You are on page 1of 13

Additive Manufacturing 46 (2021) 102187

Contents lists available at ScienceDirect

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

Research Paper

A novel high-efficient finite element analysis method of powder bed fusion


additive manufacturing
Yang Cao a, b, Xin Lin a, b, *, Nan Kang a, b, *, Liang Ma a, b, Lei Wei a, b, Min Zheng a, b, Jun Yu a, b,
Dongjian Peng c, Weidong Huang a, b
a
State Key Laboratory of Solidification Processing, Northwestern Polytechnical University, Xi’an, Shaanxi 710072, PR China
b
Key Laboratory of Metal High Performance Additive Manufacturing and Innovative Design, MIIT China, Northwestern Polytechnical University, Xi’an, Shaanxi
710072, PR China
c
Xi’an Space Engine Company Limited, Xi’an, Shaanxi 710100, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: This paper aims to develop a high-efficient finite element (FE) model by combining the equivalent boundary
Powder bed fusion additive manufacturing condition method (EBCM) and dynamic mesh method (DMM) for accelerating the thermo-mechanical simulation
Finite element analysis of additive manufacturing by powder bed fusion (PBF). EBCM and DMM take advantage of the strong non-linear
High- efficient computation
phenomenon in the thermo-mechanical affected zone (TMAZ) by simplifying the temperature boundary condi­
Multi-scaled simulation
tions and optimizing the local mesh density, respectively. In detail, EBCM can reduce the difficulty of conver­
gence in computing, while DMM can reduce the scale of stiffness matrix. The experimentally calibrated lattice
Boltzmann method (LBM) firstly is used to produce the equivalent temperature function in TMAZ as the thermal
input conditions of the FE model. The thermo- mechanical responses of several PBF deposits under different
process parameters are predicted by the developed model, presenting good agreement with experiment data in
terms of in-situ temperature and residual stress. Compared with the normal model, the novel high-efficient model
notoriously cuts the computation cost without compromising precision. This study provides an important
simulation concept for high-efficient FE analyses with high fidelity of the AM technology.

1. Introduction molten pool and powder also increase the difficulty for the under­
standing of the building processes [8]. Hence, both experiments and
Additive manufacturing (AM) enables to produce high-performance simulation are employed to investigate AM technology allowing for a
parts using a layer-by-layer deposition technique according to a virtual large number of variables. Normally, experiments hardly elucidate such
CAD model [1]. According to the powder feeding system, the metal AM complex process due to its limitations in terms of time-consuming,
processes mainly consist of directed energy deposition (DED) and expensive and parameter measurement. Alternatively, numerical simu­
powder bed fusion (PBF) [2]. Notably, the latter has been widely used to lation provides an efficient and feasible approach to study the
fabricate more complex components in the aviation, aerospace, and thermo-metallurgical-mechanical behaviors in different scales during
automobile industries due to the processing characteristics of the the 3D-printing processes. However, validation and calibration of nu­
self-support by the power bed and ultra-small layer thickness. In AM, the merical models by experiments are very necessary before predictions.
rapid movement of the heat source by a laser (or electron beam) causes The literatures have reported copious in-situ measurements and nu­
steep temperature gradients in the small molten pool and the merical simulations to study the thermo-mechanical responses in AM [4,
already-deposited layers further undergo several rapid heating and 5,7–13]. Lu et al. launched an exhaustive experimental campaign to
cooling cycles during the subsequent building process [3–5]. As a result, measure the temperature field by thermocouples and infrared imaging
significant thermal stresses are induced in the deposit, contributing to and the full-field strain by digital image correlation (DIC) technique in
part distortion, delamination or even build failure if the stress cannot be the DED process [5,7,8]. In PBF, the model validation is usually carried
appropriately controlled [6,7]. out according to the bending deformation of the cantilever beam
During AM processes, the complex interactions among the laser, structures when supports are removed [9,14]. Besides, residual strains

* Corresponding authors at: State Key Laboratory of Solidification Processing, Northwestern Polytechnical University, Xi’an, Shaanxi 710072, PR China.
E-mail addresses: xlin@nwpu.edu.cn (X. Lin), nan.kang@nwpu.edu.cn (N. Kang).

https://doi.org/10.1016/j.addma.2021.102187
Received 9 December 2020; Received in revised form 27 May 2021; Accepted 12 July 2021
Available online 26 July 2021
2214-8604/© 2021 Elsevier B.V. All rights reserved.
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

