You are on page 1of 9

Virtual and Physical Prototyping

ISSN: 1745-2759 (Print) 1745-2767 (Online) Journal homepage: http://www.tandfonline.com/loi/nvpp20

Experimental investigation of process parameters


on layer thickness and density in direct metal
laser sintering: a response surface methodology
approach

S. A. Fatemi, J. Zamani Ashany, A. Jalali Aghchai & A. Abolghasemi

To cite this article: S. A. Fatemi, J. Zamani Ashany, A. Jalali Aghchai & A. Abolghasemi (2017):
Experimental investigation of process parameters on layer thickness and density in direct metal
laser sintering: a response surface methodology approach, Virtual and Physical Prototyping, DOI:
10.1080/17452759.2017.1293274

To link to this article: http://dx.doi.org/10.1080/17452759.2017.1293274

Published online: 15 Mar 2017.

Submit your article to this journal

Article views: 4

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=nvpp20

Download by: [University of Newcastle, Australia] Date: 19 March 2017, At: 08:48
VIRTUAL AND PHYSICAL PROTOTYPING, 2017
http://dx.doi.org/10.1080/17452759.2017.1293274

Experimental investigation of process parameters on layer thickness and density


in direct metal laser sintering: a response surface methodology approach
S. A. Fatemi, J. Zamani Ashany, A. Jalali Aghchai and A. Abolghasemi
Faculty of Mechanical Engineering, K. N. Toosi University of Technology, Tehran, Iran

ABSTRACT ARTICLE HISTORY


Direct metal laser sintering is an additive manufacturing method which allows manufacturing of Received 14 December 2016
complex parts without using dies in short time. In the present study, direct metal laser sintering Accepted 7 February 2017
of steel 316 powder has been studied. Response surface methodology has been used for the
KEYWORDS
design and analysis of experiments. Nd:YAG laser with a maximum power of 75 W has used to Direct metal laser sintering;
sinter the powder. The effect of current, scan speed and laser frequency has been investigated steel 316 powder; response
on layer thickness and density of fabricated samples. In addition, by using multi response surface methodology; layer
optimisation; appropriate setting to maximise layer thickness and density and also minimise thickness; density
applied energy was suggested. It can be seen that increases in current and frequency lead to
increases in thickness and density of samples. Also increasing scan speed decreases layer
thickness and density of samples.

Introduction
dependent on proper selection of the process par-
The manufacturing companies are always seeking for ameters (Živčák et al. 2016, Kruth et al. 2007, Mumtaz
new methods to improve quality of their productions et al. 2008, Olakanmi et al. 2015, Wang et al. 2016).
while reducing time and cost (Yu 2005). Additive manu- Scan speed, layer thickness, powder thickness, hatch
facturing (AM) is a good solution to fabricate customised size, scan path pattern and laser power are the most
parts with a lower time and cost. AM is a kind of process important parameters which have significant effects on
which has evolved from the rapid prototyping method the sintering process (Tolochko et al. 2003, Tolochko
(Paul 2013). AM provides advantages which makes it et al. 2004, Yu 2005, Negi et al. 2015). These parameters
attractive for automotive, aerospace and medical indus- directly affect process time, mechanical properties, geo-
tries (Bertol et al. 2010, Yan et al. 2014, Olakanmi et al. metrical accuracy and surface roughness (Ning et al.
2015). Suitable advantages of AM include: geometrical 2004). As mentioned, the quality of fabricated parts by
flexibility, no requirement for tool and fixture and the DMLS process is entirely dependent on the process
finally energy, cost and time efficiency (Manfredi et al. parameters. Therefore, several researchers have
2014, Cabrini et al. 2016, Hagedorn 2017, Lindemann focused on process parameters of DMLS. Pogson et al.
and Jahnke 2017). Stereolithography (SLA), direct metal studied the effect of laser scan speed in the DMLS
laser sintering (DMLS) (also known as selective laser sin- process (Pogson et al. 2003). They pointed out that
tering (SLS)), electron beam melting (EBM) and fused scan speed is the parameter which controls the geome-
deposition modelling are some of the most popular pro- try of molten pool. Higher scan speeds lead to thinner
cesses of AM (Paul 2013). Among these methods, powder and longer molten pool and in lower scan speeds the
bed fusion AM processes such as DMLS or EBM have materials stay more time in the molten state. Therefore,
attracted more attention (Wang et al. 2002, Simchi in the constant power, fabricated parts by a higher
et al. 2003, Rossi et al. 2004, Sing et al. 2016). In the scan speed will have less thickness. Zhu et al. reported
DMLS process, a powerful energy source (LASER) scans that decreasing distance between laser scans leads to
powder in a continuous line and makes a molten pool. higher surface roughness and denser parts (2005).
By cooling and solidification of the molten pool, one Morgan et al. expressed that increasing power and
layer of part can be built. The generation of sequential scan speed lead to irregular structures because of
layers on top of each other results in the fabrication of strong surface tension forces. They noted that decreases
a complex 3-D part (Kruth et al. 2005). It is reported in scanning speed leads to increasing melt flow and con-
that successful implementation of DMLS is highly sequently intricate surfaces (Morgan et al. 2001). Wang

