Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Alexandria Engineering Journal (2019) 58, 67–74

H O S T E D BY
Alexandria University

Alexandria Engineering Journal


www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

Finite element modelling for part distortion


calculation in selective laser melting
Samer M. Tawfik *, Mohamed N.A. Nasr, Hassan A. El Gamal

Department of Mechanical Engineering, Faculty of Engineering, Alexandria University, Alexandria 21544, Egypt

Received 21 October 2018; revised 27 November 2018; accepted 12 December 2018


Available online 21 December 2018

KEYWORDS Abstract Additive manufacturing (AM) is presently a strong candidate for automotive, medical
Additive manufacturing and aerospace applications. The present work focuses on the influence of laser scanning speed on
(AM); part distortion, when depositing AISI 304 stainless steel using selective laser melting (SLM). The
Direct metal deposition commercial software ABAQUS was used to develop a 3D finite element model to simulate the depo-
(DMD); sition process and predict thermal gradients and part distortion. The user subroutine USDFLD was
Finite element modelling used to model powder to solid phase transformation, and the laser beam effects were modelled using
(FEM); the user subroutine DFLUX. In addition, heat convection within the molten pool was considered
Selective laser melting by enhancing the thermal conductivity of molten material. Also, the element birth technique was
(SLM); used to simulate material deposition of successive layers. The predicted results were validated by
Part distortion
comparing them to the available literature, where a good match was found. It was concluded that
higher laser scanning speeds results in lower surface temperatures and higher vertical deflections. In
addition, the effect of scanning speed on temperature gradients and part distortion was found to be
more evident by the deposition of subsequent layers.
Ó 2018 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria
University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction techniques are selective laser melting (SLM) and selective laser
sintering (SLS). The laser beam scans over a powder bed cre-
Additive manufacturing (AM) has emerged as a promising ating a molten/sintered pool (typically referred to as molten
manufacturing technique, which provides significant degree pool), which is the first step in creating a structural product
of flexibility and opened new horizons that deemed not possi- [2,3].
ble beforehand [1]. Direct metal deposition (DMD), as one of In an effort to characterize different DMD process, the
the primary AM currently available techniques, is a powder- majority of the available literature focused on examining
based process where metal powder is either melted or sintered how different process parameters affect thermal gradients,
using a laser source. Currently, the most widely used DMD and molten pool size and shape. Finite element modelling
(FEM) played an important role in these studies, as it helps
cutting down the required experimental work and provides
* Corresponding author.
insight monitoring of the ongoing processes.
E-mail address: s_tawfik@alexu.edu.eg (S.M. Tawfik).
A 3D Finite element model was developed by Hussein et al.
Peer review under responsibility of Faculty of Engineering, Alexandria
[4] to predict how the molten pool size and temperature gradi-
University.
https://doi.org/10.1016/j.aej.2018.12.010
1110-0168 Ó 2018 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
68 S.M. Tawfik et al.

