Sie sind auf Seite 1von 7

Received: 30 November 2018 Revised: 21 December 2018 Accepted: 26 December 2018

DOI: 10.1002/mdp2.28

SPECIAL ISSUE ARTICLE

Finite element–based numerical modeling framework for


additive manufacturing process

Farshid Hajializadeh | Ayhan Ince

Department of Mechanical, Industrial and


Abstract
Aerospace Engineering, Concordia
University, Montreal, Quebec, Canada Additive manufacturing (AM) process has extensively been used to fabricate
metal parts for large variety of applications. Residual stresses are inevitable
Correspondence
Ayhan Ince, Department of Mechanical,
in the AM process since material experiences heating and cooling cycles.
Industrial and Aerospace Engineering, Implementing finite element (FE) analysis tool to predict residual stress distri-
Concordia University, Montreal, Quebec, butions could be of great importance in many applications. Developing an
Canada.
Email: ayhan.ince@concordia.ca FE‐based modeling framework to accurately simulate residual stresses in a rea-
sonably reduced computational time is highly needed. The FE‐based modeling
approach presented here to simulate direct metal deposition (DMD) of AISI
304 L aims to significantly reduce computation cost by implementing an adap-
tive mesh coarsening algorithm integrated with the FE method. Simulations
were performed by the proposed approach, and the results were found in good
agreement with conventional fine mesh configuration. The proposed modeling
framework offers a potential solution to substantially reduce the computational
time for simulating the AM process.

KEYWORDS
additive manufacturing, finite element analysis, mesh coarsening

1 | INTRODUCTION

The additive manufacturing (AM) process refers to repeated addition of material onto an existing substrate which has
already been deposited or existed as the base. This method has been used for fabricating and forming net and near
net shape components for different industries such as aerospace, biomedical, and automobile applications. The AM pro-
cess normally includes the heating, melting, and solidification cycles of a material in the form of a feeding wire or pow-
der using an electron or laser beam which provides high density energy to melt the material in a fraction of time.
Basically, the AM processes resemble the multi‐pass welding methods that were extensively studied and reported in
the literature to join metallic parts, and they also have inherited most of their features.1-3 Commercial finite element‐
based packages were developed by different companies in order to simulate welding processes and study different
parameters that have significant impact on the geometrical distortions of the weld joints.4 Cyclic heating and cooling
of material may impose significant distortions that eventually results in formation of residual stresses in the AM parts;
as it is seen in the welding process of metallic alloys.5,6 The temperature history evaluation during the deposition
of material is a very complex phenomenon and directly affects the whole modeling procedure.7,8 A number of
researchers have focused on the thermal source modeling and its effects on the temperature distribution around the
scanning line.8-10 As it has been widely reported in the literature, the tensile residual stresses have detrimental effects
on the fatigue performance of a structural component. A component produced using welding and the AM process

Mat Design Process Comm. 2019;1:e28. wileyonlinelibrary.com/journal/mdp2 © 2019 John Wiley & Sons, Ltd. 1 of 7
https://doi.org/10.1002/mdp2.28
2 of 7 HAJIALIZADEH AND INCE

