ARTICLE IN PRESS

International Journal of Machine Tools & Manufacture 49 (2009) 916–923

Contents lists available at ScienceDirect

International Journal of Machine Tools & Manufacture

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

A three-dimensional finite element analysis of the temperature field during

laser melting of metal powders in additive layer manufacturing
I.A. Roberts a,, C.J. Wang a, R. Esterlein b, M. Stanford a, D.J. Mynors a
a
Department of Engineering and Technology, University of Wolverhampton, Telford TF2 9NT, United Kingdom
b
University of Dortmund, 44221 Dortmund, Germany

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

Article history: Simulating the transient temperature field in additive layer manufacturing (ALM) processes has presented
Received 15 April 2009 a challenge to many researchers in the field. The transient temperature history is vital for determining the
Received in revised form thermal stress distribution and residual stress states in ALM-processed parts that utilise a moving laser
2 July 2009
heat source. The modelling of the problem involving multiple layers is equally of great importance because
Accepted 2 July 2009
the thermal interactions of successive layers affect the temperature gradients, which govern the heat
Available online 15 July 2009
transfer and thermal stress development mechanisms. This paper uses an innovative simulation technique
Keywords: known as element birth and death, in modelling the three-dimensional temperature field in multiple
Multilayer layers in a powder bed. The results indicate that the heated regions undergo rapid thermal cycles that
Additive layer manufacturing
could be associated with commensurate thermal stress cycles. Deposition of successive layers and
Element birth and death
subsequent laser scanning produces temperature spikes in previous layers. The resultant effect is a steady
Laser melting
temperature build-up in the lower layers as the number of layers increases.
& 2009 Elsevier Ltd. All rights reserved.

1. Introduction laser sintering of titanium powder. The FEM model considered

sintering potential and specific heat capacity as functions of
Laser melting (LM) of metal powders is an additive layer temperature. They also carried out temperature validation
manufacturing (ALM) technology that enables freeform fabrica- measurements on the top surface of the powder bed with an
tion of complex three-dimensional metal parts utilising numerical infra-red camera.
control to bond defined layer geometries in a powder bed by a Although models have been proposed to investigate the
moving laser beam. The LM process is similar to the direct metal temperature distribution in laser-based ALM processes, few have
laser sintering (DMLS) process; only in the former the metal simulated the practical layer-wise building process and have been
powder is completely melted by a continuous wave Nd:YAG laser restricted by the complexity of the problem. In this paper, a
to produce parts of near full density. The process heat transfer computational model for determining the temperature history in
mechanisms are complex and have been investigated in previous a practical LM process involving the scanning of multiple layers is
studies [1,2]. Mathematical models have been proposed to explore developed. This model builds on previously developed concepts
the heat transfer mechanisms in single laser scan tracks [3,4]. by Cervera and Lombera [8] and Kolossov et al. [7], taking into
These have provided an insight into the flow of matter and energy account the process variables, multiple layers, as well as non-
in the laser sintering process. JunChang et al. [5] evaluated the linearities produced by temperature-dependent material proper-
thermal field created on a titanium sample during pulsed Nd:YAG ties and phase changes. The advances of the present work are:
laser treatment using an analytical model. (1) a three-dimensional model is employed in the study rather
Finite element analyses (FEA) using commercial packages have than making assumptions for two-dimensional idealisations, (2) a
been employed to perform temperature field analysis of various detailed understanding of the temperature history resulting from
ALM processes. Nisar et al. [6] developed a model to study the the addition of layers can be obtained using this technique.
temperature fields generated during laser enamelling using a
high-power diode laser for a single track scan. Kolossov et al. [7] 1.1. Background to heat transfer analysis in laser melting
developed a finite element model to predict the temperature
distribution on the top surface of a titanium powder bed during Fig. 1 shows a schematic diagram of laser melting the surface
of a powder bed. The laser scans the powder surface according to a
defined scanning pattern. During the process, the metal powder
 Corresponding author. Tel.: +441 902323908; fax: +441 902323843. undergoes state and phase changes from powder to liquid and
E-mail address: [email protected] (I. A. Roberts). finally to solid.