ents are affected by laser scanning speed, during depositing a Based on the available literature, the laser scanning speed
single layer austenitic stainless steel AISI 316L. It was found would strongly affect the part distortion and the surface finish
that higher speeds resulted in smaller molten pools in terms of the produced structure. Therefore, the current work exami-
of width and depth; however, the pool length (in the scanning nes the effect of laser scanning speed on part distortion and
direction) increased. Foroozmehr et al. [5] focused on examin- temperature distribution in SLM of AISI 304 stainless steel,
ing how the laser beam penetration depth relates to the molten This has been achieved by building a 3D FE model, using
pool geometry and the generated temperature gradients in the commercial FE software Abaqus/Standard, to simulate
stainless steel AISI 316L, during multi-layer SLM, at different the deposition process and comparing the results to the work
scanning speeds. Steady state conditions were found to be of Liu [9], under similar conditions, for model validation.
reached after the third layer. Amine et al. [6] examined how
laser power and scanning speed affect the thermal gradients 2. Finite element modelling
in the deposited layers, when SLM of stainless steel AISI
316L. The peak temperature was found to increase with higher In the current work, a 3D nonlinear and sequentially coupled
laser power and lower speed. Zhang et al. [7] built a 3D finite thermo-mechanical FE model was built, using the commercial
element (FE) model to investigate the dependence of tempera- software ABAQUS/Standard, to simulate SLM of AISI 304
ture gradients on laser power and scanning speed, during SLM stainless steel. First, a pure thermal analysis was run, and
of W–Ni–Fe powders. It was shown that higher laser powers the resulting temperature field was imported into a pure
and lower scanning speeds led to greater heat input and higher mechanical analysis, which was used to predict part distortion.
maximum temperature. The influence of scanning speed on Three successive layers were simulated and the dependence of
temperature gradients and molten pool dimensions were exam- part distortion on laser scanning speed was examined. The cur-
ined by Song et al. [8]. A 3D FE model was developed, when rent process parameters were selected similar to the experimen-
building a part of Ti-6Al-4V using SLM. The maximum mol- tal conditions used by Liu [9] for model validation. The used
ten pool depth was found to occur at the lowest speed. machine has a laser beam that has a scanning speed in the
Although controlling part performance in AM processes is range of 250 mm/min to 625 mm/min. In this paper, three dif-
highly dependent on part distortion, quite a little attention has ferent speeds were used; 250 mm/min, 375 mm/min and
been paid to how the process parameters affect part distortion 500 mm/min. Tables 1 and 2 illustrate the different process
and thermal stresses. Liu [9] examined the validity of FEM for parameters that were used in the current work. The deposited
simulating temperature gradients and part deformations dur- layers were assumed to be flat, and vaporization was neglected.
ing SLM of AISI 304 stainless steel, on a substrate of the same
material. The predicted results were validated by comparing 2.1. Governing equations
them to experimental ones. Multi-scale modelling was used
by Li et al. [10] in order to predict part distortion and residual
stresses, while SLM of an iron-based multi-layer part. It was A well tested three dimensional, transient and sequentially
found that the upper layers experienced lower longitudinal coupled thermo-mechanical FE model was used in order to
strains in comparison to the lower layers. Cheng et al. [11] compute temperature fields. The three-dimensional heat con-
examined different scanning strategies and how they affect duction equation, Eq. (1), was solved in purpose to obtain
part distortion and internal stresses in multilayer SLM of the temperature field T(x,y,z,t) thru the domain. Where, T rep-
IN718. It was reported that higher stresses were generated at resents the temperature, q refers to the density, C represents
the interface between the substrate and deposited layers, par- the specific heat, k refers to the heat conductivity, and Q rep-
ticularly close to the edges. resents the internal heat per unit volume [9].
     
The effect of molten pool heat convection, which is typi- @T @ @T @ @T @ @T
cally referred to as the Marangoni effect, on temperature dis- qC ¼ k þ k þ k þQ ð1Þ
@t @x @x @y @y @z @z
tribution during SLM was examined by several researchers.
Alimardani et al. [12] and Kumar et al. [13] accounted for Then, thermal gradients as a function of time were
the Marangoni effect by multiplying the molten material ther- imported from the thermal model to the mechanical one.
mal conductivity, over the melting temperature, by a constant. Where the total strain increment (eij) was calculated by solving
The constant was set to be 2.5 in consistency with the experi- Eq. (2). Where, eEij refers to the elastic strain, ePij represents the
mental work of Lampa et al. [14]. Instead of magnifying the plastic strain, eTij refers to the thermal strain, eDV
ij represents the

thermal conductivity of molten material, fluid flow in the mol-


ten pool was simulated by Manvatkar et al. [15] and Mukher-
jee et al. [16]. It was reported that the differential surface Table 1 SLM process parameters [9].
tension plays a vital role in the convective heat flow within Parameter Value
the molten pool.
Power, P (W) 607
Although the powder material properties are absolutely dif-
Scanning speed, V (mm/min) 250, 375, 500
ferent than solid material properties, there are few simulation
Number of layers 3
researches who consider this fact. In the available literature, Layer thickness, k (mm) 0.5
the temperature profiles and part deformation with different Beam radius, r (mm) 1.25
parameters was studied during the experimental and FE stud- Absorption coefficient, g 0.4
ies. However, the distribution of temperature and displace- Room temperature (K) 298
ments in different positions along the fabricated part were Emissivity, e 0.9
seldomly investigated.
Finite element modelling for part distortion calculation 69

