International Journal of Heat and Mass Transfer 60 (2013) 309–321

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer

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

About the influence of a steady magnetic field on weld pool dynamics

in partial penetration high power laser beam welding of thick aluminium parts
Marcel Bachmann ⇑, Vjaceslav Avilov, Andrey Gumenyuk, Michael Rethmeier
BAM Federal Institute for Materials Research and Testing, Unter den Eichen 87, 12205 Berlin, Germany

a r t i c l e i n f o a b s t r a c t

Article history: A multi-physics numerical model was developed to investigate the influence of a steady magnetic field
Received 17 October 2012 aligned perpendicular to the welding direction during partial penetration high power laser beam welding
Received in revised form 10 December 2012 of aluminium in downhand position. Three-dimensional heat transfer, fluid dynamics including phase
Accepted 9 January 2013
transition and electromagnetic field partial differential equations were successfully solved with the finite
Available online 4 February 2013
element differential equation solver COMSOL Multiphysics 4.2. The implemented material model used
temperature-dependent properties up to evaporation temperature. Marangoni convection in the surface
Keywords:
region of the weld pool, natural convection due to the gravitational field and latent heat of solid–liquid
Electromagnetic weld pool control
Hartmann effect
phase transition were taken into account. Solidification was modelled by the Carman–Kozeny equation
Laser beam welding for porous media morphology. The flow pattern in the melt as well as the weld bead geometry were sig-
Lorentz force nificantly changed by the induced Lorentz force distribution in the liquid metal. It reveals that the appli-
Marangoni flow cation of a steady magnetic field to laser beam welding with corresponding Hartmann numbers
Natural convection Ha2  104 allows for a suppression of the characteristic wineglass-shape of the weld cross section caused
by thermocapillary flow. The numerical results are in good agreement with experimental results obtained
with welding of AlMg3 with a 16 kW disc laser. The steady magnetic field was delivered by permanent
magnets mounted on both lateral sides of the weld specimen. The maximum magnetic flux density
was around 500 mT. It shows, that the applied magnetic field has a predominant dissipating effect on
the weld pool dynamics independently of its polarity.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction Nowadays, disc solid state and Yb fibre lasers are available.
Their wavelength of around 1 lm is ten times shorter compared
The development of modern laser systems for welding applica- to CO2 lasers whose higher wavelength is the limiting factor for
tions with output powers of up to 20 kW for CO2 and 50 kW for fi- deep penetration welding processes above 20 mm due to intensive
bre lasers lead to the fact that the deep penetration welding found scattering and absorption of the laser light by the welding plasma
its way into a lot of industrial applications. Thick aluminium alloy plum and the plasma in the keyhole. In the case of laser systems
plates to be welded are widespread in aerospace industry, ship- with a wavelength of around 1 lm the interaction of the beam
yards and also for the construction of large vessels for the energy with the plasma plume is very low and a larger penetration into
and food industry. the material can be reached. It was shown, that full-penetration
The advantages of laser beam welding are obvious as there are welding of 30 mm AlMg3 plates is possible with 15 kW laser power
the high quality of the welds at a high welding speed and its low at relatively low welding speed to reach the desired penetration
heat input leading to less distortions of the workpiece compared [2].
to traditional arc welding methods [1]. The laser beam intensity The heat conductivity of aluminium especially in its liquid state
is beyond the threshold to enforce the so-called keyhole mode is very high. This leads to comparatively large weld bead widths in
welding, where a vapour-filled cavity develops by the evaporation particular near the surfaces where thermocapillary (Marangoni)
of metal melt. convection is the dominating driving mechanism for convective
The metal parts in thick-plate welding applications are very motions. A further consequence of that behaviour is the typical
large and thus sometimes cannot be welded with the electron wineglass-shape of the weld cross sections. The strong curvature
beam as that would require technical vacuum. of the weld bead ends up in an inhomogeneous solidification front
leading to heavy stresses in the workpiece and as a consequence in
⇑ Corresponding author. Tel.: +49 3081042756. large bending and buckling distortions or residual stresses after
E-mail address: [email protected] (M. Bachmann). cooling. That well-known characteristic [3] is of importance

0017-9310/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2013.01.015
310 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321

Nomenclature

A magnetic vector potential T temperature

B magnetic flux density u fluid velocity
c1 large computational constant x, y, z spatial coordinates
c2 small computational constant
C eff
p effective heat capacity Greek symbols
E electric field c surface tension
fL liquid fraction c0 = @ c/@T surface tension coefficient
FL Lorentz force e turbulent dissipation rate
F hydrodynamic volume force g dynamic viscosity
g gravitational acceleration k heat conductivity
Hf latent heat amount l0 permeability of vacuum
Ha2 Hartmann number lr relative magnetic permeability
j current density q mass density
k turbulent kinetic energy qel electrical resistivity
L characteristic length (half width of the weld pool) r electric conductivity
n normal vector
Pk turbulence production term Sub- and superscripts
PL laser power evap evaporation
p hydrodynamic pressure liq, sol liquidus, solidus
t time melt melting

