Assessing Influencing Factors of Residual Stresses in SLM using a Novel Analysis

Method

J.-P. Kruth, J. Deckers* , E. Yasa, R. Wauthlé

Department of Mechanical Engineering , Catholic University of Leuven, Celestijnenlaan 300B BE-3001
Heverlee, Leuven, Belgium
*Email：jan.deckers@mech.kuleuven.be
Abstract
Selective Laser Melting (SLM) is a layered manufacturing process that allows to build complex, lightweight
and customized parts by consolidating successive layers of powder. Since during the SLM process the
powder material is completely molten, a low degree of porosity and good mechanical properties can be
achieved.
However, during the melting and solidification, the material experiences large temperature fluctuations in a
short time. This causes high thermal stresses which can introduce part warpage, cracks or an unwanted
decrease in strength of the parts. The goal of the reported investigation is to reduce the thermal stresses by
changing standard parameters like the scan pattern or process parameters like the preheating temperature
of the base plate, the layer thickness, pre-scanning a deposited powder layer or post-scanning an already
scanned layer. To achieve this goal, a new method is defined which enables assessing thermal stresses in a
fast way by measuring the curvature of test parts when releasing them from the base plate on which they are
built.
A review of the existing literature on this topic and the results of own experiments show that changes which
lower high temperature gradients reduce the thermal stresses. Also changing the orientation of the scan
vectors can reduce the thermal stresses, or can reduce the fatal consequences of the thermal stresses.
Keywords:
Rapid manufacturing, Selective Laser Melting (SLM), residual stresses

