Computational Materials Science 126 (2017) 360–372

Contents lists available at ScienceDirect

Computational Materials Science

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

An improved prediction of residual stresses and distortion in additive

manufacturing
T. Mukherjee a, W. Zhang b, T. DebRoy a,⇑
a
Department of Materials Science and Engineering, The Pennsylvania State University, University Park, PA 16802, USA
b
Department of Materials Science and Engineering, The Ohio State University, Columbus, OH 43221, USA

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

Article history: In laser assisted additive manufacturing (AM) an accurate estimation of residual stresses and distortion is
Received 6 September 2016 necessary to achieve dimensional accuracy and prevent premature fatigue failure, delamination and
Received in revised form 30 September buckling of components. Since many process variables affect AM, experimental measurements of residual
2016
stresses and distortion are time consuming and expensive. Numerical thermo-mechanical models can be
Accepted 1 October 2016
Available online 20 October 2016
used for their estimation, but the quality of calculations depends critically on the accurate transient
temperature field which affects both the residual stresses and distortion. In this study, a well-tested,
three-dimensional, transient heat transfer and fluid flow model is used to accurately calculate transient
Keywords:
Additive manufacturing
temperature field for the residual stress and distortion modeling. The calculated residual stress
3D printing distributions are compared with independent experimental results. It is shown that the residual stresses
Laser deposition can be significantly minimized by reducing the layer thickness during AM. Inconel 718 components are
Fatigue failure found to be more susceptible to delamination than Ti-6Al-4V parts because they encounter higher
Thermo-mechanical modeling residual stresses compared to their yield strength.
Finite element analysis Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction in two steps in sequence. First, the transient temperature field in

the entire work piece is calculated. The computed temperature
The additive manufacturing (AM) process involves heating, results are then used for the mechanical calculations. Such sequen-
melting and solidification of an alloy by a moving heat source such tial calculations of temperatures and stresses make the computa-
as a laser or an electron beam in a layer by layer manner [1,2]. As a tions tractable but the accuracy of the calculations depends
result, different regions of the work piece experience repeated critically on the quality of the transient temperature field and
heating and cooling [2]. The spatially varied thermal cycles result the thermo-physical property data of the alloy.
in residual stresses and distortion in the additively manufactured As the laser or electron beam energy impinges on the work
components [3]. The residual stresses, whose magnitude can piece surface, the powder melts quickly to form a molten pool.
exceed the yield strength of the alloy, affect corrosion resistance, The highest temperature on the molten pool surface is attained
fracture toughness, crack growth behavior and fatigue perfor- directly below the heat source and the temperature decreases with
mance [4–8]. Moreover, the residual stresses are associated distance away from this location [16,17]. Inside the molten pool,
with pronounced deformations especially for thin-walled features the liquid alloy recirculates rapidly at very high velocities driven
[9–13]. Because AM involves many process variables as is the case by the spatial gradient of surface tension. The convective flow
with fusion welding, experimental measurements of stresses and mixes the liquid metal in different regions and enhances the trans-
strains are expensive and time consuming [4,14]. Moreover, port of heat within the molten pool. The circulation pattern
experimental measurements depend on the shape and size of the strongly affects the temperature distribution in the liquid alloy,
components, nature of the stresses measured, sample preparation heating and cooling rates, solidification pattern, and subsequently
and accuracy of X-ray or neutron diffraction [7,14]. A recourse is the evolution of various solid phases that make up the final
to undertake calculations of residual stresses and strains in all microstructure of the part [18,19].
locations of the work piece [15]. These calculations are often done Simulation of complex physical processes that affect the tem-
perature field is computationally intensive, and many of the previ-
ous calculations of the temperature field involved various
⇑ Corresponding author. simplifications and assumptions to make the calculations tractable.
E-mail address: [email protected] (T. DebRoy). These include several two-dimensional models [20,21], or an

http://dx.doi.org/10.1016/j.commatsci.2016.10.003
0927-0256/Ó 2016 Elsevier B.V. All rights reserved.
 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372 361

Table 1
Thermo-physical properties of Ti-6Al-4V and IN 718 [30].

Properties Ti-6Al-4V IN 718

Liquidus temperature (K) 1928 1609
Solidus temperature (K) 1878 1533
Thermal conductivity (W/m K) 1.57 + 1.6  102 T  1  106 T2 0.56 + 2.9  102 T  7  106 T2
Specific heat (J/kg K) 492.4 + 0.025 T  4.18  106 T2 360.24 + 0.026 T  4  106 T2
Density (kg/m3) 4000 8100
Viscosity (kg/m s) 4  103 5  103
dc/dT (N/m K) 0.37  103 0.26  103

assumption that the entire deposit is heated and then cooled [22], and liquid metal velocity fields. The model solves the following
or building a part by a single layer deposition [23]. In some equations of conservation of mass, momentum and energy
instances, heat sources have been simplified as surface flux to [18,19] in three dimensions.
achieve computational speed [3,23–25]. Another common difficulty
@ðqui Þ
is that the calculations ignore the convective heat transfer which is ¼0 ð1Þ
@xi
the main mechanism of heat transfer within the liquid, as discussed
previously [16–19]. This simplification can lead to the use of inac-

