Вы находитесь на странице: 1из 9

G Model

ARTICLE IN PRESS

ADDMA-93; No. of Pages 9

Additive Manufacturing xxx (2016) xxxxxx

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Controlling of residual stress in additive manufacturing of Ti6Al4V by


nite element modeling
G. Vastola, G. Zhang , Q.X. Pei , Y.-W. Zhang
Department of Engineering Mechanics, A*STAR Institute of High Performance Computing, 1 Fusionopolis Way, Connexis #16-16,
Singapore 138632, Singapore

a r t i c l e

i n f o

Article history:
Received 8 October 2015
Received in revised form 14 March 2016
Accepted 8 May 2016
Available online xxx
PACS:
44.35.+c
46.15.x
81.05.Bx
81.20.Ev
81.40.Lm

a b s t r a c t
Minimizing the residual stress build-up in metal-based additive manufacturing plays a pivotal role in
selecting a particular material and technique for making an industrial part. In beam-based additive manufacturing, although a great deal of effort has been made to minimize the residual stresses, it is still elusive
how to do so by simply optimizing the manufacturing parameters, such as beam size, beam power, and
scan speed. With reference to the Ti6Al4V alloy and manufacturing by electron beam melting, we perform
systematic nite element modeling of one-pass scanning to study the effects of beam size, beam power
density, beam scan speed, and chamber bed temperature on the magnitude and distribution of residual
stresses. Our study elucidates both qualitative and quantitative features of the residual stress elds originated by these manufacturing parameters. Our ndings can serve as useful guidelines for engineers and
designers to deal with residual stress build-up during additive manufacturing of Ti6Al4V.
2016 Elsevier B.V. All rights reserved.

Keywords:
Residual stress
Electron beam melting
Ti6Al4V
Additive manufacturing
Powder metallurgy

1. Introduction
Successful transition from rapid prototyping to additive manufacturing is an exciting challenge that requires careful control of the
integrity, mechanical properties, and geometrical accuracy of the
parts to be put into service. For the case of metallic materials and
beam-based additive manufacturing, this challenge is particularly
important because of the residual stresses that emerge at the manufacturing step [13]. Therefore, the quantitative study of residual
stress formation, and the inuence of the additive process parameters on those stresses, becomes a key step in the design work-ow.
Powder-based additive manufacturing is a technique where a laser
or an electron beam selectively scans and melts a thin layer of powder in order to reproduce the geometrical cross-section of the part
to be built. Advancing in a layer-by-layer fashion, after the scan of
the existing layer is completed, a new layer of powder is deposited
on top of it, and then the build of a new cross-section begins. The

Corresponding authors.
E-mail addresses: zhangg@ihpc.a-star.edu.sg (G. Zhang),
peiqx@ihpc.a-star.edu.sg (Q.X. Pei).

process continues until every cross-section is scanned and the part


is complete. Scanning is performed by a laser or an electron beam,
where power is supplied by a coherent light or electron current
source, respectively. Because the intensity prole of both beams is
Gaussian (TEM00 mode), the beam is characterized by the beam spot
radius. This is the radius at which the intensity has decreased by 1/e
times the maximum value. For the case of laser beams, the spot size
is controlled by optical focus lenses. For the case of electron beam,
on the other hand, the spot size is adjusted by the so-called focus
offset as a correction to the electron beam current that effectively
acts in controlling the beam spot size [4]. The beam is highly focused
and, while each machine supplier offers slightly different ranges for
the beam size, diameters are usually in the range of 200500 m
[1]. Because the temperatures required to melt a metal are very
high, signicant thermal gradients originate around the melt pool
[5]. Moreover, during solidication of the molten metal, the upper
layer is in a hotter, softer state with respect to the lower layer,
which is in a colder, stiffer state. Therefore, as the material on top
cools down and shrinks, plastic residual stresses develop because
the thermal strains exceed the yield point of the material, and are
mainly controlled by the magnitude of the thermal gradients in the
solidied metal [6].

http://dx.doi.org/10.1016/j.addma.2016.05.010
2214-8604/ 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model