CONTACT A. Jalali Aghchai jalali@kntu.ac.ir


© 2017 Informa UK Limited, trading as Taylor & Francis Group
2 S. A. FATEMI ET AL.

et al. reported that the laser scanning speed and laser


power are the most effective parameters on density of
the fabricated parts (2016). DMLS is a complex process
which can be affected by several parameters. Therefore,
accurate analytical modelling is impossible. However,
empirical modelling can be assessed to study the
DMLS process.
Response surface methodology (RSM) was introduced
by Box and Wilson in 1951 (Bezerra et al. 2008). RSM is a
collection of mathematical and statistical techniques
which is a powerful method to build up empirical
models (Ning et al. 2004). This method can be assessed
to fit empirical models to the experimental results. There-
fore, linear or square polynomial functions are utilised to
describe the relationship between a response (output
variable) which is influenced by independent variables
(input variables) (Myers et al. 2016). According to the
benefits of RSM, researchers focused on the application
of RSM in the experimental investigations. Beal et al.
studied the effects of pulse energy, pulse width, scan
speed and frequency on defects (porosity and crack) of
fabricated samples by using RSM in the laser fusion of Figure 1. Implemented DMLS system.
H13/Cu materials (Beal et al. 2006). They optimised
process parameters by using RSM and reported that opti-
mised parameters lead to about 83% reduction in cracks to Figure 1. This system consists of: (a) Nd:YAG laser with
and porosity. Lynn and Rosen studied the effect of SLA a maximum power of 75 W, wavelength of 1.064 nm and
process parameters on geometrical tolerances by using focal length of 185 mm (b) controller computer, (c)
RSM. They presented a process planning method to optical window, (d) vacuum chamber, (e) controller, (f)
select proper values of process parameters in order to parts storage, (g) power supply and (h) vacuum pump.
achieve intended geometrical tolerances (Lynn-Charney Adjustable parameters of laser include the laser current
and Rosen 2000). Kumar et al. performed DMLS exper- (5–10 A), scan speed (0–1000 mm/s) and the output fre-
iments and analysis of variance (ANOVA) to study the quency (0–20 kHz).
influence of part orientation, laser power and tempera- In this study, steel 316 powder with the average diam-
ture on geometrical accuracy and hardness of fabricated eter of 20 µm (purity of 97.17) was investigated. 10 × 10
parts in the SLS process (Kumar et al. 2016). They found mm2 square samples were produced by single parallel
that the mentioned parameters have a significant impact path pattern scanning in different levels of process par-
on geometrical accuracy and hardness. They also ameters as shown in Figure 2. A power meter was used
expressed that temperature has a major effect on hard- to achieve laser power in different combinations of
ness; also, part orientation and laser power are the process parameters. In order to prevent oxidation of sin-
major effective parameters on geometrical accuracy. tered layers, all the experiments were performed in the
It can be seen that despite appropriate benefits of vacuum chamber. It should be mentioned that the
RSM, there is a low number of researches in this field. oxide film on the sintered layers leads to delamination
In the present study, the effects of frequency, current of layers and consequently geometrical defects.
and laser scan speed on layer thickness and density of
steel 316 powder have been investigated by using RSM
Design of experiments
and appropriate settings have been presented.
In this study, a face-centred central composite design
(CCD) was used to design the experiments (Ghoreishi
Experimental procedures 2006). A face-centred CCD has three levels (in coded
terms: −1, 0 and + 1) which represent low, medium
DMLS experiments
and high level of a parameter. For three input par-
In order to investigate the effect of process parameters, a ameters, this design requires 15 runs. In addition, in
DMLS system was designed and implemented according order to avoid singularities during regression and
VIRTUAL AND PHYSICAL PROTOTYPING 3