@ðquj Þ @ðquj ui Þ @ @uj
curate temperature field for thermo-mechanical calculations [26– þ ¼ l þ Sj ð2Þ
28], and the computed residual stress and strain fields do not
@t @xi @xi @xi
always agree well with the corresponding experimental data. where q is the density, ui and uj are the velocity components along
The errors in the transient temperature fields and heating and the i and j directions, respectively, and xi is the distance along the i
cooling rates resulting from heat conduction calculations that direction, t is the time, l is the effective viscosity, and Sj is a source
ignore the molten metal convection are well documented in the lit- term for the momentum equation. The energy conservation equa-
erature. Svensson et al. noted that ‘‘. . .the heat conduction equa- tion is:
tion has been found to be inadequate in representing

experimental cooling curves” [29]. Manvatkar et al. [16] showed @h @ðqui hÞ @ k @h @ DH @ðui DHÞ
q þ ¼ q q ð3Þ
that by ignoring the effect of convection, the cooling rates in addi- @t @xi @xi C p @xi @t @xi
tive manufacturing were over-estimated by about twice of the cor-
where h is the sensible heat, Cp is the specific heat, k is the thermal
rect values. Therefore, the temperature distribution calculated
conductivity, and DH is the latent heat content. Table 1 shows the
using heat conduction models without extensive experimental cal-
thermo-physical properties of the alloys used for the calculations.
ibrations is not accurate, which in turn, can adversely affect the
The temperature field as a function of time calculated from the
accuracy in calculations of residual stresses and distortion. What
heat transfer and fluid flow model is then imported to a mechani-
is needed and not currently available is a numerical model that cal-
cal model based on AbaqusÒ, a commercial finite element analysis
culates residual stress and strain fields from the transient temper-
(FEA) code [31]. The total strain increment (Detot lm ) with respect to
ature distribution considering convective heat transfer.
time is contributed by the several effects shown below:
Here we combine a well-tested three-dimensional transient
heat transfer and fluid flow model of additive manufacturing with Detot
lm ¼ Delm þ Delm þ Delm þ Delm
E P Th V
ð4Þ
a thermo-mechanical model to accurately calculate the tempera-
ture fields, residual stresses and distortion. The calculated temper- where DeElm , DePlm and DeTh
lm are the elastic, plastic and thermal strain

ature and residual stress distributions are tested using increments respectively. DeVlm is the strain induced due to the solid
independent experimental results. After validation, the model is state phase transformation and creep, which is assumed to be zero
used to quantitatively study the effect of a wide variety of AM vari- in the present model. The resulting stress increment estimated from
ables such as heat input and layer thickness on residual stresses the elastic strain as [5]:
and distortion. Although the results shown in this article are for
DrEij ¼ Dijlm  DeElm ð5Þ
direct energy deposition process, the findings will be useful to
make dimensionally compliant components and assess residual where Dijlm is the elastic stiffness matrix calculated from Young’s
stresses for all laser assisted powder based AM processes. modulus (E) and Poisson’s ratio (t) as,
 
E 1 m
2. Modeling Dijlm ¼ ðdij dlm þ dil djm Þ þ dij dlm ð6Þ
1þm 2 1  2m
2.1. Assumptions where d is a Dirac delta function [32] whose value is one only for
i = j and l = m, and is zero otherwise. Temperature-dependent plas-
Some simplified assumptions are made in both the heat transfer ticity with the von Mises yield criterion [5] is utilized to model the
and fluid flow model and the thermo-mechanical model. The den- flow stress and plastic strain. The thermal strain increment is calcu-
sities of the solid and liquid metals are assumed to be constant. The lated as:
surfaces of the deposited layers are considered to be flat. The loss
of alloying elements due to vaporization and its effects on both
DeTh
lm ¼ bdlm DT ð7Þ
the heat loss and composition change are not incorporated in the where b is the volumetric thermal expansion coefficient and DT is
present calculations. Finally, the effects of strains induced by the temperature increment. The temperature-dependent mechani-
solid-state phase transformation and creep are also neglected. cal properties used for the calculations for Inconel 718 and Ti-
6Al-4V are given in Tables 2 and 3, respectively. The step-by-step
2.2. Governing equations procedure for calculating temperature distribution, residual stres-
ses and distortion is illustrated in Table 4. A Python script was
A well tested, three dimensional, transient, heat transfer and developed to facilitate mapping the transient temperature fields
fluid flow model for AM [16,17] is used to compute temperature from the heat transfer and fluid flow model to the Abaqus-based
362 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372

Table 2
Temperature dependent mechanical properties of IN 718 [33].

Temperature (K) Young’s modulus (GPa) Temperature (K) Volumetric expansion co-efficient (/K) Temperature (K) Yield stress (MPa)
300 156.3 300 1.17E05 300 308.9
366.5 151.8 477.6 1.28E05 588.7 246.3
477.6 144.9 588.7 1.34E05 810.9 226.1
588.7 138 922 1.46E05 1033.2 207.7
699.8 131.4 1033.2 1.51E05 1255.4 114
810.9 124.7 1144.3 1.57E05
922 124 1366.5 1.66E05
1033.2 123.4 1672 1.66E05
1144.3 107.7 1900 1.42E05
1255.4 92.05 2400 1.08E05
1366.5 68.95 2700 9.47E06
1672 23.79 3200 7.84E06

Table 3
Temperature dependent mechanical properties of Ti-6Al-4V [34,35].

Temperature (K) Young’s modulus (GPa) Temperature (K) Volumetric expansion co-efficient (/K) Temperature (K) Yield stress (MPa)
300 125 300 8.78E06 300 955
533 110 533 9.83E06 573 836
589 100 589 1.00E05 773 732
700 93 700 1.07E05 1023 581
755 80 755 1.11E05 1073 547
811 74 811 1.12E05 1173 480
923 55 923 1.17E05 1273 405
1073 27 1073 1.22E05 1373 330
1098 22 1098 1.23E05
1123 18 1123 1.24E05
1573 12 1573 1.30E05
1873 9 1873 1.63E05

Table 4
Outline of the sequentially-coupled model combining the heat transfer and fluid flow model with the FEA model.