0890-6955/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmachtools.2009.07.004
 ARTICLE IN PRESS
I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923 917

Nomenclature r radial distance (m)

H enthalpy (J/m3)
Q vector heat quantity (J) r density (kg/m3)
I laser irradiance (W/m2) k thermal conductivity (W/m K)
q laser heat flux (W/m2) h convection heat transfer coefficient (W/m2 K)
P laser power (W) T temperature (1C)
a absorptance j porosity

The general energy balance equation in the closed system in 1.2. Laser energy modelling
accordance with the first law of thermodynamics can be written as
~ ¼Q~ þQ~ þQ~ The Gaussian model, which is the most widely adopted model,
Q L CD CV R ð1Þ
utilises the symmetrical distribution of laser irradiance across the
~ ,Q~ ,Q ~ and Q~ are the vector heat quantities for the beam [9]. It assumes that the laser irradiance is symmetric about
where Q L CD CV R
their propagation direction and in most cases the maximum
laser flux, conduction, convection losses and radiation losses,
irradiance (power per unit area) Io is at the centre of the beam
respectively.
pattern, as shown in Fig. 2 [10].
The beam irradiance at any point for the fundamental mode
(TEMoo) is defined as
2
=d21
IðrÞ ¼ Io e2d ð2Þ

where d1 is the beam diameter corresponding to the point where

the irradiance diminishes to 1/e2 and d the diametrical distance of
a point from the centre. Similarly, the thermal heat flux follows by
definition [11] as
2P 2
=r02
qðrÞ ¼ e2r ð3Þ
pr02

where P is the laser power, r0 the spot radius and r the radial
distance.
The laser energy absorptance of a material a is known to
depend on a number of factors—nature of surface, level of
oxidation, wavelength of laser, surface temperature, etc. [12–14].
The absorptance of metal powders is known to vary between the
in-coupling absorptance as proposed by Kruth et al. [14] to a few
percent when the metal melts [10]. This variation in absorptance
is not yet certain; therefore, the absorptance of pure titanium
powder at the Nd:YAG wavelength (1.06 mm), as reported by
Fischer et al. [15], is taken as the representative absorptance value
Fig. 1. Schematic representation of heat transfer. for the bulk material.

Fig. 2. Gaussian beam mode profiles [10].

 ARTICLE IN PRESS
918 I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923

The heat flux input is, therefore, expressed as With Eq. (5), most thermal processes can be described
q ¼ a  qðrÞ ð4Þ mathematically. According to the specific problem, additional
terms must be included to reflect the particular loads and
boundary conditions [16] as follows:
1.3. Governing equations
Preheat temperature : T ¼ T0
The thermal problem to be solved involves a progressive laser Input heat flux : ðkrTÞ  n^ ¼ q
beam traversing the surface of the powder bed at a constant
Surface convection : ðkrTÞ  n^ ¼ hðT ¼ T0 Þ
velocity. Bejan and Kraus [16] pointed out that the heat transfer
equation for an isotropic material, taking the centre of the beam
as the reference co-ordinate, can be written as where r is the density, DH the change in enthalpy, k the thermal
 
@H conductivity and h the film convection coefficient; n^ is the unit
r ¼ r ðkrTÞ ð5Þ
@t vector normal to the surface of the domain.

Fig. 3. Variation of thermophysical properties for Ti-alloy (TiAl6V4) with temperature [23].

Fig. 4. (a) Finite element model and (b) scanning pattern.

 ARTICLE IN PRESS
I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923 919

2. Simulation technique for multilayer part

2.1. Element birth and death

The model in this study employs the method of element birth

and death to simulate the addition of layers with time. The
deposition of powder material and the laser scanning of geometry
cross-sections on a powder bed to create solids is analogous with
the activation of new elements at desired time points [17]. Before
they are activated the elements are visually present (in a top-
down model); however, they do not add to the overall stiffness of
the matrix. This simulation technique has been employed by Gan
et al. [18] in the residual stress analysis of deposited plasma-spray
coatings on a metal substrate.

2.2. Temperature-dependent material properties

The thermophysical properties of materials can be classed as

field and non-field properties [19]. The thermal conductivity,
which is dependent on porosity, is a field property; however, the
enthalpy and density, which are controlled by the mass of a given
composition, are non-field properties. A number of models have
been proposed for the relationship between the powder bed and
solid thermal conductivities [8,20]. Thummler and Oberacker [19]
proposed that the field properties depend, in a complicated
manner, on porosity and pore geometries but can be represented
by a simplified generic relationship
lp ¼ lð1  jÞ ð6Þ

where lp, l and j are the powder bed, solid metal and the
representative porosity terms, respectively.
The effective thermal conductivity of loose metallic powders is
controlled by gas in the pores. Rombouts et al. [21], in their study
of light extinction in powder beds, found that the effective
thermal conductivity of a powder bed is essentially independent
of material but depends on the size and morphology of the
particles and the void fraction, as well as on the thermal
conductivity of the gas. It was found that for 10–50-mm-size
powders, the effective thermal conductivity is typically from 0.1 to
0.2 W/mK in air at room temperature. The thermal conductivity of
Fig. 5. (a) Infra-red temperature contour plot [26] (T in Kelvin) and (b) Simulation a gas increases approximately as the square root of temperature,
model (Quarter symmetrical model) temperature contour plot (1C). so that a value of about 0.3 W/mK is expected about the melting