Figure 2. Fabricated samples.

evaluating pure errors, five replications of center run are Table 4 shows ANOVA for density of sintered parts.
required (Ghoreishi 2006). Scan speed, current and fre- Results show that scan speed, current and frequency
quency were selected as independent input parameters have a significant effect on density of sintered parts. It
to investigate the layer thickness and density in the can be seen that scan speed–frequency and current–fre-
DMLS process. In order to select the range of parameters, quency interactions have a significant effect on density
preliminary experiments have been performed. It was of sintered parts. Also, quadratic terms show significant
seen that in the scan speeds lower than those selected, effects on density of fabricated parts which satisfies
a large molten pool was created and it led to increasing usage of the quadratic regression model.
defects such as shrinkage and consequently lower Statistical modelling was carried out to develop a
dimensional accuracy. Also in the scan speeds higher mathematical relationship between the selected
than those selected, it was seen that molten pool was process parameters and response of experiments (layer
not created. Similar trends occur by changing current thickness and density). The face-centered CCD can fit
and frequency. The selected range of these parameters quadratic models. Therefore, any nonlinearity in the
is shown in Table 1. response can be considered. The second-order
regression model is presented by the following equation:

Results and discussion 


k 
k 
k
h = b0 + bi xi + bii xi2 + bij xi xj + 1, (1)
The effects of scan speed, current and frequency have i=1 i=1 1≤i≤j

been investigated on layer thickness and density in the


DMLS process of steel 316 powder and the results are Table 2. Layer thickness, density and laser energy in different
presented in Table 2. levels of process parameters.
The significant parameters were detected by ANOVA. Scan layer
Next, the insignificant parameters were removed from speed Current Frequency thickness Density Power
(mm/s) (A) (kHz) (mm) (g/cm3) (W)
the model by a stepwise regression method. Table 3
500 8 5 0.100 3.550 2.56
shows refined ANOVA for layer thickness. It can be 500 10 5 0.220 3.650 4
seen that scan speed, current and frequency have a sig- 300 9 5 0.198 3.600 5.4
500 9 10 0.119 4.100 6.48
nificant effect on layer thickness. Table 3 shows that all 500 8 15 0.105 4.500 7.68
the quadratic terms are significant which means that 300 8 10 0.153 3.800 8.53
300 9 10 0.208 4.400 10.9
the selection of the quadratic model was sufficient. It 300 9 10 0.210 4.410 10.72
can be seen that scan speed–current and also scan 300 9 10 0.206 4.409 10.8
speed–frequency interactions have a significant effect 300 9 10 0.206 4.420 10.97
300 9 10 0.205 4.410 11
on layer thickness. 300 9 10 0.204 4.410 10.7
500 10 15 0.230 6.100 12
100 8 5 0.250 3.620 12.8
Table 1. Selected variables for DMLS process. 300 10 10 0.288 5.000 13.33
300 9 15 0.243 5.750 16.20
Variables Actual levels (coded levels)
100 10 5 0.450 3.900 20
Current (A) 8 (−1) 9 (0) 10 (+1) 100 9 10 0.350 5.300 32.4
Frequency (kHz) 10 (−1) 15 (0) 20 (+1) 100 8 15 0.300 5.900 38.4
Scan speed (mm/s) 100 (−1) 300 (0) 500 (+1) 100 10 15 0.490 7.000 60
4 S. A. FATEMI ET AL.

Table 3. ANOVA for layer thickness.


Source DF Sum of squares Mean square F value Prob > F Status
Model 8 0.192925 0.024116 419.18 0.000 Significant
A 1 0.113636 0.113636 1975.24 0.000 Significant
B 1 0.059290 0.059290 1030.59 0.000 Significant
C 1 0.002250 0.002250 39.11 0.000 Significant
A*A 1 0.012152 0.002483 43.15 0.000 Significant
B*B 1 0.001558 0.000708 12.31 0.005 Significant
C*C 1 0.000708 0.000708 12.31 0.005 Significant
A*B 1 0.002628 0.002628 45.68 0.000 Significant
A*C 1 0.000703 0.000703 12.22 0.005 Significant
Residual Error 11 0.000633 0.000058
Note: A = Scan speed, B = Current, C = Frequency.