especially for high penetration depths of the laser source associ- Fig. 2 for a magnetic flux density of 0.5 T. The high values of Ha2
ated to high power deep penetration welding. mean that the drag component due to electromagnetic braking is
Another issue is the low viscosity of the aluminium melt which much larger than the viscous component.
can promote highly dynamical processes near the surfaces leading The application of steady, travelling and oscillatory magnetic
to an unstable weld surface come along with strong spattering and fields is a common practice in many industrial processes (continu-
the ejection of melt [4], see Fig. 1(a). ous casting, crystal growth, electrolysis) to obtain various aims, e.g.
For weld bead half widths of around 10 mm which are common grain refinement, stabilization or even destabilization of liquid sur-
for the welding of large aluminium alloy parts it shows that the faces, acceleration or braking the flow of electrically-conducting
Marangoni effect which occurs in a surface-near region becomes fluids [6].
dominant. Additionally, natural convection in the melt due to the In welding, electromagnetically stirring metal melts by oscillat-
thermal expansion of the material when exposed to high tempera- ing magnetic fields to obtain a homogeneous dilution in the weld
ture differences also plays an important role especially in deep re- pool [7,8] is well-known. Electromagnetic forces in the weld pool
gions of the weld pool. Both together can lead to an unstable were also used to suppress the evolution of excessive porosity in
behaviour of the weld pool surface. the weld [9].
It is well-known that the application of magnetic fields in pres- It was also shown, that oscillating electromagnetic fields can
ence of electrically conducting liquids can have a significant effect excess a magnetic pressure on the weld bead counteracting gravi-
on the dynamics of the flow behaviour [5]. The movement of such a tational forces in downhand position [2,10,11]. Here, a suppression
fluid perpendicular to a magnetic field induces an electric current of the typical wineglass-shape was observed at the root surface
density within the penetration depth of the applied fields for aluminium.
Although the Hartmann coefficient Ha2 in [2,11] was around 500
ju  ru  B: ð1Þ
in the melt pool based on the effective magnetic flux value mean-
The induced electric currents interact with the externally applied ing that a notably effect of the electromagnetic influence was ex-
magnetic field and a Lorentz force distribution arises that is directed pected, computer simulations showed that strong convective
against the fluid flow. Hence, a dissipating effect appears, see Fig. 1(b). motions in the weld pool could not be suppressed.
In this paper the influence of the deceleration of the liquid metal An overview of different electromagnetic applications in arc and
dynamics in laser beam welding of aluminium by a steady magnetic laser beam welding is presented in [12]. The importance of the
field perpendicular to the processing direction (Hartmann effect) is Hartmann effect in further applications can be seen from [13,14].
investigated. This configuration works with permanent magnets The influence of a 200 mT steady magnetic field coaxially to the
without externally applied electric current. That means that no slid- laser beam was studied in [15] for the welding of pure aluminium
ing contacts are necessary which is preferable for the welding of alu- with a penetration of the keyhole into the material of around
minium with a thick oxide layer and also for complex geometries. 5 mm. A notably effect in terms of reducing the weld pool dimen-
Permanent magnetic fields unlike ac magnetic fields can only gener- sions and also a damping of the original melt flow was reported
ate a decelerating force following the second law of thermodynamics. independently of the direction of the applied magnetic field.
A measure for the relative strength of magnetically induced and One of the first attempts to use the Hartmann effect in laser
viscous drag gives the Hartmann number Ha2 beam welding applications were experimentally conducted in
ðjBjLÞ2 r [16] with a CO2 laser. A distinct smoothing of the weld seam for
Ha2 ¼ ; ð2Þ the welding of an aluminium alloy was reported for a weld half
g
width of around 1 mm and a steady magnetic field of 40 mT with
where B is the magnetic flux density and r the electric conductivity. corresponding Hartmann numbers Ha2  100. However, a depen-
The Hartmann number for increasing weld bead widths is shown in dence of the polarity of the magnetic fields was observed which
 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321 311

(a) (b)
Fig. 1. (a) Sketch of the half-model of the partial penetration welding process influenced by permanent magnetic fields on both sides of the workpiece and (b) mechanism of
induced Lorentz forces acting against the melt velocity.

Experiments were conducted for the welding of AlMg3 with a

16 kW disc laser leading to typical weld bead half widths of
10 mm. In contrast to the experiments in [16], the use of a disc la-
ser eliminates the source of thermoelectric currents in the work-
piece due to the CO2 laser-induced plasma. Furthermore,
comparatively high magnetic fields of around 500 mT were used.
Therefore, the Hartmann number becomes very large and domi-
nates over all other effects in the weld pool, see Fig. 2.
It is expected for high power disc laser beam welding, that the
application of a steady magnetic field reduces the dynamics in the
weld pool in terms of reduced weld pool extents and a smoother
surface of the weld. This hypothesis is sketched in Fig. 3.