successively deposited on a 12.7 mm  50.8 mm  3.2 mm


Table 2 Test matrix of the current work (Table 1 presents the
substrate. A fine mesh of 0.25  0.25  0.25 mm elements
other parameters).
was used for the deposited layers, while a coarser mesh was
Parameter Set 1 Set 2 Set 3 used for the substrate.
Scanning speed, V (mm/min) 250 375 500 The analysis was divided into two subsequent decoupled
Power, P (W) 607 607 607 steps. First, a pure thermal analysis was performed, where
an initial temperature of 298 K (room temperature) was first
assigned to all elements; and after that, the laser beam heating
effects were modelled using the user subroutine DFLUX.
strain resulted from the volumetric change in phase transfor- Throughout the analysis, nodal temperature (NT11) was
mation, and eTrpij refers to the strain due to transformation recorded in order to be used as an initial condition for the sub-
plasticity. eDV Trp
ij and eij were assumed to be zero [9]. sequent mechanical analysis. The pure mechanical analysis was
performed in order to predict part distortion. In order to sim-
eij ¼ eEij þ ePij þ eTij þ eDV
ij þ eij
Trp
ð2Þ
ulate the deposition process, the element birth technique was
Moreover, thermal strain eTij was calculated from the equa- used to activate elements when they are subjected to the flux
tion of thermal expansion constitutive (Eq. (3)). Where, a induced by the laser beam.
refers to the coefficient of thermal expansion, DT refers to
the difference in temperature between two material points, 2.3. Material properties
and dij represents a Dirac delta function, which is assumed
to be one for (i = j) and zero otherwise [16]. Temperature-dependent thermal and mechanical properties
eTij ¼ aDTdij ð3Þ were used, as obtained from [9]. The powder thermal conduc-
tivity and density were set to 0.2 W/m K and 3930 kg/m3,
respectively [17]. The subroutine USDFLD was used to model
2.2. Analysis procedure phase change from powder to solid state. Material phase distri-
bution during the deposition of the second layer is shown in
Figs. 1 and 2 schematically present the details of the built FE Fig. 3, as an example.
model, where three layers of 7.5 mm  2.5 mm  0.5 mm were
2.4. Laser beam modelling

The simulated laser beam was assumed to have a circular


cross-sectional area, and was perpendicular to the scanned
powder surface, having a uniform and constant power density.
Accordingly, the surface heat flux (I) was assumed to be uni-
form, as given by Eq. (4), where P refers to the laser power,
r is the laser beam radius (1.25 mm in the current case), and
g represents the absorption coefficient that was found to be
0.4 for the current conditions based on the experimental results
of Liu [9].
gP
I¼ ð4Þ
p  r2
The Abaqus user subroutine DFLUX was used in order to
implement the laser heating effects in terms of a surface heat
flux (Eq. (4)). A zigzag pattern was followed, as shown in

Fig. 1 (a) Solution domain of the FE model (not to scale), and


(b) Meshing scheme. Fig. 2 Quarter model.
70 S.M. Tawfik et al.

SLM model. Table 3 presents the values of laser power and


laser scanning speed which were used in validation case.
Fig. 6 presents the thermal validation, where the results of
Set 1 were compared to the FE results of Liu [9]. It is worth
noting that Liu [9] did not present experimental temperature
measurements. Fig. 6 illustrates the thermal history at point
A (shown in Fig. 2). As shown, it can be observed that the pre-
sent predicted results matched very well those of Liu [9].
After thermal validation, the presently predicted vertical
displacement of point D, shown in Fig. 2, was compared to
the experimental and FE results of Liu [9] in order to validate
the mechanical aspect of the SLM model. The predicted deflec-
tion results, for the mentioned case in Table 3, are presented in
Fig. 7, where a quite good match was found between the pre-
Fig. 3 Material phase distribution during the deposition of the sently predicted results and those of Liu [9]. On the other hand,
second layer. the difference between the predicted and experimental values
can be attributed to the assumptions and simplifications in
both of mechanical and thermal analysis.
Fig. 4, between subsequent layers. Fig. 5 shows the tempera-
ture distribution half way through the deposition of the first
3.2. Effect of laser scanning speed
layer, as an example.

