Journal of Materials Processing Technology 230 (2016) 131142

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

Analytical approach for magnetic pulse welding of sheet connections

Marlon Hahn , Christian Weddeling, Joern Lueg-Althoff, A. Erman Tekkaya
Institute of Forming Technology and Lightweight Construction (IUL), TU Dortmund University, Baroper Str. 303, 44227 Dortmund, Germany

a r t i c l e

i n f o

Article history:
Received 18 July 2015
Received in revised form
20 November 2015
Accepted 21 November 2015
Available online 2 December 2015
Keywords:
Magnetic pulse welding (MPW)
Lightweight structures
Analytical model
Impact velocity

a b s t r a c t
An analytical model to calculate the acting forming pressure in magnetic pulse welding by determining
the magnetic eld strength between the yer sheet and a one-turn coil was presented. By neglecting
plastic deformation of the yer, the model allows to calculate the transient velocity and displacement
behavior, too. The electromagnetic acceleration of 5000-series aluminum alloy sheets was investigated
under various experimental parameters. Utilizing Photon Doppler Velocimetry revealed that the analytical model appropriately describes the inuence of current amplitude, coil geometry, and, especially,
discharge frequency on the velocity-displacement curve of the yer and hence on the impact velocity.
The model introduced was applied to compute the impact velocity for the welding of long lap joints of
5000-series aluminum alloy sheets and 6000-series aluminum alloy hollow proles. Through peel tests
it was shown that the weld strength at least complied with the strength of the weaker base material as
failure always happened in the yer sheet. The wavy interface pattern typical for impact welding was
identied with the help of metallography.
2015 Elsevier B.V. All rights reserved.

1. Introduction
There is a rising demand for lightweight structures in transportrelated applications with the aim of reducing energy consumption
to minimize costs as well as environmental pollution so that more
and more light metals are applied in the automotive industry. As a
consequence thereof, manufacturers face the challenge of joining
different grades of aluminum alloys. If welding is the joining process of choice, conventional fusion-based techniques often reach
their limits due to the occurrence of microstructural and mechanical changes in the weld bead and heat affected zone (HAZ) reducing
the strength of the joint and frequently causing hot cracks especially in welds between 5000- and 6000-series aluminum alloys
(Praveen and Yarlagadda, 2005). These problems may be avoided by
utilizing high velocity impact welding processes such as magnetic
pulse welding (MPW). It is a solid-state welding process, which
also allows to minimize or even eliminate the formation of continuous intermetallic phases when joining dissimilar metals (Zhang
et al., 2011). MPW is therefore well suited for creating strong metallurgical bonds between both similar and dissimilar metals and its
alloys.
The general working principle of impact welding is illustrated in
Fig. 1. Besides MPW, further impact welding processes are (Zhang

Corresponding author.
E-mail address: marlon.hahn@iul.tu-dortmund.de (M. Hahn).
http://dx.doi.org/10.1016/j.jmatprotec.2015.11.021
0924-0136/ 2015 Elsevier B.V. All rights reserved.

et al., 2011): explosive welding (EXW), laser impact welding (LIW),

and the lately by Vivek et al. (2013) introduced vaporizing foil actuator welding (VFAW).
As outlined by Mori et al. (2013), the two joining partners, commonly named yer and target, collide under the angle at velocities
vim in the range of several hundred m/s producing impact pressures
of the order of GPa. This process is accompanied by the so-called
jetting effect that leaves behind chemically pure surfaces allowing
a metallic bond to be formed. The atoms of the involved materials are impacted to such an extent that they share and exchange
valence electrons. As a result, a wavy interface morphology is often
observable (see Fig. 1). A common explanation for the evolution
of these waves was given by Ben-Artzy et al. (2010). The authors
stated that reected shock waves in the joining partners lead to a
KelvinHelmholtz instability. For a given material combination, the
domain of the two crucial parameters (impact angle) and vc (collision velocity) necessary for a successful weld may be plotted in the
form of a welding window, which originates from EXW (Mousavi
and Sartangi, 2009). In contrast to EXW though, both and vc do
not remain constant during MPW (Verstraete et al., 2011). A compilation of welding windows as well as different bonding criteria
available in literature so far was presented by Kapil and Sharma
(2015). By means of X-ray diffraction analysis and scanning electron
microscopy, Kore et al. (2009) found that neither melted zones nor
intermetallic phases may be present in magnetic pulse welds, while

132

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

Nomenclature
Symbol/meaning/unit
a
Length of the pressure lead of the tool coil in mm

B
Magnetic ux density (vector) in G
Bg
Magnetic ux density in the gap between the yer
and the tool coil in G
C
Capacitance of the pulse generator in F
c1 , c2 , c3 Constants in the analytical model
Flyer displacement in mm
D
d1 , d2
Distances from a two-sided tool coil in mm
Dch
Critical yer displacement in the analytical model
in mm
Initial charging energy in J
E0
EL
Total magnetic energy in J
f
Frequency of the discharge circuit in Hz
FL
Lorentz force (vector) in N/mm3
f0 , fd , fb Initial (0), Doppler-shifted (d), and beat (b) frequency of the Photon Doppler Velocimeter in Hz
FPeel
Test force during peel test (index max for the maximum) in N
FUTS
Ultimate tensile strength for a specic specimen
geometry in N
h
Height of the tool coil in mm