2. Mathematical modelling

2.1. Governing equations

Since the physics of laser welding are highly demanding, it is

necessary to break down the process to the most important phys-
Fig. 2. Dimensionless Hartmann number for aluminum for increasing half width L ical aspects for the numerical calculations. The basic assumptions
of the weld bead and a magnetic flux density of 0.5 T.
for the fluid flow and temperature field simulation were similar
to those justified in detail in [11] for the case of an electromagnetic
was explained by a net electric current flowing towards the weld weld pool support. These are repeated briefly as follows.
pool end due to thermoelectric effects between cold base material
and the liquid metal. As that current does not depend on the direc-  Temporal oscillations of the weld pool or the keyhole were
tion of the imposed magnetic field, it was reported to be the reason neglected [15,18,19].
for the different effects according to the polarity of the magnetic  The geometry of the free surfaces were fixed, cf. [18,19]. The
field. Later, results of the same group showed that thermoelectric curvature of the surface is less sensitive to strong changes in
currents in the workpiece existed due to its interaction with the la- partial penetration welding of thick-walled workpieces than
ser plume [17]. In this publication, it was found that this effect only in the case of full penetration welding, where the melt can
exists for CO2 lasers with 10.6 lm wavelength whereas no electric drop-out driven by gravitational forces.
current was measured for Nd:YAG lasers with 1.06 lm wave-
length. That means that the influence of the Hartmann effect on la-
ser deep penetration welding still remains unanswered.
In the present investigation, the starting point are computer
simulations with the FE package COMSOL Multiphysics to evaluate
the critical Hartmann number for a suppression of convective mo-
tions in the weld pool. A steady magnetic field was applied to a
partial penetration weld in a 500 mm  50 mm  40 mm alumin-
ium plate. The magnetic flux density in the simulations was in-
creased up to a value of 2 T. The goal was to clarify the extent of
the influence of the Hartmann effect on the laser beam welding
process of aluminium for high Hartmann numbers, where the elec-
tromagnetically induced drag component becomes dominant over Fig. 3. Sketch of effects of the application of a steady magnetic field perpendicular
viscous effects. to the welding direction on the weld seam geometry.
312 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321

 The recoil pressure along the keyhole was assumed to be ideally  Mass conservation
balanced by surface tension forces. Therefore the keyhole geom-
r  ðquÞ ¼ 0: ð3Þ
etry was fixed. The chosen keyhole radii were taken from two-
dimensional heat transfer calculations and scaled to obtain Here, q and u = (u, v, w) are the mass density and fluid velocity.
comparable results in terms of weld pool width, see Fig. 4.  Momentum conservation
The radii depend on the depth according to the used laser type, h
laser power as well as its energy density distribution taking into qðu  rÞu ¼ r  pI þ ðg þ gT Þðru þ ðruÞT Þ

account the heat fluxes due to absorption of laser energy, evap- 2 2
oration of material and heat conduction. Thus, oscillations of  ðg þ gT Þðr  uÞI  qkI þ F ð4Þ
3 3
the keyhole and their possible influence on the flow field are
neglected. The keyhole is subjected to free slip conditions. The with source term F
keyhole diameter is much smaller than the width of the weld
ð1  fL Þ2
pool. Therefore, the influence of the keyhole shape on the flow F ¼ qgbðT  T melt Þ  c1 ðu  uweld Þ þ j  B: ð5Þ
fL3 þ c2
field is relatively small (low Peclet number) whereas the heat
input and the resulting amount of molten material depend on In (4) and (5), p, g, g, and b are pressure, dynamic viscosity, gravita-
the keyhole radii. The temperature at the keyhole surface was tional constant and the coefficient of thermal expansion. Two fur-
set to evaporation temperature. In that configuration, the key- ther equations for the turbulent kinetic energy k and the
hole radii are a kind of free model parameter to obtain a good turbulent dissipation rate e were solved
agreement with the experimental observations because of the   
gT
chosen Dirichlet boundary conditions at the keyhole wall. qðu  rÞk ¼ r  rk þ Pk  qe;
gþ ð6Þ
 A turbulent flow pattern was assumed. Although there are no
rk
  
velocity boundary layers at the surfaces of the weld pool as well g e e2
qðu  rÞe ¼ r  g þ T re þ C e1 Pk  C e2 q ð7Þ
as at the keyhole wall there is a distinct influence of turbulence, re k k
especially at the rear weld pool wall, where the liquid metal
with the turbulent viscosity gT and the turbulent production term
solidifies and a no-slip condition applies. Therefore, the
Pk
steady-state Reynolds averaged Navier Stokes (RANS) equations
2
were solved combined with the standard k–e turbulence model. k
 The material properties were temperature-dependent up to gT ¼ qC g ; ð8Þ
 e 
evaporation temperature [20,21]. 2 2
 The natural convection influence was accounted for using Bous- P k ¼ gT ru : ðru þ ðruÞT Þ  ðr  uÞ2  qkr  u: ð9Þ
3 3
sinesq approximation.
 The solid–liquid phase transformation was modelled by an The remaining constants were set according to Table 1.
enthalpy-porosity approach [22] that uses a solidification range  Energy conservation
of the material. qC eff
p u  rT ¼ r  ðkeff rTÞ: ð10Þ
 The heat effect of the plasma was neglected due to the low heat
absorption coefficient for the wavelength of around 1 lm of the Here, C eff
p ; T, and keff are heat capacity, temperature and heat con-
used disc laser [23] and also the plasma temperature being ductivity.
close to the evaporation temperature of the material [24,25]. The latent heat of melting and solidification was accounted for
using an effective heat capacity formulation:
The governing equations for mass conservation, momentum   2 
and energy transport in steady-state formulation are as follows exp  TTdTmelt
as they are implemented within the simulation framework of C eff
p ¼ C 0p þ pffiffiffiffi  Hf ; ð11Þ
COMSOL Multiphysics. pdT
where C 0p is the temperature-dependent heat capacity and Hf is the
latent heat amount which is normalized around the melting tem-
perature with half width dT = 50 K.
The effective heat conductivity keff accounts for the turbulent heat
conductivity based on Kays–Crawford heat transport turbulence
model [27].

The first term on the right-hand side (RHS) of (5) is the

Boussinesq approximation accounting for the buoyancy force.
The coefficient of thermal expansion reads
1 @q
b¼ : ð12Þ
q @T
The second term refers to the Carman–Kozeny equation accounting
for the resistance of the flow through the mushy region in the

Table 1
Model constants for the k–e turbulence model, see [26].