Fig. 6. (a) Spot temperature contour plot and (b) spot temperature variation.
 ARTICLE IN PRESS
920 I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923

point [22]. It follows, therefore, that the thermal conductivity of interfering with the temperature distribution in the scanned
TiAl6V4 starts off from this low conductivity value for powder region. The recoating time is the time taken to add new layers
material but sharply rises at the material melting point. prior to laser scanning, and for the purpose of this simulation, a
Fig. 3 shows the variation of the solid thermophysical time delay of 1 s has been chosen. The chamber (preheat)
properties with temperature for TiAl6V4 [23]. temperature is maintained at 80 1C. The scanning pattern
(striping format) is shown in Fig. 4(b). The finite element mesh
for the scanned region is fine enough to accurately simulate the
2.3. Simulation model steep temperature gradients with element length divisions of
0.025 mm (i.e. one-fourth of the laser beam diameter). The mesh
A simulation model consisting of five 30 mm layers of TiAl6V4 is of the surrounding powder bed coarsens, the further away it is
presented in this work. Numerical simulation is carried out using from the scanned region. A 10% convergence test was conducted
the ANSYS Multiphysics finite element package. A rectangular to ascertain the suitability of the chosen mesh divisions.
block of 1 1  0.15 mm built on a 3  3  3 mm mild steel base For the simulation model, the energy from the laser is applied
plate using 8-node brick elements is shown in Fig. 4. The as heat flux on the powder bed surface at a spot location for a time
simulation region is large enough and has been selected to interval defined by the scanning speed and spot size [24]. The
allow the heat from the laser source to dissipate without nodal temperatures are monitored in sub-time steps to determine

Fig. 7. Temperature distribution plots at various times from start of laser scanning (1C).
 ARTICLE IN PRESS
I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923 921

which elements have reached the melting point. The material Fig. 6(a) shows a rapid temperature rise leading up to
properties are then updated for the appropriate elements (by the melting point 1665 1C (1938 K), above which, due to the
interpolation where necessary), and the process is repeated until emergence of the very conductive molten material, the
the irradiation time is reached. On completion, the heat flux temperature change rate drops significantly as the temperature
moves to the next spot. This routine is continued until all the rises to its peak value.
regions defining the geometry on a layer have been analysed. A The temperature variation plot in Fig. 6(b) shows a region with
new layer is then activated (via element birth) and the necessary radial distance, xo50 mm, where the temperature exceeds the
boundary conditions and inputs applied. The entire process is melting point, a region 50 mmoxo100 mm with a very steep
again repeated until the last layer is complete. temperature gradient and a region x4200 mm that appears to be
The process parameters for the simulation model for the relatively unaffected by the laser irradiation in agreement with
power, spot size, scan speed and powder size are 120 W, 100 mm, the experimental measurements and observations.
220 mm/s and 30 mm, respectively. These values are typical
process parameters for building TiAl6V4 parts as advised by the 3.2. Thermal analysis results for multilayer model
original equipment manufacturer [25].
The LM process involves rapid melting and solidification. Fig. 7(a–d) show the temperature distribution plots at different
Frewin and Scott [13] suggested that convective redistribution of times as the laser scans across the surface of the first layer. The
heat within the small molten pool is not as significant as it is in arrows indicate the direction of travel of the heat flux.
processes where a liquid molten pool is permanent. Also radiation The results indicate that the heating and subsequent cooling to
losses are not included in the simulation model; although such the ambient (chamber) temperature occur within a few tenths of a
losses at the surface of the powder bed occur, the bulk of heat millisecond of each other, thus suggesting that the irradiated
transfer is by conduction into the bed. spots are subjected to rapid thermal cycles. These rapid cycles are
associated with commensurate thermal stress changes [27].
3. Results and discussion The temperature profile for path (A–A) through the centre of
the spot in Fig. 7(c) shows a pronounced skewed temperature
3.1. Simulation validation distribution towards the rear of the laser as shown in Fig. 8(a).
This can be attributed to the fact that the rapidly cooling molten
In order to determine the suitability of the simulation model, a material has greater conductive properties than the untreated
comparative study with the experimental work was conducted. powder in front of the laser. This trend can be observed in Fig. 5(a)
The simulation results were compared with published experi- and also from temperature field simulations by Shen et al. [28].
mental findings by Fischer et al. [26] in measuring the peak skin The temperature profiles for consecutive spots in the multi-
temperature of laser irradiation spot on a pure titanium powder layer model are shown in Fig. 8(b). The spots exhibit very similar
bed. The process variables for the experiment were 3 W, 100 mm, thermal cycles, indicating that every point on a layer experiences
1 mm/s and 30 mm for the average laser power, laser spot size, very similar temperature histories.
scan speed and powder size, respectively. The temperature profiles at the centre of the multiplayer
The temperature on the surface was measured with a Raytheon model for the first four layers generated by continuous wave
infra-red camera. Fig. 5 (a) shows the temperature contour plot on Nd:YAG laser are shown in Fig. 9(a–d).
the surface [26]. Fig. 9(a) indicates a number of lower secondary peaks
Fig. 5(a) shows a maximum temperature range of 2500oTmax corresponding to the laser, scanning the same position on
o3000 K, and the reported measured value is 24007200 1C [7]. successive upper layers. This is also exhibited for the upper layers
The temperature distribution for the simulation model using and suggests that the addition of layers and subsequent scanning
the process parameters from the experiment after 0.1 s for a are of significance to the temperature field in the model.
quarter symmetry model is shown in Fig. 5(b). The temperature The results show significantly higher maximum temperatures
contour results show strong similarities with the published for the upper layers. This is due to the fact that the first
experimental measurements, giving a maximum predicted tem- consolidated layer has a lower conductivity than the steel base
perature of 2380 1C (2653 K). plate. The upper layers retain the heat from the laser for long