and stresses in AM parts also can be determined by means of X-ray and stresses before the LPBF procedure. The block parts with dimensions of
neutron diffraction measurements [10,15] as well as hole-drilling 10 × 10 × 0.9 mm3 were then deposited at the processing parameters
method [16]. listed in Table 1.
The AM numerical models can be classified into micro, meso and In addition, single-track deposition with different processing pa­
macro scales [17]. Micro models (e.g. phase-filed model [18]) focus on rameters was carried out to calibrate the heat source model of the
the microstructure prediction. Meso models are interested in capturing simulation. An optical microscope was used to characterize the size of
the grain structure using cellular automata [19] and kinetic monte carlo the molten pool of a single track.
[20] methods, and the fluid dynamic phenomenon by the collision and
streaming of fluid particles, based on the lattice Boltzmann method 2.2. In-situ temperature measurement
(LBM). It is worth mentioning that LBM can effectively and efficiently
address the melting and solidification problems when using The placement and operation status of the in-situ measurement are
high-performance computation (HPC) based on parallel architectures, shown in Fig. 1(a) and (b), respectively. To provide a sufficient space for
especially for the study of the powder melting in PBF process [21–23]. setting a data recorder (using Graphtec GL-900) and uninterruptible
Macro models, consisting of finite element (FE) method and computa­ power supply, the substrate plate was supported to the processing plane
tional fluid dynamics [24], mostly concern the evolution of the and restrained by stainless-steel brackets without other modifications to
thermo-mechanical field of the built part, being responsible for tem­ the LPBF machine and forming process. Armored thermocouples (type
perature, residual stress and distortion [4,5,7]. K) with the diameters of 1 ± 0.01 mm were used to measure the tem­
In order to understand and further control the stress problems in AM, perature inside the substrate. Fig. 1(c) shows the geometric character­
the FE models usually use small mesh size less than the heat source istics of the substrate and locations of two thermocouples, TC1 and TC2,
radius in order to achieve more accurate prediction [25]. However, the 1 mm below the upper surface of the substrate.
mesh number dramatically rises as the build size increases, leading to
very expensive calculation cost, especially in PBF involving thousands of
2.3. Residual stress measurement
scanning tracks (layers). Hereby, many researches have carried out
high-efficiency simulations [14,26–34]. Based on the inherent strain
A microindentation method [35–38] was used to characterize the
method (ISM), Liang et al. [14] extracted the mean inherent strain
residual stress of the LPBF parts. The relationship between the von Mises
vector from a small representative volume and then applied to a
residual stress (σres ) and strain (εres ) can be expressed by:
part-scale layer-by-layer model to predict the accumulative de­
formations in a double cantilever beam and complex canonical parts by H = Cσres (εrepr + εres ) (1)
laser bed powder fusion (LBPF). Also, they performed a model validation
on both the accuracy and efficiency by experiments. Chiumenti et al. σres
c2 = c20 − 0.32ln⁡[1 + ] (2)
[26] utilized an equivalent heat source with several cladding track­ σ (εres )
s/layers as a basic unit to calculate the temperature field of LPBF blocks,
which improves the efficiency over five times due to the use of coarse where H is the hardness, C is a constant related to the material, εrepr is the
mesh. However, the scanning details are ignored so that the actual equivalent plastic strain (in general, εrepr = 0.08 and C = 3 [38,39]
thermal history of the cladding tracks cannot be predicted by their when the Vickers indenter is used), c20 is approximately 1 [40], and c2 is
model. Carraturo et al. [27] used the equivalent thermal load to capture defined as:
the deflection of LPBF part after support removal, based on the finite cell /
c2 = A Anom (3)
method (FCM) modified from standard FE method, allowing a very
coarse discretization [28]. Similarly, residual stresses in AM parts also
where A and Anom are the real and nominal projected contact areas
can be roughly predicted by adopting coarser meshes and larger time
(Fig. 1(a)), respectively, and σ (ε) characterizes the power-law relation­
increment to simulate the real LPBF process [29,30]. In addition, the
ship between the stress and strain in the uniaxial tensile experiment,
dynamic adaptive mesh refinement (AMR) technology [31–33] provides
a good choice to improve the computing efficiency because it can σ = K εn (4)
optimize the mesh density and distribution in real time according to
varying temperature gradients [34]. Nevertheless, the AMR algorithm is where both K and n are obtained by fitting the stress–strain curve (Fig. 1
relatively complex. (b)) of the previous related study [41], 722.58 and 0.15573,
This given, the objective of this work is to develop a high-efficient FE respectively.
thermo-mechanical model by combining the equivalent boundary con­ Thus, εres and σ res can be calculated using:
ditions with the dynamic meshing method, allowing for calculation ef­ H
(5)
1
ficiency and prediction accuracy. Firstly, an in-situ measurement εres = ( )0.15573 − 0.08
2167.74
platform was established and used to calibrate the model. Next, the LBM
model is employed to extract the local thermal response used to feed the (
1 − c2
)
high- efficient model. Finally, the effectiveness of the proposed model is σ res = 722.58|εres |0.15573 [exp − 1] (6)
0.32
validated by comparing the predicted thermo-mechanical results with
the experimental data. A LECO-AMH43 automatic hardness testing system was used to
evaluate H and c2 of the LPBF samples with a load of 300 g and dwell
2. Experimental details time of 15 s (see in Fig. 2). To reduce the error and mutual influence

2.1. LPBF experiment Table 1


LPBF processing parameters.
An AlSi10Mg powder (Baohang Advanced Materials, China) with an Case 1 Case 2 Case 3 Case 4 Case 5
average particle size of 30 µm was used for the present analysis. The Laser power (W) 300 340 380 340 340
LBPF experiments were carried out in a solution SLM 280 system. The Scan speed (mm/s) 1500 1500 1500 1000 2000
block parts were fabricated using the reciprocating scanning strategy on Layer thickness (μm) 30
a cast AlSi10Mg substrate plate with dimensions of 50 × 50 × 20 mm3. Hatching distance (μm) 100
Radius of laser beam (μm) 70
The substrate plates were heated to 300 ◦ C for 2 h to relieve residual

2
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

Fig. 1. In-situ measurement:(a) placement status, (b) work status; (c) geometric characteristic of substrate; (d) location of indention on substrate.

Fig. 2. (a) Schematic diagram of A and Anom ; (b) tensile stress-strain curve and fitted curve.

3
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

between the test points, the data at each location of the top surface cooling stage. During the multi-layer multi-track LPBF process, the di­
(positions I, II, and III in Fig. 1(d)) were the averages of three nearby mensions of the molten pool are roughly steady after a certain period of
points. The distance between adjacent points was 200 µm, as shown in deposition [44,45]; i.e., the temperature field of the region surrounding
Fig. 1(d). the molten pool tends to maintain a quasi-steady state. Based on this
approximation, the temperature of the material in the ith cladding track
3. Computational modeling can be approximately expressed by the integral of thermal response:
⎧ ∫t ⎫
3.1. High efficient model ⎪

⎨ Ti0 + Kih dt ti ≤ ti1 ⎪


(7)
ti0
Ti (s, xi , yi , zi , ti ) = ∫ ti1 ∫t
The local temperature distribution of the molten pool during the ⎪

⎩ Ti0 + Kih dt +

Kic dt ti ≥ ti1 ⎪

LPBF with a rapidly varying and large gradient [3] has a negative effect ti0 ti1

on the solution efficiency. When the heat source moves along a straight
line at a constant speed, after a period of time, the temperature field where Ti0 is the temperature when the current track starts to deposit and
surrounding the cladding track reaches the quasi-steady state [42,43]. ti0 and ti1 are the start and end times of the heating stage, respectively.
The materials near the current cladding track have the similar response Considering the multi-track process and its scanning strategy, the tem­
to heat source. The variation rate of the temperature near the molten perature function Ti (s, xi , yi , zi , ti ) combined with the superposition
pool can be expressed as a function of the location (x, y, z), time (t), and method [46] can be applied to the multi-track deposition process,
state of the material (s), where the state includes powder, liquid, and ∑
TBC (s, x, y, z, t) = Ti (s, xi , yi , zi , ti ) (8)
solid. The thermal response includes the variation rate of the local i
temperature Kh (x, y, z, t, s) at the heating stage and Kc (x, y, z, t, s) at the
This temperature function can be predefined as the Dirichlet

Fig. 3. Schematic diagram of (a) EBCM and (b) DMM.