Constant
Cg Ce1 Ce2 rk re
Value 0.09 1.44 1.92 1.0 1.3
Fig. 4. Keyhole geometry used for the simulations.
 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321 313

Fig. 5. Computational domains for the calculations.

region between the liquidus and solidus temperatures Tsol < T < Tliq. see Fig. 5. The keyhole penetrates around 21 mm into the material.
c1 is a large-valued constant accounting for the mushy region mor- The complete mesh consists of around 900,000 tetrahedral ele-
phology whereas c2 is a small constant to avoid division by zero in ments and a subsequent number of degrees of freedom to be
the solid region. For pure materials with a fixed solidification tem- solved of 8,100,000. The calculation used linear elements for the
perature, it is necessary to introduce artificial solidus and liquidus solution of the heat equation as well as for the Navier–Stokes equa-
temperatures Tsol and Tliq. The liquid fraction fL is assumed to vary tions and quadratic elements for the Maxwell equations. The sim-
linearly with temperature and then smoothed to avoid numerical ulation was done by using a segregated solver that solves for the
issues: variables in groups: (1) magnetic vector potential and electric po-
8 tential, (2) velocity and pressure, (3) turbulence variables, and (4)
> 0 T < T sol ;
< for the temperature. A two-iteration V-cycle geometric multigrid
TT sol
fL ¼ T liq T sol
T sol 6 T 6 T liq ; ð13Þ algorithm was used with the generalized minimal residual method
>
:
1 T > T liq : (GMRES) iterative solver and a direct solver (PARDISO) on the
coarsest grid level.
The half interval between the numerical solidus and liquidus tem-
perature for pure aluminium was chosen to be 50 K for easier con-
2.3. CFD boundary conditions
vergence. The last term on the RHS of (5) is the Lorentz force
contribution coupling the electromagnetic forces with the hydrody-
The boundary condition set-up used in this study was as
namics in the weld pool. It builds up due to the magnetic flux den-
follows:
sity B of the applied permanent magnets and the electric current
density j in the weld specimen that forms due to the movement
 The keyhole surface was fixed, its temperature was set to the
of the electrical conducting media through the magnetic field. The
constant evaporation temperature of the material
Maxwell equations in stationary form for the magnetic field B and
the electric field E are as follows: T ¼ T evap : ð17Þ

r  B ¼ l0 j; ð14Þ Flow normal to the keyhole wall was not allowed. The keyhole sur-
r  E ¼ 0: ð15Þ face was subjected to a slip condition
u  n ¼ 0: ð18Þ
The movement of conducting particles in the magnetic field,
driven by the welding velocity and in particular by the highly  At the upper and lower surfaces: Surface tension variations with
dynamical processes in the pool of molten metal, induces a current temperature were modelled by Marangoni shear stresses.
density according to Ohm’s law. In absence of other sources, it
@u @ c @T
reads: g ¼ ; ð19Þ
@z @T @x
j ¼ rðE þ u  BÞ; ð16Þ @ v @ c @T
g ¼ : ð20Þ
@z @T @y
which couples fluid dynamic and electrodynamic processes. The
current density j multiplied with the magnetic field B gives the Lor-
entz force FL. The resulting Lorentz force in that configuration with
the magnetic field perpendicular to the welding direction has a
component that acts against the original melt velocity thus reduc- Table 2
ing the dynamics in the weld pool. It is favourable to use a magnetic Material properties of pure aluminium at melting temperature Tmelt [20,21].
field that is homogeneously distributed over the weld pool and its
Material property Value Unit
vicinity to avoid the occurrence of non-uniform Lorentz forces.
Therefore the magnet poles must be large enough to cover the Melting temperature Tmelt 933 K
Evaporation temperature Tevap 2700 K
whole weld pool length.
Mass density q 2380 kg m3
Heat capacity Cp 1180 J kg1 K1
2.2. Computational domains Latent heat of fusion Hf 3.97  105 J kg1
Heat conductivity k 91 W m1 K1
The fluid dynamics as well as the heat equation were solved Dynamic viscosity g 1.1  103 Pa s
only in the weld specimen using the finite element solver COMSOL Surface tension c 0.871 N m1
Marangoni coefficient @ c/@T 1.55  104 N m1 K1
Multiphysics. The electromagnetic field equations were solved in
Electrical resistivity qel = r1 24.77  10 8 Xm
the plate and additionally in the surrounding air environment,
314 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321

transfer in the melt near the surfaces occurs mainly by convective

energy transport. These surfaces were also subjected to a slip
condition.
 The inlet temperature was set to room temperature and the
welding velocity uweld was constantly 0.5 m min1.
 The solid–liquid phase transition was accounted for using the
Carman–Kozeny equation [22]. Hereby, additional volume
forces were introduced in the solid phase region to brake the
fluid motion down to the welding speed of the process.
 The symmetry plane was adiabatic and flow in wall-normal
direction was not allowed.