ARTICLE IN PRESS

ADDMA-93; No. of Pages 9

G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

Residual stresses play a crucial role in manufacturing. For example, it is well-known that tensile residual stresses are able to
decrease the fatigue life of a component, because they can provide
additional driving force for crack initiation and propagation [7].
Moreover, they also induce loss of geometrical accuracy which,
for the case of additive manufacturing, can create mistting issues
among different components [8]. For these reasons, tremendous
efforts have been placed in order to understand and control residual
stress formation during additive manufacturing. Several materials
and techniques have been studied, including electron beam melting (EBM) of Ti6Al4V [911], laser engineered net shaping (LENSTM )
of Ti6Al4V [11] and steel [5], EBM [12] and selective laser melting
[13,14] of nickel-based superalloys, and wire arc additive manufacturing (WAAM) [15,16]. Moreover, modeling has been useful
in investigating machine setups that could reduce residual stress
and part deformation. Aggarangsi and Beuth [6] modeled the effect
of a secondary beam during thin-wall laser manufacturing of 304
stainless steel, showing the smaller effect of a secondary beam
in reducing the stresses compared to uniform heating. Bai et al.
[17] studied the effect of induction heating before or after scanning during weld-based additive manufacturing using mild steel,
revealing a positive effect of induction heating in reducing stresses.
However, since the layer-by-layer manufacturing is a complex process involving the coupled effect of the beam heat transfer, the
phase transformations between powder, liquid and solid [9], and
temperature-dependent material plasticity, understanding residual stress distribution necessarily requires the qualitative and
quantitative understanding of residual stress buildup during each
beam scan.
In this paper, we perform nite element method (FEM) calculations of single-scan electron beam melting of Ti6Al4V with the aim
of understanding how residual stresses are related to the process
parameters, including beam size, beam power, scan speed, and bed
pre-heating temperature. We begin by validating our model with
existing experimental and numerical studies of the same material and additive technique, and then show the residual stress eld
under nominal processing conditions. Subsequently, we proceed
by altering the process parameters and show how the residual
stresses are affected by the change. Ultimately, our modeling provides comprehensive quantitative understanding of residual stress
formation in powder-based manufacturing and suggests guidelines
for designers to control or minimize them.
2. Method, model validation and work plan
A nite element method (FEM) framework was designed based
on the Abaqus commercial software where the specic features of
additive manufacturing were implemented through user subroutines. The beam was modeled as a Gaussian heat source traveling
across the domain, with absorption prole tailored to the electron
beam melting technique. For a beam traveling along the x direction,
the heat ux (energy/volume/time) is given by:

2AiV

y, z) =
e
q(x,
2 d0

2((xvt)2 +y2 )
2

 

1
z
3
5
d0

2

z
2
+5
d0

(1)

where i is the beam current, V is the beam voltage, is the beam


spot radius, d0 is the penetration depth, v is the scan speed and A is
the powder absorptivity. In our calculations, the values i = 14 mA,
V = 60 kV, d0 = 28 m were used [9], while , v and A were varied
during different calculations as shown in detail in the next section.
Temperature boundary conditions were assigned to keep the lateral
as well the bottom boundaries at the xed temperature of 650 C,
consistent with EBM manufacturing [18]. In particular, the choice
of 650 C was made to take into account the preheating scan that
precedes each layer scan in EBM. While it is difcult to precisely

identify the temperature of the powder bed after the preheating


scan, values of 650700 C are common in EBM [4], as measured
by the thermocouple below the start plate. The boundary temperature was kept at this value in each calculation except where the
pre-heating temperature was systematically varied. Boundary conditions for the solid mechanics problem were set as the following:
the right and left boundary nodes were xed in the x coordinate,
and free on y and z (see the axis labeling in Fig. 3), while the front
and back boundary nodes were xed in the y coordinate and free
in x and z. Moreover, the bottom boundary nodes were xed in all
coordinates. Similar to the previous work on selective laser melting of stainless steel [19], a multi-phase model was implemented
for Ti6Al4V. Here, the phases of powder, liquid, and solid were
explicitly considered in the FEM model, where material properties
depend on both temperature and phase. In particular, the specic
heat, thermal conductivity and the mechanical properties of the
bulk and powder material were set by following Arce [20]. In our
continuum framework, powder and liquid Ti6Al4V were modeled
as a perfectly plastic materials with negligible yield strength as
compared to the bulk metal, to accommodate for the fact that the
powder is loose and the liquid is perfectly deformable. In particular,
a small yield stress of 0.1 MPa was chosen for our calculations so
that the FEM solver could reach convergence. Notice that, in actual
EBM manufacturing, the preheating scan induces a certain degree of
powder sintering and consolidation within the powder. However,
in our experience, the pre-heated powder is still less bind compared
to the bulk metal. As a result, a low value for the yield stress of the
powder seems appropriate to model this phase. The density  of
each phase was taken as bulk = 4430 kg/m3 , powder = 2700 kg/m3 ,
and liquid = 4133.19 kg/m3 [20]. The thermal expansion coefcient
for bulk Ti6Al4V was taken from Ref. [21], while that for liquid and
powder phases was set to zero since these phases were already perfectly plastic. Latent heat of fusion Hf was set to 289,000 J/kg [21],
with solidus and liquidus temperatures being 1605 C and 1650 C,
respectively [21]. Emissivity of Ti6Al4V was set to 0.35 [22]. Overall,
Fig. 1 summarizes the material properties that were implemented
in this study while the thermal and solid mechanics equations that
describe the problem are the following. The thermal problem is
described by the equation
(, T )Cp (, T )