are regarded to have degraded fatigue performance compared with parts produced by conventional manufacturing
methods such as machining and forging. Furthermore, the formation of residual stresses deteriorates the corrosion resis-
tance of the material in corrosive environment. Therefore, it is crucial to evaluate the extent of the residual stresses
induced inside of an additively manufactured component.11
The AM process involves a very large number of parameters that have significant impact on the properties of the
manufactured part related to the material properties, boundary conditions, and geometry complexities of the component
as well as the deposition features.10,11 Implementing experimental procedures to determine impact of every variable on
the distortion and residual stress distribution inside the built medium appears to be very expensive and also requires
enormous time and effort. Therefore, numerical analysis methods such as a finite element (FE) method offer a very suit-
able alternative in order to measure and quantify the effect of process parameters. In the recent decades, different strat-
egies and procedure have been introduced to investigate the thermomechanical phenomena during welding and AM
processes.12-18 Generally, the FE analysis of a thermomechanical process includes thermal analysis and introduces
the temperature history to the mechanical, ie, structural analysis as the thermal load which is known as weakly coupled
or uncoupled method. In this approach, the mechanical response has no impact on the thermal behavior or response of
the material. On the other hand, the FE analysis of a thermomechanical process can also be accomplished by
implementing fully coupled approach, which takes into account all the effects of the structural response on the thermal
properties' variations. The latter is considered to be more accurate representation of the AM process. However,
implementing the fully coupled approach increases the computational time and cost considerably by about four to five
times which is not considered as beneficial in the numerical modeling approach.12 A simplified method known as the
inherent strain method was proposed by Ueda19 in order to evaluate the residual stresses and distortions in the welded
joints in a short period of time compared with conventional FE analysis. The aforementioned method was also imple-
mented in simulating the powder‐bed fusion applications.20 The drawback of the aforementioned method is that it can-
not take into account the effect of different laser scanning paths and geometrical effects since the inherent strain is
constant across the whole volume.21 Bugattin and Semeraro22 reported that even though employing the inherent strain
method demonstrated very high potential in evaluating the residual stresses and distortions of a specific geometry in
AM processes, the calibration strategy fails to predict the distortion and consequently the residual stress of different
geometries.
In spite of the material‐built process similarity with multi‐pass welding, the AM process requires additional consid-
erations unlike multi‐pass welding. In welding, it is usually assumed that the material already exists ahead of the heat
source. However, in the AM process modeling, the material is added in every increment simultaneously with motion of
the heat source which is known as the element activation procedure. The element activation mostly includes three main
methods: quiet element, inactive element, and hybrid activation methods. More details on these approaches could be
found in Denlinger et al.13 In the present study, the hybrid element activation method is adopted to deal with deposition
of the material in every single increment for different layers since it inheres advantageous of the other approaches. The
main drawback of implementing the FE analysis in modeling the AM processes is its high computational cost and time.
In order to get accurate results, the mesh used for thermal and mechanical analysis has to be fine enough that could
capture the high temperature gradients and consequently high stress gradients during deposition. Therefore, an adap-
tive mesh‐based modeling framework is introduced in the present study to reduce computational time while obtaining
accurate results in the ABAQUS software package. The adaptive meshing procedure is represented, and the results are
compared with the conventional FE analysis.

2 | MODELI N G APPR OACH FO R SI M UL A T I NG ADDI T I V E


M A N U F A C T U R I N G PR O C E S S

The FE analysis approach to simulate direct metal deposition (DMD) of AISI 304 L is briefly reviewed and the material
properties and procedure of the proposed modeling approach is presented. The FE analysis of the AM process should
provide accurate simulation results in order to capture high stress gradients with a minimum possible run time. The
adaptive mesh‐based modeling procedure is accomplished based on a layer‐by‐layer mesh coarsening concept.
The material is added on the substrate with a very fine mesh, then in the next steps, mesh of the deposited layer is coars-
ened to lower the number of elements (or degrees of freedoms) of the entire model. Afterwards, results obtained from
previous fine meshed layer(s) is systematically imposed to the new model upon which a deposition of the new layer to
be built.
HAJIALIZADEH AND INCE 3 of 7

2.1 | Thermal analysis

The modeling framework developed here begins with a thermal analysis of the deposited layer using a fine mesh in
ABAQUS/STANDARD. Temperature‐dependent thermal properties are introduced into the model to realistically simu-
late the heat transfer process using UMATHT subroutine. The hybrid element activation approach was implemented in
UMATHT subroutine which limits the heat flow induced by the heat source to only those elements that are in
contact/passed by with the heat source by putting the heat conductivity of the material to zero. The thermal properties
of the AISI 304 L were given in Table 1.
Goldak double‐ellipsoid heat source model was also implemented using DFLUX subroutine to apply the body heat
flux into the depositing material16 as shown in Equation 1:

pffiffiffi  
6 3Pη −3 ðx−v2x tÞ2 þy22 þz22
Q¼ pffiffiffie a b c ; (1)
abc π π

where P is the source power; η is the efficiency; a, b, and c are the dimensions of the ellipsoid; x, y, and z are coordinate
system variables of the heat source center; vx is the traverse speed of the heat source in x‐direction, and t is time. The
thermal boundary conditions were assumed to have an accumulative coefficient of 10 W·m2·K, based on the simulations
performed in the literature.16 The energy loss due to the radiation heat transfer was ignored for the analysis.

