Trans Indian Inst Met

https://doi.org/10.1007/s12666-018-1440-7

TECHNICAL PAPER

Model Development in OpenFOAM for Laser Metal Deposition-

based Additive Manufacturing Process
Arvind Chouhan1 • Akash Aggarwal1 • Arvind Kumar1

Received: 30 July 2018 / Accepted: 12 September 2018

Ó The Indian Institute of Metals - IIM 2018

Abstract Laser metal deposition (LMD) process is an 1 Introduction

additive manufacturing process that uses a high-power
laser to melt the metal powder supplied coaxially to the Laser metal deposition (LMD) process has been the focus
focus of the laser beam through a deposition head. The of study of researchers due to its extraordinary features,
LMD process involves various interconnected physical which include: low heat-affected zone, reduced distortion,
phenomenon that are challenging to model numerically. It improved surface hardness, improved wear resistance, the
includes laser–material interaction, fall of molten droplets ability to deposit material on a localised area and the
in a melt pool and the associated heat, momentum and added benefit of being suitable for automation [1].
mass transfer, free surface flow, bubble formation and air Although substantial developments have been made dur-
entrapment, and rapid solidification. A numerical solver in ing the last decades, the LMD process is still facing some
open-source platform OpenFOAMÒ using the VOF problems due to low production rates, low material usage
approach is developed for the spot LMD process. Flows efficiency and particularly high variability, which results
due to natural and Marangoni convection are modelled by in generally poor geometrical stability in the deposited
adding source terms in momentum equation. Energy tracks [2]. The selection of optimum parameters can be
addition in the substrate due to laser irradiation is added as achieved by using experimental trials or most recently by
a source term in energy equation. Enthalpy-porosity tech- the use of closed loop control systems [3]. Unfortunately,
nique is used to model the melting and solidification phase experimental trials are costly and time-consuming; and
change. Mass and energy addition due to falling droplets control systems use expensive equipments. Modelling and
are modelled by adding molten droplets in the melt pool. simulation can be a very useful alternative tool for pre-
The role of melt pool convection on heat transfer and melt dicting the process behaviour and hence can be a cost-
pool geometry is established by an analysis using relevant effective approach for an adequate selection of processing
dimensionless numbers. Ratio of Marangoni number and parameters.
Grashof number suggest that the fluid flow within the melt The thermal-fluid transport phenomena in the LMD
pool during laser heating stage is mainly driven by Mar- process can be numerically modelled by two ways:
angoni force. macroscopic method and mesoscopic method. In macro-
scopic modelling method, the interaction of falling powder
Keywords Laser metal deposition  Volume of fluid  particles with melt pool is not directly considered, rather
OpenFOAM  Marangoni number  Grashof number than this interaction is modelled using analogous approa-
ches. In such approaches, one approach is ‘‘element’s death
and birth technique’’ in which mass and energy addition
due to falling droplets is modelled by activating and
& Arvind Kumar deactivating elements in the computational domain by
[email protected]
using a multiplying factor [4, 5]. Another approach is
1
Department of Mechanical Engineering, Indian Institute of modelling of powder particle interaction with melt pool by
Technology Kanpur, Kanpur 208016, India a moving computational mesh based on an arbitrary