2.4. Electromagnetic boundary conditions

The weld specimen in the simulations consisted of aluminium

with associated magnetic permeability lr = 1. The permanent mag-
nets with a side length of 50 mm were shifted 20 mm in vertical
direction and 10 mm in processing direction to ensure the maxi-
mum magnetic field strength in the region of the weld pool. The
Fig. 6. Normalized thermophysical properties of pure aluminium.
lateral distance between the permanent magnets was 40 mm. For
simplicity, the magnetic field strength was chosen to be constant
Here, c is the surface tension and u = (u, v, w) are the velocity com- at the upper boundary of the magnet.
ponents in the correspondent directions. Heat transport through the Further boundary conditions for the electromagnetic part were
upper and lower walls was neglected (adiabatic walls) as heat as follows:

(a) (b)

(c)
Fig. 7. Temperature and velocity distributions as well as a sketch of the main flow directions in the symmetry plane for the reference case without applied magnetic fields.
The solid line denotes the melting temperature, the dotted lines the artificial solidus and liquidus temperatures, respectively. Note that velocities above 0.1 m/s were cut.
 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321 315

(a) (b)
Fig. 8. Temperature and velocity distributions in the symmetry plane for B = 0.5 T.

(a) (b)
Fig. 9. Temperature and velocity distributions in the symmetry plane for B = 1.0 T.

 The symmetry plane acted as a perfect magnetic conductor – 3. Numerical results

the magnetic field at that plane was solely oriented normal to it.
 At all the other outer air boundaries, the electric and magnetic The present investigation was done with a welding speed of
fields were insulated, that means 0.5 m min1. The keyhole was assumed to intrude around 21 mm
into the workpiece. The denoted magnetic flux density values were
n  j ¼ 0; ð21Þ constant at the upper side of the permanent magnets. The applied
n  A ¼ 0; ð22Þ magnetic flux density was increased stepwise up to 2 T.

where n denotes the normal vector on the outer surfaces. 3.1. Case I: magnet system off

2.5. Material model The temperature and velocity fields for the partial penetration
laser beam welding process of an aluminium plate are presented
Table 2 as well as Fig. 6 give a summary of the material proper- in Fig. 7(a) and (b). The solid line represents the melting isotherm
ties used in this study. The mass density is constant at the value of of the material. It is clear that the surface tension variations at the
the melting point as its differences with temperature are neglected upper surface cause a strong vortex flowing from hot regions near
except for the buoyancy term itself. the laser spot associated with low values of the surface tension to
316 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321

Fig. 10. Hartmann number Ha2 for increasing magnetic flux densities.

regions near the boundaries of the weld pool with corresponding weld bead at the surface. This mechanism is increased by an up-
higher values of the surface tension. This mechanism is well- ward-directed flow near the keyhole due to natural convection
known and is the main reason for the formation of an elongated caused by density differences in the melt. These effects lead to a
 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321 317

typical shape of the weld bead in both the longitudinal as well as much lower than in the solid material. As the Hartmann number
the transversal cross section similar to a wineglass. Moreover, only depends on the strength of the magnetic field and not on its
the vertical flow near the keyhole causes convective heat transfer spatial orientation, it should only be seen as a rough estimate for
mainly to the upper surface in its vicinity followed by a down- the quality of flow control.
wards directed flow at the weld pool rear. Thus, the temperature For example, in the case of an oscillatory magnetic field perpen-
gradients become very high in welding direction. From Fig. 7(b), dicular to the welding direction, which can be used to push the li-
it becomes evident that the maximum velocities occur near the quid melt (cf. [10,11]) at working frequencies of some hundreds to
surfaces under the influence of thermocapillary convection. Fur- thousands Hertz, the induced electric current density due to the
ther local maxima of the velocity in the symmetry plane are situ- movement of conducting media perpendicular to the magnetic
ated at the place where the flow reversal of the surface jet occurs field (similar to Fig. 11) is superimposed by eddy currents which
and also near the keyhole under the influence of gravity-driven can be an order of magnitude larger thus forcing the net current
natural convection. In the lower part of the weld bead, the flow density on streamlines being parallel aligned to the welding
pattern is dominated by the flow of molten metal around the key- direction.
hole in horizontal direction and a downwards directed flow at the It is worth remarking, that the current density distribution in
rear of the melt. the solid material in Fig. 11 especially in the solid phase is caused
by the inlet boundary conditions in the numerical calculations, that
3.2. Case II: magnet system on means a constant relative velocity between the workpiece and the

Figs. 8 and 9 show the temperature distribution and flow veloc-

ities in the symmetry section of the weld bead for increasing mag-
netic flux densities. The upper part of the weld bead is still
dominated by the strong thermocapillary convection present at
the upper surface although its influence is minimized to a compa-
rably thin layer, cf. Fig. 7. The influence of the Lorentz forces causes
changes in the solidification front at the rear side of the weld pool
and also in the cross sectional size. The reason for this modification
lies in the braking of the dynamics especially in the rear of the
weld pool. This can clearly be observed in the course to higher
magnetic flux densities and corresponding higher Lorentz forces
in Figs. 7(b), 8(b) and 9(b) showing the velocity distributions.
The magnitudes and the extent of the local maximum of the veloc-
ity behind the keyhole is clearly diminished under the action of the Fig. 11. Exemplary electric current distribution in the weld specimen for B = 1 T.
applied magnetic field. In the end, the velocities inside the weld
pool remain on welding speed level. Furthermore, the region at
the surface that is highly influenced by Marangoni stresses be-
comes thinner. Increasing the extent of electromagnetic control
leads to a less pronounced convexity of the weld bead shape in
the cross section and a shortened weld bead at the rear side. The
same tendencies can be observed at the front side of the weld pool.
Fig. 10 shows, that a qualitative effect of the electromagnetic
forces on the weld pool shape in this configuration is observable
not before Ha2  104. Here, the Hartmann number Ha2 was calcu-
lated using the effective dynamic viscosity consisting of the (lam-
inar) viscosity gL from the material properties and an additional
contribution accounting for the turbulent flow pattern gT thus
increasing the effective viscosity:
g ¼ gL þ gT : ð23Þ
It shows, that up to a magnetic flux density of around B = 0.3 T, the
flow pattern is dominated by strong turbulence. Accordingly, the
Hartmann number is relatively small in regions of high turbulence
compared to its respective maximum value for each simulation
case. This occurs especially in the upper rear of the weld pool where
the flow reversal of the Marangoni vortex happens. Beginning with
B = 0.4 T the Hartmann number jumps to higher values predomi-
nantly homogeneously distributed in the whole weld pool, which
indicates a more laminar flow. Furthermore, a shortening of the
weld pool dimensions in terms of width and length begins.
The need for high applied magnetic flux densities to produce a
remarkable effect on the weld pool shape (cf. [7]) might also be
caused by the orientation of the induced electric current density.
The location of the permanent magnets perpendicular to the weld-
ing direction leads to two counter-rotating vortices of the electric
current density in the half-plane of the weld specimen whose rota-
tional direction depends on the direction of the applied magnetic
field, see Fig. 11. The electric conductivity inside the weld pool is Fig. 12. Velocity streamlines in the specimen for increasing magnetic flux densities.
318 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321