2.2 | Mechanical (structural) analysis

The thermal history of the nodes obtained from the heat transfer analysis is applied to the mechanical analysis to eval-
uate the residual stress distribution after each built layer. The temperature‐dependent material properties for AISI 304 L
are given in Table 1. The UMAT subroutine was developed to account for the element activation and introduce the
temperature‐dependent material definition. The scaling factor of KE = 10−12 was considered to scale down the Young
modulus for inactive elements. Furthermore, the Von Mises isotropic plasticity was considered in the subroutine UMAT
and UHARD. A constant Poisson ratio of 0.3 and material density of 7800 (kg/m3) were considered for analysis.

2.3 | Adaptive mesh‐based modeling framework

Adaptive mesh coarsening technique during the analysis of the already deposited layers is not supported in the
ABAQUS software package. Indeed, there are available options for mesh refinement in ABAQUS such as adaptive
meshing and ALE method.14 However, these techniques cannot be applied to the thermomechanical analysis, espe-
cially, when the weakly coupled approach is adopted. The ABAQUS has also developed a useful tool called solution

TABLE 1 Thermal and mechanical properties15 of AISI 304 L

Temperature, Specific Heat, Conductivity, Thermal Expansion, Yield Stress, Young Modulus,
°
C J·kg·°C J·m·°C ×10−5/°C MPa GPa

20 462 14.6 1.70 319 198.5


100 496 15.1 1.74 279 193
200 512 16.1 1.80 238 185
300 525 17.9 1.86 217 176
400 540 18.0 1.91 198 167
600 577 20.8 1.96 177 159
800 604 23.9 2.02 112 151
1200 676 32.2 2.07 32 60
1300 692 33.7 2.11 19 20
1480 700 120 2.16 8 10
4 of 7 HAJIALIZADEH AND INCE

mapping that can only be used for mesh refinement manually when the present mesh is incapable of handling high dis-
tortions. However, the solution mapping cannot be applied to AM process simulations due to the continuous addition of
material to the built part.14 Therefore, a modeling framework in Python script was developed to implement the pro-
posed adaptive mesh coarsening in the AM processes.
The mesh coarsening modeling approach is performed in a layer‐by‐layer, as the AM process itself. After mechanical
analysis for a particular layer, eg, the first layer is completed, the developed Python code is run to map the results of the
fine mesh to the coarser mesh. The mapping process begins with extrapolating the solution results to the nodal point of
the fine mesh. Then, based on spatial locations of nodes of coarse mesh, the results are interpolated to the nodes of the
coarse mesh. Afterwards, the values for stress and strain tensor are interpolated to the integration point of elements in
the coarse mesh configuration using a linear shape function. Afterwards, the field values, eg, stress components, are
imported as predefined field variables for those coarsened layers, then, the new layer deposition begins. Based on the
temperature or stress gradients, the coarsening technique can be adjusted to be performed in certain intervals. If the
temperature during the deposition is so high that could melt the deposited layers, the coarsening better begins from
the third layer or even fourth layer deposition keeping the top deposited layers with fine mesh to be able to handle
the high temperature or stress gradients. A schematic representation for implementing adaptive mesh coarsening tech-
nique in the present study is illustrated in Figure 1.