1 INTRODUCTION inhibited as a consequence of the plastic deformation

A material in its liquid phase doesn’t contain thermal developed during heating yielding a residual tensile
stresses. In the SLM process, thermal stresses arise in stress ( σ tens ) condition at the radiated zone and a
solidified material, subjected to temperature gradients, or counter-bending of the part being produced towards the
subjected to shrinkage caused by the solidification of radiated zone. According to the equilibria of force and
laser molten material. momentum of the part, the radiated zone is surrounded
by a zone of compressive stress ( σ comp ) (Figure 3).
Residual stresses arise by nonuniform plastic
deformations in solid material. As long as thermal
stresses are elastic, no residual stresses will remain
when the temperature comes back to its original value. If thermal
the thermal stresses inside a material become higher as crack
the yield stress, residual stresses will remain after
cooling. The sign of the remaining residual stress will
always be the opposite of the sign of the thermal stress.
As a decrease of thermal stress will reduce the residual
stresses, the terms ‘thermal stress’ and ‘residual stress’
are somewhat coupled. In the remaining of the paper,
these terms will be used interchangeable.
1.1 Mechanisms of thermal stresses in SLM
During the SLM process, the laser beam melts the Figure 1: SLM part with cracks induced by thermal
material and causes a temperature gradient in the solid streses.
material being radiated wherever it passes, resulting in
residual stresses. If the residual stresses are too high, the
produced part will crack (Figure 1).
Mercelis uses two descriptive models to explain the
mechanism of the thermal stresses in SLM [1],[2]: The
Temperature Gradient Mechanism (TGM) model and the
cool-down phase model. (a)
The TGM model states that during the SLM process the
laser beam heats up the solidified material being
radiated, which as a result tends to expand (Figure 2a).
The thermal expansion ( εth ) is partially inhibited by the
surrounding solid material, which bends away from the
laser beam, yielding a compressive stress/strain condition
in the radiated zone. If the compressive stress exceeds
the compressive yield stress ( σ yield ) of the material, the
compressive strain will be partially elastic ( εel ) and (b)
partially plastic ( εpl ) (Figure 2b). Figure 2: (a) Bending of an SLM part during laser beam
After the laser beam leaves that area, the radiated zone heating [2]. (b) Simplified representation of the formation
will cool and tend to shrink. The shrinkage is partially of thermal strains in the radiated zone.
th
16 International Symposium on electromachining (ISEM-XVI)
 Mechanical models simplify the heat input of the laser. In
order to have more accurate models, it is necessary to
calculate the exact temperature distribution in the part
(a) during processing. Many authors have discussed thermal
modeling of other laser material processing techniques
like laser welding and laser bending [4],[5],[6],[7].
Thermal modeling of the SLM process is similar, but the
powder characteristics and the scanning pattern of the
laser beam also have to be taken into account.
Since residual stresses do not only depend on the
temperature, thermal models cannot predict the resulting
residual stresses without mechanical considerations.
(b) Thermomechanical models are able to calculate a
resulting stress profile from a thermal model. Many
Figure 3: (a) Counter-bending of an SLM part after a laser authors use ABAQUS finite element package for creating
beam has passed [2]. (b) Simplified representation of the thermomechanical models [8],[9],[10],[11]. Others use an
formation of residual stresses in the radiated zone. own method [12],[13] or another package [14].
The cool-down phase model describes the formation of Many material properties, like the heat capacity, the
residual stresses during the SLM process from a more thermal expansion, the latent heat and the yield strength
macroscopic point of view. It states that when the laser are temperature dependent and are well understood.
beam has passed, the molten and resolidified top layer of However, all this knowledge is not been found
the part will tend to shrink. This shrinkage is partially exstensively implemented yet for the calculation of
inhibited by the underlying material, thus introducing thermal stresses [15].
tensile stresses in the added top layer.
1.3 Observations from experiments
1.2 Modelling of thermal stresses in SLM Besides modeling the SLM process, many researchers
Mercelis and Shiomi have defined a similar simplified have performed experimental tests
mechanical model which quantifies the descriptive cool- [2],[3],[11],[14],[16],[17],[18] to measure the residual
down phase model by using the equilibria of force and stresses. The purpose of these tests is to verify the
momentum [2],[3]. They assume the residual stresses in results of the theoretical models, or to examine the
the latest added layer equals the yield strength of the influence of non simulated phenomena.
material. This causes high tensile stresses in the SLM Many measuring techniques for measuring the residual
part, compressive stresses in the upper part of the base stress in a part are available [19],[20]. Usually a
plate and again lower tensile stresses in the lower part of classification is made in two groups, whether the
the base plate as illustrated in Figure 4. The high tensile measurement is destructive or not. Destructive methods
stresses inside the SLM part partially relax when include the layer removal method [19],[20], the crack
removing the part from the base plate. The occuring compliance method [2], the contour method [21] the hole
relaxation consists of a constant term, which effects a drilling method [19],[20]. Non destructive methods
uniform shrinkage of the part produced, and a linear term, include X-ray diffraction, neutron diffraction, ultrasonic
which effects the bending of the part produced. The and magnetic measuring methods [19],[20]. Since these
stresses in the part which remain after relaxation are measuring techniques are time consuming, and only
considered as the residual stresses (Figure 5). qualitative information of thermal stresses is needed in
this study (for comparing stress magnitude and analysing
trends), a new and faster method for analysing the
residual stresses is defined in the next section.
2 DESCRIPTION OF EQUIPMENT AND THE NOVEL
ANALYSIS METHOD
2.1 Equipment
In this study Ti-6Al-4V test parts are produced on an in-
(a) (b) house developed SLM machine and 316L stainless steel
test parts are produced on a Concept Laser Linear M3
Figure 4: (a) Residual stresses after laser melting two machine. Table 1 and Table 2 give an overview of the
layers on a base plate. (b) Residual stresses of a SLM properties of these machines. For the measurements a
part on a base plate after production. (simplified model) non-contact vision measuring system (Mitutoyo Quick
Vision Pro 202) was used.
Laser type Yb:YAG
Laser wavelength 1085 nm
Maximum power 300 W
(a)
Build envelope Ø 100mm x 100mm
Table 1: Properties of the KUL-SLM machine.
Laser type Nd:YAG
Laser wavelength 1064 nm
Maximum power 100 W
(b) (c) Build envelope 250mm x 250mm x 250mm
Figure 5: (a) Residual stresses of a SLM part on a base Table 2: Properties of the Concept Laser M3 Linear
plate.(b) Constant and linear relaxation terms.(c) Residual Machine.
stresses of a SLM part removed from the base plate.
2.2 Bridge curvature method 3 EXPERIMENTS AND RESULTS
The novel analysis method, the bridge curvature method, The first section of this paragraph describes the influence
is simple, fast, accurate and resembles the layer removal of the scan pattern on the measured curling angle of the
methods (§ 1.3, [20]). When the test part is cut off the test parts. Scan strategies can be adapted in a relatively
base plate by wire EDM, the bridge curls because of the easy way by changing the scanning software. The
relaxation stresses [2]. Residual tensile stresses at the second section describes the influence of more
top of the part cause this bending. Figure 6a and Figure fundamental process changes which are not yet
6b illustrate the method whereas Figure 6c shows the commonly used for SLM.
geometry of the test parts. The planes at the bottom of
the pillars deviate from their normal position and form an 3.1 Change in scan pattern
angle α which is a measure for the thermal stresses in Length of the scan vectors
this method. For each test part, the curling angle is
measured 10 times by a Mitutoyo Quick Vision Pro 202 The parallel bi-directional scan vectors, often used in
system (5 times on each side of the part). From the SLM, do not always have the same length. The first test
measured data the mean value and standard deviation checks the influence of the vector length (l) in the x-
with a confidence interval of 95% are calculated. direction on the bending of the part and is performed on
the KUL-SLM machine. Figure 7 shows the scan pattern
A detailed uncertainty analysis of the method (e.g. the used to investigate the influence of shorter scan vectors
influence of a bad EDM cut on the measured angle) can on the thermal stresses. The sequence of the vectors
be found in [22]. Table 3 and Table 4 show the standard scanned by the laser is determined by the path of the
parameters used to produce the reference parts and dotted line.
show the obtained values for the curling angle α.
The results for shorter vector lengths are shown in Figure
8. Three test parts with shorter vector lengths are
compared to the reference part. There is almost no
difference in measured angle between vectors of 20 mm
and 10 mm. At lower vector lengths, the measured angle
(a) decreases. This effect is more pronounced if the vector
length is reduced further. Vector lengths of 2 mm record
the largest improvement of 13%.