Magnetic eld strength (vector) in A/mm
H
h
Effective height of the trapezoidal coil in mm
Hg
Magnetic eld strength in the gap between the yer
and the tool coil in A/mm
Hh
Magnetic eld strength at the sidewall of the tool
coil in A/mm
Hh0 , Hy0 Coefcient functions in the analytical model in
A/mm
Magnetic eld strength due to the skin effect in
HS
A/mm
Coil current (index a for amplitude or peak value) in
I
A
Ih
Current at the sidewall of the tool coil in A
Ip
Current due to the proximity effect in A
Current due to the skin effect in A
IS
j
Imaginary unit
J
Current density (vector) in A/mm2
k
Complex propagation constant (indices F and T for
yer and tool coil, respectively) in 1/mm
l
Length in mm
L
Total inductance of the discharge circuit in H
Li
Inner inductance of the pulse generator in H
Magnetic pressure (index hf for the high-frequency
p
limit) in MPa
Plastic collapse pressure in MPa
pc
R
Total resistance of the discharge circuit in 
Inner resistance of the pulse generator in 
Ri
s
Sheet thickness in mm
t
Time in s
Current rise time in s
trise
v
Flyer velocity (index m for measured velocities) in
mm/s
Collision velocity in mm/s
vc
vim
Impact velocity in mm/s
w
Width of the tool coil in mm
w
Width of the bottom of the trapezoidal coil in mm
Impact angle in

Skin depth in mm

Electrical conductivity in 1/
0
Operating wavelength of the Photon Doppler
Velocimeter in mm


b
Y

Magnetic permeability (index 0 for air) in Vs/Am

Density of the yer material in kg/mm3
Flow stress of the yer material in MPa

Goebel et al. (2010) similarly showed that these phenomena cannot

be completely avoided for some materials. In MPW the electromagnetic forming (EMF) technology is used to plastically accelerate the
yer plate. Jablonski and Winkler (1978) stated that the forming
pressure in EMF is generated by penetration of a pulsed magnetic
eld into a conductive workpiece to be formed. The magnetic eld
in turn results from a rapid discharge of a capacitor through the
tool coil (see Fig. 2a). Materials of low electrical conductivity can be
formed with the help of thin high-conductivity driver plates (Gies
et al., 2014). Such drivers are positioned between the workpiece
and the coil to provide the forming pressure.
Neglecting the nonlinearity of circuit parameters due
to workpiece deformation, Jablonski and Winkler (1978)
described the coil current I by a simple series RLC (equivalent
resistanceinductancecapacitance) circuit yielding an exponentially damped sine wave with frequency f and initial charging
energy E0 :

 R 
E0
I(t) =
exp t sin (2ft)
2L
2CfL
where

1
f =
2

1
R2
2
LC
4L

(1)

. (2)

In order to simplify the analysis, Buehler and Bauer (1968)

approximated the frequency based on the time trise until peak current Ia as
f
=

1
4trise

. (3)

The transient magnetic eld in the vicinity of the workpiece

(yer plate) induces eddy currents in it that oppose the coil current
implying the appearance of the Lorentz volume force FL (Lorentz,
1895):

FL = J B

. (4)

J and B
 are the vectors of current density and magnetic ux
density. Following Aizawa (2003), this volume force can be mathematically transformed into a pressure p, also referred to as magnetic
pressure, acting on both the workpiece and the coil. It can be calculated as
p=

Bg2
2

 2s 

1 exp

. (5)

Here, s is the yer thickness and Bg is the magnetic ux density tangential to the yer surface near the tool coil. The presence
of a transient magnetic eld between yer and coil leads to the
evolution of two related effects, the internally caused skin and

Fig. 1. Schematic of impact welding (Mori et al., 2013).

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

133

Fig. 2. Schematic of MPW: (a) one-sided accessibility as illustrated in Weddeling et al. (2014), (b) two-sided accessibility.

the externally caused proximity effect, inducing current crowding mainly to the surfaces opposite each other in case of reverse
current ow. Leastwise the skin effect can be characterized by an
equivalent conductor thickness, the skin depth (Heaviside, 1951):
= (f)

1/2

. (6)

The parameters f, , and  stand for frequency, electrical conductivity, and magnetic permeability. Low inductances
and capacitances facilitate high discharge frequencies, which are
required for attaining an appropriate magnetic eld and thus
high forming pressure (Daehn, 2010). Generally, tool coils can be
divided into three basic categories after Harvey and Brower (1958):
compression coils, expansion coils, and at coils for sheet metal
forming. Certainly hybrids exist, also for welding tasks. Weddeling
et al. (2014), for instance, used a modied expansion coil introduced
by Kamal (2005) (called uniform pressure electromagnetic actuator) to manufacture at lap joints. A tool coil for such weld types
frequently resembles a single rectangular conductor the pressure
lead having a wider return path away from the weld area (see
Fig. 2a). As can be seen in Fig. 2b, the return path can serve as a
second pressure lead if it is narrow enough and properly placed
below the target plate so that both joining partners are accelerated
against one another (Aizawa, 2003). The mathematical description
of the magnetic ux density in Eq. (5) strongly depends on the
coil geometry, among other factors. Formulae relating the magnetic
eld to the discharge current were reviewed by Psyk et al. (2011) for
rotationally symmetric geometries. For a double-sided conductor
conguration with two sheets as shown in Fig. 2b, Aizawa (2003)
provided the following equation:
Bg =

I
tan1
w

 w 
2d1

+ tan1

 w 
2d2

Fig. 3. Flyer segment interpreted as fully clamped beam.

. (7)

In this, w is the width of the two pressure leads, d1 and d2 in

each case represent the distance between the coil surface facing
the sheet and the point where magnetic ux density is observed.
All else being equal, Eq. (7) does not consider the variation of the
magnetic eld with frequency and workpiece conductivity, meaning it always yields the same magnetic ux density for a given
current value independent of the chosen frequency and conductivity. Moreover, Eq. (7) is only valid for a symmetric conguration
consisting of two one-turn coils and thus not applicable for the
welding with one-sided accessibility (e.g., welding of sheets onto
larger proles). Since an analytical approach that overcomes the
disadvantages mentioned above has not yet been found in literature, an approach which eventually allows to calculate the impact
velocity when one-sidedly using one-turn coils is proposed and
veried in the present paper.