4
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

boundary condition to replace the Neumann boundary condition, i.e.,


( ) ∆t ( eq )
the heat flow condition from the laser energy input. This method is fj x + ej , t + ∆t − fj (x, t) = f j (x, t) − fj (x, t) + Fj (9)
τf
referred to as equivalent boundary condition method (EBCM), as shown
in Fig. 3(a). It should be mentioned that the model without high-efficient ( ) ∆t (( eq )
method is regarded as normal model in this study. hj x + ej , t + ∆t − hj (x, t) = hj (x, t) − hj (x, t) (10)
τh
In addition, the quality of the mesh in the LPBF model strongly in­
fluences the convergence, accuracy, and computational efficiency of the where τf and τh are the relaxation times of the velocity and temperature
numerical model [34]. The dynamic mesh method (DMM) has the ad­ fields, respectively. fj (x, t) and hj (x, t) represent the density and tem­
vantages of low computational cost and high accuracy. In this study, perature functions in the j-direction, respectively. Fj is the body force,
according to the different stages in the LPBF process, the corresponding such as gravity. Both f eq eq
j (x, t) and hj (x, t) are the equilibrium distribu­
models with different meshes and geometries are established automat­
tion functions [47]. The laser heat input is modeled using a Gaussian
ically before the analysis and are fixed during the simulation. Very fine
heat source,
mesh is set in the thermos-mechanical affected zone (TMAZ) while the
mesh is coarsened as far away the TMAZ, as shown in Fig. 3(b). The 2AP − 2r2
q= e R2 (11)
model information (including mesh conditions) is stored in the corre­ π R2
sponding input files of computational software, which will not require
memory consumption until calculation. According to the processing where A, P, and R are the absorptivity, laser power, and laser spot radius,
sequence, the input files are submitted for calculation in turn. Temper­ respectively. More detailed information on the used LBM model has
ature interpolation is used to maintain the continuity of the simulation. been presented by Zheng et al. [48,49].
Notably, the DMM in this study is essentially the preprocessing tech­ The transient heat transfer analysis in the FE model is governed by:
nology. This is different from the dynamic AMR used in 3DSIM [32] and ∂T ∂
(
∂T
)

(
∂T
)

(
∂T
)

(
∂T
)
Pan computing [31] which can change the mesh density depending on ρC p = k + k + k + k +Q (12)
∂t ∂x ∂x ∂x ∂x ∂y ∂y ∂z ∂z
the thermal gradient during the calculation.
The calculation frame of the high-efficient model in this study is where ρ, Cp , k, Q, and t are the density, specific heat, thermal conduc­
presented in Fig. 4. The temperature field obtained using the high- tivity, inner heat source, and time of heat transfer, respectively.
efficient model is then set as the predefined field of the mechanical FE The initial condition can be expressed as:
model to calculate the thermal stress.
T(x, y, z)t=0 = T0 (13)
3.2. Thermal analysis The boundary conditions can be expressed as:

The LBM can solve melting and solidification problems because it can ∂T
k − q + qc + qr = 0 (x, y, z) ∈ S (14)
conveniently describe the fluid dynamic phenomenon by the collision ∂n
and streaming of fluid particles[21,22]. Hence, the behavior of powder
where S is the boundary surface. T and n are the temperature and normal
particles in LPBF can be studied by LBM is and TBC (s, x, y, z, t) also can be
vector of S, respectively. The heat losses by convection and radiation can
accurately calculated.
be computed by:
In detail, the LBM model couples the temperature with the velocity
field. The density and velocity of particles are obtained with a distri­ qc = h(T − Ts ) (15)
bution function f(x, t). h(x, t) [47] is temperature distribution function.
The collision and streaming of distributions are considered by: qr = σB εr (T 4 − T 4s ) (16)

Fig. 4. The calculation frame of high efficient model.

5
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

where h, σB , and εr are the thermal convection coefficient, Stefan- Table 3


Boltzmann constant, and thermal radiation coefficient, respectively, Thermophysical parameters of AlSi10Mg used for thermal analysis [50,51].
and Ts is the temperature of the environment (set to 25 ℃ in this study). Material property Value Unit
In the high-efficient model, the heat flux q is replaced by the Dirichlet
Solidus temperature, Ts 557 ℃
boundary condition, as mentioned above.
Liquidus temperature, TL 587 ℃
Boiling temperature, Tb 2470 ℃
3.3. Mechanical analysis Density, ρ 2680 3
Kgm−
− 4 − 1 1
∂σ − 0.31 × 10 Nm K−
The relationship between the stress (σ) and total strain (ε) can be Temperature coefficient of surface tension,
∂T
expressed by: Surface tension at liquidus temperature, σL 0.824 Nm− 1
Latent heat of fusion, ∆Hf 389 × 103 J Kg− 1
dσ = Dε − CdT (17) Latent heat of vaporization, ∆Hv 10700 × 103 J Kg− 1
Linear thermal expansion coefficient, αL 23 × 10− 6 K− 1
where D is a fourth-order stiffness tensor. C is a stress matrix related to D Convective heat transfer coefficient, hc 82 Wm− 2 K− 1
Radiation emissivity, εr 0.4
and thermal expansion coefficient [12]. The strain component can be
expressed by:

εlp = εElp + εPlp + εTlp (l, p = 1, 2, 3) (18) computed straightforwardly as:


ρs − ρp
where εElp , εPlp , and εTlp are the elastic strain, plastic strain, and thermal ϕ= (20)
ρs
strain, respectively. The material in the plastic state confirms the von
Mises yield criterion, where the equivalent stress (σm ) can be expressed where ρs and ρp are the densities of the solid and powder, respectively.
by: The thermal conductivity of the powder (kp ) can be calculated as [52].
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ks (1 − ϕ)
σm =
(σ 1 − σ2 )2 + (σ2 − σ 3 )2 + (σ 3 − σ1 )2
(19) kp = (21)
2 1 + ψ kkgs

where σ 1 , σ 2 and σ3 are principal stress in different directions. ψ = 0.02 × 102(ϕ− 0.3)
(22)