Fig. 8. (a) Temperature distribution for profile path A–A and (b) spots temperature profiles.
 ARTICLE IN PRESS
922 I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923

Fig. 9. Variation of temperature at the centre of the model with number of layers: (a) first layer, Tmax = 20281C, (b) second layer, Tmax = 21761C, (c) third layer, Tmax = 21551C
and (d) fourth layer, Tmax = 21501C.

Fig. 10. Steady-state temperature variation with number of layers added.

because they are shielded, to a great extent, from the steel base 4. Conclusion
plate. The recoating time is significantly greater than the spot
irradiation time; therefore, dissipation of heat to the newly added A three-dimensional finite element model for predicting the
layer by the previous one provides ample time to reach the steady transient temperature field for multiple layers of parts produced
state (i.e. near ambient temperature) as shown by the horizontal by the LM process has been developed. The model has taken into
sections in Fig. 9(a–d). account the laser energy absorption including the non-linearities
The results also indicate a slight and consistent rise in first produced by temperature-dependent material properties and
layer steady-state temperature as the number of added layers phase changes. An innovative simulation technique known as
increases as shown in Fig. 10(b) for data points P1–P3 in Fig. 10(a). element birth and death has been employed in capturing the
This rise in final temperature in the lower layer could result in the complexities presented by the addition of multiple layers. A study
development of residual stresses during cooling on completion of validating the simulation method was conducted and the results
the process. were in agreement with published experimental measurements.
 ARTICLE IN PRESS
I.A. Roberts et al. / International Journal of Machine Tools & Manufacture 49 (2009) 916–923 923

