Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 45 (2016) 171 – 174

3rd CIRP Conference on Surface Integrity (CIRP CSI)

Prediction of Residual Stress and Part Distortion in Selective Laser Melting

C. Li, J.F. Liu, Y.B. Guo*
Dept. of Mechanical Engineering, The University of Alabama, Tuscaloosa, AL35487, USA
* Corresponding author. Tel.: +1-205-348-2615; fax: +1-205-348-6419. E-mail address: yguo@eng.ua.edu

Abstract

Selective laser melting (SLM) is widely used to make functional metal parts. The high-temperature process will produce large tensile residual
stress (RS) which leads to part distortion and poor product performance. Traditional modeling approaches are not practical to predict residual
stress and part distortion due to the exceedingly high computational cost. In this study, two efficient multiscale modeling methods have been
developed to across microscale laser scan, mesoscale layer hatch, and macroscale part buildup for fast prediction of residual stress and part
distortion. A concept of equivalent heat source has been developed from the microscale laser scan model. In the “stress-thread” method, the
local residual stress field was predicted by the mesoscale layer hatch model using the equivalent heat source, then the residual stress field is
imported, i.e., “stress-thread”, to the macroscale part buildup model to predict residual stress and part distortion. In the temperature-thread
method, the powder–liquid–solid material transition has been incorporated. A body heat flux obtained from the microscale laser scan model is
applied, i.e., “temperature-thread”, to the hatch layer. Then multiple hatches are sequentially “deposited” in the macroscale part buildup model
with different scanning strategies. The predicted part distortions by both methods were compared and compared with the experimental data.
© 2016 The Authors. Published by Elsevier
Elsevier B.V.
B.V.This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the scientific committee of the 3rd CIRP Conference on Surface Integrity (CIRP CSI)
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 3rd CIRP Conference on Surface Integrity (CIRP CSI)
Keywords: residual stress; selective laser melting; distortion, multiscale modeling; selective laser melting; additive manufacturing

accuracy but also and detrimentally affects the functional

1. Introduction performance of the end-use parts. Simulation work [7, 8] has
been done to predict residual stress and distortion of SLM
1.1. SLM process and part distortion parts on the micro or mesoscale, in which only single tracks
or multi-tracks were considered. Some studies [9, 10]
Selective laser melting (SLM) is capable of producing predicted residual stress and part distortion in SLM on the
functional parts in a layer upon layer fashion directly from macroscale. In these models a constant heat flux was applied
CAD data [1]. Parts by SLM are near full density and have to heat an entire scan line or an entire layer at the same instant
mechanical properties comparable to bulk materials [2, 3]. In to predict the temperature and residual stress distribution of
a typical SLM system, a fine powder layer is placed on a macro part.
substrate inside an inert chamber and is fully melted by a
laser. After one layer is deposited, another layer will be 1.2. Pressing issues and research objective
placed and melted until a part is produced. Several defects
usually exist in a SLM part. High thermal stress would lead to A coupled thermal-mechanical analysis for several single
part distortion and cracks. The balling effect may result in scans with a fine mesh model is very time consuming. A
poor surface finish [4]. Also, residual gas content, unmelted practical SLM part on the macroscale requires millions of
powder, and oxidized particles may lead to porosity of the microscale laser scans which dramatically increases the
component [5, 6]. computational load for the calculation of a coupled analysis.
The uneven heat input as well as the rapid cooling of the Thus, it is impossible in practice to predict the distortion of a
process generates large amounts of tensile residual stresses in practical SLM part if every scan is simulated.
the component. It not only reduces the part geometrical

2212-8271 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 3rd CIRP Conference on Surface Integrity (CIRP CSI)
doi:10.1016/j.procir.2016.02.058
172 C. Li et al. / Procedia CIRP 45 (2016) 171 – 174

The objectives of this study are to develop and compare based on the thermal data in the microscale scan model for
temperature-thread based and stress-thread based multi-scale both methods, the heat load is directly applied to the
approaches for efficient prediction of part distortion and mesoscale hatch model. Third, the thermal history of one
residual stress by: (a) developing a novel equivalent heat hatch layer is applied to the macroscale part model, and each
source in mesoscale layer hatch from a microscale laser scan; hatch layer is activated one by one until the whole part is
(b) calculating the residual stress field and distortion of a built.
practical part at macroscale; and (c) compare the predicted
distortion results with the experimental data. 3.1. Micro laser scan model