3.4. Material properties where ks and kg are the thermal conductivities of the solid and gas,
respectively. The specific heats of the powder and solid are considered
The peak temperature of the material under the irradiation of the equal.
high-energy laser may considerably exceed the melting point during the The precise response of the mechanical model depends on the
LPBF process. Thus, the temperature-dependent material thermal appropriate mechanical properties of the material. However, large gaps
physical parameters of AlSi10Mg covering a wide temperature range are between the material properties in previous studies and experimental
used in this study. The thermophysical parameters of AlSi10Mg alloy, values exist [15,53–55]. Therefore, the mechanical parameters of the
such as the viscosity, thermal conductivity, and specific heat are shown LPBF parts used in the FE analysis, such as the yield strength and elastic
in Table 2 [50]. Other parameters are presented in Table 3 [50,51]. modulus, were evaluated by our group [41]. The mechanical properties
Considering the porosity of the powder in the computational domain, of the as-cast substrate plates were obtained from reference [15], as
the thermophysical properties of the powder bed are established in shown in Table 4. The linear thermal expansion coefficient and Poisson’s
terms of solid material parameters and porosity (ϕ). The porosity is ratio (0.33) of the substrate and parts were obtained from the literature
[50]. In addition, the powder bed was ignored in the mechanical
analysis.
Table 2
Temperature-dependent thermophysical parameters of AlSi10Mg [50].
Material property Temperature (℃) Value 3.5. Simulation conditions

Viscosity, ν (Pa s) 600 0.0022


725 0.00125
3D LBM single-track models with laser powers of 300–380 W and
1175 0.0007 scanning speeds of 1000–2000 mm/s were developed in this study. The
1575 0.00058 substrate had dimensions of 2500 × 500 × 200 µm3. The melted
1975 0.00045 powder layer had a thickness of 30 µm. Hexahedral elements of LBM
2425 0.0004
models with a uniform mesh of 4 × 4 × 4 µm3 were used. Ti (s, xi , yi , zi ,
Thermal conductivity (Wm− 1 ℃− 1 ) 25 160
125 160 ti ) was fitted using the LBM temperature field.
225 160 Further, typical three-layer 3D FE models were developed to verify
325 160
425 160
Table 4
555 110
595 90 Temperature-dependent mechanical properties of AlSi10Mg.
925 100 Temperature Yield strength YS (MPa) Elastic modulus (GPa) Linear
1225 110 (℃) thermal
1525 115 Deposition Substrate Deposition Substrate
expansion
1825 120 [41] [15] [41] [15]
coefficient
Specific heat (JKg− 1 ℃− 1 ) 25 900 [15]
100 960 (10− 6 /℃)
300 1020
500 1125 25 314 195 76 69 21.7
665 1040 100 298 150 65 67 22.5
685 1040 200 236 105 59 62 23.5
825 1075 300 143 70 49 53 23.2
1287 1075 400 25 30 18 41 25.5

6
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

Fig. 5. FE model of (a) typical 3-layer and (b) 30-layer.

the efficiencies and accuracies of the EBCM and DMM, as shown in Fig. 5 in good agreement with the experimental results. These results provide
(a). The substrate had dimensions of 10 × 10 × 1 mm3. The scanning confidence in the fitting of the EBCM using the LBM model. It should be
region had a length of 2 mm and width of 2 mm. mentioned that the simulation results slightly overestimate the pene­
A 30-layer high-efficient model corresponding to the in-situ experi­ tration while slightly underestimates the width of the molten pool. As
ment was then developed. The laser scanning area was 10 × 10 mm2. Kollmannsberger et al. reported, this might be attributed the adopted
The substrate had a length of 30 mm, width of 30 mm, and thickness of Gaussian heat source with limited parameters does not accurately
2 mm, as shown in Fig. 5(b). Hexahedral elements with dimensions of represent the molten pool morphology [56].
50 × 50 × 30 µm3 were employed in the fine mesh zone with di­
mensions of 2 mm × 1200 µm × 120 µm and 10 mm × 4.2. Thermo-mechanical results
1200 µm × 120 µm in three-layer and 30-layer FE models,
respectively. This section mainly shows the comparison between the high-efficient
The models with different meshes and geometries were developed by model and the normal model. The EBCM used in the FE model is fitted by
the Abaqus scripting interface of Python. The Dirichlet boundary con­ the thermal history of the LBM model combined with the scanning
dition was set through another scripting interface of Fortran: user sub­ strategy. Fig. 7(a) shows the similar temperature distribution of a typical
routine DISP. The FE model simulations were carried out on a computer three-layer FE model using the EBCM and DMM. A nonsymmetric
with an Intel Corel(TM) i7 8700 CPU and RAM of 16 GB. thermal field was observed owing to the different material properties
between the solid and powder. The heat flow transfers mainly through
4. Results and discussion the previously scanned consolidated region, which has a higher con­
ductivity than that of the powder in the unmelted region. Hence, at the
4.1. Validation of LBM thermal model top surface of the deposited parts, the temperature of the consolidated
region is considerably higher than that of the powder bed [57,58]. Fig. 7
The temperature field calculated by the LBM model is shown in Fig. 6 (b) present the comparison of the temperature history of points P1 and
(a) at a power of 340 W and scanning speed of 1500 mm/s. The tem­ P2 for normal and EBCM + DMM model, respectively. The results show
perature gradient in the front of the laser irradiation zone is steeper than a good consistency.
that in the rear, and thus the cloud map of the temperature field exhibits The calculation costs, including the average number of iterations in
a comet tail appearance. Fig. 6(b) shows the cross- sectional geometries an increment, the number of increments, the system time and the total
of the molten pool obtained by the experiments. With the increase in the CPU time, for the different models are presented in Table 5. FE method
laser power, the volume of the molten pool gradually increases. The splits the total time period of a nonlinear analysis into several in­
simulated width and depth of the molten pool under different powers are crements, as shown in Fig. 7(c). Automatic time increment is the default

7
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

Fig. 6. (a) Temperature field of LBM at 340 W and 1500 mm/s; (b) dimensions of molten pool from simulation and experiment at various power and 1500 mm/s.