Table 3
Laser beam welding processing parameters.

Laser type Disc laser Trumpf TruDisc 16002

Laser power PL 16 kW
Welding speed uweld 0.5 m min1
Fibre diameter 200 lm
Wave length 1030 nm
Laser torch angle 18°
Focal length 300 mm
Focus position 4 mm
Shielding gas nozzle angle 45°
Shielding gas 30 l min1 Ar

stresses at the surface remains only noticeable in a very thin layer

near the surface. The rest of the liquid metal in the weld bead
moves with the processing speed synchronous to the solid phase.
The extent of damping the dynamics especially near the upper
surface which is dominated by strong Marangoni convection can
be expressed in terms of Weber number We:

Fig. 13. Weber number for different magnetic flux densities applied at 3 mm qU 2 L
behind the keyhole at the upper surface. We ¼ ; ð24Þ
c
where U is the velocity magnitude and c is the surface tension of the
magnetic field, that is not present when using permanent magnets liquid aluminium. This number relates the kinetic energy and the
mounted on the workpiece. In real experiments that case corre- surface tension of the fluid at the interface between the liquid metal
sponds to the application of a steady magnetic field delivered by and the gaseous phase. As deep-penetration aluminium weld sur-
a DC electromagnet being moved synchronous with the laser face quality is very susceptible to suffer from the dynamics in the
beam. For the electric currents inside the weld pool, the influence weld pool, the Weber number gives an estimate of its stability.
of this assumption is small as the maximum liquid metal velocities Fig. 13 shows the Weber number at the upper surface at a posi-
in the weld pool especially at the surface can be up to two orders of tion 3 mm behind the keyhole for different magnetic flux densities
magnitude larger than the processing speed [11] and the largest ef- applied. It yields that the dynamics of the weld pool corresponding
fect of the Lorentz forces in the weld pool is expected in that to high values of We mitigates the higher the applied magnetic
region. field is which is because of the much lower flow velocities.
The influence of the magnetic fields becomes large enough to
cause significant changes in the flow pattern and also the weld 3.3. Case III: comparative case with constant velocity
pool geometry. Arbitrarily chosen three-dimensional streamlines
of the velocity field are presented in Fig. 12. Note the pronounced Fig. 14(a) shows a theoretical case without simulation of elec-
three-dimensionality of the prevailing flow pattern without elec- tromagnetic fields and fluid flow. Instead, a constant welding
tromagnetic control system. It is clearly observable that the flow velocity is assumed throughout the whole plate and only a thermal
especially in the lower part of the weld pool becomes more uni- simulation with constant convective transport term was done. A
form heading to a purely horizontal flow around the keyhole with comparison with Figs. 8(a) and 9(a) reveals the stepwise approach
magnetic fields applied. For B = 1 T, the influence of the Marangoni from fully turbulent three-dimensional flow inside the molten pool

(a) (b)
Fig. 14. (a) Temperature distribution in the symmetry plane for the case of constant velocity in the whole calculation domain. (b) Weld pool widths and lengths for different
magnetic flux densities and for the case without calculating the fluid dynamics in the melt.
 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321 319