Fig. 4. Visualization of magnetic eld and current distribution in rectangular coil

and yer plate with respect to proximity and skin effect.

134

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

2. Analytical model
First, a setup as depicted in Fig. 2 is considered, particularly the
forming or weld area. There, the yer plate is xed between two
spacers. It is a reasonable simplication to treat a cross-section of
this yer segment as a fully clamped beam of density b , length l,
and thickness s under partial, uniform loading p over the coil width
w (see Fig. 3).
With ow stress  Y , the corresponding static plastic collapse
pressure pc of the beam as known from rigid-plastic theory can be
written as (Jones, 1989)
2s2 Y
wl

pc =

. (8)

It is assumed that this collapse pressure is much smaller than

the acting magnetic pressure (pc  p) so that the inuence of pc
on the yer acceleration is ignored. Here, a one-dimensional rigidbody motion according to Newtons second law is taken to describe
the velocity v of the yer plate and its displacement D :
1
v=
b s

pdt

, (9)

vdt.

Accordingly, the transition from the top to the bottom along the
outer vertical surface of the coil (height h) may be expressed by the
residual current Ih as

h
Ih =

Hh dy

. (15)

The function Hh is discussed in more detail later in this section. Applying Ampres law globally around both the coil and the
corresponding yer segment, such that Ip cancels out, yields

 s

D=

caused by the skin effect, is present. For the bottom side of the coil,
it is supposed that only the skin effect and thus IS remains because
the yer is too far away to have an inuence on the magnetic eld
HS there. The magnetic eld and the current density in the inner
area of the coil are assumed to be negligibly small. Locally employing Ampres law at the bottom of the coil may then simply result
in
w
IS = |HS |
. (14)
2

(10)

2 (2IS + Ih ) = Hg exp

h
w + |HS |w + 2

Hh dy

, (16)

Once the temporal evolution of the magnetic ux density Bg is

determined, magnetic pressure, yer velocity, and displacement
may be computed according to Eqs. (5), (9), and (10). For that reason, the electrical part of the model is established in what follows.
 as well as good conducIf an harmonic magnetic eld strength H
 are assumed, Maxwells equations (Maxwell, 1865)
tors (J = H)
may be put in the form of the second-order partial differential
equation below:
 = k2 H

2H
n

(11)

which, after inserting Eqs. (14) and (15), eventually leads to

 s

|HS | = Hg exp

 = [Hx Hy ]
For the two-dimensional magnetic eld distribution H
in the coil as described above, the following functions, that satisfy
Eq. (11), are proposed here:
Hx (y, t) = Hx1 exp (kT y) + Hx2 exp (kT y) with H x (0, t)
= Hg , Hx (h, t) = HS

where
1
1
+j
n
n

kn =

. (17)

, n = F, T . (12)
Hy (x, y, t) = Hy0 sinh

k 

 k 
T

c1

c2

It is noted that j is the imaginary unit, F and T represent the skin

depth in the yer and the tool coil, respectively. Furthermore, the
total current can be expressed by Ampres circuital law as

I=

 dr
H

. (13)

Now, solutions of Eq. (11) and boundary conditions that adequately relate to the current distribution indicated in Fig. 4 must
be found. For the determination of the magnetic eld strength in
tube compression or expansion, it is a common simplication to
neglect the workpiece movement (Psyk et al., 2011). As the displacements in MPW are generally low in comparison to sheet metal
forming tasks in EMF, the same simplication is made here as well.
Statements given in the following explicitly refer to Fig. 4, where H
7 N/A2 in air.
g = Bg / applies with 0 = 4 10
The magnetic eld strength Hg in the small gap g between the
yer and the tool coil is assumed spatially constant as the yer
remains in close proximity to the coil until the impact. Regarding
the yer plate, a one-dimensional
eld with an exponential decay


from Hg to Hg exp s/F at the side facing the target is already
implicated in Eq. (5). In the yer, only the proximity effect plays a
role since only the induced eddy current Ip emerges there (no forced
current as in the coil). This differs from the two-dimensional distribution in the current-carrying coil, where the total current I may
be split abstractly as follows. On the surface close to the yer, the
proximity effect, as an antimirror-image of Ip , as well as IS , which is

(18)

x exp

with

1
c12

1
c22

= 1.
(19)

With a time-dependent function Hh0 and the constant c2 , the

real part of Hy at the vertical coil surface (x = w/2) may be written
as
Re

w
Hy

, y, t

= Hh0 exp

y
T c2

cos

y
T c2

0 for 0 y h

(20)

Concerning boundary conditions, it is assumed that the vertical

magnetic eld strengths at the edges of the coil (points P1 and P2 )
are specied by:

Re

Hy


Re

Hy

w
2

w
2

, 0, t


, h, t

= Hg Hh0 = Hg

, (21)

= |HS |

. (22)

As Eq. (22) has no closed-form solution when solving for c2 , the

function Hh with constant c3 is taken to roughly approximate the
regarded real part in the form

Hh = Hg exp

y
T c3

Re

Hy

w
2

, y, t

for 0 y h

. (23)

Taking this into account in Eq. (22) ultimately results in

Hh = Hg exp

sy
F h

. (24)

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

135

It can be shown that this simplication in the end leads to

slightly higher values of Hg compared to using a numerical solution for c2 in Eq. (22). In this way, Eq. (23) indirectly even takes
account of edge-effects (current concentration at sharp edges of a
conductor). Again utilizing Ampres law, but with an integration
path just around the coil now gives

 s

I = Hg w + Hg exp

h
w+2

Hh dy

, (25)

which, after solving the integral and rearranging, can be rewritten

as
Hg =

w 1 + exp s/F

+ 2F h 1 exp s/F

/s

. (26)

Eq. (26) is put into Eq. (5) to complete the model providing

p=

0 I 2 1 exp 2s/F

2 w 1 + exp s/F

+ 2F h 1 exp s/F

2 (27)

Fig. 5. Exemplary analytically calculated pressures for rectangular one-turn coils

and a yer conductivity of 30.16 MS/m according to Eq. (27).

