polishing step, the surface is etched using a 30 second nittal bath (5 % solution)
and a 20 second step with Marshall's Reagent. The combination of these two
attacks all the grain boundaries and makes the grains clearly visible.
The recipe of the Marshalls reagent is as follows, constituted by two parts.
Part A consists of 5 ml sulfuric acid (concentrated), 8 g oxalic acid and 100
ml of water. Part B is a 30% solution hydrogen peroxide solution. Before use,
mix part A and B in equal parts and use the mixed solution fresh. For better
results, 1ml of hydrouoric acid per 100ml of solution can be added. Part A
can be stored, the mixture cannot.
Nano indentation
Nano indentation test are carried out on thickness reduced specimen and spec-
imen of the as received material. For each position 21 indentations where done
over the thickness of the specimen with a Berkovitch tip, with a 50 m spacing
over the thickness and a nal displacement of 500m into the material. The
sets of measurements perpendicular to the thickness have been made 100 m
from each other. The location of the indents is shown in gure 4.5. The hard-
ness that is used to compare the microstructure over the thickness of the sheet
and to compare the original material with the thickness reduced material is
computed by the indentor software. Preparation of the specimen was carried
out with the Struers Target System.
37
Figure 4.5: Positions where nano indentation was done to determine local hardness.
Tensile tests
The tensile tests as done to characterize the eect of EDM are carried out with
the same specimens as the earlier mentioned tests. The specimens for this test
have been thickness reduced using EDM or wire EDM though, to study the
eect of these processes.
Thickness treduced is 200m, the transition from the clamping area is a smooth
one due to rounding by the EDM process.
treduced
t
6,5 12
10 4
31
Figure 4.6: Tensile bar specimen with thickness reduction for tensile testing
38
Chapter 5
Results
For the out-of-plane set-ups, the most important challenge about the three
studied set-ups is the possibility to build a miniaturized version for use with
microscopy. This is analyzed for the dierent set-ups, and it will play an
important role in the end to give an answer to the question which test is the best
choice for biaxial testing with microscopy tools. For the cruciform specimen
however, miniaturization is not the most important property to analyze, as
the biggest challenge here lies not in miniaturization but in deformation up to
fracture with as little distortions as possible.
39
CHAPTER 5. RESULTS
5.1 Bulge test
The advantage of the bulge test is the availability of an analytical model, as

was shown in the literature review. This model is valid for large set-ups, but
for smaller set-ups it might not be as usable as bending will become more
inuencing. For the bulge test the stress distribution in a specimen near fracture
is shown in gure 5.1, as predicted by the FE model. The distribution shows
only a small gradient over the thickness of the sheet in the top of the dome. For
smaller set-ups however, the role of bending stresses increases, as is analyzed
in the next section.
Figure 5.1: Computed Von Mises stress eld for the bulge test with an aperture
diameter of 50mm close to fracture
A second disadvantage of the bulge test originating in the concept itself, is to

keep top of the bulge, where fracture occurs, in focus for use with microscope
techniques. This means the distance between microscope and specimen has to
be kept constant by moving the microscope along with the growing bulge. A
last problem is the explosive burst that is to be expected when the material
fractures, resulting in both dangerous situations and deformation of the fracture
zone.
5.1.1 Miniaturization
Miniaturization of a bulge set-up means the aperture which allows the sheet
to deform, will be decreased in diameter. For apertures of 20 to 200mm the
results are given in table 5.1.
The decreasing height for a miniaturized set-up is a good sign, as for microscopy
only limited height is available in most set-ups. The rising pressures on the
other hand are a big challenge, as this means the set-up needs to be reinforced to
cope with the forces acting on it. High pressure is dangerous under microscopes
as well, especially in a SEM vacuum chamber where an sudden increase of the
internal pressure will damage the microscope.
The numerical computations made with MSC.Marc have been compared with
40
CHAPTER 5. RESULTS
Table 5.1: Numerical results at fracture of bulge test simulations with MSC.Marc
for varying cavity diameters for stresses close to f racture
Cavity diameter Height Pressure Normalized height1

dc [mm] hdmax [mm] Pmax [MPa] hd /dc [-]
20 6.60 30.00 0.330
25 7.85 25.00 0.314
30 9.05 21.10 0.302
40 11.60 16.40 0.290
50 13.95 13.25 0.279
60 16.55 11.25 0.276
80 21.30 8.50 0.266
100 26.30 6.90 0.263
150 38.65 4.65 0.258
200 50.75 3.50 0.254
the analytical solution, as shown in gure 5.2. The analytically determined

pressure at the point of necking can be calculated by rearranging equation 2.8:
2
Pmax = (5.1)
( Rtdd + 1)
Figure 5.2: Comparison of numerical and analytical solution for bulge pressure ex-
periment near fracture
The analytical model follows the numerical model for large apertures, but for
small set-ups the analytical model predicts a lower pressure then is found nu-
merically. The reason for this change originates from the neglection of bending
stresses in the analytical model, which makes the analytical model overestimate
the thickness reduction. The advantage of the bulge test having an analytical
model to determine stresses therefore does not hold for a miniaturized set-up,
which means numerical computing is still needed.
41
CHAPTER 5. RESULTS
For the results as shown above a miniaturized version of the bulge test small
enough to t under a microscope would need to have cavity diameter with a
maximum of 50mm, considering the total set-up with clamps will then easily
measure 100mm in diameter. The height dierence of approximately 15mm is
relatively large, especially for use with microscopes.
5.1.2 Summary
Pressure for testing up to fracture is predicted to reach 12M P a or more,
which can lead to dangerous situations in case of fracture. The explosive
character of the test also can lead to alteration of the fracture surface,
which is unwanted.
The increasing pressure results in a second disadvantage, being the need

for a robust set-up. Miniaturizing now leads to the need for a stronger set-
up, to cope with the higher pressures, and for very small bulge diameters
this results in an increased size of the test apparatus.
For use with microscopes the area where fracture will occur has to be
kept in the same plane, which is dicult with a bulge test.
One of the biggest advantages of the bulge test normally found, is the
good description of the forces and stresses by the analytical model, but
for a miniaturized set-up the analytical model is found not to be correct
anymore.
42
CHAPTER 5. RESULTS
5.2 Punch test
The punch test, in several forms, is the most used material characterization
test in industry. The reason for this is simple, as the test can be used over a
large range of strain paths and test specimens are easy to make. A disadvantage
that is taken for granted though is that friction inuences the measured material
data.
When looking at the stress distribution as shown in gure 5.3, the punch test
and the bulge test look to behave similar, but this is not entirely true. This
originates in the fact that for a sheet on a punch, localization is postponed due
to restriction of material ow. This is not the case in the bulge test, where the
maximum allowed strains thus are found to be larger.
Figure 5.3: Numerically found stress eld for a 50mm spherical punch test at fracture
Data obtained from the punch test simulations is given in gures 5.4 where the
inuence of friction is shown. In table 5.2 the location of necking relative to the
centre is given in relation to friction, where higher friction results in fracture
away from the centre. Experimental work conrms this problem, e.g. at Corus
RD&T punch tests frequently have to be repeated several times before fracture
at the centre of the contact area of punch and specimen is found. [40]
This inuence of friction leads to a fundamental problem of the punch test:
Friction is known to inuence the measured material data, but the friction
coecient itself is often unknown. This results in measured material data that
does not represent the actual material behaviour, but material behaviour in a
certain set-up. This eect was also found in literature, as was shown in gure
2.5.
43
CHAPTER 5. RESULTS
Table 5.2: Properties of the miniaturization study FE model
Friction coecient Necking Distance Sheet Thickness

[-] from centre [mm] in centre [mm]
0 0 0.490
0.01 0.05 0.492
0.025 0.1 0.495
0.05 0.4 0.499
0.10 1.0 0.504
0.20 3.5 0.507
0.30 5.0 0.511
0.40 8.6 0.525
0.50 10.7 0.546
0.75 15.4 0.581
1.00 16.5 0.616
Figure 5.4: The inuence of friction between punch and sheet on the stress distribu-
tion in a 50mm punch near fracture
A logical result for miniaturizing a punch set-up is the growth of the inuence
of bending stresses. Therefore a normal industrial set-up with a punch diameter
of 100mm is compared with a miniaturized version with a punch diameter of
30mm. In the standard set-up, as shown in gure 5.5 b), the gradient of the
stresses and strains over the thickness of the element is small. The stresses on
the outer layer are only 1.5 % higher then on the inner layer. The strains vary
even more, measuring approximately 7 % increase from outer to inner layer.
For the same region in a miniaturized set-up we nd much steeper gradients,
as shown in gure 5.5 a), with variations up to 20 % for the plastic strain.
Figure 5.5 clearly shows how a miniaturized punch set-up gives diculties in
determining material properties, as the assumption that neglecting bending
44
CHAPTER 5. RESULTS
Figure 5.5: Comparison of large punch (D = 100mm) (a) and small punch (D =
30mm) (b) stress- and strain distributions (Von Mises) in a radius of
2.5mm round the centre at fracture; Note that the eect is exaggerated
due to scaling.
stresses is not inuencing the results, does not hold anymore. A useable punch
size of 50mm gives strain gradients of 8 % in the failure region, which is still
unwanted.
Table 5.3: Maximum punch forces at necking, obtained via FE-modelling in
MSC.MARC for dierent set-up sizes
Punch diameter [mm] Punch Force [kN]