2.5. Heat transfer to the surroundings In order to examine how temperature distribution and part
distortion are affected by laser scanning speed, three different
scanning speeds were used; 250, 375 and 500 mm/min, that
The user subroutine FILM was used to simulate the heat loss,
correspond to Set 1, Set 2, Set 3, respectively, indicated in
via convection and radiation, to the surroundings. An empiri-
Table 2. The predicted results of the first, second and third
cal equation, Eq. (5) [9], was used to represent the combined
layer top surfaces are illustrated below.
effect of both mechanisms. In Eq. (5), h represents the overall
heat transfer coefficient (W/m2 K) and e refers to emissivity,
3.2.1. Temperature distribution
which was assumed to be 0.9 based on the work of Liu [9].
Fig. 8 illustrates the temperature distribution along the laser
h ¼ 2:41  103 eT1:61 ð5Þ scanning path on the top surface of the first layer, at the end
In addition, following the same procedure used by Alimar- of the deposition of the first layer. As can be seen, the scanning
dani et al. [12] and Kumar et al. [13], a modified thermal con- speed has an insignificant effect on temperature, where the
ductivity was used in the current work in order to account for maximum difference of 100 K (5% of the maximum surface
the Marangoni effect. Above the liquidus temperature, a factor temperature) is very localized underneath the laser beam
of 2.5 was used based on the work of Lampa et al. [14]. (shown by the red arrow in the Figure). That is attributed to
the little difference in the deposition time of a single layer
3. Results and discussion between different speeds. Such difference did not exceed 1 s,
when using a scanning speed of 500 mm/min instead of a scan-
3.1. Model validation ning speed of 250 mm/min. Therefore, the difference in the
available time for heat absorption from the heating source to
the deposited layer, between different speeds, was negligible.
The presently predicted results were compared to the experi- Besides, the difference in the allowable time for heat dissipa-
mental and FE results of Liu [9] in order to verify the present tion from the deposited material to the substrate was quite lit-
tle between different speeds.
Fig. 9 presents the temperature distribution on the top sur-
face of the second layer along x-direction (laser scanning direc-
tion), after the deposition of the second layer. By comparing
Figs. 9 to 8, it can be noticed that the effect of scanning speed
was more significant along the second layer as compared to the
first layer. This is because, compared to the first layer, the dif-
ference in layer deposition time was doubled between different
speeds. This results in an increase in the difference in available
time for heat dissipation from the deposited material to the
substrate. As the same time, the difference in the allowable
time for heat absorption from the laser beam to the deposited
material also increases. Far from the laser beam, the impact of
heat dissipation was more dominant; and accordingly, higher
surface temperatures were generated using higher scanning
speeds. On the other hand, under the laser beam, both effects
Fig. 4 Section (E-E) as per Fig. 1 (not to scale).
Finite element modelling for part distortion calculation 71

Fig. 5 Nodal Temperature (NT11) distribution, in Kelvin, half way during the deposition of the first layer.

Table 3 Validation case [9].


Parameter Set 1
Power, P (W) 607
Scanning speed, V (mm/min) 250

Fig. 8 Temperature distribution along the first layer top surface


after the deposition of the first layer only.

Fig. 6 Thermal history of point A.

Fig. 9 Temperature distribution along the second layer top


surface after the deposition of the second layer (reversed
Fig. 7 Vertical deflection of point D, shown in Fig. 2, for Set 1. direction).
72 S.M. Tawfik et al.

almost balanced out resulting in an almost insignificant effect


of scanning speed.
Temperature distribution on the top surface of the third
layer along the laser scanning direction (x-direction), after
the deposition of the third layer, is presented in Fig. 10. It
was found that the laser scanning speed effect on temperature
distribution was more obvious along the third layer top surface
than such along the second layers top surface. It was found
that the use of lower scanning speeds resulted in higher surface
temperatures under the laser beam, and lower surface temper-
ature far from the laser beam. Additionally, it can be observed
that the heat accumulated in the previously deposited layers
resulted in higher maximum temperatures, and more gradual
temperature gradients along the third layer top surface as com-
pared to the second and first layers. This is attributed to the
accumulated heat from previous layers, which resulted in a
reduction in the cooling rate, especially far from the laser
beam.
Fig. 11 Displacement in depth direction of the top surface of the
3.2.2. Part distortion first layer, after the deposition of the first layer.