3 | R E S U L T S AN D D I S C U S S I O N

An 18‐layer L‐shape part made from AISI 304 L was built using both the conventional and adaptive mesh‐based
models in order to assess the capability and accuracy of the proposed approach. Both models were generated with
the same thermal and mechanical features and also the same process parameters. Each L‐shape layer with length of
equally 12 (mm) and thickness and width of 1 (mm) was built using laser power of 250 (W), beam radius of 0.5
(mm), and traverse speed of 11.25 (mm/s). All the layers were built considering a cooling time approximately equal
to deposition time of each layer. The adaptive meshing is used in four steps to coarsen the fine mesh (five elements
in thickness) to coarse mesh (three elements in thickness) of each layer. Eight‐node linear heat transfer and structural
brick element were used for thermal and structural analysis, respectively. The bottom layer is fixed to act as a substrate
for the model.
Figures 2 and 3 represent S11 (or σxx) and S22 (or σyy) contour for the coarsened and fine meshes, respectively.
The numbers 1, 2, and 3 correspond to X, Y, and Z directions in the global coordinate system, respectively. The adap-
tive meshing on areas with consistent stress distribution demonstrates reasonable performance shown in Figures 2
and 3. The stress pattern remains almost the same for those areas, and the stress values are also found to be very
close to the fine mesh ones. On the other hand, very local stresses with high stress gradients do not show mapping
capability since the new coarse mesh is incapable of capturing very localized stress gradients. The S11 is almost near

FIGURE 1 A, Schematic representation of analysis sequences and B, sequence of finite element approach for adaptive mesh‐based
modeling
HAJIALIZADEH AND INCE 5 of 7

FIGURE 2 S11 contour of A, coarsened and B, fine mesh of modeling 18‐layer AISI 304 L

FIGURE 3 S22 contour of A, coarsened and B, fine mesh of modeling 18‐layer AISI 304 L

zero (−58 to 58 MPa) for both approaches in the middle of the wall. However, stress distribution in the stacking
direction (S22) illustrates high compressive values which could be important in fatigue performance of the AM proc-
essed parts.
Another important factor that justifies using the adaptive meshing in modeling of AM process is the computation
run time for the FE model. Analysis were performed by a Core i7 desktop PC with 16 GB RAM. Table 2 represents
the significant difference in the computation run time of the fine uniform mesh and adaptive mesh. It shows that the
analysis time to achieve an 18‐layer build with the fine uniform mesh is almost three times the time required for adap-
tive mesh approach. It should be noted that adaptive meshing run time converges to a specific value after several pairs
of layers. This could be beneficial especially when dealing with the simulation of a very large component. The total
mapping time of approximately 2 hours is added upon the computational time for the coarsening approach.

TABLE 2 Comparison of run time between fine mesh and adaptive mesh.

Layer No. 1‐3 4‐6 7‐9 10‐12 13‐15 16‐18 Total

Coarsening approach Run time, h 1:50 2:35 3:20 3:50 4:15 4:20 20:10 + 2:00 (mapping time)
Layer No. 1‐18

Fine mesh Run time, h 58:30 58:30


6 of 7 HAJIALIZADEH AND INCE

4 | CONCLUSION

In this study, a new modeling framework for implementing coarsening approach in simulation of the AM processes was
presented. The simulation results showed that the adaptive mesh coarsening approach is capable of capturing very
important aspects of the AM parts regarding the formation and distribution of compressive and tensile stresses in the
body of the component. Furthermore, the results showed significant improvement of the computation time in compar-
ison to the conventional FE analysis.

CONFLICT OF INTEREST
Farshid Hajializadeh and Ayhan Ince declare that they have no conflict of interest.