dT
= q(r, t) + Q (r, t)
dt

(2)

where (, T) is the density, which is a function of the local phase


 and temperature T, Cp (, T) is the specic heat capacity, q(r, t)
is the heat ux vector, r = (x, y, z) is the coordinate vector, Q is the
heat source and t is time. Notice that the heat ux follows from the
temperature eld as q = k T where k is the thermal conductivity.
In Eq. (2), the heat source term is given by Eq. (1) when multiplied by the time step. In our simulations, the time step was 3 s.
Equilibrium for the solid mechanics problem was given by

 = 0

(3)

where  is the stress. Here, because the problem is elasto-plastic


and also involves thermal strains, the total strain could be written
as = e + pl + T , where e , pl and T are the elastic, plastic and
thermal strains, respectively. The mechanical constitutive law for
the elastic problem was given by  = Ce , where C is the material
stiffness tensor. Plasticity was modeled with isotropic hardening
where yield stress was related to true plastic strain t,pl as shown
in Fig. 1(c). In particular, the nominal strain was converted to
true strain by t = ln (1 + ), while nominal stress was converted
to true stress by t =  (1 + ). In this manner, true plastic strain
was obtained as t,pl = t Et , where E is the Youngs modulus
(E = 113 GPa at 30 C, and the full dependence of E upon temperature was taken from Ref. [21]). Moreover, the increment in thermal
strain dT was given by dT = dT, where = (, T) is the thermal

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model
ADDMA-93; No. of Pages 9

ARTICLE IN PRESS
G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

Fig. 1. Temperature and phase-dependent material properties implemented in the model. (a) Specic heat-temperature; (b) thermal conductivity-temperature; (c) true
stresstrue plastic strain; (d) linear thermal expansion coefcient-temperature. In (c), the true stress for powder, liquid, and solid above 950 C is shown by horizontal yellow
line at 0.1 MPa. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

expansion coefcient and dT is the change in temperature during


the time step dt.
2.1. Model validation
A
FEM
domain
of
size
2 mm 1.5 mm 0.65 mm
(length width height) was devised using linear, 4-integration
points brick elements. The top surface was allowed to exchange
heat with the surrounding through radiation, but not through
convection because the manufacturing chamber is in vacuum
[18]. Specically, radiation was implemented using the equation
q =  ( 4 04 ), where q is the heat ux,  is the StefanBoltzmann
constant,  is the surface emissivity, is the local temperature
and 0 is the chamber temperature. Because the powder bed is a
non-concave surface and the chamber walls, including the heat
shields are relatively distant from the part, this formulation is
a reasonably good approximation for the actual heat exchange
between the powder bed and the chamber. Our modeling was
compared to that of Jamshidinia et al. [9] in order to accurately
take into account the effect of the uid ow within the melt pool.
In particular, the same process parameters (scan speed, beam
size, beam power) as in Ref. [9] were employed. This comparison
allowed us to nd effective beam parameters and latent heat
ux to accurately account for the uid effects within the melt
pool, even if our calculations were purely thermo-mechanical
without explicitly considering the uid dynamics of the melt pool.
Specically, our calculations were tted to reproduce the melt
pool length, width, and ratio between the length of the mushy
zone and length of the melt pool. Fig. 2 illustrates this tting
as compared to the results by Jamshidinia et al. [9]. Moreover,
this approach was chosen in order to reduce the computational

demand of the calculations, which was already high because of the


coupled thermo-mechanical scheme required to compute residual
stresses. Comparison with Ref. [9] allowed us to nd that the uid
effects in the melt pool can be reproduced using an asymmetric
beam shape, in particular, a beam with elliptic shape. Specically,
the asymmetric beam was modeled with Eq. (4) where x and y
are the beam radius along and perpendicular to the scan direction,

Fig. 2. Validation of our modeling in terms of melt pool length and width (a), as
compared to the calculations of Jamshidinia et al. [9] (b), for the same processing
conditions of 0.5 m/s scan speed, i = 14 mA, V = 60 kV. In (a), the colors represent the
local phase, in particular red means liquid, yellow means mushy, light blue means
solid and dark blue means powder. The simulation domain size for this calculation
was 2 mm 1.5 mm 0.65 mm (length width height), where Dirichlet boundary
conditions of 650 C were imposed at the outer boundaries of the domain (see Section 2.1 for details). In (b), the contour of the melt pool (i.e., where the temperature
is above the liquidus) is highlighted in red. The white arrow shows the scan direction. Panel (b) is reprinted with permission from ASME. (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of
this article.)