10 3.8
20 8.3
30 12.7
40 17.3
50 21.8
60 26.3
80 35.1
100 43.9
A last property determined via the FE-modelling of the punch test is the nec-
essary force for a given size of the set-up. Table 5.3 shows the punch forces
obtained from the MSC.MARC model. Punch forces are directly related to
the size of the set-up, decreasing with miniaturization. For a set-up of 50mm
forces of approximately 22kN are needed, which is still a very high force to be
achieved by a small set-up. Smaller set-ups are not desirable, as bending starts
playing a larger role, inuencing the results even more.
45
CHAPTER 5. RESULTS
5.2.2 Summary
Fracture in the punch test is not similar to fracture in a free surface under
biaxial tension, due to friction and bending. This results in an error in
the material behaviour that is measured.
An increase or decrease in friction results in a change in the measured

data. As the inuence of friction cannot be measured, the eect of it on
the measured material data is unknown.
Miniaturizing results in an increased bending stress, due to the decrease in

the punch radius. Therefore the material data measured in a miniaturized
set-up is even more inuenced by external factors then in a large set-up.
Forces needed for a miniaturized set-up with a diameter of 50mm are

found to be around 22kN .
46
CHAPTER 5. RESULTS
5.3 Marciniak test
The Marciniak test makes it possible to transfer a vertical displacement in to a

horizontal load, which can be used to deform a sheet biaxially in the horizontal
plane. The modication proposed by Raghavan enables a sample to be loaded
not only biaxially, but in multiple strain paths, as has been described in the
literature study. This makes the Marciniak test an interesting set-up to use, if
it can be miniaturized for use with microscope equipment.
5.3.1 Working principle of the Marciniak test
Figure 5.6: Overview of the stress-eld in the Marciniak test under loading up to
necking
The numerical model is rst used to verify the claims that no stress gradient is
to be found in the biaxially loaded region of the specimen, which can be seen in
gure 5.6. The stress eld is found to be similar as found in literature. When
look at the stress distribution away from the centre, a problem occurs. The
stress (centre ) is found to be lower then the stress on the outer radius of the
punch (edge ) and for all numerically tested set-ups it was found that:
edge
1 (5.2)
centre
This indicates failure is expected to always occur away from the centre of the
sheet. However, when looking at the test results of the set-up that was made
available by Corus RD&T (see 4.1.1) three modes of failure are found. An often
observed failure mode is cutting behaviour, where the edge of the washer hole
cuts into the tested sheet, thereby initiating a crack and guiding it in a circular
direction. A second failure mode is a typical deep drawing failure, where the
sheet and washer fracture in the sidewall after stretching. The wanted failure
mode is fracture in the centre under biaxial loading, which is found in some
of the tests. In gure 5.7 (a) cutting, (b) deep drawing and (c) random crack
mode are shown.
47
CHAPTER 5. RESULTS
Figure 5.7: Three modes of failure found when testing sheet metal with the
Marciniak test; a) Cutting; b) Deep drawing; c) Random crack
The experimental results show that the set-up does result in fracture in the
centre, under the right circumstances. The numerical computations and exper-
imental results are found to do show similarities, as both seem to be critical for
reaching fracture in the centre. Therefore the experimental results have been
analyzed in more detail to nd a possible explanation. To determine whether
the stresses occurring in the experimental Marciniak test reach higher values
outside of the biaxially loaded region as well, the local strains were measured
using a regular pattern on the undeformed sheet and comparing this to the re-
sulting pattern on the tested sheet. This system, known as Argus, needs every
dot to be visible in at least 3 and preferably 5 photos taken from dierent an-
gles, but this was found to be an impossible task with the small test specimen.
A photo and a set of microscope pictures of the sample as shown in 5.8 have
been used instead to approximate the strains manually instead.
Figure 5.8: Marciniak test with regular pattern on top to determine the strain eld
By comparing the increase in distance between the undeformed pattern and the
pattern after the test, local strains even can be computed as = (L L0 )/L0 .
For several fractured specimen, this results in strains of (28 +/- 2%) and (29
+/- 2%) for the respectively the centre and edge of the test specimen. This
results in a measured value of approximately 1.04 for edge /centre , again larger
then one. The numerical model is veried with another geometry as well, as
used in the article by Raghavan [47].
It seems that with the quotient of edge /centre is larger then one experimentally
48
CHAPTER 5. RESULTS
as well,but the specimen can still fracture in the centre. An explanation for
this is that interaction between washer and sheet restricts the tested sheet from
localizing in the contact area. This result is in agrement with experimental
results, where necking is only found in the free centre of the specimen and not
on the washer. The critical balance that is observed is likely to originate from
this phenomenon as well, as edge can not be a lot larger then centre to still
get fracture in the centre.
Figure 5.9: Close-up of the random fractured sheet with necks stopping at the
washer-sheet interface for the 7.5mm washer hole
An interesting observation is made on the randomly fractured sheets and shown

in gure 5.9). It can be seen that several necks have formed, that all stop at
the point where the washer is in contact with the sheet. This leads to the
hypothesis that the deformation away from the centre can indeed be larger
then in the centre, but due to restriction of localization the sheet cannot fail in
the edge and therefore eventually fails in the centre.
From the above it can be concluded that the restriction of localization is a
key factor in the Marciniak test. It is therefore important to get a better
understanding how several design parameters inuence the stresses. This infor-
mation can then be used when miniaturizing. The numerical model is found to
be useable for this, as the simulations seem to predict the (absolute) very well.
The point where localization starts can not be predicted yet, and therefore the
location of the nal crack can not be predicted.
5.3.2 Numerical - experimental study of the Marciniak test
For better understanding of possibilities to optimize the Marciniak test for

miniaturizing, the important design parameters will be analyzed numerically.
The parameters that are studied are the friction between punch and washer
(Mpw ), the friction between washer and sheet (Mws ), the punch edge radius,
the washer hole radius, inner die radius and die edge radius. These numerical
computations will be complemented with a set of experimental tests, to verify
the numerical results. These test have been carried out with a set-up available
at the Corus RD&T. The results are shown in table 5.4.
In table 5.4 the washer hole radius is the size of the hole in the washer, punch
49
CHAPTER 5. RESULTS
Table 5.4: Marciak Tests at Corus - Results

Nr. Washer hole Punchspeed Stops Roughened Load Displ. Failure mode
radius [mm] [kN] [mm]
1 11 10 mm/min No No 65.0 14.9 Cutting
5 7.5 10 mm/min No No 73.0 16.8 Random
6 5 10 mm/min No No 78.1 12.1 Deepdraw
7 10 10 mm/min No Yes 67.4 14.8 Random
8 10 10 mm/min No Yes 66.1 15.7 Random
9 10 10 mm/min Yes Yes Cutting 2
10 10 1 mm/min Yes Yes 70.0 14.7 Random
11 7.5 1 mm/min No Yes 85.2 15.7 Deepdraw
12 10.5 1 mm/min No Yes 72.7 14.6 Random
13 7.5 10 mm/min No Yes 79.4 14.7 Deepdraw
speed is the speed of the punch moving up. Stops were used in some test
to take photos during the test, for use with the Aramis strain measurement
system. Roughening was done by grinding of the washer, to increase the amount
of friction between washer and sheet. All tests but the rst one were done
with lubrication of the punch, as it was not possible to vary the lubrication
conditions.
Friction
The eect of friction was studied by varying the friction coecient in the sheet-
washer (Msw ( and punch-washer contact (Mpw ). It is important to know that
friction coecients for unlubricated steel on steel contact go as high as 0.65,
while lubricated contact friction coecients can be as low as 0.04, in the case
of perfect lubrication with teon for instance [18].
Figure 5.10 shows the inuence of (Msw ). The results show a rather small but
clear inuence, where low friction between the sheets results in higher stresses
on the outside of the tested sheet as the washer can not transfer its deformation
on to the tested sheet. For high friction coecients the stresses in washer and
sheet grow to become of the same size, resulting a relatively higher stress in
the centre.
Figure 5.10: Inuence of friction between the washer and the sheet; Dpunch =
50mm; Rpunchedge = 10mm; Rwasherhole = 11mm; Mwp = 0.05;
50
CHAPTER 5. RESULTS
Experimentally the eect of increasing friction between washer and sheet can
clearly be seen that for a washer with a hole radius of 10mm. This results in
a cutting mode type of failure for low friction (test nr. 4), while for increased
friction a random crack forms in the biaxially loaded region (tests nr. 7, 8 and
10). The increased friction is does not result in a huge improvement of the
edge /centre value, but due to high sensitivity to this value a small change in
friction can be of inuence. This delicate balance is also found when comparing
tests nr. 5 and 11, where a 7.5mm washer hole was used.
The second set of models shows the inuence Mpw , as shown in gure 5.11. The
friction coecient in the washer-sheet contact is set to 0.8, as this high value
was found to be needed. As is expected, a low friction results in a better stress
distribution, but again the inuence seems to be small when only checking the
stress quotient. From gure 5.11 it can not be seen how the stress distribution
changes for friction coecients higher then approximately 0.5, where the biaxial
loaded region never reaches the maximum tensile stress due to failure in a deep
drawing mode.
Figure 5.11: Inuence of friction between the washer and the punch; Dpunch =
50mm; Rpunchedge = 10mm; Rwasherhole = 11mm; Msw = 0.8;
The experimental results have not been used to verify the inuence of Mwp , as
changing lubrication conditions in a controlled manner was not possible.
Washer hole size
The radius of the washer hole is a simple property to adjust, so for an experi-
mental set-up it is important to know its inuence to take advantage of it. In
gure 5.12 it can be seen how the washer hole diameter has an optimum at
around 10mm. The innitesimal small washer hole equals a deep drawing test
of a sheet that is twice the thickness of the tested material.
Figure 5.12: Inuence of the washer hole radius; Dpunch = 50mm; Rpunchedge =
10mm; Mwp = 0.05; Msw = 0.8;
51
CHAPTER 5. RESULTS
The experiments conrmed that for small washer holes the failure mode indeed
becomes similar to a deep drawing test (e.g. test nr. 6). Tests nr. 11 and 13
show failure in a deep drawing mode as well for the 7.5mm washer hole size.
This shows both friction and washer geometry can cause this failure mode, as
the 7.5mm washer hole of test nr. 5 fails in a random fracture due to lower
friction between washer and sheet.
A larger washer hole can result in the washer sliding of the punch and thereby
initiating a cutting type of failure. This is found with the 11mm washer hole in
tests nr. 1, 2 and 3. The fact that for 10mm and 10.5mm washer holes random
failure is possible shows once more the sensitivity of the Marciniak test.
Punch edge radius
The radius of the punch edge is important when miniaturizing, as it inuences