Computational model Inputs Outputs Description

3D transient heat Process parameters and temperature Temperature and velocity fields The model solves the equations of conservation of mass,
transfer and fluid dependent thermo-physical properties of momentum and energy in a 3D discretized solution
flow model the alloys domain consisted of the substrate, and deposited layers
Importing the nodes, Nodes, elements and temperature data from A temperature field data file A Python script that combines all nodes, elements and
elements and the heat transfer and fluid flow model (ODB) that can be imported in corresponding transient temperature data and generates
temperature data to the FEA solver for mechanical an ODB file that can be directly used in the Abaqus-based
the FEA model analysis FEA solver to calculate stress and strain fields
Abaqus-based FEA Nodes, elements, boundary conditions, Transient stress and strain fields Finite element solution of static force equilibrium
model for stress and transient temperature data, and equations in the discretized solution domain for the
strain calculations temperature-dependent mechanical temperature data calculated using the heat transfer and
properties of the alloys fluid flow model

FEA model. The script used the Abaqus Scripting Interface, an appli-
cation programming interface (API), to create an ODB file that con-
tained the transient temperature fields. Compared to the other
option using Abaqus user subroutine UTEMP to load temperature
fields into the stress model, the ODB file had the advantage of more
easily handling large dataset of temperature fields.

2.3. Heat source model and boundary conditions

The volumetric heat source term for laser beam used in the heat
transfer and fluid flow model is expressed by the following equa-
tion [16]:


Pd r2
Si ¼ ½ g þ ð1  g Þ g  exp d ð8Þ
pr2b k P P l
r2b
Fig. 1. Solution domain for the heat transfer and fluid flow and mechanical analysis.
where gp is the fraction of laser energy absorbed by the powder Due to the symmetry, a half of the solution domain is considered to reduce time and
during flight from nozzle to substrate, gl refers to the absorption computer memory requirements for the analysis. The two path along which
coefficient of the deposited layer, P is the laser power, d is the laser residual stress values are calculated are shown.
 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372 363

Table 5
Process parameters used for calculations.

Parameter set Laser power (W) Beam radius (mm) Scanning speed (mm/s) Layer thickness (mm) Substrate thickness (mm) Powder flow rate (g/s)
1 2000 1.5 10.5 0.90 10 0.432
2 600 0.8 4 0.67 11 0.358
3 200–400 0.5 15 0.4–0.8 4 0.416

Fig. 3. Temperature variation with time at different locations on the top surface for
different y-values but at x = 4.5 mm. This temperature distribution is for a single
layer laser assisted deposition of IN 718 powder on IN 718 substrate using the
power of 250 W and 15 mm/s scanning speed.

The stresses along x, y and z directions are referred to as the

longitudinal, transverse and through-thickness stresses, respec-
tively. From the calculated stress field, these individual residual
stress components are extracted along paths 1 and the 2, as shown
in Fig. 1. The longitudinal residual stress along path 1 is important
because it is a driving force for crack propagation, buckling and dis-
tortion. The through-thickness stress along path 2 can be responsi-
ble for delamination [23–25].

Fig. 2. Temperature and velocity distributions during the deposition of the 2nd
layer for (a) IN 718 and (b) Ti-6Al-4V. The simulation is done for 250 W laser power, 3. Results and discussions
15 mm/s scanning speed and 0.4 mm layer thickness.
Fig. 2 shows the computed temperature and velocity fields, and
the shape and size of molten pool for IN 718 and Ti-6Al-4V during
beam intensity distribution factor, k is the layer thickness, rb is the
deposition of the second layer. Each color1 band in Fig. 2 represents
focused beam radius and r is the radial distance from the beam axis.
a temperature range shown in the legend and the molten pool
Other boundary conditions for the thermal analysis include heat
lengths are indicated in the figures. It shows that the molten pool
loss by convection and radiation to the surroundings. The convec-
size for Ti-6Al-4V is slightly larger than that for IN 718 for the same
tive flow of molten metal inside the pool is driven by the surface
laser power and scanning speed mainly because of the lower density
tension gradient on the top surface of the pool resulting from the
of Ti-6Al-4V as shown in Table 1. The computed velocity fields in
spatial variation of temperature [18,19]. The resulting Marangoni
Fig. 2 result from the spatial gradient of surface tension on the top
stress can be expressed as [19]:
surface of the molten pool. A reference vector is shown by an arrow
dc dT du and a comparison of the length of this arrow with the vectors in the
s¼ ¼l ð9Þ plots indicates the magnitudes of the computed velocities. The
dT dr dz
results show a strong recirculatory flow of liquid metal from the
where T is the temperature, c is the surface tension, s is the Maran- middle of the pool to the periphery. For the processing conditions
goni stress, and r is the radial distance from the axis of the heat considered here, the values of Peclet numbers are significantly
source. Fig. 1 represents the solution domain of a 14-mm-long higher than 1 (between 9 and 13). Therefore, the calculated Peclet
deposit on a 5-mm-thick and 16-mm-long substrate. Calculations numbers and the magnitude of the velocities clearly show the dom-
are done over a half of the geometry taking advantage of symmetry. inance of convective heat transfer over heat transfer by conduction.
The laser beam travels along the positive x-axis. Positive z-axis rep- The temperature field in AM is highly transient as well as spa-
resents the build direction vertically upward. The boundary condi- tially non-uniform. Fig. 3 represents the temperature variation as
tions for the Abaqus-based mechanical analysis include fixed a function of time for different locations along the deposit width
bottom surface, i.e., the displacements of all nodes of the bottom (y-distance) starting from the pool center (y = 0) for IN 718 deposit.
surface along the x, y and z directions are zero. Table 5 summarizes
the additive manufacturing process parameters used in the 1
For interpretation of color in Fig. 2, the reader is referred to the web version of
calculations. this article.
364 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372