2. SLM experiment conditions Model dimensions

In the microscale scan model, a thermal analysis was
The laser source in this study is a continuous Nd:YAG conducted using the FEA package ABAQUS/Standard to
laser with a wavelength of 1064 nm. The process parameters predict the temperature field in the melt pool. The powder
are listed in Table 1[3]. layer with thickness of 0.15 mm was placed on the substrate.
This study aims to predict part distortion and validate with The powder layer was 5 mm in length, 0.3 mm in width and
the experimental data using the lab-made iron-based powders, 0.15 mm in thickness. The substrate was 5 mm in length, 0.3
but the material properties are not available, so a commercial mm in width and 5 mm in height. The initial temperature of
powder with similar chemical composition is used instead as the whole model was set to room temperature 20 °C.
an approximation, it would be a future research subject to Laser heat source modeling
investigate the effect of temperature-dependent material data A moving Gaussian distributed heat flux is developed to
on model accuracy. The temperature-independent mechanical model the heat input of the scanning laser in the microscale
and thermal material properties are [11]: the elastic modulus model. The temperature field of the melt pool was captured
is 130 GPa, the Possion’s ratio is 0.41, the tensile and yield when the laser travelled to the center point of scan track, and
stress is 600 and 400 MPa, the melting point is 1330 oC, the this field was used to develop equivalent heat source for
coefficient of thermal expansion is 9 × 10-6/K, the thermal stress-thread based method. The thermal history (heating and
conductivity is 13 W/m·K, and the specific heat is 375 cooling) of the center point of scanning track was recorded for
J/kg·K. equivalent heat source development for temperature-thread
based method.
Table 1 SLM process parameters [3] Temperature data output
Laser power Beam dia. Scan speed Scan spacing Layer thickness Fig. 2(a) shows the temperature field of the cross-section
300 W 600 μm 50 mm/s 100 μm 150 μm of the melt pool when the laser is located at the center of the
scan track. This temperature field will be extended on a larger
scale with scan spacing considered, used for mesoscale model
3. Multiscale simulation approaches
in stress-thread method. Fig. 2(b) shows the thermal history of
the melt pool center which will be used for developing an
Two types of multi-scale finite element methods have been
equivalent heat source for temperature-thread method.
developed and compared, i.e., the stress-thread based and
temperature-thread based methods. Detailed procedure for
3.2. Meso layer hatch model
both methods is shown in Fig. 1. For both methods, the
powder material is melted by a moving heat flux to capture
Scan spacing is an important SLM process parameter and
the thermal data of a melt pool in the microscale laser scan
defined as the length of overlap between two neighboring
model. In the stress-thread method, the temperature field of
scan tracks. In this study, the scan spacing was incorporated
the melt pool is recorded. In the temperature-thread method,
in the equivalent heat source in the mesoscale hatch model for
the entire thermal history of the melt pool is recorded.
both the stress-thread and temperature-thread methods.
Second, two types of equivalent heat sources are developed
Output Micro laser scan model Output
x Melt pool geometry x Thermal analysis x Melt pool geometry
x Temp. field of the melt pool x Powder heated by moving heat flux x Temp. history of the melt pool

Stress-Thread Temperature-Thread

Output Meso layer hatch model Meso layer hatch model Output
x Temp. field of x Thermal-mechanical analysis x Thermal analysis x Temp. history of
one hatch x Equivalent heat input x Equivalent heat input one hatch
x Local RS field x Whole hatch heated at same time x Whole hatch heated at same time
Stress-Thread Temperature-Thread
Output Macro part model Macro part model Output
x RS field x Mechanical analysis x Thermal-mechanical analysis x RS field
x Part distortion x Apply local RS field x Apply equivalent heat source x Part distortion
x Applied hatch by hatch x Applied hatch by hatch

Fig. 1 Stress-Thread based and Temperature-thread based method for the prediction of distortion of SLM parts.
 C. Li et al. / Procedia CIRP 45 (2016) 171 – 174 173

from the mesoscale hatch model was assigned to each element

Temperature field in the
c
cross-section of melt pool
5000 as an initial stress tensor in the corresponding patch.
4000
Macro part model of the temperature-thread based method
is a coupled thermal-mechanical analysis. Three heat loss
Heating Cooling

Temp. (°C)
Temp. 3000
(°C)
mechanisms were considered in the macro part model, i.e.,
(a)Stress-thread 2000 heat conduction to the substrate, the heat convection of the
1000
melt pool to the surrounding powder bed and atmosphere, and
the heat radiation to the atmosphere. For both methods, the
0
0 0.05 0.1
two sides of the substrate was fixed during the process and
Melt pool center Time (s) released at the last step.
(b) Temp.-

Part dimensions: Substrat

35 mm × 15 mm × 0.15 mm Part
Fig. 2 Output of microscale laser scan model. (a) Stress-thread: temperature
45 mm
field in the cross-section of melt pool; (b) Temperature-thread: temperature
contour in the melt pool and temperature history at the center of melt pool.