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model

ARTICLE IN PRESS

ADDMA-93; No. of Pages 9

G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

Table 1
Parameters for including melt-pool uid effects in a purely thermo-mechanical calculation.
Scan speed (mm/s)

Nominal beam size


(m)

Actual beam size


along scan speed
(m)

Actual beam size


perpendicular to
scan speed (m)

Powder absorptivity

Effective latent
heat

100
500

200
200

90
2300

325
360

0.080
0.203

0.2252 * Hf
0.5855 * Hf

respectively, where the Gaussian characteristic of the beam was


retained along each direction.


2
2AiV

q(x,
y, z) =
e
x y d0

(xvt)2
2
x

y2

+ 2

1
z
3
5
d0

2

z
2
+5
d0

(4)

Moreover, the correct amount of heat ow from the melt pool to


the solid was reproduced by using an effective latent heat, adopting
the strategy of Ref. [23]. The parameters that were obtained for the
elliptic beam, as well as the effective latent heat and the powder
absorptivity at different scan speeds are reported in Table 1. After
this step, our simulations were further compared to the experiments of Yadroitsev et al. [22] in terms of the surface temperature
of the melt track. Our computed surface temperature was found to
be around 2300 C, in good agreement with the experimental measurement [22]. Specically, this temperature value was obtained
using the parameters v = 0.5 m/s, i = 14 mA, V = 60 kV and the beam
size and powder absorptivity as in Table 1 second line. Indeed,
the surface temperature is difcult to measure experimentally, and
there is discrepancy in existing reports [22,24]. Here, the data by
Yadroitsev et al. [22] was used because these data are closest to
what we are simulating. The surface temperature of typical scanning, using a beam scan speed of 0.5 m/s and our asymmetric beam
parameters, is shown in Fig. 2. After the satisfactory comparison
with existing literature, including with experimental results, our
model was considered validated and the investigation of residual
stresses was initiated.
Each simulation was designed such that a single scan was made
and then the beam was turned off to allow cooling of the material
to the powder pre-heating temperature. In particular, each simulation was designed as depicted in Fig. 3 as a half-layer of powder
placed side-by-side with an already scanned material, sitting on top
of bulk Ti6Al4V. By performing one-pass scan on such domain, our
simulation addresses the addition of one scan segment. Here, the
heat coming from the previous raster scan (of the already solidied
material) was not considered. As a result, our simulations address
specically the low-scan speed, short scan-segment regime, where
heat coming from the previous scan track plays a smaller role
in the melting of the current track [4]. Powder layer thickness
was set to 50 m [18] and mesh size was particularly ne within
this layer, with each element size of 62.5 m 62.5 m 12.5 m
(length width height). After cooling to 650 C was completed,
residual stresses were extracted by computing the stress prole
along the vertical line as shown in Fig. 4, moving from the free
surface into the bulk material at the center of the domain. These vertical plots were chosen because they closely resemble the data that
can be experimentally measured using for example the central-hole
drilling method [25].
A rst calculation was run using nominal process parameters
of scan speed of 500 mm/s, nominal beam size of 200 m, and
boundary temperature of 650 C, allowing us to elucidate the
qualitative and quantitative features of the stress eld that will be
discussed in detail in the next section. Subsequently, the process
parameters were systematically varied in a range centered around
the nominal values, in order to avoid processing parameters that
would result in signicant defects such as porosity and overmelting [26]. In particular, the beam size was varied systematically

Fig. 3. Illustrative snapshots of the one-pass scan simulation, where the domain is
made of bulk Ti6Al4V with an extra half-layer already built (light blue), where the
remaining half layer is powder ready to be processed (dark blue). See Section 2.1
for the complete description of the domain, mesh size and boundary conditions.
Panel (a) shows the initial condition; panel (b) shows a snapshot half-way during
scanning; panel (c) shows the nal state where the beam is off and new material
has solidied. In (b), red is the melt pool and yellow is the mushy zone. The arrow
in (b) shows the scan direction. (For interpretation of the references to color in this
gure legend, the reader is referred to the web version of this article.)

in the range 40% to +40% around the nominal value. In these


calculations, the beam power was also changed in order to keep
the beam energy density constant. Specically, by considering the
beam volume

V =

x2
2
x

exp

y2

+ 2

d0

x2
x2

d0

2

exp

z
+ 5 d

d0

dx

 z 2

3
0

1
z
3
5
d0

z
2
+5
d0

y2
y2

dz =

dy

3

x y d0 ,
2
5

both x and y were multiplied by a factor f , f [0.6 : 0.8],


while the power was multiplied by a factor Pf = f2 to maintain

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model
ADDMA-93; No. of Pages 9