well with the experimentally measured values [36] for a 5-layer-

high deposition of a 150-mm-long Ti-6Al-4V build. The tempera-
ture was experimentally measured using a thermocouple located
on the top of the substrate at the mid length of the deposit. Ther-
mal cycles are simulated for the entire build using a corresponding
solution domain of 150 mm length. The excellent agreement
between the experimental and theoretical calculations shows the
importance of considering both conduction and convection in the
simulation. The agreement also indicates that the computed tran-
sient temperature field can be used for the residual stress and
strain calculations with confidence.
Fig. 5 shows the temperature distribution for different layers
during a 10-layer-high deposition of IN 718 powder. In AM, the
substrate acts as a heat sink. Therefore, for the upper layers, heat
transfer through the substrate decreases, which in turn, effectively
increases the peak temperature for the upper layers. Because of the
rapid scanning of laser beam the temperature contours are elon-
Fig. 4. Comparison of experimentally measured [36] and numerically computed gated behind the heat source and compressed in front of the beam.
thermal cycle for a 5 layers laser assisted deposition of 150 mm long Ti-6Al-4V The highly transient and spatially non-uniform temperature distri-
deposit on Ti-6Al-4V substrate using process parameter set 1 in Table 5. The bution, such as that shown in Fig. 5, is responsible for the genera-
monitoring location is at the mid length of the deposit on the top surface of the tion of the stress and strain fields.
substrate.
Fig. 6 shows the evolution of the longitudinal strain field during
The peak temperature is the highest at the pool center and the deposition of IN 718. With the progress of the deposition pro-
decreases gradually with distance away from the center. The vari- cess, more heat accumulates in the work piece. Also, both the peak
ation of the temperature with respect to time represents the ther- temperature and the pool volume increase slightly while at the
mal cycle. Fig. 4 shows that the calculated thermal cycle agrees same time the stiffness drops for the upper layers because of

Fig. 5. Temperature distribution during the deposition of (a) 1st (b) 3rd (c) 5th (d) 7th (e) 9th and (f) 10th layer deposition of IN 718 powder on IN 718 substrate. The laser
beam position is 10 mm from the starting point of the deposition. Laser beam scanning direction is along the positive x-axis. This simulation is done for 300 W laser power
and 15 mm/s scanning speed for the solution domain in Fig. 1.
 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372 365

Fig. 6. Longitudinal strain field at the end of depositing (a) 1st (b) 3rd (c) 5th (d) 7th (e) 9th and (f) 10th layer deposition of IN 718 powder on IN 718 substrate. Laser beam
scanning direction is along the positive x-axis. This simulation is done for 300 W laser power and 15 mm/s scanning speed for the solution domain in Fig. 1. Deformation is
magnified by 10.

Fig. 7. Variation in (a) temperature distribution and (b) longitudinal stress distribution along path 1 with time after the laser beam extinguishes at the end of 2nd layer. This
simulation is done for IN 718 powder deposited on IN 718 substrate with 300 W laser power and 15 mm/s scanning speed. ‘‘t” represents the time after the laser beam
extinguishes at the end of the deposition.

higher temperatures. Therefore, the strain value increases continu- Evolution of the stresses depends on the transient temperature
ously for the upper layers as shown in Fig. 6(a)–(f). The results distribution especially during cooling of the deposit. Fig. 7(a)
indicate that thin-walled, taller structures are more susceptible shows the temperature distribution along path 1 during cooling
to deformation than the shorter ones. of the build. Fig. 7(b) represents the corresponding longitudinal
366 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372

Fig. 8. Residual stress along x-direction (longitudinal) at the end of the deposition of (a) 2nd (b) 4th (c) 6th (d) 8th and (e) 10th layer of IN 718 powder on IN 718 substrate.
Laser beam scanning direction is along the positive x-axis. This simulation is done for 300 W laser power and 15 mm/s scanning speed for the solution domain in Fig. 1.
Deformation is magnified by 10.

stress profile developed along path 1. At t = 0 s, i.e., just after the of the 4th layer the maximum stress accumulates near layer 4.
laser beam traverses the entire length and is switched off tem- Similar observations can be made after the depositions of 6th,
porarily to prepare for the deposition of the next layer, the peak 8th and 10th layers. Secondly, in both Figs. 8 and 9, the residual
temperature of the deposit along path 1 (located at the stresses change from tensile to compressive at the interfaces of
substrate-deposit interface) is about 950 K. This high temperature the two successive layers as indicated in the figures. Finally, the
softens the material locally. Therefore, the magnitude of the stress through-thickness stress in Fig. 10 is compressive in the center of
is relatively low as it is limited by the yield strength at the high the deposit and tensile near the start and stop at the substrate
temperature. The stress field evolves as the deposit continues to deposit interface. These computed stress results can be useful in
cool down further. After 50 s, the temperature of the deposit the future to study the delamination of deposit from substrate,
almost cools down to the room temperature and the longitudinal separation of layers and warping.
stress along path 1 is highly tensile, as shown in Fig. 7(b). If there Fig. 11 shows a fair agreement between the calculated residual
was no additional layer deposited, the stress field at the end of stresses with the corresponding experimentally measured values
cooling would not change further, thus corresponding to the resid- [37] for a dissimilar metal deposition of IN 718 on a Ti-6Al-4V sub-
ual stress field in the part. strate. Fig. 11(a) and (b) represents the longitudinal and through-
Figs. 8–10 show the distribution of the residual stresses along x thickness components of the residual stresses, respectively. The
(longitudinal), y (transverse) and z (through-thickness) directions, stresses were measured at different locations along path 1 marked
respectively. The following major observations can be made from in Fig. 1. Several measurements were taken at the same location to
these results. The location of the maximum longitudinal and trans- estimate the error bar [37]. The reasons for the slight mismatch
verse (tensile) residual stresses are near the top of the deposit. For between the experimental and calculated values could be caused
example, after the deposition of the 2nd layer the maximum stress by the measurement difficulty and the assumptions used in
accumulation is observed near the 2nd layer. However, this stress numerical calculations; both are susceptible to some errors.
is relieved partially because of the reheating and cooling effects To understand the effect of alloy on the residual stresses, Fig. 12
while depositing the upper layers. Therefore, after the deposition compares the distribution of the residual stress components along
 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372 367