4. Experimental conditions

The welding experiments were performed with a disc laser Tru-

Disc 16002 with an output laser power of 16 kW in downhand po-
sition. To avoid reflections to the optical system, the laser incident
angle was chosen to be 18° from the vertical. The experiments
were done with AlMg3 with a cross sectional width of 40 mm
and a height of 50 mm. An overview of the parameters of the weld-
ing process can be found in Table 3.
Two permanent magnets were mounted stationary on both lat-
eral sides of the workpiece, a water-cooled plate made from ferro-
Fig. 15. Experimental setup. magnetic steel was attached in between to prevent the magnets
from being heated above the Curie temperature, see Fig. 1(a). The
magnets have a cross section of 50 mm  50 mm. Their vertical off-
set to the upper surface of the specimen was 20 mm to ensure the
maximum magnetic flux density in the region of the weld pool.
Their lateral distance was 40 mm according to the width of the
workpiece. The welds were performed by traversing the laser torch
linearly along the long axis of the specimen. Shielding gas was pro-
vided at the upper side behind the laser beam with a nozzle angle
of around 45°, see Fig. 15.
A splash guard was used to avoid heavy spattering of the pro-
cess, especially in the reference case without magnetic field ap-
plied. They were made of the same material as the workpiece to
avoid thermoelectric effects on the contact surfaces between the
sample and the splash guard.
The permanent magnets are made of neodymium iron boron
with a nickel finish. As aluminium and air have the same magnetic
properties (lr = 1), it was possible to measure the magnetic flux
density before starting the welding tests by mounting the perma-
nent magnets on a test stand in the welding configuration but
without the workpiece. The results can be seen in Fig. 16 proving
Fig. 16. Magnetic flux density measured in the middle plane between the that the magnetic flux distribution is approximately constant over
permanent magnets. The black frame denotes the magnet geometry. the inner 20 mm  20 mm where the weld pool is developing thus
justifying the assumption of a constant magnetic field boundary
to the case without any convection except from processing speed condition in the calculations.
inside the melt seen in Fig. 14(a). This yields that an increase of
the magnetic field above Hartmann number of around 105 would
5. Comparison between simulation and experimental results
not result in further qualitative changes of the weld pool geometry.
For higher Hartmann numbers, the weld pool width and length
The comparison between the experimental results and the com-
converge to the theoretical case without additional convection in
puter simulations of the process shows a good agreement of the
the melt, see Fig. 14(b).
observed trends in terms of a reduction of the weld pool extent

Fig. 17. Comparison of the macro sections of the welds produced experimentally and in computer simulations. The solid line marks the melting temperature whereas the
region inside the dotted lines is the solidification interval.
320 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321

Fig. 18. Macro sections at different positions relative to the mounted permanent magnets.

This mechanism increases drastically for the welding with high

power laser systems as the corresponding Hartmann number rises
quadratically with the half width of the welds. This yields that the
magnetic braking of the metal melts works the better the more la-
ser output power was used and the lower the welding speed was
both ending in an increase of the weld pool size.
To confirm that the observed effects are really result of the
Hartmann effect, test welds were performed with different polarity
of the applied magnetic field using Ar shielding gas at a flow rate of
20 l min1, see Fig. 19. The very wide weld pools are very suscep-
tible to the shielding conditions and show a more curved upper
surface than the welds in Fig. 18. High speed videos of the process
show the formation of a very thick oxide layer on the surface.
Therefore, the pores cannot escape from the weld pool and remain
in the vicinity of the surface.
Nevertheless, the curvature of the magnetically controlled
welds is much lower independently of the polarity of the applied
magnetic field which indicates a strong influence of the applied
braking Lorentz forces in the melt.
In contrast to [16], in the experiments with a disc laser system
with around ten times smaller optical wavelength of the laser radi-
ation shown in this paper, an influence of the polarity of the mag-
Fig. 19. Macro sections of two welds, shown are both the reference case without
magnetic field (left) as well as the case with magnetic field applied and changed netic field on the weld shape was not observed. That means that
polarity (right). the reason for the difference between the cases with and without
applied magnetic fields can only be the Hartmann effect with suf-
ficient high Hartmann numbers. Independently on the direction of
and also the attenuation of the influence of the thermocapillary the applied magnetic field, the resulting Lorentz force is directed
flow at the upper surface, see Fig. 17. against the original melt velocity thus lowering the dynamics in
The weld surface as well as macro sections at different positions the weld pool significantly.
from an experimentally obtained weld are shown in Fig. 18. The
welding parameters were according to Table 3. The macro sections
far away from the permanent magnets show a strong convexity 6. Conclusions
due to strong Marangoni convection. Furthermore, the weld seam
surface is very wavy as it is highly influenced by the dynamics in- The purpose of the present investigation was to work out the
side the weld pool and no reasonable weld was obtained. The influence of a magnetic field on a weld pool during high power
macro sections under the influence of the magnetic field show a thick section partial penetration laser beam welding of aluminium
regular weld pool without strong convexity of the solidification based on the Hartmann effect. It was shown, that the application of
front and with an approximately flat upper surface, cf. Fig. 3. The an external magnetic field perpendicular to the welding direction
weld pool depth remains the same for every position at around induces electric currents, which in combination with the applied
25 mm whereas the width at the upper surface becomes smaller magnetic field built a Lorentz force distribution that is directed
when a magnetic field perpendicular to the welding direction is ap- against the melt flow. Thus a braking of the flow velocities in the
plied. The reason is the induced Lorentz force distribution that mit- weld pool was observed.
igates the strong thermocapillary surface flow as it is directed The simulations as well as the experimental work show a severe
against the original flow direction in wide areas of the weld pool. influence of the applied magnetic fields beginning with Hartmann
 M. Bachmann et al. / International Journal of Heat and Mass Transfer 60 (2013) 309–321 321