Table 4. ANOVA for density of sintered parts. observed. Therefore, it can be concluded that the pre-
Sum of Mean F Prob
Source DF squares square value >F Status
sented model is adequate (Montgomery and Myers
Model 7 16.9490 2.4213 99.02 0.000 Significant
1995).
A 1 1.4592 1.4592 59.67 0.000 Significant It is reported that while the interaction between par-
B 1 1.8318 1.8318 74.91 0.000 Significant ameters is significant, the effects of factors should be
C 1 11.9465 11.9465 488.54 0.000 Significant
A*A 1 0.4533 0.4533 5.11 0.043 Significant explained by interaction plots (Montgomery and Myers
C*C 1 0.0953 0.0953 3.90 0.072 Significant 1995, Montgomery 2008). Therefore, interaction plots
A*C 1 0.4901 0.4901 20.04 0.001 Significant
B*C 1 0.6728 0.6728 27.51 0.000 Significant are presented. Figure 4 shows the effects of scan speed
Residual 12 0.2934 0.0245 and frequency on layer thickness while the current is
Error
kept at a medium level. It can be concluded that at a con-
Note: A = Scan speed, B = Current, C = Frequency.
stant scan speed, increasing laser frequency leads to
increasing layer thickness. For example, at a constant
where k is the number of variables, b0 is the constant scan speed of 100 mm/s, 66% increases in laser fre-
term, bi represents the coefficients of the linear term, quency leads to a 20% increase in layer thickness.
bii represents the coefficients of the quadratic par- Laser wave can be produced either in a continuous
ameters, bij represents the coefficients of the interaction wave (CW) mode or pulsating mode. Pulsating mode
term, xi.j represents the variables i and j and 1 is the can be produced by Q-switch. Q-switch breaks CW into
residual associated with the experiments. Coefficients smaller wavelengths. By increasing frequency of Q-
of the regression model for layer thickness and density switch, the numbers of wave increases but the power
are presented in Tables 5 and 6, respectively. of each wave decreases. Therefore, the output power
It is always necessary to check the fitted model to of laser is always constant and the energy of laser can
ensure that it provides an adequate approximation to be inserted uniformly to the surface. The ability of
the experiments (Montgomery and Myers 1995). In pulsed lasers is the production of solid layers in higher
order to examine the model adequacy, residual analysis scan speeds and preventing destructive thermal effects.
can be applied. Normal probability plot of residuals is Increased laser frequency leads to higher power
shown in Figure 3. It can be seen that residuals are density in a location. By increasing power density, the
approximately placed along a straight line; therefore, penetration of laser in the powder increases and conse-
the normality assumption is satisfied. Also, it can be quently the layer thickness increases. It can be concluded
seen that residuals are randomly distributed and clearly that at a constant frequency, an increase in scan speed
structured distribution of residuals (i.e. S shape) is not leads to decreases in layer thickness. For example, at a

Table 5. Coefficient of regression model for layer thickness.


Term Coefficient
Table 6. Coefficient of regression model of density.
Constant 0.801235
A −7.43068E−05 Term Coefficient
B −0.184631 Constant 4.57365
C −0.00702386 A −0.00239844
A*A 7.51136E−07 B −0.152000
B*B 0.0160455 C −0.367200
C*C 0.000641818 A*A 4.93906E−06
A*B −9.06250E−05 C*C 0.00690250
A*C −9.37500E−06 A*C −2.47500E−04
B*C Insignificant B*C 0.0580000
VIRTUAL AND PHYSICAL PROTOTYPING 5

Figure 3. Normal probability plot of residuals (a) layer thickness and (b) density.

constant frequency of 5 kHz, 80% increases in scan speed Similar to the effect of frequency, increasing current
leads to 70% decreases in layer thickness, due to redu- leads to more power density and higher layer thickness.
cing power density. At lower currents and higher scan speeds, samples show
Figure 5 shows the effects of scan speed and current limited connection through sintering.
on layer thickness while the frequency is kept at a Figure 6 shows the effects of scan speed and fre-
medium level. It can be concluded that at a constant quency on density of fabricated parts. It can be seen
scan speed, higher current leads to higher layer thickness. that increasing frequency and decreasing scan speed
For example, at a constant scan speed of 100 mm/s, increase density of the fabricated parts. For example, at
20% increases of current leads to 66% increases of layer a constant scan speed of 100 mm/s, 66% increases in
thickness. Scan speed has two major effects on the pulse frequency of laser leads to 45% increases in
DMLS process. Scan speed controls the geometry of the density. Also, at a constant frequency of 15 kHz, 80%
molten pool. Higher scan speeds lead to thinner and increases in scan speed lead to 15% decreases in
longer molten pool. Also, in the lower scan speeds, density of fabricated samples. Increasing frequency and
materials stay more time molten. These effects lead to decreasing scan speed lead to higher power density
lower thicknesses in the higher scan speeds. and more melted material. In facts, at lower scan
6 S. A. FATEMI ET AL.