From the simulation results we can conclude that [8] G.B.M. Cervera, G. Lombera, Numerical prediction of temperature and density
distributions in selective laser sintering processes, Rapid Prototyping Journal
5 (1) (1999) 21–26.
 the laser-scanned regions within a layer undergo rapid [9] R.C. Crafer, P.J. Oakley, Laser Processing in Manufacturing, Chapman & Hall,
temperature cycles that are similar, discounting edge locations London, 1993.
 the upper layers reach higher stable temperatures than the [10] J.F. Ready, Effects of High-Power Laser Radiation, Academic Press, New York,
London, 1971, p. 433.
first layer in contact with the base plate [11] K. Dai, L. Shaw, Finite element analysis of the effect of volume shrinkage
 the addition of more layers and subsequent laser scanning alter during laser densification, Acta Materialia 53 (18) (2005) 4743–4754.
the temperature distribution in the preceding layers, resulting [12] W.-L. Chen, Y.-C. Yang, H.-L. Lee, Estimating the absorptivity in laser
processing by inverse methodology, Applied Mathematics and Computation
in secondary peaks in their temperature histories 190 (2007) 712–721.
 although the recoating phase provides ample time for cooling, [13] M.R. Frewin, D.A. Scott, Finite element model of pulsed laser welding,
there is a small but incremental temperature build-up in the Welding Research Supplement (1999) 15–22.
[14] J.P. Kruth, et al., Lasers and materials in selective laser sintering, Assembly
part and base plate. Automation 23 (4) (2003) 357–371.
[15] P. Fischer, et al., Sintering of commercially pure titanium powder with a
Nd:YAG laser source, Acta Materialia 51 (6) (2003) 1651–1662.
The simulation results provide a better understanding of the LM
[16] A. Bejan, A.D. Kraus, Heat Transfer Handbook, Wiley, New York, 2003, p. 1480.
heat transfer process. Further research will focus on the use of this [17] ANSYS10.0, ANSYS Release 10.0 Documentation, 2007.
method in the study of the development of thermal and residual [18] Z. Gan, H.W. Ng, A. Devasenapathi, Deposition-induced residual stresses in
stresses in multi-layer LM parts. plasma-sprayed coatings, Surface and Coatings Technology 187 (2–3) (2004)
307–319.
[19] F. Thummler, R. Oberacker, An Introduction to Powder Metallurgy, The
References University Press, Cambridge, London, 1993.
[20] S.S. Sih, J.W. Barlow, The prediction of the emissivity and thermal
conductivity of powder beds, Particulate Science and Technology 22 (2004)
[1] N.K. Tolochko, et al., Mechanisms of selective laser sintering and heat transfer 427–440.
in Ti powder, Rapid Prototyping Journal 9 (5) (2003) 314–326. [21] M. Rombouts, et al., Light extinction in metallic powder beds: correlation
[2] A. Simchi, Direct laser sintering of metal powders: mechanism, kinetics and with powder structure, Journal of Applied Physics 98 (1) (2005) 013534 (12
microstructural features, Materials Science and Engineering A 428 (1–2) pages).
(2006) 148–158. [22] A.V. Gusarov, et al., Heat transfer modelling and stability analysis of selective
[3] B. Xiao, Y. Zhang, Laser sintering of metal powders on top of sintered layers laser melting, Applied Surface Science (2007).
under multiple-line laser scanning, Journal of Physics D: Applied Physics 40 [23] K.C. Mills, Recommended Values of Thermophysical Properties for Selected
(21) (2007) 6725–6734. Commercial, Woodhead Publishing, 2002, p. 320.
[4] T. Chen, Y. Zhang, Three-dimensional modeling of laser sintering of a two- [24] K. Dai, L. Shaw, Thermal and stress modeling of multi-material laser
component metal powder layer on top of sintered layers, Journal of processing, Acta Materialia 49 (20) (2001) 4171–4181.
Manufacturing Science and Engineering 129 (2007) 575–582. [25] Electrical Optical Systems, EOS M250Xtended, M270 Basic Operation Manual,
[5] L. JunChang, C. Langlade, A.B. Vannes, Evaluation of the thermal field 2005.
developed during pulsed laser treatments: semi analytical calculation, [26] P. Fischer, et al., Temperature measurements during selective laser sintering
Surface and Coatings Technology 115 (1) (1999) 87–93. of titanium powder, International Journal of Machine Tools and Manufacture
[6] A. Nisar, et al., Three-dimensional transient finite element analysis of the laser 44 (12–13) (2004) 1293–1296.
enamelling process and moving heat source and phase change considerations, [27] K. Dai, L. Shaw, Distortion minimization of laser-processed components
in: Proceedings of the Institution of Mechanical Engineers—Part B— through control of laser scanning patterns, Rapid Prototyping Journal 8 (5)
Engineering Manufacture, vol. 217(6), 2003, pp. 753–764. (2002) 270–276.
[7] S. Kolossov, et al., 3D FE simulation for temperature evolution in the selective [28] X.-F. Shen, et al., Finite element simulation of thermal distribution in direct
laser sintering process, International Journal of Machine Tools and Manu- metal laser multi-track sintering, Journal of Sichuan University (Engineering
facture 44 (2–3) (2004) 117–123. Science Edition) 37 (1) (2005) 47–51.