ORCID
Ayhan Ince https://orcid.org/0000-0002-2892-8551

R EF E RE N C E S
1. Chiumenti M, Cervera M, Salmi A, Agelet de Saracibar C, Dialami N, Matsui K. Finite element modeling of multi‐pass welding and
shaped metal deposition processes. Comput Methods Appl Mech Eng. 2010;199(37–40):2343‐2359.
2. Xiong Y, Hofmeister WH, Cheng Z, Smugeresky JE, Lavernia EJ, Schoenung JM. In situ thermal imaging and three‐dimensional finite
element modeling of tungsten carbide–cobalt during laser deposition. Acta Mater. 2009;57(18):5419‐5429.
3. Zheng J, Ince A, Tang L. Modeling and simulation of weld residual stresses and ultrasonic impact treatment of welded joints. Procedia
Eng. 2018;213:36‐47.
4. Ferro P, Berto F. Residual Stress Analysis on Welded Joints by Means of Numerical Simulation and Experiments. London, UK: IntechOpen
Limited; 2018.
5. Song L, Bagavath‐Singh V, Dutta B, Mazumder J. Control of melt pool temperature and deposition height during direct metal deposition
process. Int J Adv Manuf Technol. 2012;58(1–4):247‐256.
6. Heralić A, Christiansson A‐K, Lennartson B. Height control of laser metal‐wire deposition based on iterative learning control and 3D
scanning. Opt Lasers Eng. 2012;50(9):1230‐1241.
7. Svensson L, Gretoft B, Bhadeshia H. An analysis of cooling curves from the fusion zone of steel weld deposits. Scand J Metall. 1986;15(97):
e103.
8. Chae HM. A numerical and experimental study for residual stress evolution in low alloy steel during laser aided additive manufacturing
process; 2013.
9. Crespo A, Vilar R. Finite element analysis of the rapid manufacturing of Ti–6Al–4 V parts by laser powder deposition. Scr Mater.
2010;63(1):140‐143.
10. Baufeld B, Van der Biest O, Gault R. Additive manufacturing of Ti–6Al–4 V components by shaped metal deposition: microstructure and
mechanical properties. Mater Des. 2010;31:S106‐S111.
11. Trelewicz JR, Halada GP, Donaldson OK, Manogharan G. Microstructure and corrosion resistance of laser additively manufactured 316 L
stainless steel. Jom. 2016;68(3):850‐859.
12. Jayanath S, Achuthan A. A computationally efficient finite element framework to simulate additive manufacturing processes. J Manuf Sci
Eng. 2018;140(4):041009.
13. Denlinger ER, Irwin J, Michaleris P. Thermomechanical modeling of additive manufacturing large parts. J Manuf Sci Eng.
2014;136(6):061007.
14. Hibbett, Karlsson, Sorensen. ABAQUS/standard: User's Manual. (Vol. 1). Pawtucket, RI, USA: Hibbitt, Karlsson & Sorensen; 1998.
15. Moshayedi H, Sattari‐Far I. Numerical and experimental study of nugget size growth in resistance spot welding of austenitic stainless
steels. J Mater Process Technol. 2012;212(2):347‐354.
16. Michaleris P. Modeling metal deposition in heat transfer analyses of additive manufacturing processes. Finite Elem Anal Des.
2014;86:51‐60.
17. Denlinger ER, Heigel JC, Michaleris P. Residual stress and distortion modeling of electron beam direct manufacturing Ti‐6Al‐4 V. Proc
Inst Mech Eng B J Eng Manuf. 2015;229(10):1803‐1813.
18. Mukherjee T, Zhang W, DebRoy T. An improved prediction of residual stresses and distortion in additive manufacturing. Comput Mater
Sci. 2017;126:360‐372.
HAJIALIZADEH AND INCE 7 of 7

19. Ueda Y, Kim YC, Yuan MG. A predicting method of welding residual stress using source of residual stress (report I): characteristics of
inherent strain (source of residual stress) (mechanics, strength & structural design). Trans JWRI. 1989;18(1):135‐141.
20. Keller N, loshikhin V. New method for fast predictions of residual stress and distortion of AM parts. In: Solid Freeform Fabrication Sym-
posium (SFF); Aug. 2014; Austin, TX.
21. Parry LA. Investigation of residual stress in selective laser melting. University of Nottingham; 2018.
22. Bugatti M, Semeraro Q. Limitations of the inherent strain method in simulating powder bed fusion processes. Addit Manuf.
2018;23:329‐346.

How to cite this article: Hajializadeh F, Ince A. Finite element–based numerical modeling framework for
additive manufacturing process. Mat Design Process Comm. 2019;1:e28. https://doi.org/10.1002/mdp2.28

Das könnte Ihnen auch gefallen