Figure 4. The effects of scan speed and frequency on layer thickness.

Figure 5. The effects of scan speed and current on layer thickness.

Figure 6. The effects of scan speed and frequency on density.


VIRTUAL AND PHYSICAL PROTOTYPING 7

Figure 7. The effects of scan speed and current on density.

speeds, a higher amount of power density reaches to


molten bead and it leads to more powder being
melted and produces denser pieces.
Figure 7 shows the effects of scan speed and current
on the density of the fabricated parts. It can be con-
cluded that increasing current results in increased
density due to increased power density.
For example, at a constant scan speed of 100 mm/s,
20% increases of current leads to 18% increases in the
density of fabricated parts.
Finding optimal conditions that will produce the best
values for the responses is one of the main goals of using
design of experiments. It is clear that minimising the
required power for performing the DMLS process leads
to decreasing manufacturing cost. Also maximising the
layer thickness leads to decreasing manufacturing time
and cost. In addition, maximising density of fabricated
parts leads to an increase in the quality of manufactured
parts. Response optimisation is a method that allows for
compromise among the various responses. In order to
maximise layer thickness and density and minimise Figure 8. Multi response optimisation results.
required power, Minitab response optimiser has been
used. Figure 8 shows response optimiser results. (1) Scan speed, current and frequency have high signifi-
Results show that the best values for scan speed, cant effects on layer thickness.
current and frequency are 121 mm/s, 10 and 13 kHz (2) Interactions between factors are significant; there-
respectively. fore, the influence of each factor cannot be investi-
gated separately.
(3) At a constant scan speed, an increase in laser fre-
Conclusion
quency leads to an increase in layer thickness.
In the present study, influences of scan speed, current (4) At a constant frequency, an increase in scan speed
and frequency on layer thickness in DMLS of steel 316 leads to a decrease in layer thickness.
powder have been studied. RSM has been used for the (5) Increasing frequency, current and decreasing scan
design and analysis of the experiments. The most impor- speed leads to fabrication of parts with higher
tant results are summarised as follows: density.
8 S. A. FATEMI ET AL.