the stress eld and can thus be used to lower the value of edge /centre . To
analyze the inuence, dierent radii edge radii have been numerically tested.
The washer hole size was adjusted to be approximately Rpunch Rpunche dge 0.9,
so the washer sheet stays on top of the punch during the test.
Figure 5.13: Inuence of the punch edge radius; Dpunch = 50mm; Rwasherhole =
11mm; Mwp = 0.05; Msw = 0.8;
As is shown in gure 5.13 the maximum stress in the central region of the tested
sheet cannot reach its maximum for small radii of the punch edge. This can
easily be explained, as for an innitesimal small radius the punch has become
a cutting tool, where the tested sheet will fail due to shear.
Other design parameters
Numerical computations have been done to study the inuence of the inner die
radius and die edge radius as well, showing no inuence on the stress distribu-
tion.
Conclusions
The numerical and experimental results show similar trends, which means the
model can be used for a qualitative analysis. The small dierences between the
52
CHAPTER 5. RESULTS
centre and edge stresses show that it is not easy to lower the stress quotient,
which means optimizing the set-up for miniaturization will not be easy.
5.3.3 Minaturization
All the earlier studied properties work together in a miniaturized set-up. When
scaling down the set-up in total, the curve in gure 5.14 is the result. This
shows how the quotient of the stresses raises when miniaturizing, which is
to be expected due to increasing bending stresses. Also it can not predict a
minimum size that will still work, but it does show a clear trend.
Figure 5.14: Inuence of miniaturizing the Marciniak set-up; Rwasherhole = 11mm;

Mwp = 0.05; Msw = 0.8;
To nd the best set-up, several design parameters have been analyzed and from
the results obtained it is possible to construct a numerical model of a miniature
Marciniak test choosing all the best values. This perfect set-up has large friction
on the sheet-washer interface and low friction on the punch interface and a
washer with a hole scaled to the optimal size found for a 50mm punch.
Table 5.5: Geometry for the miniaturized Marciniak set-up
Property Value
Punch edge diameter [mm] 8
Washer hole size [mm] 15
Inner die radius [mm] 42
Die edge radius [mm] 3.5
The result for this optimized set-up was found to be marginally worse (edge /centre =
1.015) then for the simply scaled set-up as shown in gure 5.14 (edge /centre =
1.012). This again shows how optimizing the Marciniak test set-up is not as
trivial as for other set-ups. With a more extensive optimization it is thought
to be possible to nd a better geometry, and several other design parameters
can be considered as well.
Additional washer on top
The use of an additional washer on top of the standard set-up is thought to

result in better restriction of localization, and more force transferring between
the washers and sheet. The expected decrease of edge /centre is not found
though, as the computational model results in worse behaviour then the original
test. Experimentally adding an extra washer might still work due due to an
53
CHAPTER 5. RESULTS
increase in restriction of localization that the washer and additional washer

have on the tested sheet, but at the expense of increased forces.
Washer material
An alternative approach for changing the stress eld is to change the material
of which the washer is made. Finding a suitable material might prove dicult
though, as the following requirements have to be met:
The material needs to have high friction with the sheet material.
The material needs to have low friction in the contact with the punch.
The material must be roughly as formable as the tested sheet, so the

washer can drag the sheet outward.
The material must be capable of reaching stresses comparable to the

stresses in the tested sheet so it can transfer forces and restrict localiza-
tion.
Some materials that come to mind are polymers, rubbers and other metals.
The last group will result in higher forces, as another metal has to be as strong
or stronger then the tested material. Polymers and rubbers can easily undergo
the large deformations, but due to their generally lower strength might not
able to transfer enough stress to and from the tested sheet. As friction and
the area of contact play an important role in the transfer of stresses as well,
optimization of a set-up with another material proves to be a complicated task,
but there is room for improvement.
Other design restrictions
When the miniaturizing both the needed forces and maximum height of the
set-up are important, therefore these values where computed for ve dierent
sizes of the punch. When comparing the values in 5.6 with those found for the
punch test, it immediately becomes clear that the forces in the Marciniak test
are a lot higher. The maximum height is found to be relatively low, as 11mm
is not expected to cause problems with microscopy set-ups. The forces can
prove to be a problem though, as a set-up strong enough to cope with forces
of approximately 70kN might be impossible to be build on a small scale. For
high strength steel or thicker sheets the forces will even be higher.
The computed punch forces as shown in table 5.6 were compared with the forces
found in the experimental set-up. For a washer with a 11mm hole radius it
was found that the computed maximum force equals 62kN , the measured force
lays between 63 and 65kN . For a smaller washer of 7.5mm the computed force
is 70kN , which is again a little lower then the measured punch force of 73kN .
54
CHAPTER 5. RESULTS
Table 5.6: Maximum punch forces at necking for a Marciniak test set-up;
Rwasherhole = 11mm; Mwp = 0.05; Msw = 0.8;
Punch diameter [mm] Force [kN] Maximum height [mm]

25 28 6.2
35 41 8.2
50 62 11.1
75 92 15.6
100 125 20.3
An idea to decrease the needed force for the Marciniak set-up is to use a thinner
washer sheet. It was numerically veried that reducing the thickness of the
washer up to 50% is possible without a signicant inuence on the stress eld.
This results in a force that is approximately 20% lower, depending on washer
hole radius as well.
Conclusion
The results for miniaturization show that a challenge lies ahead when wanting
to miniaturize the Marciniak set-up. There are several possibilities to improve
the used set-up though, which leaad to opportunities to miniaturize.
5.3.4 Summary
The experimental results clearly show the Marciniak test works, although an
exact prediction of the failure mode is dicult. Friction between washer and
sheet is found to be important, as is the washer size.
The numerical results obtained with the FE model showed lower stresses in the
centre of the cup then on the edge. This is conrmed experimentally and leads
to the hypothesis that necking or localization is restricted on the washer-sheet
interface.
The Marciniak test is found to experimentally work and can be qualita-

tively described with a numerical model up to necking.
The Marciniak test shows a high sensitivity to friction and tool geometry,
resulting in failure modes other then fracture under biaxial tension. For
miniaturization purposes optimization is needed to nd the best tooling
geometry, which probably varies with material properties of the tested
specimen.
Building a Marciniak test of dimensions small enough to t under a mi-

croscope is found to be possible. However, a disadvantage is the need
for high forces up to 50kN or more, which ask for a clever design when
miniaturizing.
55
CHAPTER 5. RESULTS
Many researches studied the use of cruciform specimen over the last decades,
with most of the studies focussing on yielding properties. One step further is
studying necking and fracture with cruciform test specimen, which leads to the
new challenge of having to raise the stresses in the centre of the cruciform so
failure occurs there.
A study performed by Roel Vos [54] showed the cruciform specimen can be
used up to fracture, but a thickness reduction is needed in the central part to
make it fail before one of the arms does. After analyzing several geometries
a bowl shaped thickness reduction was suggested, which will be the geometry
used for further numerical and experimental work. The study by Vos predicts a
reduced thickness of 0.2mm should be sucient to reach fracture in the centre
of a cruciform with 10mm width arms.
5.4.1 Optimization of the cruciform design
To verify the earlier results and analyze the resulting stress and strain elds,
new simulations have been run. The resulting strain distributions for Von
Mises, major and minor strains are shown in gure 5.15. The two important
properties that are found from the numerical analysis are the biaxial straining
of the centre and the strain band developing from the centre outwards. The rst
indicates failure can occur in the centre, the shows geometrical constraining.
The diagonal strain band shows how a crack can only develop in this direction,
which means the orientation of the crack is not dened by the material but by
the geometry. This results in the need for testing with specimen of dierent
orientation to fully characterize a material that is anisotropic.
Figure 5.15: Strain elds as determined numerically showing Von Mises strains, ma-
jor strains and minor strains
For an experimental set-up the clamping area of the specimen is important to

prevent slip in the arms. As the width of the clamps in the biaxial testing
set-up is 12mm, this is the maximum width the specimen can have. To reduce
the forces on the set-up and reduce the change of clamp slip, the width of the
arms for experimental testing has been reduced to 6mm. This results in the
dimensions as shown in table 5.7. The thickness of the centre is adjusted to
180m to increase the chance of fracture in the centre.
56
CHAPTER 5. RESULTS
Table 5.7: Dimensions of the cruciform specimen as shown in gure 4.4