Equivalent heat source modeling 1 mm

For the stress-thread method, an equivalent heat source
was developed based on the temperature field of the cross- Element size:
22 mm
section of the melt pool from the micro scan model. Multiple 250 × 250 × 50 μm
scan tracks were deposited simultaneously with the fixed scan
spacing taken into consideration. The mesoscale hatch model Fig. 3 Mesh design and dimensions in the macro part model.
for the stress-thread method was a thermal-mechanical
coupled model. After the powder and substrate were cooled Part build-up modeling
down to room temperature, the residual stress field of the part The scanning strategy in the SLM experiment [3] is a
will be averaged as , and this averaged residual stress sequential scan pattern (totally near 70 scans) as shown in
tensor was applied to each corresponding patch, detail method Fig. 4(a). In this study, the sequential pattern was simplified
could be found in [12]. by dividing the part into seven zones: from #1 to #7 shown in
For the temperature-thread method, the melt pool shape Fig. 4(b). Number in each zone represents the scanning order
was simplified to a cuboid with a length of laser spot diameter for each patch.
ds, a width of scan spacing H, and a depth of melt pool dm. In the stress-thread method, the local residual stress field
The equivalent body flux was defined as the power density was assigned as an initial stress tensor to all elements in the
(W/m3) which represents the input power for unit volume of corresponding patch. At first, all the element of the part was
melted material. An equivalent heat source model was deactivated. As the process was started, separated zone will be
developed based on the thermal history of the melt pool from activated one by one according to the scanning order.
the microscale model [13]. The heat input was modeled by a
body heat flux q which is associated with laser power P, laser
absorption coefficient A, laser spot diameter ds, melt pool ᬅᬆᬇᬈᬉᬊᬋ
Y
depth dm, and hatch spacing H. The body heat flux q was
given by Eqn. (1):
X Scanning order
A˜ P (a) Sequential pattern (vertical scan) (b) Equivalent heat input model
q (1)
ds ˜ dm ˜ H
Fig. 4 Scanning pattern applied to the macro part model.

3.3. Macro part buildup model

In the temperature-thread method, zone #1 was first heated
by the equivalent body heat flux and then cooled down for ten
Model dimensions and mesh
seconds. Same heating and cooling conditions were applied to
In the macro part model, one powder layer (highlighted by
other 6 zones sequentially.
the red dashed line in Fig. 3) was deposited on a 1 mm thick
steel substrate. The dimensions of the substrate were shown in
4. Model validation and discussion
Fig. 3. The part dimensions were 35 mm (length) × 15 mm
(width) × 0.15 mm (height) (Fig. 3). The same mesh was used
4.1. Part distortion
for both stress-thread and temperature-thread based methods.
The initial temperature of powder and substrate was set to 20
The normalized distortions in the Z direction along this
°C.
nodal path as shown in Fig. 5 for both of the methods were
Boundary conditions
compared. The predictions and experimental data show the
Macro part model of the stress-thread based method is a
same bending trend. And the distortion predicted by
mechanical analysis. The local residual stress field calculated
temperature-thread based method is much closer to
174 C. Li et al. / Procedia CIRP 45 (2016) 171 – 174

measurement than that of stress-thread based method. The 5. Conclusions

possible reason could be that the part distortion was caused by
tensile residual stress which is the mechanical response of Two multiscale modeling approaches, namely, the stress-
SLM part to the thermal load. In temperature-thread based thread and the temperature-thread based methods, have been
method, the thermal history of the whole part is closer to the developed for efficient prediction of part distortion in SLM.
real process compared to the stress-thread based method. Simulation methodology and results for the two approaches
Thus, a closer part distortion compared to the measurement have been compared. The key findings are as follows:
was predicted using temperature-thread method. x Using the stress-thread method, the local residual stress
For both of the two methods, a concave up shape curve predicted in the mesoscale model is incorporated in the
was observed. The formation of this curve is due to the macroscale part to predict part distortion. Using the
thermal history of the part and the substrate. At first, the temperature-thread method, an equivalent heat source has
material located in the upper layer of the part expanded been applied to the mesoscale model, and incorporated in
because of the laser heating. As the material cools down, the the macroscale part to predict part distortion.
plastic strain in the upper layers becomes smaller than the x Similar bending trends were found for both methods. The
lower layers. Finally, a concave shaped distortion was formed. stress-tread method underestimated part distortion
compared to the temperature-thread method.
x Tensile residual stress with a magnitude near the yield
point of the work material was found on the top surface of
the part. A typical residual stress profile in depth direction
for a SLMed part was predicted.