Mumtaz, K.A., Erasenthiran, P., and Hopkinson, N., 2008. High


Disclosure statement density selective laser melting of Waspaloy®. Journal of
No potential conflict of interest was reported by the authors. Materials Processing Technology, 195 (1–3), 77–87.
Myers, R.H., Montgomery, D.C., and Anderson-Cook, C.M.,
2016. Response surface methodology: process and product
optimization using designed experiments. Hoboken, NJ:
References Wiley.
Negi, S., Dhiman, S., and Sharma, R.K., 2015. Determining the
Beal, V., et al., 2006. Optimisation of processing parameters in effect of sintering conditions on mechanical properties of
laser fused H13/Cu materials using response surface laser sintered glass filled polyamide parts using RSM.
method (RSM). Journal of Materials Processing Technology, Measurement, 68, 205–218.
174 (1), 145–154. Ning, Y., et al., 2004. An intelligent parameter selection system
Bertol, L.S., et al., 2010. Medical design: direct metal laser sinter- for the direct metal laser sintering process. International
ing of Ti–6Al–4 V. Materials & Design, 31 (8), 3982–3988. Journal of Production Research, 42 (1), 183–199.
Bezerra, M.A., et al., 2008. Response surface methodology (RSM) Olakanmi, E.O., Cochrane, R.F., and Dalgarno, K.W., 2015. A
as a tool for optimization in analytical chemistry. Talanta, 76 review on selective laser sintering/melting (SLS/SLM) of alu-
(5), 965–977. minium alloy powders: processing, microstructure, and prop-
Cabrini, M., et al., 2016. Evaluation of corrosion resistance of Al– erties. Progress in Materials Science, 74, 401–477.
10Si–Mg alloy obtained by means of Direct Metal Laser Paul, R., 2013. Modeling and optimization of powder based addi-
Sintering. Journal of Materials Processing Technology, 231, tive manufacturing (AM) processes. Thesis (PhD). Ohio:
326–335. University of Cincinnati.
Ghoreishi, M., 2006. Statistical analysis of repeatability in laser Pogson, S.R., et al., 2003. The production of copper parts using
percussion drilling. The International Journal of Advanced DMLR. Rapid Prototyping Journal, 9 (5), 334–343.
Manufacturing Technology, 29 (1–2), 70–78. Rossi, S., Deflorian, F., and Venturini, F., 2004. Improvement of
Hagedorn, Y., 2017. 6 – Laser additive manufacturing of ceramic surface finishing and corrosion resistance of prototypes pro-
components: materials, processes, and mechanisms A2 – duced by direct metal laser sintering. Journal of Materials
Brandt, Milan. Laser Additive Manufacturing. Woodhead Processing Technology, 148 (3), 301–309.
Publishing, Vol 1, 163–180. Available from: http://dx.doi.org/ Simchi, A., Petzoldt, F., and Pohl, H., 2003. On the development
10.1016/B978-0-08-100433-3.00006-3. of direct metal laser sintering for rapid tooling. Journal of
Kruth, J.-P., et al., 2005. Binding mechanisms in selective laser Materials Processing Technology, 141 (3), 319–328.
sintering and selective laser melting. Rapid Prototyping Sing, S.L., et al., 2016. Laser and electron-beam powder-bed
Journal, 11 (1), 26–36. additive manufacturing of metallic implants: A review on
Kruth, J.-P., et al., 2007. Consolidation phenomena in laser and processes, materials and designs. Journal of Orthopaedic
powder-bed based layered manufacturing. CIRP Annals- Research, 34 (3), 369–385.
Manufacturing Technology, 56 (2), 730–759. Tolochko, N.K., et al., 2003. Mechanisms of selective laser sinter-
Kumar, N., Kumar, H., and Khurmi, J.S., 2016. Experimental investi- ing and heat transfer in Ti powder. Rapid Prototyping Journal,
gation of process parameters for rapid prototyping technique 9 (5), 314–326.
(selective laser sintering) to enhance the part quality of proto- Tolochko, N.K., et al., 2004. Balling processes during selective
type by Taguchi method. Procedia Technology, 23, 352–360. laser treatment of powders. Rapid Prototyping Journal, 10
Lindemann, C.F.W., and Jahnke, U., 2017. 11 – Modelling of laser (2), 78–87.
additive manufactured product lifecycle costs A2 – Brandt, Wang, X., et al., 2002. Direct selective laser sintering of hard
Milan. Laser Additive Manufacturing. Woodhead Publishing, metal powders: experimental study and simulation. The
Vol 1, 281–316. Available from: http://dx.doi.org/10.1016/ International Journal of Advanced Manufacturing
B978-0-08-100433-3.00011-7. Technology, 19 (5), 351–357.
Lynn-Charney, C., and Rosen, D.W., 2000. Usage of accuracy Wang, X., et al., 2016. Densification of W–Ni–Fe powders using
models in stereolithography process planning. Rapid laser sintering. International Journal of Refractory Metals and
Prototyping Journal, 6 (2), 77–87. Hard Materials, 56, 145–150.
Manfredi, D., et al., 2014. Direct Metal Laser Sintering: an addi- Yan, C., et al., 2014. Evaluation of light-weight AlSi10Mg peri-
tive manufacturing technology ready to produce lightweight odic cellular lattice structures fabricated via direct metal
structural parts for robotic applications. La metallurgia itali- laser sintering. Journal of Materials Processing Technology,
ana, 1 (10), 15–24. 214 (4), 856–864.
Montgomery, D.C., 2008. Design and analysis of experiments. Yu, N., 2005. Process parameter optimization for direct metal laser
Hoboken, NJ: Wiley. sintering (DMLS). Thesis (PhD). Singapore: National University
Montgomery, D.C. and Myers, R.H., 1995. Response surface meth- of Singapore.
odology: process and product optimization using designed Zhu, H.H., Fuh, J. Y. H., and Lu, L., 2005. Microstructural evolution
experiments. Hoboken, NJ: Wiley. in direct laser sintering of Cu-based metal powder. Rapid
Morgan, R., Sutcliffe, C.J., and O’Neill, W., 2001. Experimental Prototyping Journal, 11 (2), 74–81.
investigation of nanosecond pulsed Nd:YAG laser re- Živčák, J., Šarik, M., and Hudák, R. 2016. FEA simulation of
melted pre-placed powder beds. Rapid Prototyping Journal, thermal processes during the direct metal laser sintering of
7 (3), 159–172. Ti64 titanium powder. Measurement, 94 (1), 893–901.

You might also like