3. Experimental procedure

/s

for the transient pressures based on a measured or calculated coil

current I if its decaying time course is interpreted as a sequence
of harmonic half-waves. The characteristics of the proposed model
are plotted exemplarily in Fig. 5 for an arbitrary point in time.
It can be seen that the pressure theoretically increases till innity for an innitely large current and that it decreases to zero for a
very wide coil. The most conspicuous point, though, is the existence
of a high-frequency limit, which is phf = p (f ) = 0.5  (I/w)2 and
equivalent to the hypothesis that the current entirely ows on the
coil surface near the yer plate. Naturally, the pressure becomes
zero for a frequency of zero. The physical explanation for that is
given by the fact that eddy currents are only induced in the yer
in case of a temporally varying magnetic eld. A higher yer conductivity mathematically equals a higher frequency with regard to
magnetic pressure since both parameters similarly affect p via the
skin depth of the yer. It is further noted that the assumed magnetic eld distribution is only valid for yers situated close to the
pressure lead of the coil. This assertion will be further discussed
later.

The MPW experiments conducted within the scope of this work

can be divided into two major parts: velocity measurements (part
I) and welding experiments (part II). Firstly, data obtained from
part I was used to verify the analytical model introduced above
and, secondly, to identify suitable parameter settings for the actual
welding part. Every experiment was repeated three times for reasons of statistical certainty. The basic setup of both experimental
parts is schematically shown in Fig. 6. The proposed model assumes
a two-dimensional eld distribution in the tool coil, which theoretically implies an innite coil length along the axis of the current
ow. That is why relatively long coils (effective length 300 mm)
were used for the experimental part of the work reported here.
In part I, a yer plate was accelerated over a distance of
5 mm by a one-turn coil without a real target, but with a hole
drilled into the opposing clamping xture to allow for recording
the transient yer velocity at the central point between the two
5 mm-spacers by means of a Photon Doppler Velocimetry (PDV)
system, which is addressed later in this section. On the basis of
a copper-chrome-zirconium coil designed by Poynting GmbH (coil
type: F-VWB-300-10), two different pressure lead geometries were

Table 1
Experimental design of part I.
Velocity measurements (PDV, 5 mm travel)
Each experiment: 3 repetitions

Approx. frequency
CMaxwell = 504 F: 20 kHz

Charging
energy

3.25 kJ

4.09 kJ

5.75 kJ

6.68 kJ

8.25 kJ
Experiments were conducted for both coils (RE, TR) with 1 mm thick EN AW 5005A yers

Energy variation:

, frequency variation (with Ia 207 kA):

CSMU = 80 F: 55 kHz

CSMU = 40 F: 65 kHz

136

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

Fig. 6. Schematic of experimental setup for part I and II of MPW experiments.

Table 2
Experimental design of part II.
Welding experiments
Each experiment:
3 repetitions

Standoff distance

1 mm

Charging
energy

2 mm

5.75 kJ

8.25 kJ

9 kJ
TR coil on SMU capacitor bank at approx. 55 kHz
Target: EN AW 6060 hollow prole, yer: 1 mm thick EN AW 5005A sheet

With mandrel in prole:

, without mandrel:

tested (see Fig. 6). Strictly speaking, the cross-section of the tool coil
with the chamfers was a hexagon consisting of a trapezoid and an
adjacent rectangle. To emphasize the geometry of the pressure lead
in close proximity to the yer, this coil is called trapezoid (TR) in
Fig. 6 and hereafter; the completely rectangular one is called RE coil
from now on. In part II, yer plates were accelerated onto a rectangular hollow prole to create magnetic pulse welds in the form
of a lap joint having a predetermined standoff distance, which was
ensured by two insulating spacers. In some cases a massive steel
mandrel was put into the hollow prole to prevent deformation of
the prole upon yer plate impact. In both experimental parts the
horizontal distance between the two insulating spacers amounted
to 50 mm. The parameters varied are listed in the ensuing tables for
each part.
The experimental design of part I is summarized in Table 1 and
may be further divided into the two subparts frequency variation
and energy variation. When changing the discharge frequency of
the circuit, it is useful to keep the peak current Ia constant in order
to retain comparability. Therefore, depending on the capacitor bank
conguration, various charging energies needed to be employed

Fig. 7. Schematic of the PDV system used for experimental part I (Lueg-Althoff et al.,
2014).

(RLC analysis). Regarding the two coil types, however, the same
charging energies were applied for the frequency variation yielding peak currents that were not perfectly identical but in the same
range (approx. 207 kA, see Table 1). Two different pulse generators
(9 kJ Poynting SMU 0612 FS and 32 kJ Maxwell Magneform 7000
series) were used to cover a frequency range from 20 kHz till 65 kHz.
The second subpart, the variation of the charging energy, comes
along with a variation of the peak current; in this case, at a relatively constant frequency of about 55 kHz. 1 mm thick sheets made
from the aluminum alloy EN AW 5005A were chosen as yer material. The rolling direction was always perpendicular to the length
of the pressure lead of the coil. Density and electrical conductivity of the yer plates were taken to be 2.70 g/cm3 and 30.16 MS/m,
respectively (N.N., 2015).
The experimental design of part II is compiled in Table 2. Without anticipating results, it can be seen that only one capacitor bank
conguration (55 kHz: SMU pulse generator with a capacitance of
80 F) and only the TR coil were utilized for the magnetic pulse
welding of the EN AW 5005A yer plates onto extruded rectangular EN AW 6060 proles with a wall thickness of 5 mm. Moreover,
charging energies ranging from 5.75 kJ to 9 kJ as well as standoff distances of 1 mm and 2 mm were deployed with the aim to provide
different impact velocities and angles.