selection. Abaqus uses the initial increment size at the beginning of the distribution patterns and magnitudes; high S11 and S22 are observed near
increment. The size of the subsequent time increments is automatically the edge of the deposition zone in the X and Y directions, respectively.
adjusted based on the solution converges. System time represents the The magnitude of S33 with a uniform distribution is considerably lower
time required to execute all I/O requests, including data storage after than those of S11 and S22. The potential cracks at the edge of the deposit
each increment and data transfer, etc. Total CPU time contains both the could partly be attributed to these factors. This phenomenon is similar to
system time and computational time. those in the previous studies [13,58] which also verifies that the tem­
The model with EBCM can significantly reduce the number of in­ perature field obtained by the model with the EBCM and DMM can also
crements and the calculation time. Compared to other regions, the high- be used to predict the residual stress field reasonably, compared to the
speed moving heat source with a high- energy input leads to a wide normal model. Owing to the difference in the temperature field, the
quick temperature variation [59] in the region near the molten pool distributions and magnitudes of the stress components are slightly
which was predefined by EBCM(as shown in Fig. 3). Owing to the different, which may cause errors in the prediction, particularly S11 and
relatively fine temperature changes (lower temperature gradient and S22. For example, the maximum S22 values of both EBCM and
heating/cooling rates), the solution of the other regions more easily EBCM + DMM are approximately 15 MPa lower than the normal value.
converges at each increment which means that each increment requires
an average of 4 iterations. So Abaqus allows an extension of the incre­
ment in the model with EBCM during the simulation. Consequently, for a 4.3. Validation of high-efficient model
certain thermal simulation of the LPBF process, the total number of
required increments is reduced which leads to a lower computational This section mainly shows the comparison between the high-efficient
cost and system time. model and the experiments. Fig. 9 shows the temperature evolution
The model using the DMM can reduce the number of meshes and size obtained by the in-situ measurement and simulation using EBCM+DMM.
of the stiffness matrix of the FE analysis. However, it could be concluded Good agreements between the measurement and simulation were
that DMM hardly favors the iteration convergence but improves the observed at different laser powers. As the variation trend of TC2 with the
iteration efficiency if the average number of iterations in each increment deposition time is very similar to that of TC1, only the temperature data
keeps steady. DMM needs more system time for transferring data be­ of TC1 are presented in this manuscript. Notably, the heating and
tween old and new mesh. However, the less the mesh data, the less the cooling processes alternately occur during the deposition of each layer.
system time. The slight decrease in system time indicates its weak in­ In the LPBF process, the thermal accumulation of the substrate increased
fluence on the efficiency of DMM. the temperature. When the deposition was completed, owing to the
Fig. 8 presents the residual von Mises stress and stress components dissipation of heat accumulation, the substrate gradually cooled. The
(S11 in the X direction, S22 in the Y direction, and S33 in the Z direction of thermal histories of TC1 obtained by the experiment and simulation at
the coordinate system) obtained using the sequential coupling temper­ different scanning speeds are also shown in Fig. 10. The variation trends
ature fields with the different models. The tensile residual stress at the of the temperature with the deposition time comply with those in Fig. 9.
top surface originates mainly from shrinkage restriction by surrounding With the increase in the power and decrease in the scanning speed, the
materials during the cooling [60]. All models provide similar temperature increases, owing to the higher energy input. The rate and
range of temperature increase are positively correlated with the heat

8
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

Fig. 7. (a) Temperature field of typical 3-layer model at first deposition layer at 0.004 s, (b) temperature histories of points P1 and P2, and (c) time scheme of
nonlinear analysis.

experiment can be calculated as:


Table 5 ⃒ ⃒
The calculation time of various 3-layer model. ∑n ⃒ ⃒
⃒(Tsim )q − (Texp )q ⃒
100 ⃒ (Texp ) ⃒
Normal DMM EBCM EBCM+DMM q=1⃒
q ⃒
%Error = (23)
Average number of iteration in 8 8 4 4 ns
an increment
Number of increment 5646 5843 618 624 where (Tsim )q and (Texp )q are the simulation and experimental values in
System time 2838 s 2682 s 276 s 164 s the qth increment, respectively, and ns is the number of increments. The
Total CPU time 32,457 s 21,812 s 2187 s 1274 s
relative error of TC1 at different processing parameters are presented in
Table 6.
input. For example, the heating rate at 340 W and 2000 mm/s was The computation costs (wall clock time) of the normal and high-
reduced. efficient model at a power of 300 W and scanning speed of 1500 mm/
In the cooling stage, the simulated temperature decreased more s are 12,000 h and 18 h, respectively. Because the normal model
rapidly than the in-situ experimental data. A possible reason for this required large computation times, in this study, the time of the first-10-
behavior is that the environment temperature in the FE model is set to a tracks deposition is used to linearly estimate the calculation cost of the
constant of 25 ◦ C. The lower sink temperature leads to a faster heat 30-layer deposition. In the LPBF process, the calculation time increases
dissipation. The relative temperature error between the simulation and nonlinearly with the number of elements in the FE model. Thus, the

9
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

Fig. 8. Residual stress (a) von Mises stress and stress components, (b) S11, (c) S22, (d) S33 of various model.

Fig. 9. Temperature evolution of TC1 obtained from in-situ measurement and simulation at (a) 300 W, (b) 340 W, (c) 380 W with constant scanning speed
1500 mm/s.

high-efficient model could save more time. The high-efficient model deposit section with a height of 0.50 mm at 340 W/1500 mm/s. They
provides more obvious benefits for the larger model, as the larger exhibit similar distributions, in which the residual stress in the center
normal model is more demanding in terms of computing resources (such and later depositions of the as-fabricated parts are higher than those in
as CPU and RAM). the other regions. Fig. 11(b) and (c) present the residual stress obtained
In the microindentation test, the LPBF samples were initially ground by the numerical simulation and measurement at a scanning speed of
and polished, which led to a reduction in height and partial release of 1500 mm/s and power of 340 W, respectively. In general, the value
stress. Fig. 11(a) shows the simulated and tested residual stress of the obtained by the numerical simulation using the LPBF material properties

10
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

Fig. 10. Temperature evolution of TC1 obtained from in-situ measurement and simulation at (a) 1000 mm/s, (b) 1500 mm/s, (c) 2000 mm/s with constant laser
power of 340 W.

is 13 MPa higher than the value obtained by the microindentation


Table 6
method on average.
Error of TC1 between simulation and experiment.
Case 1 Case 2 Case 3 Case 4 Case 5 5. Conclusions
Error 4.5% 4.8% 6.3% 6.6% 4.3%
In this study, a 3D high-efficient FE model was developed to predict
the temperature of the PBF process and validated by in-situ experiments.
The simulated temperature fields obtained by the high-efficient model
and normal model were used for the sequential coupling statics analysis.
The stress field was then verified with the microindentation method. The
conclusions of this study can be summarized as follows.

Fig. 11. Comparison of residual stress prediction and experiment: (a) the case of 340 W/1500 mm/s; (b) using different powers under 1500 mm/s); (c) using
different scanning speeds under 340 W.