Fig. 9. Residual stress along y-direction (transverse) at the end of the deposition of (a) 2nd (b) 4th (c) 6th (d) 8th and (e) 10th layer of IN 718 powder on IN 718 substrate.
Laser beam scanning direction is along the positive x-axis. This simulation is done for 300 W laser power and 15 mm/s scanning speed for the solution domain in Fig. 1.
Deformation is magnified by 10.

x, y and z directions at the end of the cooling of 2nd layer of IN 718 path 1 is higher for IN 718 than that of Ti-6Al-4V. This is because
and Ti-6Al-4V deposits. All three residual stress components are of the lower yield strength of IN 718 than Ti-6Al-4V (Tables 2
highly non-uniform, as expected. The residual stresses in the sub- and 3). Therefore, under same processing conditions, IN 718 is
strate are mostly compressive. There also exists a sharp gradient of more susceptible to warping and delamination from the perspec-
stresses at the substrate-deposit interface. For both alloys the lon- tive of residual stresses. The high susceptibility to warping and
gitudinal stress (x-direction) reaches the maximum at the mid delamination of additively manufactured Inconel 718 components
length of the deposit and exhibits a sharp decrease toward both is also reported by Prabhakar et al. [40]. It should also be men-
ends (free surfaces). A high gradient in through-thickness residual tioned that the through-thickness stress along path 2 is always
stress (z-direction) at substrate-deposit interface can potentially lower than the yield strength of the alloy. For example, under the
cause the separation of the component from the substrate (i.e., same conditions of Fig. 13, the maximum through-thickness stres-
delamination). The yield strength of Ti-6Al-4V at room tempera- ses for IN 718 and Ti-6Al-4V are 100 MPa and 450 MPa,
ture is much higher than that of IN 718 (see Tables 2 and 3). There- respectively.
fore, the residual stresses are much higher for Ti-6Al-4V as shown The formation of tensile residual stresses has significant effects
in Fig. 12. on the mechanical properties of the product. For example, Fig. 14, a
Fig. 13(a) also shows that the longitudinal stress along path 1 plot based on independent literature data [41,42], shows that the
for Ti-6Al-4V is higher than that for IN 718. However, susceptibility fatigue cracks in additively manufactured Ti-6Al-4V parts grow
to warping and delamination depends not only on the magnitude faster in the presence of a higher tensile residual stress in the com-
of the residual stresses but also the yield strength of the alloy. ponent. The maximum tensile residual stress values are estimated
Therefore, a normalized residual stress [38,39] expressed as the from the reported stress distributions [41,42]. The y-intercept of
ratio of the longitudinal residual stress along path 1 to the room- the curve is about 8  106 mm/cycle which corresponds to the
temperature yield strength of the alloy is used for assessment of crack growth rate with very low residual stresses (such as the tra-
such problems. Fig. 13(b) shows that the normalized stress along ditionally processed Ti-6Al-4V parts) [43]. In the subsequent para-
368 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372

Fig. 10. Residual stress along z-direction (through-thickness) at the end of the deposition of (a) 2nd (b) 4th (c) 6th (d) 8th and (e) 10th layer of IN 718 powder on IN 718
substrate. Laser beam scanning direction is along the positive x-axis. This simulation is done for 300 W laser power and 15 mm/s scanning speed for the solution domain in
Fig. 1. Deformation is magnified by 10.

Fig. 11. Comparison of experimentally measured [37] and numerically computed (a) longitudinal residual stress and (b) through-thickness residual stress during the laser
deposition of IN 718 powder on a Ti-6Al-4V substrate using process parameter set 2 in Table 5.

graphs the residual stresses resulting from different layer thickness Fig. 15(a) and (b) shows the longitudinal and through-thickness
and heat input are discussed to provide a preliminary assessment residual stress distributions, respectively, during the deposition of
on the effect of processing parameters on the mechanical property a 0.8-mm-high wall using 2 and 4 layers. Building the same height
of additively manufactured parts. using more layers requires a lower layer thickness. For the same
 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372 369

Fig. 12. Residual stress distribution for IN 718 along (a) x (b) y and (c) z directions, and for Ti-6Al-4V along (d) x (e) y and (f) z directions of the 2 layers deposit. Laser beam
scanning direction is along the positive x-axis. The simulations are done for 250 W laser power and 15 mm/s scanning speed for the solution domain in Fig. 1.