123
 Trans Indian Inst Met

Lagrangian–Eulerian (ALE) formulation. The macroscopic

simulation approach can provide an ample amount of
information at the process scale with reasonable compu-
tational cost and complexity, but it is unable to provide
information about free surface deformation due to parti-
cles’ impact in the melt pool, effect of particle size on melt
pool formation and grain morphology and porosity gener-
ation which are the main physical phenomena that are
typically encountered in the LMD process. These phe-
nomena can be modelled using mesoscopic approaches.
They are either Lattice Boltzmann model [6] or volume of
fluid (VOF) method [7]. These methods are used to track
the free surface between liquid and gas regions, so the
interaction of each falling particle with the melt pool, and
addition of mass and energy in the melt pool can be
modelled. In most of the research works reported,
researchers consider simulation of full track or multi-track
during the deposition process. Due to high computational
cost, they have neglected many small-scale phenomenon
that takes place during the actual deposition process, such
as ripples formation on the free surface, bubble formation Fig. 1 Schematic of the spot LMD process
in the melt pool due to droplet impact and resulting air
entrapment which can give rise to porosity formation in the being ejected from the nozzle, while laser starts acting on
final product. The complex physics associated with the the substrate immediately when the laser is turned on. The
local melting and solidification phase change during the computational model is implemented in OpenFOAM by
LMD process is yet not completely understood. There is a adding the energy transport equation in the interFOAM
strong need for accurately describing the transient thermal solver which is readily available in the OpenFOAM.
field and the flow patterns in the melt pool and their effect Firstly, phase fraction equation is solved using the multi-
on the melt pool geometry. Further, any work on spot dimensional universal limiter with explicit solution
deposition, which is the most fundamental level work, is (MULES) algorithm. In order to correctly predict pressure
not performed. In this study, a transient numerical model and velocity, PISO algorithm is used. Finally, temperature
considering VOF approach has been used to simulate the is determined using the energy equation in which latent
heat transfer and fluid flow during the melting and the heat is accounted with the enthalpy-porosity method.
solidification stage for spot LMD process of 304 stainless The thermo-physical properties of 304 stainless steel are
steel. An analysis of the temperature and velocity fields and listed in Table 1.
their influence on melt pool geometry is performed. An axisymmetric domain of size 600 lm in x direction
and 800 lm in z direction is considered in simulations.
Substrate size is taken such that there is no heat accumu-
2 Computational Modelling lation due to laser heating, because heat accumulation will
alter the melt pool size and shape. Accordingly, substrate
Figure 1 shows the schematic of the LMD process on a dimensions are taken as four times of maximum melt pool
spot. A laser with power P and beam radius R acts upon a width and depth (obtained during simulation) in x and
304 stainless steel substrate kept initially at room temper- z direction, respectively. The grid density is large in the
ature. The entire process is carried out in the inert envi- vicinity of the region where most of the physical activities
ronment of argon gas. In the present model, it has been are encountered and is lower elsewhere in the domain.
assumed that the spot deposition process is symmetric From the grid independence study, it has been found that
about the central axis, and hence an axisymmetric domain the mesh independent solutions are obtained at grid size of
approach is used for simulation. The laser first acts for (100 9 100) for substrate and (100 9 100) for the gas
1 ms till the formation of a considerable sized melt pool, region and with the refinement of the grid, further solutions
then droplet mass (material of droplet) is added in the form will not change and it will only increase computational
of particles of diameter D. This assumption is made after cost. Hence, in this study, grid size of (100 9 100) for
considering the fact that in actual process it takes some substrate and (100 9 100) for the gas region has been
time for the powder particles to reach the substrate after

123
Trans Indian Inst Met

Table 1 Thermophysical properties of 304 stainless steel oc  !

þ r  ðc!
u Þ þ r  ð1  cÞc Ur ¼ 0 ð1Þ
Parameter Value ot

Solid and liquid metal density, qm 7200 kg/m3

where c denotes the phase fraction, t is the time, !
u is the
!
Effective dynamic viscosity, lm 0.06 kg/(m s)
velocity vector and Ur is the compression velocity [9].
The compression velocity compresses the interface by
Surface tension, r 1.9 N/m
decreasing the numerical diffusion of the phase fraction c
Solidus temperature, Ts 1697 K
and preserving its boundedness.
Liquidus temperature, Tl 1727 K
Specific heat of solid and liquid, Cp 712 J/(Kg K)
2.1.2 Mass Conservation
Thermal conductivity of solid, ks 15 W/(m K)
Thermal conductivity of liquid, kl 33 W/(m K)
The continuity equation for incompressible flow is given
Latent heat, L 2.47 9 105 J/Kg
by
Thermal expansion coefficient, bT 1.96 9 10-5 1/K  
Temperature coefficient of surface tension, - 4.3 9 10-4 N/(m r ! u ¼0 ð2Þ
dr
dT K)
2.1.3 Momentum Conservation
Emissivity, e 0.45
Absorption coefficient, A 0.27
The governing momentum conservation equation is given
Convective coefficient 10 W m-2/K
by
oðq!uÞ !
þ u  rðq!u Þ ¼ rp þ r
chosen to run further simulations. This enables optimal ot    T  !
utilisation of the computer resources. The process param-  l r!u þ r! u þF
eters used in simulations are taken from the experiment ð3Þ
conducted by Crespo [8] for the LMD process (Table 2). ! ! !
~
F ¼ FS þ FN þ F M ð4Þ
2.1 Governing Transport Equations !
The source term FS in Eq. (4) is defined in Eq. (5). It
brings down the velocity of the molten fluid at the liquid–
The momentum and energy conservation equations are solid phase transition interface and makes the fluid motion
coupled with the VOF transport equation. Assuming the in the unmelted zone zero.
flow to be Newtonian, incompressible and laminar, the 2
governing equations for mass, momentum and energy ! ð1  fl Þ
FS ¼ 3 ~
Cu ð5Þ
conservation are devised. The density of liquid and solid fl þ b
phase of the metal is assumed to be same. In the simulation,
The fl represents the fraction of liquid and is given by
the thermo-physical properties of the two immiscible 0 1
phases (metal and inert gas) such as density are calculated 0 T\Tsolidus
using the continuum formulation based on the classical B T  Tsolidus Tsolidus  T\Tliquidus C
fl ¼ @ A ð6Þ
Tliquidus  Tsolidus
mixture theory.
1 T  Tliquidus
2.1.1 Volume of Fluid Equation The flow due to natural convection is included in this
model with the help of Boussinesq approximation. The
!
The modified VOF equation for the interface tracking is source term FN in Eq. (4) is given as
given by !
FN ¼ qg ~bT ðT  Tref Þ ð7Þ
where q, bT and Tref are the density, coefficient of thermal
Table 2 Parameters for simulation
expansion and reference temperature, respectively.
Parameter Value !
The source term FM in Eq. (4) is given as
 