(b)

Figure 7: Scan pattern to investigate short scan vectors.

(c)
Figure 6: Principle of the method for measuring the
residual stresses in the test parts. (a) Before and (b) after
removal from base plate. (c) Geometry of the test parts.
Material Ti-6Al-4V
Scan vectors x-direction, bi-directional
Hatch spacing 74 µm
Scan speed 225 mm/s Figure 8: Influence of vector length (l) in the x-direction
Laser power 42 W on curling of the test part (REF. = reference part).
Contour scan before fill Orientation of the scan vectors
Layer thickness 30 µm To see the influence of the orientation of the parallel scan
Reference value vectors on the measured angle α, different test parts are
αref, LM = 2,797°± 0,033° made on the KUL-SLM machine, in which the angle β
Table 3: Reference parameters used to build Ti-6Al-4V between the parallel scan vectors and the direction of
test parts on the KUL-SLM machine. measurement (x-direction) changes from 0°(reference) to
90°. Figure 9 illustrates this rotation for 45° and 90°.
Material 316L stainless steel The measured angle α decreases for rotations going
Scan vectors x-direction, bi-directional from 0° to 90° (Figure 10). The bending of the test part
reduces by 59% if the scan vectors are oriented 90° from
Hatch spacing 126 µm
the direction of measuring. If the parallel scan vectors are
Scan speed 380 mm/s layerwise alternated (β =0°-90°) after each layer, the
Laser power 100 W reduction of the measured angle is found to be 45%.
Contour scan before fill Notice that during this rotation of the scan vectors also
the length of the vectors changes. Due to the small
Layer thickness 30 µm difference in curling angle for test parts with vector
Reference value αref, LM = 1,333°± 0,024° lengths of 20 mm and 10 mm (Figure 8), this change in
vector length (l) has a negligible influence on the
Table 4: Reference parameters used to build 316L obtained results.
stainless steel test parts on the Concept Laser machine.
 Figure 9: Scan pattern to investigate the orientation β of
the scan vectors.