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

137

Fig. 8. IUL peel test setup integrated in Zwick/Roell tensile testing machine.

During every experiment, the coil current was measured with

the help of a Rogowski coil (type: PEM CWT2500B with 500 kA
peak current rating in case of the SMU pulse generator and PEM
CWT1500R with 300 kA peak current rating in case of the Maxwell
pulse generator) placed at the terminals of the capacitor bank. As
mentioned above, a PDV system was used to record velocity-time
graphs. Such an optical measurement system is illustrated in Fig. 7.
It is based on the idea that a laser beam of a known initial wave
length 0 (1550 nm here) is reected from a moving workpiece
the yer plate with a Doppler-shifted frequency so that a com-

bined time-dependent beat frequency, which is proportional to the

wanted workpiece velocity, can be detected (see Fig. 7). The functioning of a PDV system is treated in more detail by Strand et al.
(2004). Besides the RIO Grande Laser Module with an output power
of 1 W used here, the other PDV components conformed with those
described in Daehn et al. (2008). A LeCroy Waverunner 104MXi
oscilloscope having a maximum sampling rate of 10 GS/s ensured
the recording of both the PDV data and the coil currents. Further
data processing was performed with the proprietary software MATLAB.

Fig. 9. Analytical and measured yer velocities at certain displacements for varied energies: (a) for the rectangular coil, (b) for the trapezoidal coil.

138

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

For the purpose of assessing the static strength of sheet-toprole welds (and similar joints), a device for peel tests was
developed and integrated in a Zwick/Roell Z250 tensile testing
machine. With this test rig, which is pictured in Fig. 8, peel tests
can be conducted under different angles.
The principal structure largely resembled that of a conventional
tensile test, except that the upper clamping assembly (holds prole) additionally featured one degree of freedom in the horizontal
direction, accomplished by a runner on a rail. The runner was connected to the lower clamping assembly (holds bent sheet) by a wire
guided over a pulley so that the vertical force lines of the lower and
upper clamping assembly steadily coincided when peeling off the
sheet from the prole. A plot of test force FPeel versus displacement
was made during every experiment.
4. Verication of the analytical model
In case of MPW, the velocity versus displacement graph v(D) of
a yer is an important tool with regard to determining standoff distances and machine parameters for specic lap joints. Such graphs
constitute the focus of the verication of the analytical approach
developed in Section 2. For tool coils having a geometry only slightly
differing from that of a rectangular one (e.g., TR coil in Fig. 6), Eq.
(27) may be modied as

p=

0 I 2 1 exp 2s/F

2 w + w exp s/F + 2F h 1 exp s/F

(28)

/s

where w is still the width of the surface parallel and close to the
yer plate (3 mm; see Fig. 6) whereas w is the width of the opposite
surface (6 mm, also see Fig. 6). In cross-section, h now represents
the length of the open polygonal path along the lines connecting w
and w (h = 15.6 mm in case of the TR coil). Analytical velocities (Eq.
(9)) and displacements (Eq. (10)) generated from pressures according to Eqs. (27) or (28) were based on measured coil currents in
this work. In the analytical model, dv/dt 0 applies, while there is
naturally also a deceleration phase in reality. It is therefore useful
to compare the analytics with experimental results until or at the
displacement where the measured yer velocity vm achieved its
maximum, D(vm,max ). Such comparisons are shown in Fig. 9a and b
in terms of varying the charging energy for a given capacitor bank
conguration and, thus, a constant discharge frequency (approx.
55 kHz).
Maximum velocities ranged from less than 200 m/s at a charging energy of 3.25 kJ to 420 m/s at 8 kJ. The TR coil provided higher
velocities than the RE coil and these higher velocities already
occured at shorter distances compared to the RE coil, which can
be traced back to higher magnetic pressures due to the smaller
pressure lead. Calculated velocities at D(vm,max ) were in acceptable agreement with measured ones for both coil geometries. The
average deviation between model and experiment amounted to
9% for the variation of charging energy, the largest deviation was
20% at 8 kJ and a comparatively large yer displacement of 3.4 mm.
In case of the TR coil, the charging energies corresponded to peak
currents ranging from 200 kA at 4.8 kJ to 303 kA at 8 kJ. A larger
cross-sectional area comes along with a lower resistance, which
is why higher peak currents were recorded when using the RE coil
(between 214 kA at 3.25 kJ and 338 kA at 8 kJ). Maximum measured
yer velocities and associated analytical ones resulting from changing the frequency while keeping the current amplitude almost
constant (approximately 207 kA) are depicted in Fig. 10a and b for
both coils used.
Here, the deviation between model and experiment also varied
from 0% to not more than 20%, again with an average deviation of
about 9%. It is noticeable that most of the measured and calculated
velocities lay in the same area (ca. 200 m/s), only the correspond-