numbers Ha2  104. Hereby, the polarity of the applied magnetic [7] M. Gatzen, Z. Tang, F. Vollertsen, Influence of a magnetic field in laser welding
of aluminium, in: Proceedings of LAMP, 2009.
field was found to be irrelevant as the resulting Lorentz force is di-
[8] Z. Tang, M. Gatzen, Influence on the dilution by laser welding of aluminium
rected against the melt velocity as long as the magnetic field is with magnetic stirring, Physics Procedia 5 (2010) 125–137.
aligned perpendicular to the welding direction. [9] J. Zhou, H.L. Tsai, Effects of electromagnetic force on melt flow and porosity
The modifications in the flow field in the molten pool also prevention in pulsed laser keyhole welding, Int. J. Heat Mass Transfer 50
(2007) 2217–2235.
change the local temperature distribution. Therefore, also the heat [10] V. Avilov, R. Moldovan, P. Berger, T. Graf, Electromagnetic weld pool control
transfer changes in the weld pool as heat conduction gains more systems for laser beam welding of thick metal plates, in: International
influence compared to convective energy transport. The decelerat- Workshop on Thermal Forming and Welding Distortion, Bremen, April 22–23,
2008.
ing nature of the applied electromagnetic forces leads to a shorten- [11] M. Bachmann, V. Avilov, A. Gumenyuk, M. Rethmeier, Numerical simulation of
ing of the weld bead in longitudinal as well as in transversal full-penetration laser beam welding of thick aluminium plates with inductive
directions. The significant mitigation of the flow velocities espe- support, J. Phys. D: Appl. Phys. 45 (2012) 035201. 13pp..
[12] F. Vollertsen, C. Thomy, Magnetic stirring during laser welding of aluminium, J.
cially near the surface of the weld bead is favourable in terms of Laser Appl. 18 (2006) 28–34.
also reducing the evolution of spatter and melt ejections. [13] B.H. Dennis, G.S. Dulikravich, Magnetic field suppression of melt flow in crystal
It was observable that the influence of thermocapillary flow, growth, Int. J. Heat Mass Transfer 23 (2002) 269–277.
[14] O. Velde, R. Gritzki, R. Grundmann, Numerical investigation of Lorentz force
namely a wineglass shape of the weld bead, on the weld cross sec- influenced Marangoni convection relevant to aluminum surface alloying, Int. J.
tion could almost completely be eliminated. The resulting smaller Heat Mass Transfer 44 (2001) 2751–2762.
curvature of the solidification front related to a V-shape of the [15] M. Gatzen, Z. Tang, CFD-based model for melt flow in laser beam welding of
aluminium with coaxial magnetic field, Physics Procedia 5 (2010) 317–326.
weld cross section benefits a more uniform stress distribution in-
[16] M. Kern, P. Berger, H. Hügel, Magneto-fluid dynamic control of seam quality in
side the weld specimen which in effect leads to smaller deforma- CO2 laser beam welding, Weld. J. (2000) 72s–78s.
tions of the workpiece after cooling down. [17] G. Ambrosy, V. Avilov, P. Berger, H. Hügel, Laser induced plasma as a source for
an intensive current to produce electromagnetic forces in the weld pool, in:
XVI International Symposium on Gas Flow, Chemical Lasers, and High-Power
Acknowledgements Lasers, 2007.
[18] M. Beck, Modellierung des Lasertiefschweißens, Ph.D. thesis, Institut für
Financial funding of the Deutsche Forschungsgemeinschaft DFG Strahlwerkzeuge (IFSW) Stuttgart, 1996.
[19] T. Fuhrich, P. Berger, H. Hügel, Marangoni effect in laser deep penetration
(Bonn, Germany) under Grant No. DFG GU 1211/2-1 and the welding of steel, J. Laser Appl. 13 (2001) 178–186.
German Allianz Industrie Forschung (AiF), Grant No. 17.265N is [20] K.C. Mills, Recommended Values of Thermophysical Properties for Selected
gratefully acknowledged. Commercial Alloys, Woodhead Publishing Ltd, 2002.
[21] B.J. Keene, Review of data for the surface tension of pure metals, Int. Mater.
Rev. 38 (1993) 157–192.
References [22] A.D. Brent, V.R. Voller, K.J. Reid, Enthalpy-porosity technique for modelling
convection diffusion phase change: application to the melting of pure metal,
[1] J.F. Ready, D.F. Farson, LIA Handbook of Laser Materials Processing, Laser Numer. Heat Transfer 13 (1998) 297–318.
Institute of America/Magnolia Publishing/Springer, Berlin, 2001. [23] T. Hoffmann, U. Riedel, Modeling of an arc welding process using high
[2] V. Avilov, A. Gumenyuk, M. Lammers, M. Rethmeier, PA position full temperature transport properties, in: Proceedings of the European Combustion
penetration high power laser beam welding of up to 30 mm thick AlMg3 Meeting, 2009.
plates using electromagnetic weld pool support, Sci. Technol. Weld. Joining 17 [24] P.Y. Shcheglov, S.A. Uspenskiy, A.V. Gumenyuk, V.N. Petrovskiy, M. Rethmeier,
(2012) 128–133. V.M. Yermachenko, Plume attenuation of laser radiation during high power
[3] D. Radaj, Welding Residual Stresses and Distortion. Calculation and fiber laser welding, Laser Phys. Lett. 8 (2011) 475–480.
Measurement, DVS-Verlag, 2003. [25] P.Y. Shcheglov, A.V. Gumenyuk, I.B. Gornushkin, M. Rethmeier, V.N. Petrovskiy,
[4] A. Matsunawa, Problems and solutions in deep penetration laser welding, Sci. Vapor-plasma plume investigation during high power fiber laser welding,
Technol. Weld. Joining 6 (2001) 351–354. Laser Phys. 7 (2012).
[5] R. Moreau, Magnetohydrodynamics, Kluwer Academic Publisher, 1990. [26] D.C. Wilcox, Turbulence Modeling for CFD, third ed., DCW Industries, 2006.
[6] Proceedings 6th International Conference on Electromagnetic Processing of [27] W.M. Kays, Turbulent Prandtl number – where are we, J. Heat Transfer 116
Materials, Dresden, Germany, 2009. (1994) 284–295.