Fig. 13. (a) The longitudinal and (b) the normalized residual stress (longitudinal residual stress/yield strength) distributions along path 1 for IN 718 and Ti-6Al-4V for 250 W
laser power and 15 mm/s scanning speed.

laser power and scanning speed, a lower layer thickness increases increase the distortion [44]. However, the deformation can reduce
the volumetric heat flux intensity and hence the peak temperature. the residual stresses in the deposit. Therefore, both the longitudi-
At the same time, it takes longer time to build the same height nal and the through-thickness residual stresses decrease with an
using thinner layers. Therefore, the total deposition time increases. increasing number of layers as shown in Fig. 15. For the deposition
Both the higher peak temperature and more exposure time tend to conditions studied here, the maximum longitudinal and through-
370 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372

Fig. 14. Fatigue crack growth rate (da/dN) for additively manufactured Ti-6Al-4V
components as a function of maximum tensile residual stress. The data are obtained
from the crack growth rate curves reported in independent literature studies
[41,42]. The stress intensity factor ratio is 0.1 and the stress intensity factor range is
10 MPa m1/2.

thickness residual stresses can be reduced by about 20% and 30%, Fig. 16. (a) Variation of thermal strain during deposition as a function of heat input
calculated both numerically and analytically, and (b) effect of heat input on
respectively by using 4 layers to build the same height instead of
maximum longitudinal stress and strain. Both the plots are drawn for 2 layers of IN
2 layers, as shown in Fig. 15. 718 deposit on IN 718 substrate.
To understand the effect of heat input and peak temperature on
the distortion and strain, we have recently proposed [44] a strain
parameter (e⁄) as an indicator of the susceptibility to distortion. does not consider the plastic strain, and this is probably the reason
for the slight mismatch between the analytical and numerical
bDT tH3=2 results in Fig. 16(a).
e ¼ pffiffiffiffi ð10Þ Fig. 16(b) shows that doubling the heat input increases the
EI F o q
strain by almost 2.5 times. However, the maximum longitudinal
where b is the volumetric coefficient of thermal expansion, DT is the residual stresses can be reduced by about 20% by doubling the heat
maximum rise in temperature during the process, E is the elastic input. Distortion is often a crucial issue in AM as it affects the
modulus, I is the moment of inertia of the substrate, t is the charac- dimensional accuracy of the fabricated part [44,46]. On the other
teristic time, H is the heat input per unit length, Fo is the Fourier hand, the residual stresses can be reduced by post process heat
number, and q is the density of the alloy. The product, EI, is the flex- treatment or by reducing the layer thickness. Therefore, an appro-
ural rigidity of the structure. Fourier number is the ratio of the heat priate processing condition is to be carefully selected considering
storage rate to heat dissipation rate. The temperature rise and the both the distortion and residual stresses. In Fig. 16(b), for example,
Fourier number are both calculated using the heat transfer and fluid a heat input of 20 J/mm seems to provide a strain of 0.0057 and a
flow model. This strain parameter quantitatively represents the longitudinal residual stress of 540 MPa. This figure is useful to
peak thermal strain encountered by the fabricated part [44]. Higher select a heat input considering both the maximum longitudinal
heat input increases the peak temperature as well as the pool vol- residual stress and strain. In practice lower heat input can be
ume [45]. Larger pools shrink more during solidification and exhibit achieved by low laser power and faster scanning speed. The layer
more distortion. Therefore, strains increase with the heat input. thickness may also be kept low in order to minimize the residual
Fig. 16(a) compares the peak thermal strains at the end of the stresses. This finding is consistent with the current industrial prac-
deposition calculated both numerically using the thermo- tice. For example, General Electric uses low heat input (100–500 W
mechanical model as well as analytically using Eq. (10). Eq. (10) and 2000 mm/s) and low layer thickness of 40–100 lm to fabricate

Fig. 15. (a) The longitudinal and (b) the through-thickness residual stress distributions of IN 718 deposit on IN 718 substrate for the process parameter set 3 in Table 5 using 2
and 4 layers deposition to build a 0.8-mm-high wall. The simulations are done for 250 W laser power and 15 mm/s scanning speed.
 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372 371