11
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

(1) The temperature results obtained from LBM model were used to [10] R.K. Ganeriwala, M. Strantza, W.E. King, B. Clausen, T.Q. Phan, L.E. Levine, D.
W. Brown, N.E. Hodge, Evaluation of a thermomechanical model for prediction of
feed the coupled thermo-mechanical FE model by a scripting
residual stress during laser powder bed fusion of Ti-6Al-4V, Addit. Manuf. 27
interface, achieving the multi-scale simulation of AM processes. (2019) 489–502, https://doi.org/10.1016/j.addma.2019.03.034.
(2) In-situ temperature and residual stress measurements were car­ [11] M. Busi, N. Kalentics, M. Morgano, S. Griffiths, A.S. Tremsin, T. Shinohara, R. Logé,
ried out to calibrate the numerical model. Comparison of the C. Leinenbach, M. Strobl, Nondestructive characterization of laser powder bed
fusion parts with neutron Bragg edge imaging, Addit. Manuf. 39 (2021), 101848,
thermo-mechanical results between simulation and experiment https://doi.org/10.1016/j.addma.2021.101848.
presented a good agreement. [12] Z. Chen, H. Ye, H. Xu, Distortion control in a wire-fed electron-beam thin-walled
(3) A high-efficient FE model for LPBF was proposed by combining Ti-6Al-4V freeform, J. Mater. Process. Technol. 258 (2018) 286–295, https://doi.
org/10.1016/j.jmatprotec.2018.04.008.
the EBCM and DMM, achieving a significant reduction on the [13] E.R. Denlinger, M. Gouge, J. Irwin, P. Michaleris, Thermomechanical model
calculation cost without loss of accuracy. development and in situ experimental validation of the laser powder-bed fusion
(4) The EBCM displayed prominent high-efficiency since it favors process, Addit. Manuf. 16 (2017) 73–80, https://doi.org/10.1016/j.
addma.2017.05.001.
convergence and enlarges the increment size in LPBF simulation. [14] X. Liang, Q. Chen, L. Cheng, D. Hayduke, A.C. To, Modified inherent strain method
Meanwhile, DMM further improved the speed of convergence due for efficient prediction of residual deformation in direct metal laser sintered
to the size reduction of the stiffness matrix. components, Comput. Mech. 64 (6) (2019) 1719–1733, https://doi.org/10.1007/
s00466-019-01748-6.
[15] L. Wang, X. Jiang, Y. Zhu, X. Zhu, J. Sun, B. Yan, An approach to predict the
CRediT authorship contribution statement residual stress and distortion during the selective laser melting of AlSi10Mg parts,
The, Int. J. Adv. Manuf. Technol. 97 (9–12) (2018) 3535–3546, https://doi.org/
10.1007/s00170-018-2207-3.
Yang Cao: Conceptualization, Formal analysis, Investigation, [16] C. Chen, Z. Xiao, H. Zhu, X. Zeng, Distribution and evolution of thermal stress
Writing – original draft. Xin Lin: Conceptualization, Writing – review & during multi-laser powder bed fusion of Ti-6Al-4 V alloy, J. Mater. Process.
editing, Supervision. Nan Kang: Conceptualization, Methodology, Technol. 284 (2020), 116726, https://doi.org/10.1016/j.jmatprotec.2020.116726.
[17] J.H.K. Tan, S.L. Sing, W.Y. Yeong, Microstructure modelling for metallic additive
Writing – review & editing. Liang Ma: Conceptualization, Visualization.
manufacturing: a review, Virtual Phys. Prototyp. 15 (1) (2019) 87–105, https://
Lei Wei: Data curation, Methodology. Min Zheng: Visualization, doi.org/10.1080/17452759.2019.1677345.
Methodology. Jun Yu: Investigation, Methodology. Dongjian Peng: [18] V. Fallah, M. Amoorezaei, N. Provatas, S.F. Corbin, A. Khajepour, Phase-field
Investigation, Methodology. Weidong Huang: Investigation, Project simulation of solidification morphology in laser powder deposition of Ti–Nb alloys,
Acta Mater. 60 (4) (2012) 1633–1646, https://doi.org/10.1016/j.
administration. actamat.2011.12.009.
[19] J. Yang, H. Yu, H. Yang, F. Li, Z. Wang, X. Zeng, Prediction of microstructure in
selective laser melted Ti6Al4V alloy by cellular automaton, J. Alloy. Compd. 748
Declaration of Competing Interest (2018) 281–290, https://doi.org/10.1016/j.jallcom.2018.03.116.
[20] S. Sunny, H. Yu, R. Mathews, A. Malik, W. Li, Improved grain structure prediction
The authors declare that they have no known competing financial in metal additive manufacturing using a dynamic kinetic Monte Carlo framework,
Addit. Manuf. 37 (2021), 101649, https://doi.org/10.1016/j.addma.2020.101649.
interests or personal relationships that could have appeared to influence [21] D. Sun, M. Zhu, S. Pan, D. Raabe, Lattice Boltzmann modeling of dendritic growth
the work reported in this paper. in a forced melt convection, Acta Mater. 57 (6) (2009) 1755–1767, https://doi.
org/10.1016/j.actamat.2008.12.019.
[22] R. Ammer, M. Markl, U. Ljungblad, C. Koerner, U. Ruede, Simulating fast electron
Acknowledgements beam melting with a parallel thermal free surface lattice Boltzmann method,
Comput. Math. Appl. 67 (2) (2014) 318–330, https://doi.org/10.1016/j.
This work was supported by the National Key Research and Devel­ camwa.2013.10.001.
[23] A. Xu, L. Shi, T.S. Zhao, Accelerated lattice Boltzmann simulation using GPU and