Property Value
Width (w) [mm] 6.0
Radius of rounding (Rw ) [mm] 3.0
Radius of bowl (Rreduced ) [mm] 4.75
Thickness sheet [mm] 0.7
Thickness of centre [ m] 180
5.4.2 Proof of principle
Using the experimental set-up as described in section 4.2.1, a set of cruciform

specimen with dimensions as given in table 5.7 was biaxially loaded up to
failure. From six specimen tested, only one fractured in the centre, showing
either the thickness reduction predicted by the numerical analysis is not enough
or the material behaviour in the thickness reduction is not exactly the same as
the global material behaviour of the IF-steel.
A new set of specimen with a thickness reduced area of 160m was manufac-
tured and tested, resulting in ve successful tests out of ve tested specimen.
Fracture in all specimen occurred as predicted numerically, as is shown in gure
5.16. The crack clearly develops diagonal, and video footage show the crack
starts in the exact centre of the specimen.
Figure 5.16: Cruciform specimen fractured under biaxial deformation, with and
without an Aramis pattern on the surface
For comparison of the experimental results with the numerical results, the
Aramis system is used to measure the local strains on the tested specimen. All
ve experiments show similar strain distributions, and similar strain values at
fracture, as is shown in gure 5.17. These measured strains are similar to the
computed strains as shown in gure 5.15.
The reason for the dierence between measured and computed strains, espe-
cially in the centre of the specimens, is the fact that local strains are measured,
which are not numerically computed. The problem can also be originating in
the manufacturing of the cruciform specimens, which could lead to damage
or alteration of the material properties. This will be investigated in the next
section of this chapter.
With the computed and measured strain elds being so similar, the next step is
to analyze the displacements and forces acting on the cruciform. This is done by
57
CHAPTER 5. RESULTS
Figure 5.17: Strain elds measured with Aramis for three dierent cruciform speci-
mens, showing Von Mises strains, major strains and minor strains
measuring the displacement and force at the tensile machine, the Zwick 1474,
as no load cells are present in the biaxial tester. As the set-up is only for small
displacements, the displacements at the clamps can easily be determined from
this as 1/2 of the total displacement. The total elongation of the cruciform in
one direction therefore is equal to the displacement of the tensile machine. For
the forces a similar computation can be made, as due to the 45 angles the total
vertical force is equal to the total horizontal force, which means 1/4th of the
total force is acting on each clamp.
Figure 5.18: Elongation-Force diagrams for the cruciform specimens deformed up to

fracture (Total elongation of the cruciform in one direction)
The stress acting on the arms of the cruciform can now be calculated from these
results, by dividing the force acting on the arms by the area of the cross-section.
58
CHAPTER 5. RESULTS
This results in a stress of = F/A = 1200/(6 103 0.7 103 = 286M P a,

which is clearly under the maximal value of 325M P a as was found as the
maximum allowed (engineering) stress for IF-steel. To determine the stresses in
the biaxially loaded centre of the cruciform numerical computations are needed
for which the material model has to be used. A problem that arises here is to
choose the right parameters for the material model, as the material behaviour
of the cruciform is not the same as the material behaviour of the sheet metal.
This originates from the fact that sheet metals are not homogenous over the
thickness of the sheet, as is assumed in simple material models.
An eect that cannot be numerically modelled and was not experimentally
studied is the inuence of size eects. As the grains in the used IF steel are
approximately 10 15m, a thickness of 150m is expected to not be sensible
to size eects, but for material with larger grains this might become an issue
that has to be dealt with.
59
CHAPTER 5. RESULTS
5.4.3 Specimen manufacturing and characterization
As was mentioned earlier for the punch test, material data as measured in a test
is not always due to deformation of the material. In case of the punch a clear
eect is introduced by friction, in the case of the cruciform an eect might be
found due to specimen preparation. In this section the eects of manufacturing
cruciform specimen with EDM will be analyzed, while more information about
EDM in general can be found in Appendix A.
It is known from literature that the EDM process alters the surface of a metal,
as vaporized material falls back on the surface en solidies. This so-called white
layer, or recast zone, exists as a hard layer on top of the original material, with
micro cracks in it due to thermal expansion.
To get a better understanding of the inuences of manufacturing the cruciform
specimen, several characterization techniques including surface prolometry,
microscopy, grain size measurement, tensile testing and nano indentation have
been used.
Surface prolometry
To get a better idea of the inuence of EDM on the surface of the material
and on the microstructure, the rst step in characterizing the EDM process
for manufacturing of cruciform specimen is to analyze both the surface and
cross-section of the specimen.
Using the Sensofar confocal microscope, height proles where made of the cru-
ciform specimen, as shown in gure 5.19. This image shows the surface texture
of the thickness reduced area to dier from the original material, as it looks
more bumpy. A second observation is asymmetry found in the specimen, mean-
ing the top and bottom sides of the cruciform are not exactly similar. This was
found for all the tested cruciform specimen.
Figure 5.19: Height prole as measured with the Sensofar confocal microscope,
showing top and bottom side of a cruciform specimen with measure-
ments on the same scale
A last observation is the existence of variation in the thickness, although this

is found to be less then 15m for the measured specimen. This is an increase
of more then 9% for a 160m thick specimen.
60
CHAPTER 5. RESULTS
Microscopy analysis
A rst hint that reveals the EDM process might alter the material behaviour, is
found by examining the surface of the thickness reduced specimen. Figure 5.20
shows how a layer of resolidied material is left behind by the process. When
comparing with the as received material, it is clear that EDM does change the
surface of the sheet, even though the measured global roughness is the same as
for the as received material.
Figure 5.20: Microscope pictures of IF and by EDM thickness reduced IF showing

resolidied material from the EDM process
Grain size measurements
In gure 5.21 the small grains on the edge and larger grains in the centre of the
original sheet are visible. This shows that the assumption of a homogeneous
material is not correct a representation of reality.
Figure 5.21: Cross-section of IF-steel sample after etching with Nittal / Marshall
Reagent
A similar study of the cruciform sample shows slightly larger grains near the
edge of the cross-section, in the same order as the grains that where originally
61
CHAPTER 5. RESULTS
originated at those positions before the outer layer was removed. Due to the
random distribution in the grain size and the limited resources to compute the
average grain size over a large area of the cross-section, no qualitative compar-
ison can be made, but comparing several smaller areas show no indication of
grain size increasing or decreasing due to the EDM process.
Figure 5.22: Microscopic pictures comparing the thickness reduced area with the
original sheet
Scanning Electron Microscopy
When analyzing the cross-section of a sheet metal with a SEM, the original
material shows a clean surface with no irregularities while the cross-section
that was subjected to the EDM process shows deposited foreign material on the
outside of the sheet, as can be seen in gure 5.23. The size of these resolidied
droplets range from 10 to 30 m in width and 3 to 6 m in height. This so-
called white layer (or recast layer) has been investigated more, to nd out the
inuence on the biaxial tensile tests.
Figure 5.23: Close-ups of the cross-section the edge of both the unaltered sheet
and the edge after EDM processing, showing an irregular edge with
resolidied material attached to the surface
62
CHAPTER 5. RESULTS
A rst observation when analyzing the cruciform specimen in SEM after frac-
ture, is the existence of micro cracks in the mentioned recast layer. In gure
5.24 an example of such a crack in the recast layer is shown, resulting in a notch
in the under lying steel. The existence of a micro crack on the surface can lead
to fracture starting at the surface instead of failure from the middle of the
sheet, as is found normally in ductile fracture. To verify fracture did not start
at the surface, the crack tip of the cruciform specimen is compared to the crack
tip of the Marciniak test specimen. The reason to use the Marciniak specimen
as a reference is that for this specimen no external inuences are present that
alter the biaxial fracture mode.
Figure 5.24: SEM image of the fractured recast layer in a biaxially tested cruciform
specimen
When comparing the two crack tips, a similar type of fracture is observed. In
gure 5.25 the crack tips for the Marciniak specimen and the cruciform speci-
men are shown, which do not show a dierence that can lead to the conclusion
that the fracture mechanisms dier. Both the Marciniak and cruciform speci-
men are found to have no visual evidence of a crack starting from the surface
of the sheet.
Figure 5.25: SEM images of the crack tip of biaxially fractured material in the
Marciniak test and the cruciform test showing many similarities
63
CHAPTER 5. RESULTS
Tensile tests
Microscopy showed possible eects of the EDM process are to be found in the
altering of the surface roughness, as the microstructure of the material is not
changed. To analyze the inuence of this change on the biaxial measurements,
the following section shows the results of mechanical testing of the material.
The results of the tensile tests (gure 5.26) show a clear inuence of the EDM
and wire EDM processes, especially for the fracture strain and hardening shows
a minor decrease as well. The huge decrease of the fracture strain can be
explained by the micro cracks that have formed. For biaxial loading these do
not show a clear inuence, as the measured maximum stresses are found to be
roughly the same for all tests. For the uniaxial tensile tests the cracks are of
more inuence though, due to easier localization at the notch that was formed.
This is the result of the equal thickness of the tensile bar.
450
400
350
True Stress [MPa]
300
250
200
150
100
Uniform Thickness Specimen
50 Wire EDM Thickness Reduced
Sink EDM Thickness Reduced
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
True Strain []
Figure 5.26: Results of the tensile tests for thickness reduced IF-steel
The dierences between die sinking EDM and wire EDM shown in the stress-
strain curves, showing an increase of the fracture strain for wire EDM. This
shows that wire EDM and die sinking EDM can not considered as resulting in
the exact same damage to the material, as was suggested by the EDM workshop.
[10]
For hardening it is more logical to increases, due to the addition of the recast
layer, but the response of the total specimen shows a decrease. This can be
explained by the heterogeneity of the sheet through the thickness, of which
only the centre part is tested due to the thickness reduction. From the nano
indentation results it can be found that the centre of the sheet has a lower
hardness, resulting in less hardening of the thickness reduced specimen.
64
CHAPTER 5. RESULTS
Nano indentation
The hardness as determined by nano indention is visually represented in gure