The vertical displacements of the nodes located along the laser


scanning path were recorded in order to examine part distor-
tion. The effect of laser scanning speed on the vertical deflec-
tion of the top surfaces of the first, second and third layers
are demonstrated in Figs. 11–13, respectively.
Fig. 11 illustrates the vertical displacements along the laser
scanning line on the first layer top surface, after finishing the
deposition of the first layer. It can be noticed that the speed
did not have an obvious effect on the vertical displacements,
which was attributed to the little difference in the resulted tem-
perature gradients (shown in Fig. 8). In addition, it can be
obviously noticed that the whole surface deflected positively;
i.e., upwards. Because there was no deposited material over
that surface, at the end of first layer deposition, there were
no restrictions for the deformation in such direction. As seen,
the deflection curve has two peaks; the first at around
x = 1 mm and the second at around x = 6.5 mm. The first
peak could be attributed to the sudden increase in the depos-
ited material temperature at the start of deposition, which is
due to the low thermal conductivity of powder material. The Fig. 12 Displacement in depth direction of the top surface of the
second layer, after the deposition of the second layer.

second peak is aligned with the location of maximum temper-


ature along the first layer top surface, after the deposition of
the first layer, as shown in Fig. 8.
Fig. 12 presents the in-depth displacements of the nodes
located along the laser scanning line on the top surface of
the second layer, after the deposition of the second layer. It
was found that the laser scanning speed did not have a mono-
tonic effect on the vertical displacement. It was also observed
that the effect of laser scanning speed on vertical displacement
was similar to its effect on temperature distribution (Fig. 9).
Away from the laser beam, it was observed that higher scan-
ning speeds resulted in higher deformation. That is attributed
to the less available time for heat dissipation from the newly
deposited layer to the underlying material (previously depos-
ited layers and substrate) at higher speeds, which was domi-
nant at such location.
Also, it can be observed that the right end of the second
Fig. 10 Temperature distribution along the third layer top layer deflected negatively; i.e., downwards, which was attribu-
surface after the deposition of third layer. ted to the use of a zigzag scanning pattern, where the second
Finite element modelling for part distortion calculation 73

effect becomes more evident with the deposition of more


layers, which is due to the heat accumulation from previous
layers.
 The laser scanning speed effect on temperature distribution
is highly dependent on the number of added layers. Such
effect is more evident with the deposition of more layers.
This is because the difference in the deposition time between
different speeds increases by the deposition of more layer.
 The effect of laser scanning speed on part distortion in
depth direction depends on the number of the added layers.
Such effect is not obvious after the deposition of only one
single layer. This is due to the less difference in layer depo-
sition time between different speeds. However, such effect
becomes more obvious by the deposition of more layers,
where higher scanning speeds results in higher vertical dis-
placements. This is due to the less available time for heat
dissipation and cooling down at higher speeds.