ARTICLE IN PRESS
G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

the ratio P/V constant. This was necessary to ensure that powder
melting was achieved through the entire powder layer, independently of the processing conditions [19]. Using this approach,
the effect of beam size was studied for scan speed of 0.5 m/s and
0.1 m/s. Moreover, at scan speed of 0.5 m/s, the effect of the beam
power was studied by performing a set of calculations using a
beam energy density increased by 20%. Finally, the effect of bed
pre-heating temperature was investigated by changing the xed
temperature boundary conditions in the range 650900 C. Notice
that, after the annealing step is completed and the FEM domain
is thermalized to constant temperature, further slow cooling to
room temperature does not alter the residual stress distribution
as long as the thermal gradients from cooling are sufciently
low so as to not induce further plastic deformation in the part.
Indeed, post-build cooling times of 25 h are necessary to avoid
further increase of residual stress due to excessively fast cooling.
Therefore, all our calculations are quantitatively comparable with
experimental measurements done at room temperature.

3. Origin of residual stress

Fig. 4. Calculated residual stress at the end of the simulation, after the beam has
been switched off and the temperature has equilibrated to the processing temperature (see text for details). Panels (a) and (b) show the color map of the s11 and s22
stress components, respectively, at a vertical midplane cutting through the simulation domain. With reference to Fig. 3, these panels show the consolidated material
below and to the side of the new scan track. Panels (c) and (d) show the line plot of
the same stress components as previously mentioned, along the vertical line shown
in panel (a). (For interpretation of the references to color in this gure legend, the
reader is referred to the web version of this article.)

We begin by commenting on the scan using nominal processing


parameters. Fig. 4(a) and (b) shows the color map of the  11 and  22
stress components, where x = 11 is the scan direction and y = 22
is the horizontal direction perpendicular to it. A region of high stress
is found surrounding the original scan track, both in depth and
width. We dene this region as the heat affected zone (HAZ), with
both depth and width of approximately 200 m. Vertical stress
proles are shown in Fig. 4(c) and (d), where we can see that
the stress is in initially tension, then compression, and ultimately
vanishes as the bottom of the simulation domain is approached.
These features conrm the thermal origin of the residual stress
[1], where the hotter topmost region of the domain shrinks on
top of the colder region beneath it. In fact, taking for example a
yield stress  y = 150 MPa at 700 C for Ti6Al4V, and a Youngs modulus Y = 74 GPa [21], yield occurs at a strain p = 0.00255. Using
a thermal expansion coefcient = 10 106 /K, thermal strains
achieve plastic deformation if the material is subject to differences
in temperature of T = / 255 C. Assuming this difference in
temperature occurs within one powder layer of thickness 50 m,
this temperature difference corresponds to a thermal gradient of
5.1 104 K/cm. Thermal gradients of this order of magnitude
are indeed the norm in beam-based additive manufacturing of
Ti6Al4V [27] and conrm the thermal origin of the observed residual stresses. These estimates are also consistent with the work of
Vasinonta et al., who studied the correlation between yield and
thermal gradients in LENSTM manufacturing [5].
Further inspection of Fig. 4 shows that the highest stresses are
recorded at or near the free surface, in agreement with experimental measurements in laser-based additive manufacturing [25].
Quantitatively, the panels show that the stress reaches 150 MPa
in the HAZ, which is consistent with post-build measurements of
Ti6Al4V EBM components [28]. Moreover, it is seen that, while the
 11 component is relatively uniform in the HAZ, and then decays
quickly outside of it, the  22 component shows a different behavior.
In particular,  22 reaches its maximum value at the HAZ boundary, while it decreases toward the free surface. Notice that, because
material is being added to an already solidied part, heat conduction and thermal gradients are not symmetric with respect to the
scan direction. In particular, as shown in Fig. 5, thermal gradients
are milder toward the bulk material, and steeper toward the powder. This feature follows from the higher thermal conductivity of
the bulk metal compared to the powder material. The anisotropic
shape of the melt pool could also contribute to this effect, as for
example noted in similar simulations done by Dai and Shaw [29].

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model
ADDMA-93; No. of Pages 9

ARTICLE IN PRESS
G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

Fig. 5. (a) Snapshot of the scan vector with temperature data together with the lines
used to extract the thermal gradients. (b) Transverse thermal gradients along the
three line proles of panel (a); notice the asymmetry between right and left side of
the scan, due to the higher thermal conductivity of the bulk metal compared to the
powder. Here, Line prole 1 cuts across the melt pool; Line prole 2 and Line
prole 3 are positioned 400 m and 750 m behind Line prole 1, respectively. (c)
Thermal gradients along the scan direction. Notice the region behind the melt pool
where gradients stabilize, as the result of solidication (mushy zone) where latent
heat is released.