5.27, where the results for both an unaltered sheet (in red) and a thickness
reduced sheet (in black) are given. The results show an increase in hardness
from the centre of the material outwards, which for the unaltered material
increases approximately 20%, from a hardness of around 1.4 GPa to almost 1.7
GPa near the edges. The measurements on the thickness reduced sample show
a similar result, with the hardness increasing outwards from the centre. The
average hardness of the thickness reduced samples increases slightly faster then
the hardness of the original material, but in the centre part of the cross-section
no inuences are found at all. The measured lower hardening in the tensile
tests is thus found to originate from the fact that the part of the sheet that is
measured has a lower hardness, due to the removal of the harder outer layer.
Figure 5.27: Hardness distribution determined by nano indentation
In the gure the error bars show an uncertainty of up to 5% for most of the
data, which is caused by the inuences of the microstructure of the material,
like grain boundaries. A larger uncertainty is usually found near the edges,
mainly due to specimen preparation. In appendix C an example of increased
uncertainty is given, caused by the polishing step used.
Eects of thickness reduction on measured material behaviour
An eect of material removal is an inuence on the global behaviour of the sheet

metal, e.g. hardening, as was shown in gure 5.27. This change of measured
sheet properties originates in the heterogenous structure of a typical metal sheet
as shown in gure 5.28. For sheets containing dierent phases like dual phase
steels, this eect is even stronger as the microstructure vary greatly over the
thickness then for the tested IF-steel.
This eect introduces opportunities when determining material properties. Re-
moving a part of the sheet by thinning may be carried out in such a way that in-
stead of testing the sheet, testing a small part of the material or microstructure
in the sheet becomes possible by removing the other layers. Some possibilities
include material removal from one side only, as shown in 5.28 in the bottom
picture. This approach would test the material near the surface, instead of the
65
CHAPTER 5. RESULTS
Figure 5.28: Inuence of microstructural inhomogeneities over the thickness of a

sheet metal
material in the centre of the sheet, and might provide insight in the layered
structure of some sheet metals.
Combining material properties found through the thickness of the sheet can
then give more insight in the deformation behaviour of the sheet as a whole.
This combined material model should give the same result as the a model
determined for the sheet metal itself, as both average the materials behaviour
over the thickness.
A challenge that remains is to remove the material without inuencing the
measured material data. For the removal of the inuencing recast layer, the
use of ECM to polish or using ECM as only step instead are to be considered.
Research into optimizing the manufacturing process has been started, but has
not been nished during this project. There are strong indications though that
the recast layer can be fully removed, as was conrmed during a meeting with
Philips. [41]
With the earlier mentioned set-ups, the possibility for building a miniature
version of an existing set-up is the main problem. For cruciform specimens this
is not the case, as the designed set-up already is build at small scale.
The current set-up for 60x60mm specimen can be decreased in size by shorten-
ing the arms. From the numerical computations it can be found that the strain
eld is only showing a gradient up to 12mm from the centre of the specimen.
This means a specimen of 24mm, not including the clamping area, should be
possible. When reducing the width of the arms, an even smaller specimen can
be used. The limiting factor is only the material used, as large grains or other
microstructural properties can lead to size eects when miniaturizing into a
small length scale.
Tensile forces and displacements
For a cruciform with arms of 6mm width, the needed forces were found to be
less then 3kN . By decreasing the width of the arms this force can be lowered
even more. For building a set-up, the low forces are not expected to be a
problem.
Experimental and numerical results show the cruciform specimens have a max-
66
CHAPTER 5. RESULTS
Table 5.8: Numerical results for force and displacement for a cruciform with a centre
thickness of 160m
Property Value
Maximum Elongation [mm] 4.5 (2 x 2.25)
Maximum Force [kN] 2.95 (2 x 1.47)
imum displacement of 4 6mm. This value can be lowered by shortening the

arms, which gives more space for clamping as well.
Table 5.9: Comparing numerical and experimental results for the strain eld in a
cruciform with a centre thickness of 160m
Property Numerical Experimental

Maximum Strain in the Arms [-] 0.11 0.15-0.20
Maximum Strain in the centre [-] 0.41 0.46-0.65
When combining the results, no reason can be found that makes building a
biaxial testing device for cruciform specimen that can be used with SEM and
other microscopes.
5.4.5 Summary
The cruciform specimen is found to be an excellent method for testing under

biaxial loads. Both size and forces can be kept small for this set-up, making it
relatively easy to build as a miniature set-up.
The cruciform test is found to be experimentally usable to deform metal

sheets up to fracture under biaxial loading.
The numerical and experimental data up to necking is found to be remark-

ably similar for the used simple numerical model. This makes designing
and optimizing the cruciform geometry a relatively easy task.
Miniaturization is not an issue with the cruciform specimen, unless scaling

down into the domain where size eects become important. Forces and
displacements are small in comparison to other techniques.
The manufacturing of cruciform specimen is the biggest challenge, al-

though promising processes have been discovered and are currently being
investigated.
67
CHAPTER 5. RESULTS
5.5 Comparative evaluation of the set-ups
The four tested set-ups have all got their advantages, which will be discussed
in this section. The result of this is summarized in table 5.10.
Limitations of set-up and specimen
As is shown in literature, some set-ups provide better material data then others.
This is mainly inuenced by in-plane or out-of-plane testing, contact in the area
of the measurement and constraints that lead to force failure modes:
The bulge test is an out-of-plane test, resulting in bending inuences,

which make the measured data deviate signicantly from the actual ma-
terial data.
The punch test is an out-of-plane test, like the bulge test, making bending
also a problem. Contact in the area of interest makes the measured data
deviate even more for describing the material.
The Marciniak test is an in-plane test, with no contact in the area of

interest. The Marciniak test deforms the sheet under pure biaxial load,
without any distortions from contact or bending.
The in-plane cruciform test also has no contact or bending inuencing

the measured data, but due to the specimen geometry a forced failure
direction exists.
Specimen preparation
The specimen for a bulge test is cut out of the original plate, to t in the
clamping die. No special manufacturing steps are needed and because
of that all materials, as long as they are impenetrable for uids, can be
tested with the bulge test.
The punch test specimen is exactly the same as for the bulge test. Also
for the punch test there are no real limitations on what materials can be
tested.
For the Marciniak test a specimen like the one for the bulge and punch
test is needed. A washer sheet is needed as well, which is a copy of the
test sheet, with a hole cut out in the centre. Determining the exact shape
of washer is a challenge, as the test is sensitive to small changes in the
shape. For testing dierent materials, it is likely that the set-up needs
to be optimized again, resulting in dierent set-ups or washer geometries
for dierent materials. Strong heterogeneity over the thickness can result
in early fracture at the punch edge due to bending stresses.
68
CHAPTER 5. RESULTS
The in-plane cruciform test has the most challenging specimen design.
Manufacturing of a thickness reduced test specimen involves EDM or
ECM, which is both time consuming and can alter the material. In the-
ory any material can be tested with the cruciform loading test, but the
eect of EDM or ECM might not be equal for every material. Strongly
heterogenous sheets can prove to be dicult to characterize, as the dif-
ferent layers have to be taken into account.
Measuring opportunities
With the bulge test image correlation or microscopy can be used, but
the lens will have to move with the growing bulge. This results in a
challenging set-up, that involves constant measuring of the bulge height.
The measurements at the surface are not representative for the whole
sheet, as the stress- and strain elds are not uniform.
The punch test has the same problem with non uniform stress- and strain
elds as the bulge test. Keeping the top of the bulge at the same distance
from the lens is easier though, as the movement of the punch dictates the
movement of the sheet.
In the Marciniak test the area of interest stays in the same plane, which
makes digital image correlation or microscopy easy. The stress eld can
be computed, if the material model is known, which can be a challenge
as it is the material model that needs to be determined.
For the in-plane cruciform test a digital image correlation system or mi-
croscope can be used as well, as the area of interest stays in-plane. The
uncertainty in the exact geometry of the specimen makes determining the
stress eld more dicult, and involves numerical computing again for the
thickness reduced centre.
Strain paths and strain path changes
The bulge test can only be used for strain paths with 1 > 0 and 2 > 0.
To achieve dierent strain paths, elliptical dies with dierent aspect ratios
are needed. There is no possibility to change the strain path during a
test
The punch test can be used for strain paths with 1 0, while 2 can be
both positive and negative. This can be achieved easiest by changing the
geometry of the test specimen, for instance with a Nakazima test. Strain
path changes can not be made during a test
The Marciniak test has the same strain path range as the punch test,
1 0 while 2 can be both positive and negative. This is done with
dierent specimen and washer geometries [47]. Varying the strain path
during the test is not possible
69
CHAPTER 5. RESULTS
The in-plane cruciform test is the only test where the strain path can be
changed during a tensile test. All strain paths under tensile loading can
be achieved by changing the load on the dierent arms
Miniaturizing
Miniaturizing the bulge test results in two problems that make it unde-
sired to do so. The rst is the increased bending inuence for a small
bulge test. Even more problematic is the pressure of over 10M P a that
is needed to go up to fracture in a bulge test with 50mm diameter. This
results in uncontrollable and dangerous situations at fracture.
The punch test shows the same problems with bending as the bulge test,
but decreasing the size of this set-up results in a lower force needed to
go to fracture, but an increase in friction inuence due to the decrease
of the area the forces are working. The needed forces still go as high as
20kN and to be able to neglect bending a large set-up is preferred.
In the Marciniak test the bending problem is not found, which makes
miniaturizing easier. The critical balance between the dierent parame-
ters makes a miniature set-up dicult to optimize and to get it working.
Also high forces are needed to deform up to fracture, as forces go as high
as 60kN for a punch diameter of 50mm
To deform a cruciform specimen up to fracture, forces as low as 2.5kN

are sucient. The only fundamental limitation for miniaturizing is the
thickness reduction, that still has to be large enough to not feel size eects
70
Table 5.10: Comparison of the dierent biaxial test methods
Property Bulge Test Punch Test Marciniak Test Cruciform