Fig. 13 Displacement in depth direction of the top surface of the


third layer, after the deposition of the third layer. References

[1] K.V. Wong, A. Hernandez, A review of additive manufacturing,


layer was reversely deposited as compared to the first layer
ISRN Mech. Eng. 2012 (2012).
deposition. Therefore, at the start of the second layer deposi- [2] T.J. Horn, O.L. Harrysson, Overview of current additive
tion, the second layer was deposited over molten material manufacturing technologies and selected applications, Sci.
(the temperature at the end of scanning path along the first Prog. 95 (2000) 255–282.
layer was still over the melting temperature, as shown in [3] S.H. Huang, P. Liu, A. Mokasdar, L. Hou, Additive
Fig. 9). On the other hand, the rest of the second layer top sur- manufacturing and its societal impact: a literat0ure review, Int.
face deflected positively due to previously mentioned reason. J. Adv. Manuf. Technol. 67 (2013) 1191–1203.
The presented results in Fig. 12 were compared to those in [4] A. Hussein, L. Hao, C. Yan, R. Everson, Finite element
Fig. 11, and it was found that the maximum deflection along simulation of the temperature and stress fields in single layers
the top surface of the second layer was higher than that along built without-support in selective laser melting, Mater. Des. 52
(2013) 638–647.
the first layer top surface. This is because the maximum tem-
[5] A. Foroozmehr, M. Badrossamay, E. Foroozmehr, Finite
perature along the second layer top surface was higher than element simulation of selective laser melting process
that along the first layer top surface (Figs. 8 and 9). considering optical penetration depth of laser in powder bed,
Fig. 13 shows the displacement in depth direction along the Mater. Des. 89 (2016) 255–263.
laser scanning path on the third layer top surface, after the [6] T. Amine, J.W. Newkirk, F. Liou, An investigation of the effect
deposition of the third layer. Primarily, it was observed that of direct metal deposition parameters on the characteristics of
the effects of scanning speed on the vertical deflection and tem- the deposited layers, Case Stud. Therm. Eng. 3 (2014) 21–34.
perature distribution along the third layer top surface were [7] D. Zhang, Q. Cai, J. Liu, L. Zhang, R. Li, Select laser melting of
similar. Higher scanning speeds resulted in higher vertical W-Ni–Fe powders: simulation and experimental study, Int. J.
deflections, where the most significant effect was far from the Adv. Manuf. Technol. 51 (2010) 649–658.
[8] B. Song, S. Dong, H. Liao, C. Coddet, Process parameter
laser beam. That was attributed to the less allowable time
selection for selective laser melting of Ti6Al4V based on
for the heat to be dissipated from the newly added layer to temperature distribution simulation and experimental
the previously deposited layers and substrate at higher speeds. sintering, Int. J. Adv. Manuf. Technol. 61 (2012) 967–974.
Therefore, there was less available time for the deposited mate- [9] H. Liu, Numerical analysis of thermal stress and deformation in
rial to contract at higher speeds. In addition, by comparing the multi-layer laser metal deposition process, MSc thesis, Missouri
results of Figs. 13 to 12, it was found that the maximum dis- University of Science and Technology, 2014.
placement along the third layer top surface was lower than [10] C. Li, J. Liu, Y. Guo, Prediction of residual stress and part
the second layer top surface. This is attributed to the more distortion in selective laser melting, Procedia CIRP 45 (2016)
accumulated heat in the previously added layers after the 171–174.
deposition of the third layer. Also, the temperature gradients [11] B. Cheng, S. Shrestha, K. Chou, Stress and deformation
evaluations of scanning strategy effect in selective laser
along the third layer top surface were more gradual than along
melting, Addit. Manuf. 12 (2016) 240–251.
the second layer top surface. That caused lower deformation [12] M. Alimardani, E. Toyserkani, J.P. Huissoon, Three-
along the third layer top surface. dimensional numerical approach for geometrical prediction of
multilayer laser solid freeform fabrication process, J. Laser
4. Conclusions Appl. 19 (2007) 14–25.
[13] A. Kumar, C. Paul, A. Padiyar, P. Bhargava, G. Mundra, L.
Kukreja, Numerical simulation of laser rapid manufacturing of
 The use of lower laser scanning speeds results in higher
multi-layer thin wall using an improved mass addition approach,
maximum surface temperatures due to the more available
Numer. Heat Transf., Part A: Appl. 65 (2014) 885–910.
time for heat absorption by the deposited material. Such
74 S.M. Tawfik et al.

[14] C. Lampa, A.F. Kaplan, J. Powell, C. Magnusson, An analytical [16] T. Mukherjee, W. Zhang, T. DebRoy, An improved prediction
thermodynamic model of laser welding, J. Phys. D Appl. Phys. of residual stresses and distortion in additive manufacturing,
30 (1997) 1293. Comput. Mater. Sci. 126 (2017) 360–372.
[15] V. Manvatkar, A. De, T. DebRoy, Spatial variation of melt pool [17] A. Gusarov, I. Yadroitsev, P. Bertrand, I. Smurov, Model of
geometry, peak temperature and solidification parameters radiation and heat transfer in laser-powder interaction zone at
during laser assisted additive manufacturing process, Mater. selective laser melting, J. Heat Transf. 131 (2009) 072101.
Sci. Technol. 31 (2015) 924–930.

You might also like