opment Programme of China (Grant No. 2016YFB1100602 and No. OpenACC with data management, Int. J. Heat Mass Transf. 109 (2017) 577–588,
2016YFB1100100) and National Natural Science Foundation of China https://doi.org/10.1016/j.ijheatmasstransfer.2017.02.032.
(Grant No. 52005411). Thanks for many useful helps form Mr. Xufei Lu, [24] M. Bayat, S. Mohanty, J.H. Hattel, Multiphysics modelling of lack-of-fusion voids
formation and evolution in IN718 made by multi-track/multi-layer L-PBF, Int. J.
Mr. Bo Yao, Mr. Xing Wang and Mr. Wenbo Lu in our group. Heat. Mass Transf. 139 (2019) 95–114, https://doi.org/10.1016/j.
ijheatmasstransfer.2019.05.003.
References [25] A.J. Dunbar, E.R. Denlinger, M.F. Gouge, P. Michaleris, Experimental validation of
finite element modeling for laser powder bed fusion deformation, Addit. Manuf. 12
(2016) 108–120, https://doi.org/10.1016/j.addma.2016.08.003.
[1] D. Herzog, V. Seyda, E. Wycisk, C. Emmelmann, Additive manufacturing of metals,
[26] M. Chiumenti, E. Neiva, E. Salsi, M. Cervera, S. Badia, J. Moya, Z. Chen, C. Lee,
Acta Mater. 117 (2016) 371–392, https://doi.org/10.1016/j.actamat.2016.07.019.
C. Davies, Numerical modelling and experimental validation in selective laser
[2] T. DebRoy, H.L. Wei, J.S. Zuback, T. Mukherjee, J.W. Elmer, J.O. Milewski, A.
melting, Addit. Manuf. 18 (2017) 171–185, https://doi.org/10.1016/j.
M. Beese, A. Wilson-Heid, A. De, W. Zhang, Additive manufacturing of metallic
addma.2017.09.002.
components – process, structure and properties, Prog. Mater. Sci. 92 (2018)
[27] M. Carraturo, J. Jomo, S. Kollmannsberger, A. Reali, F. Auricchio, E. Rank,
112–224, https://doi.org/10.1016/j.pmatsci.2017.10.001.
Modeling and experimental validation of an immersed thermo-mechanical part-
[3] X. Lu, X. Lin, M. Chiumenti, M. Cervera, J. Li, L. Ma, L. Wei, Y. Hu, W. Huang,
scale analysis for laser powder bed fusion processes, Addit. Manuf. 36 (2020),
Finite element analysis and experimental validation of the thermomechanical
101498, https://doi.org/10.1016/j.addma.2020.101498.
behavior in laser solid forming of Ti-6Al-4V, Addit. Manuf. 21 (2018) 30–40,
[28] J. Parvizian, A. Düster, E. Rank, Finite cell method, Comput. Mech. 41 (1) (2007)
https://doi.org/10.1016/j.addma.2018.02.003.
121–133, https://doi.org/10.1007/s00466-007-0173-y.
[4] X. Lu, X. Lin, M. Chiumenti, M. Cervera, Y. Hu, X. Ji, L. Ma, W. Huang, In situ
[29] R.J. Williams, C.M. Davies, P.A. Hooper, A pragmatic part scale model for residual
measurements and thermo-mechanical simulation of Ti-6Al-4V laser solid forming
stress and distortion prediction in powder bed fusion, Addit. Manuf. 22 (2018)
processes, Int. J. Mech. Sci. 153–154 (2019) 119–130, https://doi.org/10.1016/j.
416–425, https://doi.org/10.1016/j.addma.2018.05.038.
ijmecsci.2019.01.043.
[30] M. Bayat, C.G. Klingaa, S. Mohanty, D. De Baere, J. Thorborg, N.S. Tiedje, J.
[5] X. Lu, X. Lin, M. Chiumenti, M. Cervera, Y. Hu, X. Ji, L. Ma, H. Yang, W. Huang,
H. Hattel, Part-scale thermo-mechanical modelling of distortions in laser powder
Residual stress and distortion of rectangular and S-shaped Ti-6Al-4V parts by
bed fusion – analysis of the sequential flash heating method with experimental
directed energy deposition: modelling and experimental calibration, Addit. Manuf.
validation, Addit. Manuf. 36 (2020), 101508, https://doi.org/10.1016/j.
26 (2019) 166–179, https://10.1016/j.addma.2019.02.001.
addma.2020.101508.
[6] K. Carpenter, A. Tabei, On residual stress development, prevention, and
[31] P. Michaleris, Modeling metal deposition in heat transfer analyses of additive
compensation in metal additive manufacturing, Materials 13 (2) (2020), https://
manufacturing processes, Finite Elem. Anal. Des. 86 (2016) 51–60, https://doi.
doi.org/10.3390/ma13020255.
org/10.1016/j.finel.2014.04.003.
[7] X. Lu, M. Chiumenti, M. Cervera, J. Li, X. Lin, L. Ma, G. Zhang, E. Liang, Substrate
[32] K. Zeng, D. Pal, H.J. Gong, N. Patil, B. Stucker, Comparison of 3DSIM thermal
design to minimize residual stresses in directed energy deposition AM processes,
modelling of selective laser melting using new dynamic meshing method to ANSYS,
Mater. Des. 202 (2021), 109525, https://doi.org/10.1016/j.matdes.2021.109525.
Mater. Sci. Technol. 31 (8) (2014) 945–956, https://doi.org/10.1179/
[8] C. Li, C.H. Fu, Y.B. Guo, F.Z. Fang, A multiscale modeling approach for fast
1743284714Y.0000000703.
prediction of part distortion in selective laser melting, J. Mater. Process. Technol.
[33] J. Baiges, M. Chiumenti, C.A. Moreira, M. Cervera, R. Codina, An adaptive finite
229 (2016) 703–712, https://doi.org/10.1016/j.jmatprotec.2015.10.022.
element strategy for the numerical simulation of additive manufacturing processes,
[9] I. Setien, M. Chiumenti, S. van der Veen, M. San Sebastian, F. Garciandía,
Addit. Manuf. 37 (2021), 101650, https://doi.org/10.1016/j.addma.2020.101650.
A. Echeverría, Empirical methodology to determine inherent strains in additive
manufacturing, Comput. Math. Appl. 78 (7) (2019) 2282–2295, https://doi.org/
10.1016/j.camwa.2018.05.015.