In-plane No No Yes Yes
Contact in region No Yes No No
of interest
Geometrical con- No Yes, friction dictates the No, fracture is free to Yes, direction of frac-
straints in region fracture starting point start anywhere for the ture is dictated
of interest biaxial case
Specimen prepa- Unaltered sheet Unaltered sheet Unaltered sheet and a Thinned sheet, preci-
CHAPTER 5. RESULTS
ration simple washer geometry sion job altering speci-

men geometry
Measuring op- Straineld, pressure Straineld, punch force Straineld, punch force Strain eld, tensile load
tions in arms
Strain elds Elliptical dies, only pos- Punch shapes, specimen Specimen and washer Load distribution of the
71
itive 1 and 2 shapes, friction, positive shapes, , positive 1 , machine, positive and
1 , positive and nega- positive and negative 2 negative 1 and 2
tive 2
Strain path No No No Yes
change
Heterogeneous Additional inuence be- Additional inuence be- No additional inuence, Inuence because of
materials cause of bending cause of bending besides averaging ma- thinning
terial properties over
thickness
Miniaturizing Bending Bending Bending near etches In the range of grainsize
problems for thickness
Expected forces 10M P a, corresponding 20kN 50kN 2.5kN
to 20kN for a 50mm di-
ameter test
Chapter 6
Conclusions and
recommendations
6.1 Conclusions
From the result it can be concluded that the best options for a miniature set-up
to study biaxial deformation and strain path changes are in-plane loading using
a cruciform geometry (ILCG) and the Marciniak test. Which of these is better
is dictated by the results that are wanted, as each has its own advantages and
disadvantages.
For the cruciform set-up, the following can be concluded:
The cruciform test set-up is best a better option to be used under a

microscope, as it can be miniaturized relatively easily. Necessary forces
and the specimen size are both small enough to make a miniaturized set-
up possible. Forces as low as 2kN are found to deform IF steel up to
fracture.
A disadvantage of the cruciform set-up is the challenge that lies in manu-

facturing test specimen. Manufacturing induced eects can lead to data
not representative of the tested sheet, due to either damage introduced
in the tested material or altered material properties over the thickness
of the specimen. To obtain reliable data these eects need to be kept as
small as possible.
A second disadvantage of the thickness reduction is the need to deter-

mine the needed thickness reduction to reach fracture. The ideal thick-
ness therefore depends on material behaviour and therefore the optimal
cruciform geometry changes with every material.
A concurring advantage associated with the need for precise specimen

preparation in order to achieve thickness reduction in the cruciform spec-
imen, is the possibility to test dierent layers in a sheet metal. As most
72
CHAPTER 6. CONCLUSIONS AND RECOMMENDATIONS
sheet metals are heterogeneous over their thickness, this can prove to be
a welcome addition to standard sheet tests.
A huge advantage of the cruciform test set-up, that is not found in any
other test, is the possibility to change strain paths during a test. Com-
plex strain paths can be described relatively easily, by prescribing the
displacements of the clamps. The other set-ups that have been analyzed
have no possibilities for doing this.
The Marciniak test is an interesting alternative, as it undistorted biaxial

loading results due to the nature of the test and no distortion of the
measured data by specimen preparation is found. A miniaturized test
set-up needs punch forces as high as 50kN for IF-steel, which might be
challenging when designing a small set-up.
When miniaturizing the Marciniak set-up many design parameters are
inuencing a critical balance between the biaxial fracture mode and other
failure modes. This results in a complex optimization problem, which
cannot easily be solved. It is likely that a miniaturized set-up can be
optimized to work, but no proof can be given at this moment. Also,
optimizations are needed for dierent materials so a set-up can be used
to test more then one material.
If there is need for changing strain paths, the Marciniak test cannot be
used. The only strain path changes that can be carried out involve using
pre-strained material to cut test specimen from. This gives severe limi-
tations to the variety of strain paths that can be assessed, and leads to
a more lengthy test procedure. Also when testing under dierent strain
paths dierent washer and specimen shapes are needed. This involves
more optimizations, just as the original test set-up.
6.2 Recommendations for future work
The challenges that need to be solved to develop an ideal set-up that is useable
under all circumstances are the following:
The in-plane cruciform specimen has not been optimized yet, only the
earlier found geometry of Vos has been used and reduced in thickness to
achieve fracture. To make the cruciform test smaller, an optimization of
the cruciform size, the thickness reduction and geometry of the thickness
reduction is recommended.
A study into EDM, ECM or another advanced method for material re-
moval is recommended, in order to make the cruciform specimen with
a less distorted microstructure. A method useable in a laboratory en-
vironment is desirable, to make the specimen design cycle shorter. A
study into optimizing the EDM process or using ECM has been started,
in cooperation with Philips DAP [41], but results are pending.
73
CHAPTER 6. CONCLUSIONS AND RECOMMENDATIONS
To understand the Marciniak test, studying of the local necking be-

haviour is necessary, so numerical results can be improved to describe
the Marciniak test with the localizations that were found. The improved
numerical model can then be used to analyze miniaturization further.
74
Bibliography
[1] Amorim, F. and Weingaertner, W. (2005). The inuence of generator ac-

tuation mode and process parameters on the performance of nish EDM of
a tool steel. J. Mat. Proc. Techn., 166:411416.
[2] vila, A. and E.L.S., V. (2003). Proposing a better forming limit diagram
prediction: a comparative study. Mat. Proc. Techn., 141:101108.
[3] Banadic, D. (2000). Formability of Metallic Materials, chapter 5. Springer-

Verlag.
[4] Bhattacharyya, B., Mitra, S., and Boro, A. (2002). Electrochemical ma-
chining: new possibilities for micromachining. Robot. Comp. Integr. Man.,
18:283289.
[5] Bhattacharyya, B., Munda, J., and Malapati, M. (2004). Advancement

in electrochemical micro-machining. Int. J. Machine Tools & Manufacture,
44:15771589.
[6] Bleys, P., Kruth, J., Lauwers, B., Schacht, B., Balasubramanian, V.,
Froyen, L., and Van Humbeeck, J. (2006). Surface and sub-surface qual-
ity of steel after EDM. Adv. Eng. Mat., 8:1525.
[7] Boehler, J., Demmerle, S., and Koss, S. (1994). A new direct biaxial testing
machine for anisotropic materials. Exp. Mech., 34:19.
[8] Butuc, M., Gracio, J., and Barata da Rocha, A. (2003). A theoretical study
on forming limit diagrams prediction. J. Mat. Proc. Techn., 142:714724.
[9] Campos, H., Butuc, M., Gracio J.J.AND Rocha, J., and Duarte, J. (2006).
Theorical and experimental determination of the forming limit diagram for
the AISI 304 stainless steel. J. Mat. Proc. Techn., 179:5660.
[10] de Laat, H. (2007). Personal communication.
[11] Demmerle, S. and Boehler, J. (1993). Optimal design of biaxial tensile

cruciform specimens. J. Mech. Phys. Solids, 41:143181.
[12] DiBitonto, D., Eubank, P., Patel, M., and Barrufet, M. (1989). Theoretical
models of the electrical discharge machining process. I. A simple cathode
erosion model. J. Appl. Phys., 66:40954103.
[13] Fenner, R. (1999). Mechanics of solids. CRC Press LLC.
75
BIBLIOGRAPHY
[14] Foecke, T., Banovic, S., and Fields, R. (2003). Sheet metal formability
studies at the national institute of standards and technology. JOM, 53:27
30.
[15] Foecke, T., Iadicola, M., Lin, A., and Banovic, S. (2007). A method for
direct measurement of multiaxial stress-strain curves in sheet metal. Met.
and Mat. Transactions, 38A:306313.
[16] Gan, W., Babu, S., Kapustka, N., and Wagoner, R. (2006). Microstruc-
tural eects on the springback of advanced high-strength steel. Metal. and
Mat. Transactions, 37A:32213231.
[17] Garenfeld, S. (2007). Personal communication.
[18] Giancoli, D. (2000). Physics for scientists and engineers, chapter 5, page
107. Prentice-Hall, third edition.
[19] Gozzi, J., Olsson, A., and Lagerqvist, O. (2005). Experimental investiga-
tion of the behavior of extra high strength steel. Soc. Exp. Mech., 45:533540.
[20] Gronostajski, J. and Dolny, A. (1980). Determination of forming limit

curves by means of marciniak punch. Mem. Sci. Rev. Metall., 77(4):570
578.
[21] Gutscher, G., Wu, H.-C., Ngaile, G., and Altan, T. (2004). Determination
of ow stress for sheet metal forming using the viscous pressure bulge (VPB)
test. J. Mat. Proc. Techn., 146:17.
[22] Guu, Y. (2005). AFM surface imaging of AISI D2 tool steel machined by
the EDM process. Appl. Surf. Sc., 242:245250.
[23] Guu, Y., Hocheng, H., Chou, C., and Deng, C. (2003). Eect of electrical
discharge machining on surface characeristics and machining damage of AISI
D2 tool steel. Mat. Sc. and Eng., 358:3743.
[24] Hannon, A. and Tiernan, P. (2008). A review of planar biaxial tensile test
systems for sheet metal. J. Mat. Proc. Techn., 198:113.
[25] Ho, K. and Newman, S. (2003). State of the art electrical discharge ma-
chining (EDM). J. Mach. Tools & Manufacture, 43:12871300.
[26] Hoferlin, E., van Bael, A., van Houtte, P., Steyaert, G., and de Mar, C.
(1998). Biaxial tests on cruciform specimens for the validation of crystallo-
graphic yield loci. Mat. Proc. Techn., 80-1:545550.
[27] Kelly, D. (1976). Problems in creep testing under biaxial stress systems.
J. Strain Anal., 1:11.
[28] Keskin, Y., Halkac, H., and Kizil, M. (2006). An experimental study for
determination of the eects of machining parameters on surface roughness in
electrical discharge machining (EDM). Int J. Adv. Manuf. Techn., 28:1118
1121.
76
BIBLIOGRAPHY
[29] Kiyak, M. and akr, O. (2007). Examination of machining parameters on

surface roughness in EDM of tool steel. J. Mat. Proc. Techn., 191:141144.
[30] Kuwabara, T. (2007). Advances in experiments on metal sheets and tubes