Figure 13: Results of island scanning with different island

orientations β on the Concept Laser machine.
3.2 Process changes
Post-scanning
Post-scanning means re-scanning a laser molten and
consolidated layer: the cross-section of each layer is first
molten by the reference parameters given in Table 3 or
Table 4. Then the cross-section is re-scanned with the
Figure 10: Influence of the orientation β of the scan same scan pattern and spot size. The results for post-
vectors on the curling of the test part. scanning with different scan speeds on the KUL-SLM
Island scanning machine are shown in Figure 14. In this figure, the value
at 0 mm/s is the reference value without post-scanning.
Island scanning is a scan strategy patented by Concept
Laser GmbH. This scan strategy divides the area to be
scanned into small square islands. The sequence in
which the islands are scanned is chosen randomly. The
special slicing algorithm embedded in Magics® software
allows to change the size of the islands, the orientation of
the islands, and it is also possible to shift the islands in
the x- and y-direction between different layers. Figure 11
shows a possible scan pattern for islands of 5x5 mm and
a 15° rotation β from the x-direction.

Figure 14: Results of post-scanning with different scan

speeds on the KUL-SLM machine.
Only the low scan speeds reduce the measured angle. A
scan speed of 100 mm/s gives the maximum reduction of
8% (the values at 0 mm/s are the reference values
without post-scanning). Also some test part were
Figure 11: Possible scan pattern for island scanning with produced on the Concept Laser machine, but none of
islands of 5x5 mm and a 15° rotation β. this parts showed any noticeable reduction in curling
angle [22].
The effect of the island size is shown in Figure 12 for
islands rotated 15° from the x-direction. The use of island Pre-scanning
scanning reduces the measured angle α, but the size of When pre-scanning is applied, the laser first sinters the
the islands doesn’t seem to influence the results. Figure powder material, before melting the material completely
13 illustrates the effect of the rotation β for islands of 5x5 with the reference parameters of Table 3. The results for
mm. If the rotation β is 45° instead of 15°, the measured pre-scanning with 800 mm/s and 1600 mm/s and a
angle further decreases. Island scanning with islands of constant laser power of 42 W on the KUL-SLM machine
5x5 mm, rotated 45° from the x-direction reduces the are presented in Figure 15. The value at 0 mm/s is the
measured angle by 36%. reference value without pre-scanning. Pre-scanning
reduces the measured angle α only slightly with a
maximum reduction of 6% for pre-scanning with 800
mm/s.