12
Y. Cao et al. Additive Manufacturing 46 (2021) 102187

[34] Z. Luo, Y. Zhao, A survey of finite element analysis of temperature and thermal [48] M. Zheng, L. Wei, J. Chen, Q. Zhang, C. Zhong, X. Lin, W. Huang, A novel method
stress fields in powder bed fusion additive manufacturing, Addit. Manuf. 21 (2018) for the molten pool and porosity formation modelling in selective laser melting,
318–332, https://doi.org/10.1016/j.addma.2018.03.022. Int. J. Heat Mass Transf. 140 (2019) 1091–1105, https://doi.org/10.1016/j.
[35] S. Suresh, A.E. Giannakopoulos, A new method for estimating residual stresses by ijheatmasstransfer.2019.06.038.
instrumented sharp indentation, Acta Mater. 46 (16) (1998) 5755–5767, https:// [49] M. Zheng, L. Wei, J. Chen, Q. Zhang, J. Li, S. Sui, G. Wang, W. Huang, Surface
doi.org/10.1016/S1359-6454(98)00226-2. morphology evolution during pulsed selective laser melting: numerical and
[36] R. Moharrami, M. Sanayei, Developing a method in measuring residual stress on experimental investigations, Appl. Surf. Sci. 496 (2019), 143649, https://doi.org/
steel alloys by instrumented indentation technique, Measurement 158 (2020), 10.1016/j.apsusc.2019.143649.
107718, https://doi.org/10.1016/j.measurement.2020.107718. [50] P. Wei, Z. Wei, Z. Chen, Y. He, J. Du, Thermal behavior in single track during
[37] S. Carlsson, P.L. Larsson, On the determination of residual stress and strain fields selective laser melting of AlSi10Mg powder, Appl. Phys. A 123 (2017) 604, https://
by sharp indentation testing. Part II: experimental investigation, Acta Mater. 49 doi.org/10.1007/s00339-017-1194-9.
(12) (2001) 2193–2203, https://doi.org/10.1016/S1359-6454(01)00123-9. [51] H. Kobatake, J. Brillo, J. Schmitz, P.Y. Pichon, Surface tension of binary Al-Si
[38] S. Carlsson, P.L. Larsson, On the determination of residual stress and strain fields liquid alloys, J. Mater. Sci. 50 (9) (2015) 3351–3360, https://doi.org/10.1007/
by sharp indentation testing.: Part I: theoretical and numerical analysis, Acta s10853-015-8883-6.
Mater. 49 (12) (2001) 2193–2203, https://doi.org/10.1016/S1359-6454(01) [52] H.M. Khan, M.H. Dirikolu, E. Koç, Z.C. Oter, Numerical investigation of heat
00122-7. current study across different platforms in SLM processed multi-layer AlSi10Mg,
[39] Y.L. Hu, X. Lin, K. Song, X.Y. Jiang, H.O. Yang, W.D. Huang, Effect of heat input on Optik 170 (2018) 82–89, https://doi.org/10.1016/j.ijleo.2018.05.081.
cracking in laser solid formed DZ4125 superalloy, Opt. Laser Technol. 86 (2016) [53] J. Wu, L. Wang, X. An, Numerical analysis of residual stress evolution of AlSi10Mg
1–7, https://doi.org/10.1016/j.optlastec.2016.06.008. manufactured by selective laser melting, Optik 137 (2017) 65–78, https://doi.org/
[40] F. Liu, X. Lin, G. Yang, M. Song, J. Chen, W. Huang, Microstructure and residual 10.1016/j.ijleo.2017.02.060.
stress of laser rapid formed Inconel 718 nickel-base superalloy, Opt. Laser Technol. [54] B. Chen, S.K. Moon, X. Yao, G. Bi, J. Shen, J. Umeda, K. Kondoh, Strength and
43 (1) (2011) 208–213, https://doi.org/10.1016/j.optlastec.2010.06.015. strain hardening of a selective laser melted AlSi10Mg alloy, Scr. Mater. 141 (2017)
[41] Y. Cao, X. Lin, Q.Z. Wang, S.Q. Shi, L. Ma, N. Kang, W.D. Huang, Microstructure 45–49, https://doi.org/10.1016/j.scriptamat.2017.07.025.
evolution and mechanical properties at high temperature of selective laser melted [55] N. Takata, H. Kodaira, K. Sekizawa, A. Suzuki, M. Kobashi, Change in
AlSi10Mg, J. Mater. Sci. Technol. 62 (2021) 162–172, https://doi.org/10.1016/j. microstructure of selectively laser melted AlSi10Mg alloy with heat treatments,
jmst.2020.04.066. Mater. Sci. Eng. A 704 (2017) 218–228, https://doi.org/10.1016/j.
[42] P. Levin, A general solution of 3-D quasi-steady-state problem of a moving heat msea.2017.08.029.
source on a semi-infinite solid, Mech. Res. Commun. 35 (3) (2008) 151–157, [56] S. Kollmannsberger, M. Carraturo, A. Reali, F. Auricchio, Accurate prediction of
https://doi.org/10.1016/j.mechrescom.2007.09.003. melt pool shapes in laser powder bed fusion by the non-linear temperature
[43] X. Wang, E. Pan, Three-dimensional quasi-steady-state problem of moving heat and equation including phase changes, Integr. Mater. Manuf. Innov. 8 (2) (2019)
diffusion sources in an infinite solid, Mech. Res. Commun. 35 (7) (2008) 475–482, 167–177, https://doi.org/10.1007/s40192-019-00132-9.
https://doi.org/10.1016/j.mechrescom.2008.05.005. [57] L. Parry, I.A. Ashcroft, R.D. Wildman, Understanding the effect of laser scan
[44] B. Cheng, J. Lydon, K. Cooper, V. Cole, P. Northrop, K. Chou, Melt pool sensing and strategy on residual stress in selective laser melting through thermo-mechanical
size analysis in laser powder-bed metal additive manufacturing, J. Manuf. Process. simulation, Addit. Manuf. 12 (2016) 1–15, https://doi.org/10.1016/j.
32 (2018) 744–753, https://doi.org/10.1016/j.jmapro.2018.04.002. addma.2016.05.014.
[45] I.A. Roberts, C.J. Wang, R. Esterlein, M. Stanford, D.J. Mynors, A three- [58] B. Cheng, S. Shrestha, K. Chou, Stress and deformation evaluations of scanning
dimensional finite element analysis of the temperature field during laser melting of strategy effect in selective laser melting, Addit. Manuf. 12 (2016) 240–251,
metal powders in additive layer manufacturing, Int. J. Mach. Tool. Manuf. 49 https://doi.org/10.1016/j.addma.2016.05.007.
(12–13) (2009) 916–923, https://doi.org/10.1016/j.ijmachtools.2009.07.004. [59] S. Price, B. Cheng, J. Lydon, K. Cooper, K. Chou, On process temperature in
[46] Y. Jiang, E. Li, X.Q. Zhang, Q.G. Wu, Y.H. Yap, Superposition method for the powder-bed electron beam additive manufacturing: model development and
simulation of heat transfer, Int. J. Heat Mass Transf. 120 (2018) 914–922, https:// validation, J. Manuf. Sci. Eng. 136 (6) (2014), 061018, https://doi.org/10.1115/
doi.org/10.1016/j.ijheatmasstransfer.2017.12.129. 1.4028485.
[47] Z. Guo, B. Shi, C. Zheng, A coupled lattice BGK model for the Boussinesq equations, [60] T. Simson, A. Emmel, A. Dwars, J. Böhm, Residual stress measurements on AISI
Int. J. Numer. Methods Fluids 39 (4) (2002) 325–342, https://doi.org/10.1002/ 316L samples manufactured by selective laser melting, Addit. Manuf. 17 (2017)
fld.337. 183–189, https://doi.org/10.1016/j.addma.2017.07.007.

13

You might also like