aero-engine fuel nozzles [47]. Such improved predictive models [4] P.H. Chang, T.L. Teng, Numerical and experimental investigations on the
residual stresses of the butt-welded joints, Comput. Mater. Sci. 29 (2004) 511–
are essential to help select process parameters and evaluate other
522.
strategies such as preheating for mitigating residual stresses and [5] D. Deng, H. Murakawa, Numerical simulation of temperature field and residual
distortion in the future. stress in multi-pass welds in stainless steel pipe and comparison with
experimental measurements, Comput. Mater. Sci. 37 (2006) 269–277.
[6] R.J. Moat, A.J. Pinkerton, L. Li, P.J. Withers, M. Preuss, Residual stresses in laser
direct metal deposited Waspaloy, Mater. Sci. Eng. A 528 (2011) 2288–2298.
4. Summary and conclusions [7] P. Rangaswamy, M.L. Griffith, M.B. Prime, T.M. Holden, R.B. Rogge, J.M.
Edwards, R.J. Sebring, Residual stresses in LENSÒ components using neutron
diffraction and contour method, Mater. Sci. Eng. A 399 (2005) 72–83.
The evolution of the stresses and strains for Inconel 718 and Ti- [8] G. Bussu, P.E. Irving, The role of residual stress and heat affected zone
6Al-4V and their dependence on important process parameters properties on fatigue crack propagation in friction stir welded 2024-T351
such as heat input and layer thickness are investigated using a cou- aluminum joints, Int. J. Fatigue 25 (2003) 77–88.
[9] A.S. Wu, D.W. Brown, M. Kumar, G.F. Gallegos, W.E. King, An experimental
pled thermal, fluid flow and mechanical model. This model is used
investigation into additive manufacturing-induced residual stresses in 316L
to estimate the appropriate heat input and layer thickness to fab- stainless steel, Metal. Mater. Trans. A 45 (2014) 6260–6270.
ricate dimensionally accurate components with low residual stres- [10] J. Zhang, X. Wang, S. Paddea, X. Zhang, Fatigue crack propagation behaviour in
wire + arc additive manufactured Ti-6Al-4V: effects of microstructure and
ses. Below are the specific findings.
residual stress, Mater. Des. 90 (2016) 551–561.
[11] A. Riemer, S. Leuders, M. Thöne, H.A. Richard, T. Tröster, T. Niendorf, On the
(1) Ti-6Al-4V components suffer higher residual stresses than fatigue crack growth behavior in 316L stainless steel manufactured by
IN 718 under the same processing conditions. However, IN selective laser melting, Eng. Fract. Mech. 120 (2014) 15–25.
[12] L.M. Sochalski-Kolbus, E.A. Payzant, P.A. Cornwell, T.R. Watkins, S.S. Babu, R.R.
718 parts are more susceptible to warping, delamination Dehoff, M. Lorenz, O. Ovchinnikova, C. Duty, Comparison of residual stresses in
and buckling because of their higher residual stress to yield Inconel 718 simple parts made by electron beam melting and direct laser
strength ratio. metal sintering, Metal. Mater. Trans. A 46 (2015) 1419–1432.
[13] D. Qiao, W. Zhang, T.-Y. Pan, P. Crooker, S. David, Z. Feng, Evaluation of residual
(2) The AM parts fabricated using thin layers encounter a high plastic strain distribution in dissimilar metal weld by hardness mapping, Sci.
peak temperature and long exposure time under the laser Technol. Weld. Join. 18 (2013) 624–630.
beam. As a result, the residual stresses can be decreased as [14] P.J. Withers, H.K.D.H. Bhadeshia, Residual stress. Part 1 – measurement
techniques, Mater. Sci. Technol. 17 (2001) 355–365.
much as 30% by reducing the thickness of each layer to fab- [15] G.S. Schajer, Application of finite element calculations to residual stress
ricate the component. This finding is consistent with the measurements, J. Eng. Mater. Technol. 103 (1981) 157–163.
industrial practice of 40–100 lm layer thickness to produce [16] V. Manvatkar, A. De, T. DebRoy, Heat transfer and material flow during laser
assisted multi-layer additive manufacturing, J. Appl. Phys. 116 (2014) 124905.
aero engine fuel nozzles.
[17] V. Manvatkar, A. De, T. DebRoy, Spatial variation of melt pool geometry, peak
(3) Doubling the heat input can reduce the residual stresses by temperature and solidification parameters during laser assisted additive
about 20%. However, the same condition enhances the ther- manufacturing process, Mater. Sci. Technol. 31 (2015) 924–930.
[18] S.A. David, T. DebRoy, Current issues and problems in welding science, Science
mal distortion by about 2.5 times. Therefore, an appropriate
257 (1992) 497–502.
heat input selected by trading off both distortion and resid- [19] A. Raghavan, H.L. Wei, T.A. Palmer, T. DebRoy, Heat transfer and fluid flow in
ual stresses will be helpful to fabricate a dimensionally accu- additive manufacturing, J. Laser Appl. 25 (2013) 052006.
rate part with good mechanical properties. [20] A. Vasinonta, J.L. Beuth, M. Griffith, Process maps for predicting residual stress
and melt pool size in the laser-based fabrication of thin-walled structures, J.
(4) The longitudinal residual stress exhibits a steep gradient at Manuf. Sci. Eng. 129 (2007) 101–109.
the both ends of the deposit that makes the parts susceptible [21] M.P. Mughal, H. Fawad, R.A. Mufti, M. Siddique, Deformation modelling in
to buckling and warping. The through-thickness stress that layered manufacturing of metallic parts using gas metal arc welding: effect of
process parameters, Model. Simul. Mater. Sci. Eng. 13 (2005) 1187.
is responsible for the possible delamination of a component [22] A.H. Nickel, D.M. Barnett, F.B. Prinz, Thermal stresses and deposition patterns
changes sharply at the substrate deposit interface. The resid- in layered manufacturing, Mater. Sci. Eng. A 317 (2001) 59–64.
ual stress changes from tensile to compressive at the layer [23] S. Ghosh, J. Choi, Three-dimensional transient finite element analysis for
residual stresses in the laser aided direct metal/material deposition process, J.
interfaces. In extreme cases, this behavior may result in Laser Appl. 17 (2005) 144–158.
the separation of layers. [24] P. Mercelis, J.P. Kruth, Residual stresses in selective laser sintering and
selective laser melting, Rapid Prototyp. J. 12 (2006) 254–265.
[25] L. Wang, S.D. Felicelli, P. Pratt, Residual stresses in LENS-deposited AISI 410
It is hoped that the procedure proposed here will constitute a
stainless steel plates, Mater. Sci. Eng. A 496 (2008) 234–241.
step forward in predictive modeling and results will be useful to [26] G. Vastola, G. Zhang, Q.X. Pei, Y-W. Zhang, Controlling of residual stress in
assess residual stresses and make dimensionally compliant AM additive manufacturing of Ti6Al4V by finite element modeling, Addit. Manuf.
(2016), http://dx.doi.org/10.1016/j.addma.2016.05.010.
components.
[27] Z. Nie, G. Wang, J.D. McGuffin-Cawley, B. Narayanan, S. Zhang, D. Schwam, M.
Kottman, Y.K. Rong, Experimental study and modeling of H13 steel deposition
using laser hot-wire additive manufacturing, J. Mater. Process. Technol. 235
Acknowledgement (2016) 171–186.
[28] M. Alimardani, E. Toyserkani, J.P. Huissoon, A 3D dynamic numerical approach
for temperature and thermal stress distributions in multilayer laser solid
This work was supported by the US Department of Energy NEUP freeform fabrication process, Opt. Laser Eng. 45 (2007) 1115–1130.
[grant number DE-NE0008280]. One of the authors (T.M.) also [29] L.E. Svensson, B. Gretoft, H.K.D.H. Bhadeshia, An analysis of cooling curves
acknowledges the support from American Welding Society from the fusion zone of steel weld deposits, Scand. J. Metall. 15 (1986) 97–103.
[30] K.C. Mills, Recommended Values of Thermo-Physical Properties for Selected
research fellowship [grant number 179466]. Commercial Alloys, Woodhead Publishing, Cambridge, 2002.
[31] Abaqus Documentation, Version 6.14, Dassault Systems, 2015.
[32] A. Kamara, S. Marimuthu, L. Li, A numerical investigation into residual stress
References characteristics in laser deposited multiple layer waspaloy parts, Trans. ASME-B
– J. Manuf. Sci. Eng. 133 (2011) 1–9.
[33] P. Scott, R. Olson, J. Bockbrader, M. Wilson, B. Gruen, R. Morbitzer, Y. Yang, C.
[1] D.D. Gu, W. Meiners, K. Wissenbach, R. Poprawe, Laser additive manufacturing
Williams, F. Brust, L. Fredette, N. Ghadiali Battelle, The Battelle Integrity of
of metallic components: materials, processes and mechanisms, Int. Mater. Rev.
Nuclear Piping (BINP), Program Final Report, Columbus, OH, 2005.
57 (2012) 133–164.
[34] T. Seshacharyulu, S.C. Medeiros, W.G. Frazier, Y.V.R.K. Prasad, Microstructural
[2] W.J. Sames, F.A. List, S. Pannala, R.R. Dehoff, S.S. Babu, The metallurgy and
mechanisms during hot working of commercial grade Ti–6Al–4V with lamellar
processing science of metal additive manufacturing, Int. Mater. Rev. (2016) 1–
starting structure, Mater. Sci. Eng. A 325 (2002) 112–125.
46.
[35] P. Rangaswamy, H. Choo, M.B. Prime, M.A. Bourke, J.M. Larsen, High
[3] J. Ding, P. Colegrove, J. Mehnen, S. Ganguly, P.M. Sequeira Almeida, F. Wang, S.
temperature stress assessment in SCS-6/Ti-6Al-4V composite using neutron
Williams, Thermo-mechanical analysis of wire and arc additive layer
diffraction and finite element modeling, in: International Conference on
manufacturing process on large multi-layer parts, Comput. Mater. Sci. 50
Processing & Manufacturing of Advanced Materials, Las Vegas, NV, USA, 2000.
(2011) 3315–3322.
372 T. Mukherjee et al. / Computational Materials Science 126 (2017) 360–372