References

[1] Levy, G.N., Schindel, R., Kruth, J., 2003. Rapid manufacturing and rapid
tooling with layer manufacturing (LM) technologies, state of the art and
future perspectives, CIRP Annals-Manufacturing Technology 52, p. 589-
609.
Fig. 5 Part distortion prediction: stress-thread vs. temp.-thread methods. [2] Kruth, J., Levy, G., Klocke, F., Childs, T., 2007. Consolidation
phenomena in laser and powder-bed based layered manufacturing, CIRP
Annals-Manufacturing Technology 56, p. 730-759.
4.2. Residual stress [3] Kruth, J., Froyen, L., Van Vaerenbergh, J., Mercelis, P., Rombouts, M.,
Lauwers, B., 2004. Selective laser melting of iron-based powder, Journal
Fig. 6(a) shows residual stress S11 profile comparison of of Materials Processing Technology 149, p. 616-622.
[4] Tolochko, N.K., Mozzharov, S.E., Yadroitsev, I.A., Laoui, T., Froyen, L.,
temperature-thread and stress-thread methods in depth Titov, V.I., Ignatiev, M.B., 2004. Balling processes during selective laser
direction. A similar residual stress characteristic was found treatment of powders, Rapid Prototyping Journal 10, p. 78-87.
for both of the two methods. In the temperature-thread based [5] Lewis, G.K., Schlienger, E., 2000. Practical considerations and
capabilities for laser assisted direct metal deposition, Materials & Design
method, the maximum tensile residual stress was found on the 21, p. 417-423.
top surface of the part, and the tensile stress decreases as the [6] Dadbakhsh, S., Hao, L., Sewell, N., 2012. Effect of selective laser
depth increases. In stress-thread method, a slight change of melting layout on the quality of stainless steel parts, Rapid Prototyping
Journal 18, p. 241-249.
residual stress in depth direction was observed. For both [7] Dai, K., Shaw, L., 2004. Thermal and mechanical finite element
methods, a sudden change of residual stress (S11) was found modeling of laser forming from metal and ceramic powders, Acta
on the boundary between the part and substrate, changing Materialia 52, p. 69-80.
from tensile on the part to compressive on the substrate. The [8] Aggarangsi, P., Beuth, J.L., 2006. Localized preheating approaches for
reducing residual stress in additive manufacturing, p. 709-720.
predicted residual stress profile was very similar to the typical [9] Nickel, A., Barnett, D., Prinz, F., 2001. Thermal stresses and deposition
p
simplified residual stress pprofile in depth direction [14]. patterns in layered manufacturing, Materials Science and Engineering: A
Z 317, p. 59-64.
[10] Zaeh, M.F., Ott, M., 2011. Investigations on heat regulation of additive
Part + manufacturing processes for metal structures, CIRP Annals -
Manufacturing Technology 60, p. 259-262.
_ [11] EOS GmbH, 2004. Direct Metal and DirectSteel materials for EOSINT
M 250 Xtended.
[12] Li, C., Fu, C.H., Guo, Y.B., Fang, F.Z., 2016. A multiscale modeling
approach for fast prediction of part distortion in selective laser melting,
Journal of Materials Processing Technology 229, p. 703-712.
Substrate [13] Li, C., Liu, J., Guo, Y., Li, Z., 2015. A Temperature-Thread Multiscale
Modeling Approach for Efficient Prediction of Part Distortion by
+ Selective Laser Melting, Proc. of the 26th SFF Symposium p. 1166-1181.
[14] Kruth, J., Deckers, J., Yasa, E., Wauthlé, R., 2012. Assessing and
0 σyeild σ comparing influencing factors of residual stresses in selective laser
(b) melting using a novel analysis method, Proceedings of the Institution of
Mechanical Engineers, Part B: Journal of Engineering Manufacture 226,
Fig. 6 Residual stress profile characteristics in depth direction: (a) residual p. 980-991.
stress profile prediction by stress-thread method vs. temperature-thread
method; (b) characteristics of residual stress profiles (redraw from [14]).