in support of constitutive modeling and forming simulations. Plasticity,
23:385419.
[31] Kuwabara, T., Ikeda, S., and Kuroda, K. (1998). Mesurement and analysis
of dierential work hardening in cold-rolled steel sheet under biaxial tension.
Mat. Proc. Techn., 80-1:517523.
[32] Kuwabara, T., Kuroda, K., Tvergaar, V., and Nomura, K. (2000). Use of
abrupt strain path change for determining subsequent yield surface: experi-
mental study with metal sheets. Acta Materalia, pages 20712079.
[33] Landolt, D., Chauvy, F.-F., and Zinger, O. (2003). Electrochemical micro-
machining, polishing and surface structuring of metals: fundamental aspects
and new developments. Electr. Acta, 48:31853201.
[34] Lee, S. and Li, X. (2001). Study of the eect of machining parameters on
the machining characteristics in electrical machining of tungsten carbide. J.
Mat. Proc. Techn., 115:344358.
[35] Lee, S. and Li, X. (2003). Study of the surface integrity of the machined
workpiece in the EDM of tungsten carbide. J. Mat. Proc. Techn., 139:315
321.
[36] Lee, Y.-W., Woertz, J., and Wierzbicki, T. (2004). Fracture prediction
of thin plates under hemi-spherical punch with calibration and experimental
verication. J. Mech. Sciences, 48:751781.
[37] Liu, J., Ahmetoglu, M., and Altan, T. (2000). Evaluation of sheet metal
formability, viscous pressure forming (VPF) dome test. J. Mat. Proc. Techn.,
98:16.
[38] Marciniak, Z., Duncan, J., and Hu, S. (2002). Mechanics of Sheet Metal
Forming. Butterworth Heinemann.
[39] Marciniak, Z. and Kuczyski, K. (1967). Limit strains in the process of

stretch-forming sheet metal. Int. J. Mech. Sci., 9:609620.
[40] Meeting, C. R. (2008a). Personal communication.
[41] Meeting, P. D. (2008b). Personal communication.
[42] Mller, W. and Phlandt, K. (1996). New experiments for determining

yield loci of sheet metal. Mat. Proc. Techn., 60:643648.
[43] Muzyka, N. (2002). Equipment for testing sheet structural materials under
biaxial loading. part 2. testing by biaxial loading in the plane of the sheet.
Strenght of Materials, 34:206212.
77
BIBLIOGRAPHY
[44] on Automotive Applications, C. (2006). Advanced high strength steel

(AHSS) application guidelines. Technical report, International Iron and Steel
Instiute.
[45] Ozturk, F. and Lee, D. (2005). Experimental and numerical analysis of

out-of-plane formability test. J. Mat. Proc. Techn., 170:247253.
[46] Parmar, A. and Mellor, P. (1979). Growth of voids in biaxial stress elds.
Int. J. Mech. Sci., 22:133150.
[47] Raghavan, K. (1995). A simple technique to generate in-plane forming limit

curves and selected applications. Met. and Mat. Transactions, 26A:2075
2084.
[48] Ramarao, P. and Faruqi, M. (1982). Characteristics of the surfaces ob-

tained in electro-discharge machining. Prec. Eng., 2:111113.
[49] Ranta-Eskola, A. (1979). Use of the hydraulic bulge test in biaxial tensile
testing. Int. J. Mech. Sci., 21:457465.
[50] Rees, D. (1995). Plastic ow in the elliptical bulge test. Int. J. Mech. Sci.,
37:373389.
[51] Slota, J., Spiak, E., and Stachowicz, F. (2004). Investigation of biaxial
stress-strain relationship of steel sheet metal. Int. J. Appl. Mech. and Eng.,
9:161168.
[52] Smits, A., van Hemelrijck, D., Philippidis, T., and Cardon, A. (2006). De-
sign of a cruciform specimen for biaxial testing of bre reinforced composite
laminates. Composite Sci. and Techn., 66:964975.
[53] Tadros, A. and Mellor, P. (1977). An experimental study of the in-plane

stretching of sheet metal. Int. J. Mech. Sci., 20:121134.
[54] Vos, R. (2007). A computational study of biaxial sheet metal testing:

eect of dierent cruciform shapes on strain localization. Technical Report
MT 07.03, Techn. Univerisity of Eindhoven.
[55] Wang, C., Yan, B., Chow, H., and Suzuki, Y. (1997). Cutting austempered
ductile iron using an EDM sinker. J. Mat. Proc. Techn., 88:8389.
[56] Wierzbicki, T. (2006). Fracture of advanced high strenght steels (AHSS).

Technical report, Massachusetts Institute of Technology.
[57] Wu, X., Wan, M., and Zhou, X. (2005). Biaxial tensile testing of cruciform
specimen for limit strain analysis by FEM. Mat. Proc. Techn., 168:181183.
[58] Yu, Y., Wan, M., Wu, X., and Zhou, X. (2002). Design of a cruciform
biaxial tensile specimen for limit strain analysis by FEM. Mat. Proc. Techn.,
123:6770.
78
APPENDIX A. ELECTRICAL DISCHARGE MACHINING
Appendix A
Electrical Discharge Machining
Theory
The EDM process is basically a cathode-anode set-up of two electrodes, sepa-

rated by a liquid dielectric, with a tool and workpiece. During the machining
process, a voltage is applied over the gap between anode and cathode, causing
the dielectric to break down. During the 'on-time' of the electrode a plasma
channel grows, surrounded by a vapour bubble. The dense liquid dielectric
restricts the growth of the plasma channel width, which results in the energy
being concentrated in a small volume. Local temperatures can reach as high
as 20 000 to 40 000 K, with plasma pressures as high as 3 kbar. The shape of
this plasma channel as seen in gure A.1.
Anode Melt
Anode (+) Cavity
Cathode
Compressed
Liquid
Erosion rate
Shockwave
Plasma Front Anode
Ambient
Liquid
Dielectric
Cathode (-) Cathode Melt

Cavity On-time
Figure A.1: Schematic diagram of the EDM process showing the circle heat sources,
plasma conguration and melt cavities (left);
The dierence between anode and cathode material removal rates in
respect to 'on-times' (right) [12]
High energy levels in the plasma cause both electrodes to melt, but due to the
high plasma pressures, vaporization is limited. When the pulse is turned on,
the anode will be the rst to rapidly melt because of fast moving electrons, and
melting of the cathode starts later during the process, due the lower mobility of
the positive ions. This results in dierent material removal rates for the anode
79
and cathode side, as is shown in gure A.1.

For die sinking machines the cathode is usually the workpiece, resulting in
relatively large 'on-times' in the order of 10-100 s to make sure material is
removed from the workpiece, instead of the tool. For most wire machines, the
cathode is the wire, cutting into the anodic workpiece. The pulse time for wire
machines therefore is generally much shorter, in the order of less then 10 s.
At the end of the 'on-time', the current is terminated and a violent collapse
of the plasma channel follows. This causes the superheated, molten liquid on
the surface of the electrodes to explode into the dielectric uid. Part of the
material is then carried away, the remainder resolidies on the surface. [12, 25]
White layer and Heat Aected Zone (HAZ)
Bleys et al [6] saw several dierent layers in the thermally inuenced zone
in EDM. They report formation and thickness of these layers depends on the
process conditions and work piece properties.
The 'white layer', a molten and resolidied layer, also known as the recast layer
is a result of EDM. In this layer micro cracks are found, that seldom go deeper
then the layer itself. Also present are micro-holes and droplets of resolidied
material. Below the recast layer, the heat aected zone (HAZ) is found. The
material in this zone has not been melted, but did undergo thermal inuence.
Several layers are present, although they are not easily distinguished. It is
shown that it is possible to reduce the recast layer to a thickness in the order
of micrometres, by nishing in several steps with decreasing pulse current.
Hardness test on an EDM machined surface show an increase in hardness in
the HAZ, because of either diused carbon or a ner grain structure due to
rapid cooling after the EDM process. The study shows residual stresses up to
500 M P a, present from several m to almost 10 m into the material.
EDM parameters and their inuence
Four parameters determine the material removal rate and thickness of the heat
aected zone and recast layer. These parameters are pulsed current, pulsed
voltage, pulse 'on-time' and pulse 'o-time'. The rst three determine the en-
ergy put into the process, the latter determines the cooling time in between
pulses. According to dierent sources, pulsed current and pulse 'on-time' in-
uence the surface roughness, recast layer thickness and induced stresses most
[22, 23, 48].
Varying pulse 'on-time' and pulse current has shown similar results for several
researches. Kiyak and akr [29] show a relation for pulse time and current
on AISI P20 tool steel, where pulse time clearly has more inuence then pulse
current. They conclude a better nish is obtained for low currents and short
pulse 'on-time', in combination with a relatively high pulse 'o-time'. Guu et al
80
[23] give an empirical model that describes how the surface roughness changes
in relation to pulse current (Ip ) and pulse 'on-time' (on ) Ra = A (Ip )a ( )b
that can be tted to experimental data.
Lee et al [34] investigated the inuence of pulse 'on-times' of 25 s up to
600 s, showing surface roughness, tool wear and material removal rate is
almost constant for pulse 'on-times' higher then 50 s. Keskin et al [28] show
a similar result, but for pulse 'on-times' of 100 s and higher, with dierent
power settings and materials.
Lee and Li [35] studied the eect of pulse current and pulse 'on-time' on the
thickness and composition of the recast layer. They conclude the recast layer
grows with higher pulse current or pulse 'on-time'. Denser materials where
found up to 15 s into the material for pulse 'on-times' of 12.5 s. The compo-
sition of the top layer was studied with energy dispersive X-ray method (EDX),
showing higher levels of carbon in the layer. The study of Guu et al [23] gives
a similar result, and gives an empirical model for the recast layer thickness, t
t = A (Ip )a (on )b . The experimentally found thickness varies from 7 to 31
m, for relatively long pulse 'on-times' of 20 to 180 s.
A study of Guu et al [23] shows correlation between tensile strength and machin-
ing parameters, where a lower tensile strength was found for EDM machined
materials. In a later article, Guu [22] shows the depth of micro-cracks grows
with increasing pulse current and pulse 'on-time' as well, measuring 1272 to
1873 nm in depth for currents up to 1.5 A and 'on-times' up to 6.4 s. Re-
moving the recast layer is named as a possible solution to remove this eect.
Figure A.2: Surface roughness for positive (left) and negative (right) polarity on tool
[1]
A study by Amorim and Weingaertner [1] shows clear dierences for using
positive or negative electrodes as tooling. The large dierence in removal rates
is due to the plasma channel shape as explained earlier. An optimal 'on-time'
of around 50 s is found for a positively charged tool, while an 'on-time' of
around 8-12 s is found optimal for a negatively loaded tool. This eect was
also predicted by [12]. Amorim and Weingaertner also studied the inuence of
polarity on the surface roughness, as shown in gure A.2. This gure clearly
shows how a negatively charged tool (tool as cathode) gives a smoother nish,
81
at the cost of a much lower material removal rate and greater tool wear. A
study by Bleys et al [6] into surface roughness shows very low roughness values
can only be obtained by using low currents and inverse polarity, so-called 'EDM
polishing'. A surface roughness Ra of 0.09 to 0.26 m is shown to be possible
by EDM polishing.
Tool wear
A side eect that should not be forgotten, is tool wear, as wearing of the
tool cannot be prevented. Wang et al [55] show increased tool wear for high
and short pulse currents. For extremely long pulse times, the dierence in
wear becomes less and eventually virtually negligible. As for a ne surface
roughness it was found that the pulse current should be low, this will not cause
any problems for EDM where a ne surface roughness is required.
82
APPENDIX B. ELECTRICAL CHEMICAL MACHINING (ECM)
Appendix B
Electrical Chemical Machining