[36] F. Lia, R.M. Martukanitz, J.Z. Park, T. DebRoy, T. Mukherjee, J.S. Keist, S.N. selective laser melting: fatigue resistance and crack growth performance, Int. J.
Patankar, Process and microstructural validation of the laser-based directed Fatigue 48 (2013) 300–307.
energy deposition process for Ti-6Al-4V and Inconel 625 material, Mater. Sci. [42] P. Edwards, A. O’Conner, M. Ramulu, Electron beam additive manufacturing of
Technol. (2015) (Columbus, OH, USA), The abstract can be retrieved from: titanium components: properties and performance, J. Manuf. Sci. Eng. 135
<http://www.programmaster.org/PM/PM.nsf/ApprovedAbstracts/ (2013) 061016.
A0DC8061FC54E5B785257E0A0073048E?OpenDocument>. [43] A.G. Gavras, F.C. Brendan, D.A. Lados, Effects of microstructure on the fatigue
[37] K. Shah, I.U. Haq, S.A. Shah, F.U. Khan, M.T. Khan, S. Khan, Experimental study crack growth behavior of light metals and design considerations, Matéria (Rio
of direct laser deposition of Ti-6Al-4V and Inconel 718 by using pulsed de Janeiro) 15 (2010) 319–329.
parameters, Sci. World J. 2014 (2014). [44] T. Mukherjee, J.S. Zuback, A. De, T. DebRoy, Printability of alloys for additive
[38] A.M. Al-Mukhtar, Residual stresses and stress intensity factor calculations in T- manufacturing, Sci. Rep. 6 (2016).
welded joints, J. Fail. Anal. Prevent. 13 (2013) 619–623. [45] H.L. Wei, J.W. Elmer, T. DebRoy, Origin of grain orientation during
[39] R. Jackson, I. Chusoipin, I. Green, A finite element study of the residual stress solidification of an aluminum alloy, Acta Mater. 115 (2016) 123–131.
and deformation in hemispherical contacts, J. Tribol. 127 (2005) 484–493. [46] T. Mukherjee, V. Manvatkar, A. De, T. DebRoy, Mitigation of thermal distortion
[40] P. Prabhakar, W.J. Sames, R. Dehoff, S.S. Babu, Computational modeling of during additive manufacturing, Scripta Mater. 127 (2017) 79–83.
residual stress formation during the electron beam melting process for Inconel [47] P.M. Kenney, D.E. Lindley, General Electric Company, 2013, U.S. Patent
718, Addit. Manuf. 7 (2015) 83–91. Application 14/440, 154.
[41] S. Leuders, M. Thöne, A. Riemer, T. Niendorf, T. Tröster, H.A. Richard, H.J. Maier,
On the mechanical behaviour of titanium alloy TiAl6V4 manufactured by