Laser spot size (lm) 300
~ dr 2q
F M ¼ rj^ nþ ½rT  n^ðn^  rT Þ jrcj ð8Þ
Powder particle diameter (lm) 50 dT qm þ qg
Laser power (W) 130
Mass flow rate (gm/min) 0.16956 where r is the surface tension coefficient, j ¼ ðr  n^Þ is
the mean curvature of free surface, n^ ¼ rc=jrcj is the

123
 Trans Indian Inst Met

interface normal unit vector, and dr/dT is the temperature from a certain height. The droplets fall from a height
coefficient of surface tension. 50 lm above the initial substrate position, in the laser
generated melt pool. Mass flow rate is calculated by mul-
2.1.4 Energy Conservation tiplying the number of droplets with the mass of one
individual droplet, in total 2 ms time (the time up to which
The governing heat transfer equation is given by the laser remains on). In this way, 12 droplets are added
oðqCP TÞ   (one at a time) in the system, and the mass flow rate
þ~
u  r qCp T ¼ r  ðKrT Þ þ Q þ Slatent ð9Þ (M.F.R.) comes out to be 0.16956 gm/min. After 2 ms,
ot
laser is switched off and the melt is allowed to cool down
where Cp denotes the heat capacity, K denotes the average and solidify. Figure 2 shows the evolution of the melt pool
thermal conductivity of the material according to mixture during the heating stage. The temperature distribution and
theory and Slatent is the latent heat source term. The source the flow pattern in the melt pool are represented by tem-
Slatent accounts for the evolution of the latent heat during perature colour map superimposed by streamlines. The
phase change. The latent heat is absorbed during melting direction of the flow is represented by vectors on
and released during solidification. This term is given by:
8

< o
 L ðqfl Þ for c ¼ 1
Slatent ¼ ot ð10Þ
:
0 for c\1

In (Eq. 10), L is the latent heat of fusion of the metal and

fl is the liquid fraction.
The source term Q in Eq. (9) is given as
2qCpeff
Q ¼ q00 jrcj ð11Þ
qm Cpm þ qg Cpg

The jrcj represents the gradient of phase fraction and it

00
implies that the surface heat flux q is smeared over the
interface cells. It captures the moving interface between the
powder particles and the inert gas so that laser heat flux can
be applied only to the interface. The heat flux q00 is given
by:
 
q00 ¼ q  hc ðT  T1 Þ  er T 4  T1 4
ð12Þ

where q is the input heat flux from the high intensity laser
and is approximated by a Gaussian distribution. The input
heat flux q is given by the following expression
 
2AP 2r 2
q¼ exp  2 ð13Þ
pR2 R
where A is the absorptivity, P is the power of laser beam
and R is the radial distance in which energy density equals
to e-2 times that at the centre of the laser spot.