(ECM)
Theory
The concept of using chemical solutions for material dissipation has already
been patented in 1929, but was not improved much till the late 50s and 60s,
when aircraft industry started to using ECM. In the last decade the microma-
chining step has evolved, allowing surfaces to be machined with micrometre
resolution and polished with the same machine. This specialized use in the
form of Electrochemical Micro Machining (EMM) show great potential as no
tool wear exists and high removal rates are possible. Alternatives like laser
cutting, EDM and other non conventional machining tools are mostly thermal
oriented, therefore making high precision more dicult due to heating of the
workpiece. [4, 5]
The ECM EMM process physical background lies in anodic dissolution of
workpiece material, where the workpiece and tool act as anode and cathode,
separated by an electrolyte. By applying an electric current, the anode work-
piece dissolves locally, so shaping it to become the mirrored image of the tool.
The electrolyte, often a salt solution, is then used to bind the free ions from
the dissolved material and removes it by owing through the machining gap.
[4]
The standard ECM process makes precise machining of vertical walls nearly
impossible, but with a clever tooling geometry as shown in gure B.1 it can
be done. A more complex dual pole set-up can be used to achieve even better
results. [33]
The ECM EMM process is still under development and likely to be more
enhanced over the coming years, but is being used in several commercial man-
ufacturing processes. Manufacturing steps like cutting slots in sheet metal,
machining MEMS components and shaping of surgical equipment can all be
done with ECM. [4, 33]
83
Figure B.1: Comparison of dierent anode-cathode set-ups for EMM. a) uninsulated,

b) insulated, and c) dual pole tools [33]
ECM parameters and their inuences
The most important parameters in the ECM process are again pulse 'on-
time', pulse 'o-time', pulsed current and pulsed voltage, like for the EDM
process. The combination of these parameters determines the width of the
inter-electrode gap, by resulting in equilibrium. As the material is not evapo-
rated during the ECM process, heating up of the workpiece can be controlled
a lot better, which results in a cleaner surface. [5]
For high precision material removal low currents are preferred, so the removed
material contaminating the die-electric can ow away before inuencing the
local material removal rates. The inter-electrode gap can be as small as a
few micrometre, but this limits the maximum material removal rate. Other
properties inuencing the accuracy of the ECM process are the choice of the
electrolyte and the tool, where the latter should be thermally and electrically
conductive, corrosion resistant and sti enough to withstand the electrolyte
pressure without vibrating. [5]
The electrolyte used is a last process parameter that can be used to enhance
the ECM process, as dierent materials may need dierent electrolytes. A
dierent electrolyte can also be used to favour certain chemical reactions and
thereby change the material removal process. Electrolytes can be divided in
84
two main types: passivating electrolytes, that contain oxidizing anions, like
sodium nitrate or sodium chlorate, and non-passivating electrolytes containing
more aggressive anions such as sodium chloride. The rst are used for better
machining precision, the latter for higher removal rates. [33]
Surface roughness
Surface nishing by ECM is better then for most other processes, but as gure
B.2 shows this does depend on the length scale being considered. The reason for
the worsening for larger length scales is the development of wave-like patterns
on the surface because of hydrodynamic vortex phenomena (dierences in ow
speed). Choosing the right ECM tooling helps to reduces this surface roughness
problem. [5, 33]
Figure B.2: Comparison of surface roughness for conventional polishing and ECM
[33]
Tool wear
The problem of EDM where the electrode shape changes is not found in ECM,
as only hydrogen gas is evolving at the cathode, thus no material is removed
from the electrode during the process.
85
Appendix C
Specimen Preparation: TegraPol

or Target System
An important aspect to take into account when interpreting data, is the amount
of scatter found. For data sets where the scatter is large, there might be a
problem with the acquisition of the data. This might be due to a problem with
the set-up or a problem with the used specimen.
Figure C.1: Maximum load in an indentor test over the cross-section of a metal sheet
on a specimen prepared with the TegraPol system
While using the nano indentor to determine the hardness gradient of the IF
steel sheet, a large scatter was found on specimen prepared with the TegraPol
polishing system as can be seen in gure C.1. It was found that this was related
to the atness of the specimen.
86
APPENDIX C. SPECIMEN PREPARATION: TEGRAPOL OR TARGET
SYSTEM
Figure C.2: Maximum load in an indentor test over the cross-section of a metal sheet
on a specimen prepared with the Target system, showing lower scatter
then for the TegraPol prepared specimen
The data obtained from specimen prepared with the Struers Target system,
where the surface was found to be better then for conventional polishing, showed
a lot less scatter. Even though one side of the specimen still showed rounding
of the edges, as can be seen in gure C.2 the scatter for the same measurement
was decreasing.
87

Biaxial Testing of Sheet Metal: An Experimental-Numerical Analysis

Uploaded by

Copyright:

Available Formats

Biaxial Testing of Sheet Metal: An Experimental-Numerical Analysis

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Biaxial Testing of Sheet Metal: An Experimental-Numerical Analysis

Uploaded by

Copyright:

Available Formats

Biaxial Testing of Sheet Metal:

TU/e Master Thesis

Engineering thesis committee

Eindhoven University of Technology

Met de recente populariteit van 'advanced high strength' staalsoorten (dual

6 Conclusions and recommendations 72

In the automobile industry a rst attempt to achieve weight reduction was

1.1 Formability and strain path dependency

The existence of a homogeneous stress- and strain distribution in the

dividing the force by the area it is acting on.

Figure 2.1: Uniaxial and biaxial stress states [13]

Some experimental set-ups test a material in-plane, others out-of-plane, de-

Measuring stresses, strains or forces

2.2 Bulge test

In the above relations the dome radius Rd is dened as

((dc /2) + Rc )2 + h2d 2Rc hd

and the thickness at the top of the dome as

Advantages and Disadvantages

The most important properties described extensively in literature including

2.3 Punch test

A second, somewhat similar approach, is the use of a punch to deform sheet

Figure 2.3: Punch shapes as used in the Keeler test [3]

Another observation is sheet thickness inuencing the results in all set-ups,

Overestimation of fracture strains

A second disadvantage of the punch test is the overestimation of acceptable

2.4 Marciniak test

specimen and washer geometries. With this combination, compared to earlier

Tooling geometry constraints

2.5 In-plane loading with cruciform geometry

A lot of research is focusing on cruciform shaped specimen to overcome the

Figure 2.10: Optimal design of a biaxial specimen by Demmerle [11]

An alternative to weakening the centre part of the cruciform is strengthening

A second approach found in literature is the use of a mechanism, xed in a

3.1 Material model

Figure 3.1: Plasticity curve as used in the MSC.Marc material model

Necking, damage and localization

When interpreting the numerical results, it is important to understand how

3.2 Bulge test

Table 3.2: Properties of the bulge test model

3.3 Punch test

friction is believed to inuence the measured material parameters. The model

Figure 3.3: FE Model of the punch test (MSC.MARC Mentat)

Table 3.3: Properties of the computational punch model

3.4 Marciniak test

Figure 3.4: FE Model of the Marciniak test (MSC.MARC Mentat)

The simulations to investigate the inuence of changing parameters are per-

3.5 In-plane loading with cruciform geometry

The cruciform model is computed with 3D elements to describe the thickness

Table 3.5: Dimensions and properties of the cruciform model [54]

4.1 Marciniak test

To compare the numerical results with experimental results, IF steel specimens

4.1.1 In-plane testing with the Marciniak set-up

4.1.2 Specimen manufacturing

4.2 In-plane loading with cruciform geometry

The cruciform geometry will be experimentally tested, to verify the numerical

4.2.1 Design of a test set-up

A set-up to test the cruciform specimen at the Eindhoven University of Technol-

Figure 4.2: Experimental set-up for testing cruciform specimen

In the automobile industry a rst attempt to achieve weight reduction was

In the above relations the dome radius Rd is dened as

Another observation is sheet thickness inuencing the results in all set-ups,

A second approach found in literature is the use of a mechanism, xed in a

friction is believed to inuence the measured material parameters. The model

The simulations to investigate the inuence of changing parameters are per-

the analytical solution, as shown in gure 5.2. The analytically determined

Friction coecient Necking Distance Sheet Thickness

The radius of the punch edge is important when miniaturizing, as it inuences

Figure 5.14: Inuence of miniaturizing the Marciniak set-up; Rwasherhole = 11mm;

Building a Marciniak test of dimensions small enough to t under a mi-

Table 5.7: Dimensions of the cruciform specimen as shown in gure 4.4

The hardness as determined by nano indention is visually represented in gure

Eects of thickness reduction on measured material behaviour

An eect of material removal is an inuence on the global behaviour of the sheet

Figure 5.28: Inuence of microstructural inhomogeneities over the thickness of a