3.1. Various effects on residual stress


After elucidating the stress eld with nominal processing
conditions, systematic calculations were performed starting with
altering the beam size. Here, the beam size was scaled by a factor
f and the beam power by a factor Pf from the respective nominal
values. Results are compiled in Fig. 6(a), (b) and (c) for the  11 ,  22
and Von Mises stress respectively. These calculations show that,
interestingly, the in-plane beam size affects the depth prole of
the residual stress. We notice that small beam sizes give higher
stresses within a smaller HAZ, while larger beams have a more

Fig. 6. Effect of beam size on residual stress prole at a scan speed of 0.5 m/s. With
respect to nominal processing conditions of = 200 m and i = 14 mA, the beam size
is scaled by a factor of f and the power by a factor of Pf . Notice that the power is
P = Vi, where V is constant at 60 kV. The residual stress components shown are the
s11 , s22 and the Von Mises stress in panels (a), (b), (c) respectively. The panels show
that the effect of beam size is to alter the depth of the heat affected zone (panels (a)
and (b)) as well as inuence the qualitative behavior of the Von Mises stress prole
(c).

uniform stress within a larger HAZ. Considering the  22 component, we notice a similar trend of increased HAZ depth with beam
size, coupled with an increase in maximum stress from 110 MPa
at small beams to 150 MPa at large beams. In particular, a deeper
HAZ means that the thermal gradients are sufcient to generate
yield at a larger depth from the free surface. This effect suggests
that the increase in thermal gradients with beam size is more
pronounced along this direction in comparison to the x direction.
Notice that, in panel (b) at f = 0.6, Pf = 0.36, the stress at the surface

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model
ADDMA-93; No. of Pages 9

ARTICLE IN PRESS
G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

Fig. 7. Systematic calculations performed at a beam energy density 20% higher as


compared to the results of Fig. 6, for the s11 , s22 stress components as well as for the
Von Mises stress in panels (a), (b) and (c) respectively. The effect of an increased
energy density can be noticed by comparison with Fig. 6, showing the quantitative
increase in size of the heat affected zone with the beam energy density.

Fig. 8. Systematic calculations showing the effect of scan speed on the residual
stress prole. Here, a scan speed of 0.1 m/s was used, as compared to a scan speed
of 0.5 m/s in Fig. 4. Panels (a), (b) and (c) show the s11 , s22 , and Von Mises stress
components along the vertical line shown in Fig. 4, respectively.

was found to be negative. Given that these parameters correspond


to a very small beam, sufcient to melt only few elements, we
cannot rule out possible numerical inaccuracies in the solution,
due to the large variations in mechanical properties between the
powder and the bulk (see Fig. 1 for reference). Finally, inspection
of the Von Mises stress (Fig. 6(c)) suggests that increasing beam
size induces a transition from sharp to uniform overall stresses. In
fact, stress at small beams (f = 0.6) continually decreases, while
at large beams (f = 1.4) it is mostly uniform within the HAZ. This
nding suggests a possible route toward the design and control
of residual stress, because it shows how the stress prole can be
controlled by choosing the appropriate value of beam size.

Next, we turn our attention to the effect of beam power. Here,


a power increased by 20% was used for each beam size previously
considered. Results are compiled in Fig. 7. The calculations show
that, while the qualitative features of the stress eld are similar
to those revealed in Fig. 6, the quantitative effect of an increased
beam power is to shift the curves toward deeper regions. This effect
translates into larger HAZ zones which, in turn, are consistent with
the view that higher beam powers have larger effect on the heat
transport during scanning. Quantitatively, Fig. 7 suggests that a 20%
increase in beam power induces an increase of 15% in the HAZ size.
The effect of scan speed was investigated by running systematic
calculations at a scan speed of 0.1 m/s and using the parameters

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model
ADDMA-93; No. of Pages 9
8