ing displacements partly differed to a greater extent. So, for a given

peak current and impact velocity, the discharge frequency might
serve as a parameter to adjust the desired standoff distance. Fast
capacitor banks can improve the process efciency because higher
frequencies allow for achieving the same yer velocity as with
slower capacitors but with less energy input (see Fig. 10). Certainly,
this is not a general statement due to the fact that all circuit parameters are of interest when choosing the charging energy. In case of
the 65 kHz experiments, for example, a lower capacitance facilitated the frequency increase, but, at the same time, an inductance
slightly higher than that for the 55 kHz experiments led to the need
of a higher charging energy to reach the same peak current. The
latter two gures just refer to a specic yer displacement. Representative curves of velocity versus time and displacement are
displayed in Fig. 11a and b for the RE coil as well as in Fig. 12a and
b for the TR coil.
Despite a small difference in the time domain in Fig. 11a, the
related velocity-displacement curve in Fig. 11b shows how accurate the model represented the experiment until the deceleration
phase of the yer began. The same basically applies to the graphs
in Fig. 12a and b except that there was no difference between both
graphs at the beginning of the yer acceleration, plus a slight overestimation at the maximum velocity was observed.
The analytical model provided satisfying results until the actual
yer displacement D(vm,max ) was reached. As can be seen in
Figs. 11 and 12, the model might also still be helpful in the
early deceleration phase just after vm,max . Moreover, velocity measurements are always necessary to detect D(vm,max ). An adequate
domain of denition shall be dened independent of specic velocity measurements. The magnetic eld distribution as described in
Section 2 implies that the workpiece movement is ignored and
without specifying a distance that the yer is located near the
pressure lead of the coil. A characteristic distance Dch between
tool coil and yer plate, which can be set as maximum admissible yer displacement in the analytical model, can be obtained by
considering the total magnetic energy EL :
1
0
EL = LI 2 =
2
2

 dV.
|H|

(29)

For long coils, EL may be assumed to be completely stored in

the gap between the coil and the yer. To simplify matters, the
high-frequency limit with Hg = I/w is used so that EL can be written
as
1
0
EL = LI
2
2

Hg2 dV =

0 2
0 aDch 2
H waDch =
I
2 g
2w

(30)

where a is the length of the pressure lead in the direction of current

ow (295 mm here). Neglecting the resistance in Eq. (2), the total
inductance L can be expressed in the form
L

1
42 f 2 C

. (31)

After substituting Eq. (31) into Eq. (30), solving for Dch yields
Dch =

w
4a0 2 f 2 C

. (32)

Dch represents a criterion for the maximum yer displacement at

which the analytical eld distribution remains valid. For a given
current, and under the assumptions made above, a higher value of
Dch would lead to a magnetic energy larger than the initial charging energy. Nevertheless, the actual maximum yer velocity vm,max
might occur before, at, or after Dch . Measured velocities vm (Dch ) and
analytically calculated ones v(Dch ) for both the variation of energy
(or peak current Ia ) and frequency are collected in Fig. 13a and b
for the respective coil geometry.

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

139

Fig. 10. Analytical and measured yer velocities at certain displacements for varied frequencies: (a) for the RE coil, (b) for the TR coil.

Fig. 11. Example comparison between the analytical model and an experiment performed on the Maxwell pulse generator with the rectangular coil: (a) velocity-time and
current-time graph, (b) velocity-displacement graph.

Fig. 12. Example comparison between the analytical model and an experiment performed on the SMU pulse generator with the trapezoidal coil: (a) velocity-time and
current-time graph, b) velocity-displacement graph.

The velocities vm (Dch ) usually lay in the same range as the actual
maximum velocities vm,max (compare with Figs. 9 and 10. What is
more, the analytical velocities at Dch were mostly in good agree-

ment with the corresponding experimental velocities (see Fig. 13).

Deviations between them now ranged from perfect agreement to
55%. The largest deviations, though, were outliers in that they per-

140

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

Fig. 13. Experimental and analytically calculated velocities at distance Dch : (a) trapezoidal coil, (b) rectangular coil.

tained to displacements of almost 4 mm, which equals four times

the initial thickness of the yer plate. At such large displacements,
the occurrence of considerable tensional membrane forces make
the assumption of a rigid-body motion (see Eq. (9)) seem inadmissible. Consequently, the model clearly overestimated velocities at
large displacements (several times the yer thickness) where plastic work became signicant, meaning that the yer was already in
the deceleration phase for a long time. Another reason for the inapplicability of the model at high displacements is that the model (Eqs.
(27) and (28), respectively) as well as the criterion given through Eq.
(32) presume a spatially constant magnetic eld in the gap between
the yer plate and the tool coil (see Fig. 4). This simplication connotes that the size of the gap and thus the yer movement do not
inuence the magnitude of magnetic pressure. If the gap becomes

too large in reality, though, the electromagnetic coupling and, as a

consequence thereof, the magnetic pressure diminish so that the
eld distribution illustrated in Fig. 4 ultimately becomes invalid
at high displacements. It is therefore noted that both the analytical
model as well as the displacement criterion are only feasible as long
as a good coupling is ensured. Since maximum standoff distances
are typically only of the order of a very few millimeters or less in
MPW, the formulas introduced here can support the process design
of lap joints without the need of costly velocity measurements.
5. Evaluation of weld quality
As the previous section showed, the analytical model proposed
in this article could be used to approximate the impact velocity vim

Fig. 14. MPW strength evaluation for peel tests: maximum test force versus impact velocity for different standoff distances (with and without putting a massive steel mandrel
in the EN AW 6060 hollow prole during magnetic pulse welding).

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

141

Fig. 15. Exemplary photograph and micrograph of MPW sheet-to-prole lap joint.