Figure 12: Results of island scanning with different island

size on the Concept Laser machine.
Figure 15: Results of pre-scanning with different scan
speeds on the KUL-SLM machine.
Different layer thickness Preheating of the base plate
The KUL-SLM machine and the Concept Laser machine In order to investigate the influence of preheating on
both employ a standard layer thickness of 30 µm. Larger residual stresses, the base plate is heated till 180°C. This
layer thicknesses can be used to increase the production resulted in a reduction of 10% of the measured angle.
speed of the process, but also can induce unwanted
porosities which decrease the strength of the produced Heat treatment
parts drastically. Since the stress normal to a pore is Heat treatments reduce the residual stresses developed
zero, porosities tend to decrease the residual stresses. during fabrication (stress relieving) [23]. Figure 16 shows
The scan speed needed to produce optimal density parts the temperature cycle that is applied to a reference part
was not known. Therefore multiple test parts with a layer produced on the KUL-SLM machine (Table 3). This heat
thickness of 60 µm but with different scan speeds (v) are treatment reduces the measured angle by 80%.
produced on the Concept Laser machine and
investigated. For each part produced, the curling angle is
measured (Figure 16). Also the part densities are
measured with the Archimedes method (Table 5) and the
cross-sections are investigated by an optical microscope
(Figure 17).
A low scan speed is expected to heat up more material
around the laser spot and to reduce in this way the
temperature gradient and related thermal stresses.
Contrary to this expectation, Figure 16 shows only a
decrease in curling angle for scan speeds varying from
200 mm/s to 175 mm/s. Since part n° 3 has a rather high
density value, the lower curling angle for its scan speed
cannot be due to stress relaxation caused by a large Figure 18: Temperature cycle for heat treatment.
amount of pores in the part. It seems to be difficult to
compare the influence of layer thickness on thermal 4 DISCUSSION
stresses without taking into account the combined Scanning with short scan vectors reduces the thermal
influence of part density and scan parameters (in this stresses. If the length of the scan vectors is 2 mm
case the scan speed) to achieve this density. On the instead of 20 mm, the measured angle is reduced by
other hand, it can be stated that the measured angle of 13%. Increasing the process temperature by preheating
the test part with a density closest to the density of the the base plate up to 180°C reduces the curling angle by
reference part (part n°3) and a layer thickness of 60µm 10%. Both results indicate that changes lowering the
gives a reduction of 6%. thermal gradients reduce the thermal stresses: Shorter
scan vectors experience a smaller temperature gradient
due to the heat of the previous scan line. Preheating the
base plate also reduces the temperature gradient during
scanning.
The tests in § 3.1 ‘Orientation of the scan vectors’ and
‘Island scanning’ show that the magnitude of the thermal
stresses differs as the orientation of the scan vectors to
the direction of measurement (x-direction) changes.
Thermal stresses in the direction perpendicular to the
scan vectors are lower as thermal stresses parallel to the
scan vectors.
Post-scanning and pre-scanning with the same spot size
Figure 16: Influence of layer thickness and scan speed on and the same scan pattern only reduce the measured
the curling of the test part. angle by respectively 8% and 6%. These reductions are
n° scan speed [mm/s] density [%] angle α [°] considered to be relatively low because the same
Temperature Gradient Mechanism causes the high
1 (ref.) 380 97,57 1,333 thermal stresses.
2 380 95,83 1,212 5 SUMMARY AND CONCLUSION
3 250 98,23 1,251 The paper presents a new method for assessing thermal
stresses in a simple, fast and accurate way. The method
4 200 98,82 1,374 uses the angle of curvature α of SLM-made bridge
5 175 98,88 1,323 structures cut off from the base plate as a measure for
the thermal stresses. Table 6 and Table 7 give a
Table 5: Density of the test parts produced with a layer summary of the most important results obtained.
thickness of 60 µm and different scan speeds.
The results from the experiments as well as the findings
from literature lead to two general conclusions: Changes
which reduce the high temperature gradient, like using