ARTICLE IN PRESS
G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

a result, thermal gradients, while similar in magnitude, occur at a


greater depth compared to higher scan speed, therefore resulting
in a deeper HAZ.
Overall, Figs. 68 have revealed the effect of beam size, beam
power, and scan speed in the HAZ, however, they have also shown
that the residual stress is not signicantly reduced by any particular choice of these parameters. At the same time, reduction
in residual stresses is of paramount importance for long fatigue
life of service parts, as we have discussed in Section 1. To further
search for parameters that could reduce residual stress, systematic calculations were performed by varying the bed pre-heating
temperature and using the nominal beam parameters at 0.5 m/s
scan speed. Results are compiled in Fig. 9 and show that signicant
reduction in residual stress can be achieved as the bed pre-heating
temperature is increased. Quantitatively, each increase by 50 C in
pre heating temperature showed a stress reduction by 20%. This
effect arises from the fact that thermal gradients are lower when the
powder-bed temperature is higher. Moreover, the material properties are temperature-dependent, in particular, yield corresponds
to lower stress at higher temperatures (see Fig. 1(c)). In this regard,
our calculations well support those by Aggarangsi and Beuth [6] as
well as those of Bai et al. [17].
4. Conclusions
In conclusion, a systematic computational study of residual
stress formation was performed for single-track electron beam
melting of Ti6Al4V. After careful matching of our modeling to published experimental data, residual stress was studied for nominal
process parameters, showing the existence of a well-dened heat
affected zone surrounding the scan area. Later, starting from the
nominal parameters, calculations were performed to study the
effect of beam size, beam power density, scan speed and bed preheating temperature in altering the stress eld. The simulations
have revealed that control of the HAZ size can be achieved by choosing the appropriate beam size. At the same time, the beam power
density was found to positively correlate with the HAZ size, even
at values within the limits of good powder processability. With
respect to the effect of the scan speed, our simulations showed that
a deeper HAZ results from lower scan speeds, while smaller HAZ
result from higher scan speeds. Finally, the bed pre-heating temperature was found to have the largest quantitative impact on the
residual stress, with lower stresses recorded at higher bed temperatures. Considered all together, our calculations can represent a valid
reference for designers and engineers when the need of understanding and control of residual stress in additive manufactured
Ti6Al4V arises.
Fig. 9. Residual stress prole computed at scan speed of 0.5 m/s and otherwise
nominal parameters as in Fig. 4, but applying different initial and boundary bed
temperatures. Because material plasticity is highly temperature-dependent (see
Fig. 1(c)), higher bed temperature corresponds to lower residual stress. Overall,
the gure shows the qualitative and quantitative effect of bed temperature on the
residual stress. Panels (a), (b) and (c) represent the s11 , s22 , and Von Mises stress
components along the vertical line shown in Fig. 4, respectively.

in the second row of Table 1. Results are shown in Fig. 8(a)(c) for
the  11 ,  22 and  Mises residual stress components, respectively. By
comparison with Fig. 6, the effect of scan speed can be understood.
In particular, the curves in Fig. 8 appear located at greater depth
compared to the same curves in Fig. 6, showing that a lower scan
speed has the effect to deepen the heat affected zone. At the same
time, the overall highest values of stress seem not to be affected by
the value of scan speed. Overall, this picture is consistent with the
interpretation that a lower scan speed allows more time for heat
conduction away from the melt pool and into the bulk metal. As

Acknowledgements
G.V. acknowledges helpful discussions with W. Huang at A*STAR
IHPC. Useful discussions with S.M.L. Nai, P. Wang, S. Raghavan,
W.J. Sin at A*STAR Singapore Institute of Manufacturing Technology (SIMTech), and X. Tan, Y. Kok, Prof. S.B. Tor at Nanyang
Technological University are also acknowledged. This work was
supported by the Agency for Science, Technology and Research of
Singapore through the Industrial Additive Manufacturing Program
(SERC grants 132 550 4103 and 132 550 4106).
References
[1] P. Mercelis, J.-P. Kruth, Residual stresses in selective laser sintering and
selective laser melting, Rapid Prototyp. J. 12 (2006) 254265.
[2] P. Withers, H. Bhadeshia, Residual stress. Part 2 nature and origins, Mater.
Sci. Technol. 17 (2001) 366375.

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

G Model
ADDMA-93; No. of Pages 9

ARTICLE IN PRESS
G. Vastola et al. / Additive Manufacturing xxx (2016) xxxxxx

[3] D. Hofmann, S. Roberts, R. Otis, J. Kolodziejska, R. Dillon, J.O. Suh, A. Shapiro,