(rst impact onto the prole in this case) for the welding experiments mentioned in Section 3 (experimental part II). The maximum
test force Fmax recorded during the peel test also explained in Section 3 was taken as a representative value for the weld strength.
Expressing the strength in terms of stress would not be feasible
here due to the fact that the area truly welded was not accessible
nondestructively. Hence, a diagram where Fmax is plotted versus vim
for various experimental congurations is shown in Fig. 14. Even
though the impact angles were not known here (not reliably measurable in situ), it can be stated that higher standoff distances were
accompanied by higher impact angles.
Data points lying around Fmax = 0 symbolize that no weld was
achieved in those cases. From vim = 400 m/s on, it seems that the
maximum load the joint was able to bear was reached (approx.
3 kN), independent of impact angle (or standoff distance) and
impact velocity. Certainly, this maximum force indicated a minimum peel strength of the joint because all welded specimens
failed in the base metal of the yer plate near the weld seam while
the weld seam itself remained free of failure (see photograph in
Fig. 14). It can also be seen from Fig. 14 that the force FUTS , which
corresponds to the ultimate tensile strength of the yer material,
was a little higher than Fmax of the joints. This observation may
be explained by the stress state of a yer segment in the region
where failure occured during peeling (see sketch in Fig. 14): On the
one hand, the test load FPeel acted as a tensional membrane force
within the yer. On the other hand, the test force also caused a
bending moment in a yer cross-section close to the weld seam.
This moment was characterized by tensional stresses near the target (prole here) and compressive stresses near the opposing yer
surface. Consequently, the superposition of tensional stresses generated by FPeel and its associated bending moment led to a lower
maximum test force than in pure tension. Furthermore, it can be
concluded from Fig. 14 that the usage of a mandrel to prevent defor-

mation of the prole did not affect the weld strength. Naturally, the
experimental setup is less complex and more exible if a mandrel is
not required. When no mandrel was used, the deection resulting
from the yer impact reduced the inner height of the hollow prole
from 50 mm to approximately 49 mm (1/5 of the wall thickness).
With mandrel, no deection of the prole could be detected. Yet,
Psyk et al. (2014) showed that the target deformation can signicantly inuence the joint quality if the yer and the target are more
similar in thickness than in the present study.
Finally, in Fig. 15, a welded specimen, representative of the successful welding experiments, was regarded on the macro as well
as on the micro scale to further evaluate the quality of the MPW
joints. In the etched microsection, it can be seen that there were
two small symmetric regions, where the yer was actually welded
to the prole (bigger grains in the prole due to extrusion process),
while no weld could be created in the center. Since a small fraction
of the yer surface was parallel to the target surface at the very rst
impact, the impact angle was too low for the formation of a weld in
this central region. Within the welded region, however, the typical
wavy interface could be observed. Microscopically, neither interlayers nor local melt zones are visible (see Fig. 15). Raoelison et al.
(2013) found waves of about the same amplitude (ca. 20 m) in
tubular MPW joints of aluminum alloy 6060 and claimed that such
continuous interfacial waves without voids imply a good and permanent bonding, which again endorses the peel test results shown
in Fig. 14.

6. Conclusions
For the magnetic pulse welding (MPW) of at sheets using a oneturn coil, the following conclusions can be drawn from the work
presented:

142

M. Hahn et al. / Journal of Materials Processing Technology 230 (2016) 131142

A simplied analytical model that allows to compute the magnetic pressure as well as the velocity-time and displacement-time
history of the yer plate until the rst impact onto the target
was introduced. It takes into account the geometry, the current
amplitude, and the discharge frequency.
The model was veried experimentally by utilizing Photon
Doppler Velocimetry (PDV) to record the transient yer velocities for various charging energies (3.258.25 kJ) and frequencies
(approx. 20 kHz to approx. 70 kHz). Average deviations between
the model and the experiments amounted to 9%.
Further insight into the impact welding process was gained with
the help of the model by showing that an impact velocity of about
400 m/s is necessary for the magnetic pulse welding of 1 mm thick
EN AW 5005A sheet onto an EN AW 6060 hollow prole.
Etched microsections made clear that a wavy interface morphology is present in the welded regions in which no interlayers,
voids, or melt zones could be found.
Acknowledgements
This paper is based on investigations of the Collaborative
Research Center SFB/TR 10, subproject A10 Joining by forming,
which is kindly supported by the German Research Foundation
(DFG). The peel test used for this work has been developed within
the scope of subproject A1 of the priority program SPP 1640 (joining by plastic deformation) also funded by the DFG.
References
Aizawa, T., 2003. Magnetic pressure seam welding method for aluminium sheets.
Weld. Int. 17 (12), 929933, http://dx.doi.org/10.1533/wint.2003.3199.
Ben-Artzy, A., Stern, A., Frage, N., Shribman, V., Sadot, O., 2010. Wave formation
mechanism in magnetic pulse welding. Int. J. Impact Eng. 37 (4), 397404,
http://dx.doi.org/10.1016/j.ijimpeng.2009.07.008.
Buehler, H., Bauer, D., 1968. Ein Beitrag zur Magnetumformung rohrfrmiger
Werkstcke. Werkstatt und Betrieb 110 (9), 513516.
Daehn, G.S., Zhang, Y., Golowin, S., Banik, K., Vivek, A., Johnson, J.R., Taber, G.,
Fenton, G.K., Henchi, I., Leplattenier, P., 2008. Coupling experiment and
simulation in electromagnetic forming using Photon Doppler Velocimetry.
Proceedings of the 3rd International Conference on High Speed
FormingICHSF, 3544 https://eldorado.tu-dortmund.de/handle/2003/27105.
Daehn, G.S., 2010. Energy eld methods and electromagnetic sheet metal forming.
In: Zhang, W. (Ed.), Intelligent Energy Field Manufacturing: Interdisciplinary
Process Innovations. CRC Press, pp. pp. 471504, ISBN 978-1420071016.
Gies, S., Weddeling, C., Tekkaya, A.E., 2014. Proceedings of the 6th International
Conference on High Speed FormingICHSF, 315324 https://eldorado.tudortmund.de/handle/2003/33476.
Goebel, G., Kaspar, J., Herrmannsdrfer, T., Brenner, B., Beyer, E., 2010. Insights into
intermetallic phases on pulse welded dissimilar metal joints. Proceedings of
the 4th International Conference on High Speed FormingICHSF, 127136
https://eldorado.tu-dortmund.de/handle/2003/27191.
Harvey, G.W., Brower, D.F., 1958. Metal Forming Device and Method. US-Patent No.
2976907.
Heaviside, O., 1951. Electromagnetic Theory. The Complete & Unabridged Edition.
E. & F.N. Spon.