3 Results and Discussion

The simulation of LMD process is divided into three

stages; pure laser heating stage (0–1 ms), mass addition
stage (1–2 ms) and solidification stage (2 ms to complete
solidification). Laser power of 130 W and laser beam
radius of 150 lm has been used in this study. The mass is
added in the form of falling liquid droplets of diameter
Fig. 2 Calculated temperature distribution and flow pattern during
50 lm, at 2000 K temperature with a velocity of 1 m/s particle impact

123
Trans Indian Inst Met

streamlines. Two temperature contour levels for 1697 K

and 1727 K (solidus and liquidus temperature) are marked
to show melt pool boundary.Umax and Uavg marked in
figures are the maximum and the average velocity in the
melt pool.
As the droplet falls, its temperature rises because of
laser heating. Falling droplet creates an opposite direction
of flow in the melt pool which competes with the flow due
to Marangoni convection as shown in figures. As the dro-
plet interacts with the melt pool, a small portion of gas gets
trapped between the droplet and the melt pool, and as the
droplet descends further, this bubble takes the shape of a
flat sheet and finally breaks into two bubbles. Figure 2 also
depicts that how the flow pattern in the melt pool changes
because of droplet impact. During this stage, size of the
melt pool, maximum velocity, average velocity and max-
imum temperature keep on increasing due to continuous
heating by laser. In this stage, both the surface tension
force (Marangoni convection) and the natural convection
can drive the fluid flow. In order to assess which force is
prominent, dimensionless parameters have been analysed.
Table 3 shows the comparison of surface tension force and
buoyancy using Marangoni number (Ma) and Grashof
number (Gr), respectively. It can be observed that Ma is
much larger than Gr, which suggests that surface tension is
the main driving force for the fluid flow in the melt pool.
Figure 3a, b shows the variation of melt pool half width
and depth with time. The half width is calculated as
maximum distance of the solidus isotherm (1697 K) from
the symmetry axis in radial direction, and depth is calcu-
lated as distance of the solidus isotherm from the initial
substrate position (400 lm) in z direction. A similar kind of
trend can be seen in both the plots, both half width and Fig. 3 Variation of a melt pool half width, b melt pool depth with
depth increase rapidly at the start and then the rate of their time. Laser power, P = 130 W, M.F.R. = 0.16956 gm/min, Pulse on
growth slows down. The reason for this is that, as the melt time = 2 ms
pool size grows, the area (top surface) from which the heat
losses take place increases. Also, due to continuous heat-
ing, melt pool temperature increases. Both of these factors 4 Conclusions
increase radiative and convective losses from the top sur-
face of the melt pool, which slow down the melt pool In this study, a 2D axisymmetric numerical model has been
growth rate. At 2 ms, the laser is switched off; hence, the developed in OpenFOAM to simulate the laser metal
melt pool size decreases rapidly after this time. This deposition process at the mesoscale. The model couples
rapidly reduces depth and half width after 2 ms. heat transfer and fluid dynamics with volume of fluid
method to trace the boundary of the melt pool. Analysis of
dimensionless parameters suggests that during the laser
Table 3 Surface tension force vs buoyancy force heating stage, the fluid flow within the melt pool is mainly
Time (ms) Ma Gr driven by the Marangoni convection and buoyancy force
plays a very minor role. During the mass addition stage,
0.2 15.51 0.26 9 10-5
there is a competition between flow due to droplet impact
0.4 45.06 1.44 9 10-5
and Marangoni convection. Though gas porosities are
0.6 64.01 2.66 9 10-5
formed in the melt pool, they are finally ejected. Every
0.8 81.05 3.99 9 10-5
time a droplet is added, there is a substantial change in the
1.0 91.11 5.04 9 10-5 field variables (temperature, flow pattern, velocity, etc.)

123
 Trans Indian Inst Met

References 6. Klassen A, Scharowsky T, and Körner C, J Phys D Appl Phys 47

(2014) 275303.
1. Zhong M, and Liu W, Proc Inst Mech Eng Part C J Mech Eng Sci 7. Ibarra-Medina J, Vogel M, and Pinkerton A J, in Proc.
224 (2010) 1041. ICALEO’2011, (2011), p 23.
8. Crespo A, in Convection and Conduction Heat Transfer, InTech,
2. Costa L, and Vilar R, Rapid Prototyp J 15 (2009) 264.
3. Mazumder J, Dutta D, Kikuchi N, and Ghosh A, Opt Lasers Eng New York (2011).
34 (2000) 397. 9. Samkhaniani N, and Ansari M R, Heat and Mass Transf 53 (2017)
4. Hao M, and Sun Y, Int J Heat Mass Transf 64 (2013) 352. 2885.
5. Long R S, Liu W J, Fei X I N G, and Wang H B, Trans Nonferr
Metals Soc China 18 (2008) 691.

123