Z.-K. Liu, J.-P. Borgonia, Developing gradient metal alloys through radial
deposition additive manufacturing, Sci. Rep. 4 (2014) 5357.
[4] Arcam User Manual, Arcam AB, 2011.
[5] A. Vasinonta, J. Beuth, M. Grifth, Process Maps for Controlling Residual Stress
and Melt Pool Size in Laser-Based SFF Processes, 2000, pp. 200208.
[6] P. Aggarangsi, J. Beuth, Localized preheating approaches for reducing residual
stress in additive manufacturing, SFF Proc. (2006) 709720.
[7] L. Wagner, Mechanical surface treatments on titanium, aluminum and
magnesium alloys, Mater. Sci. Eng. A 263 (1999) 210216.
[8] M. Zh, S. Lutzmann, Modelling and simulation of electron beam melting,
Prod. Eng. Res. Dev. 4 (2010) 1523.
[9] M. Jamshidinia, F. Kong, R. Kovacevic, Numerical modeling of heat
distribution in the electron beam melting of Ti-6Al-4V, J. Manuf. Sci. Eng. 135
(2013) 061010.
[10] D. Pal, N. Patil, B. Stucker, Prediction of mechanical properties of electron
beam melted Ti6Al4V parts using dislocation density based crystal plasticity
framework, SFF Proc. (2012) 507525.
[11] J. Heigel, P. Michaleris, E. Reutzel, Thermo-mechanical model development
and validation of directed energy deposition additive manufacturing of
Ti-6Al-4V, Addit. Manuf. 5 (2015) 919.
[12] P. Prabhakar, W. Sames, R. Dehoff, S. Babu, Computational modeling of
residual stress formation during the electron beam melting process for
Inconel 718, Addit. Manuf. 7 (2015) 8391.
[13] M. Labudovic, D. Hu, R. Kovacevic, A three dimensional model for direct laser
metal powder deposition and rapid prototyping, J. Mater. Sci. 38 (2003)
3549.
[14] M. Matsumoto, M. Shiomi, K. Osakada, F. Abe, Finite element analysis of single
layer forming on metallic powder bed in rapid prototyping by selective laser
processing, Int. J. Mach. Tools Manuf. 42 (2002) 6167.
[15] J. Ding, P. Colegrove, J. Mehnen, S. Williams, F. Wang, P.S. Almeida, A
computationally efcient nite element model of wire and arc additive
manufacture, Int. J. Adv. Manuf. Technol. 70 (2014) 227236.
[16] J. Gockel, J. Beuth, K. Taminger, Integrated control of solidication
microstructure and melt pool dimensions in electron beam wire feed additive
manufacturing of Ti-6Al-4V, Addit. Manuf. 1 (2014) 119126.

[17] X. Bai, H. Zhang, G. Wang, Modeling of the moving induction heating used as
secondary heat source in weld-based additive manufacturing, Int. J. Adv.
Manuf. Technol. 77 (2015) 717727.
[18] X. Tan, Y. Kok, Y.J. Tan, G. Vastola, Q. Pei, G. Zhang, Y.-W. Zhang, S. Tor, K.
Leong, C. Chua, An experimental and simulation study on build thickness
dependent microstructure for electron beam melted Ti-6Al-4V, J. Alloys
Compd. 646 (2015) 303309.
[19] G. Vastola, G. Zhang, Q. Pei, Y.-W. Zhang, Modeling and control of remelting in
high-energy beam additive manufacturing, Addit. Manuf. 7 (2015) 5763.
[20] A.N. Arce, Thermal modeling and simulation of electron beam melting for
rapid prototyping on Ti6Al4V alloys (Ph.D. thesis), North Carolina State
University, 2012.
[21] G. Welsch, R. Boyer, E. Collings, Alpha-beta alloys, in: Materials Properties
Handbook: Titanium Alloys, ASM International, 2007, pp. 516.
[22] I. Yadroitsev, P. Krakhmalev, I. Yadroitsava, Selective laser melting of Ti6Al4V
alloy for biomedical applications: temperature monitoring and
microstructural evolution, J. Alloys Compd. 583 (2014) 404409.
[23] S. Koric, B. Thomas, V. Voller, Enhanced latent heat method to incorporate
superheat effects into xed-grid multiphysics simulations, Numer. Heat
Transf. B 57 (2010) 396413.
[24] A. Klassen, T. Scharowsky, C. Krner, Evaporation model for beam based
additive manufacturing using free surface lattice Boltzmann methods, J. Phys.
D: Appl. Phys. 47 (27) (2014) 275303.
[25] C. Casavola, S. Campanelli, C. Pappalettere, Experimental analysis of residual
stresses in the selective laser melting process, in: XIth Int. Congress Proc. of
the Society Exp. Mech., 2008.
[26] C. Guo, W. Ge, F. Lin, Effects of scanning parameters on material deposition
during electron beam selective melting of Ti-6Al-4V powder, J. Mater. Process.
Technol. 217 (2015) 148157.
[27] S. Bontha, N. Klingbeil, Thermal process maps for controlling microstructure
in laser-based solid freeform fabrication, SFF Proc. (2003) 219226.
[28] P. Edwards, A. OConner, M. Ramulu, Electron beam additive manufacturing of
titanium components: properties and performance, J. Manuf. Sci. Eng. 135
(2013) 061016.
[29] K. Dai, L. Shaw, Thermal and mechanical nite element modeling of laser
forming from metal and ceramic powders, Acta Mater. 52 (2004) 6980.

Please cite this article in press as: G. Vastola, et al., Controlling of residual stress in additive manufacturing of Ti6Al4V by nite element
modeling, Addit Manuf (2016), http://dx.doi.org/10.1016/j.addma.2016.05.010

Вам также может понравиться