Jablonski, J., Winkler, R., 1978. Analysis of the electromagnetic forming process. Int.
J. Mech. Sci. 20, 315325, http://dx.doi.org/10.1016/0020-7403(78) 90093-0.
Jones, N., 1989. Structural Impact, 1st ed. Cambridge University Press, Cambridge,
ISBN 978-0521628907.
Kamal, M., 2005. A uniform pressure electromagnetic actuator for forming at
sheets. In: Ph.D. Thesis. The Ohio State University http://rave.ohiolink.edu/
etdc/view?acc num=osu1127230699.
Kapil, A., Sharma, A., 2015. Magnetic pulsewelding: an efcient and
environmentally friendly multi-material joining technique. J. Cleaner Prod.
100, 3558, http://dx.doi.org/10.1016/j.jclepro.2015.03.042.
Kore, S.D., Imbert, J., Worswick, M.J., Zhou, Y., 2009. Electromagnetic impact
welding of Mg to Al sheets. Sci. Technol. Weld. Join. 14 (6), 549553, http://dx.
doi.org/10.1179/136217109X449201.
Lorentz, H.A., 1895. Versuch einer Theorie der elektrischen und optischen
Erscheinungen in bewegten Krpern. E. J. Brill, Leiden (in German).
Lueg-Althoff, J., Lorenz, A., Gies, S., Weddeling, C., Goebel, G., Tekkaya, A.E., Beyer,
E., 2014. Magnetic pulse welding by electromagnetic compression:
determination of the impact velocity. Adv. Mater. Res. 966967, 489499,
http://dx.doi.org/10.4028/www.scientic.net/amr.966-967.489.
Maxwell, J.K., 1865. A dynamical theory of the electromagnetic eld. Phil. Trans. R.
Soc. Lond. 155, 459512.
Mori, K., Bay, N., Fratini, L., Micari, F., Tekkaya, A.E., 2013. Joining by plastic
deformation. CIRP Ann.-Manuf. Technol. 62 (2), 673694, http://dx.doi.org/10.
1016/j.cirp.2013.05.004.
Mousavi, A.S.A Akbari, Sartangi, P.F., 2009. Experimental investigation of explosive
welding of cp-titanium/AISI 304 stainless steel. Mater. Des. 30 (3), 459468,
http://dx.doi.org/10.1016/j.matdes.2008.06.016.
N.N., 2015. aluSELECT database. European Aluminium Association (accessed
18.06.15.).
Praveen, P., Yarlagadda, P.K.D.V., 2005. Meeting challenges in welding of aluminum
alloys through pulse gas metal arc welding. J. Mater. Process. Technol.
164165, 11061112, http://dx.doi.org/10.1016/j.jmatprotec.2005.02.224.
Psyk, V., Risch, D., Kinsey, B.L., Tekkaya, A.E., Kleiner, M., 2011. Electromagnetic
forminga review. J. Mater. Process. Technol. 211 (5), 787829, http://dx.doi.
org/10.1016/j.jmatprotec.2010.12.012.
Psyk, V., Lieber, T., Kurka, P., Drossel, W.-G., 2014. Electromagnetic joining of
hybrid tubes for hydroforming. Procedia CIRP 23, 16, http://dx.doi.org/10.
1016/j.procir.2014.10.063.
Raoelison, R.N., Buiron, N., Rachik, M., Haye, D., Franz, G., Habak, M., 2013. Study of
the elaboration of a practical weldability window in magnetic pulse welding. J.
Mater. Process. Technol. 213, 13481354, http://dx.doi.org/10.1016/j.
jmatprotec.2013.03.004.
Strand, O.T., Berzins, L.V., Goosman, D.R., Kuhlow, W.W., Sargis, P.D., Whitworth,
T.L., 2004. Velocimetry using heterodyne techniques. Proc. SPIE 5580, 26th
International Congress on High-Speed Photography and Photonics, http://dx.
doi.org/10.1117/12.567579.
Verstraete, J., De Waele, W., Faes, K., 2011. Magnetic pulse welding: lessons to be
learned from explosive welding. Sustainable Constr. Des. 2 (3), 458464, ISSN
2032-7471.
Vivek, A., Hansen, S.R., Liu, B.C., Daehn, G.S., 2013. Vaporizing foil actuator: a tool
for collision welding. J. Mater. Process. Technol. 213 (12), 23042311, http://
dx.doi.org/10.1016/j.jmatprotec.2013.07.006.
Weddeling, C., Hahn, M., Daehn, G.S., Tekkaya, A.E., 2014. Uniform pressure
electromagnetic actuatoran innovative tool for magnetic pulse welding.
Proceedings of the International Conference on Manufacture of Lightweight
ComponentsManuLight 2014 vol. 18, 156161, http://dx.doi.org/10.1016/j.
procir.2014.06.124, Procedia CIRP.
Zhang, Y., Babu, S.S., Prothe, C., Blakely, M., Kwasegroch, J., LaHa, M., Daehn, G.S.,
2011. Application of high velocity impact welding at varied different length
scales. J. Mater. Process. Technol. 211, 944952, http://dx.doi.org/10.1016/j.
jmatprotec.2010.01.001.