short scan vectors and preheating of the base plate,
reduce the thermal stresses. Also an optimal use of the
orientation of the thermal stresses can reduce these
(a) (b) stresses or make the orientation more beneficial.
Figure 17: Investigation of the cross-sections of 60µm
layer thickness parts by an optical microscope.
(a) v=380mm/s, (b) v=175mm/s.
 [7] Mackwood A., Crafer R., 2005, Thermal modelling
of laser welding and related processes: a literature
review, Optics and Laser Technology, 37, 99-115.
[8] Chin R. K., Beuth J. L., Amon C. H, 1996,
Thermomechanical modeling of successive material
deposition in layered manufacturing, Mech. Mater.,
24,257-271.
[9] Li L., Achara C., 2004, Chemical assisted laser
machining for the minimisation of recast and heat
affected zone, CIRP annals 2004, 53/1, 175-178.
[10] Nickel A. H., 1999, Analysis of thermal stresses in
Shape Deposition Manufacturing of metal parts,
Ph.D. thesis Stanford University.
[11] Nickel A.H., Barnett D.M., P. F. B., 2001, Thermal
stresses and deposition patterns in layered
Table 6: Summary of most important results for reduction manufacturing, Materials Science and Engineering,
of thermal stresses with Ti-6Al-4V on the KUL-SLM A317/1,59-64.
machine. [12] Matsumoto M., Shiomi M., Osakada K., Abe F.,
2002, Finite element analysis of single layer forming
on metallic powder bed in rapid prototyping by
selective laser processing, Int. Journal of Machine
Tools & Manufacture, 42, 61-67.
[13] Osakada K., Shiomi M., 2006, Flexible
manufactuting of metallic products by selective
laser melting of powder, International Jounal of
Machine Tools & Manufacture, 46, 1188-1193.
[14] Over C., Meiners W., Wissenbach K., Hutfless J.,
Lindeman M., 2003, Rapid manufacturing of metal
parts and tools using laser melting, Proc. WLT-
Conference on Lasers in Manufacturing, 265-269.
Table 7: Summary of most important results for reduction [15] Bhadeshia H. K. D. J., 2002, Material factors,
of thermal stresses with 316L stainless steel on the Handbook of Residual Stress and Deformation of
Concept Laser machine. Steel (Totten G. E., Howes M. A. H., Inoue T., eds.),
chapter ‘Effect of Materials and Processing’, 3-10.
[16] Kruth J.-P., Mercelis P., V. Vaerenbergh J., Froyen
6 ACKNOWLEDGMENTS
L., Rombouts M., 2005, Binding mechanisms in
The authors thank I.W.T. for the support through the SBO selective laser sintering and melting, Rapid
project DiRaMaP and the TUBITAK Research Council of Prototyping Journal, 11/1, 26-36.
Turkey for its financial support given.
[17] Rombouts M., 2006, Selective Laser
7 REFERENCES Sintering/Melting of Iron-Based Powders, PhD
[1] Kruth J.-P., Froyen L., Van Vaerenbergh J., Mercelis thesis K.U.Leuven – April.
P., Rombouts M., Lauwers B., 2004, Selective laser [18] Vandenbroucke B., 2008, Selective laser melting of
melting of iron based powders, Journal of materials biocompatible metals for rapid manufacturing of
processing technology, 14th international medical parts, PhD thesis K.U.Leuven (Dept. Mech.
symposium on electromachining (ISEM XIV) – 10 Eng.).
June, Volume 149, Issues 1-3, 616-622.
[19] Kandil F. A., Lord J. D., Fry A. T., Grant P. V., 2001,
[2] Mercelis P., 2007, Control of Selective Laser A review of residual stress measurement methods –
Sintering and Selective Laser Melting Processes, a guide to technical selection, tech. rep. NPL
PhD thesis K.U.Leuven – April. Materials Centre.
[3] Shiomi. M., Osakada K., Nakamura K., Yamashita [20] Withers P. J., Bhadesia H. K. D. H., 2001, Residual
T., Abe F., 2004, Residual stress within metallic stress part 1: Measurement techniques, Materials
model made by selective laser melting process, Science and Technology, 17.
Annals of the CIRP, 53/1, 195.
[21] Prime M. B., 2001, Cross-sectional mapping of
[4] Cheng P. J., Lin S. C., 2000, An analytical model for residual stresses by measuring the surface contour
the temperature field in the laser forming of sheet after cut, Journal of Engineering Materials and
metal, Journal of Materials Processing Technology, Technology, 123, 162-168.
101, 260-267.
[22] Wauthlé R., 2009, Verminderen van thermische
[5] Kruth J.-F., 2001, Modelleren van laserbuigen: een spanningen bij selectief laser smelten, Master
verkennende studie, Master’s thesis K.U.Leuven. thesis K.U.Leuven (Dept. Mech. Eng.).
[6] Duflou J.R., Callebaut B., Kruth J.-P., 2005, [23] Gilbert R., Shannon R., 1991, Heat treating of
Laserforming of 3D features, Advanced Materials titanium alloys, ASM Handbook, 4.
Research, Vols 6-8, 425-432.