You are on page 1of 12

Available online at www.sciencedirect.

com

ScienceDirect
Procedia Manufacturing 10 (2017) 779 – 790

45th SME North American Manufacturing Research Conference, NAMRC 45, LA, USA

Investigation of Sintering Shrinkage in Binder Jetting Additive


Manufacturing Process
Yujia Wang and Yaoyao Fiona Zhao *
Additive Design and Manufacturing Laboratory, Department of Mechanical Engineering, Mcgill University, Montreal, Quebec, H3A 0G4,
Canada

Abstract

Binder jetting process is an additive manufacturing technology that produces objects layer by layer from 3D digital data. After
printing process, the secondary post-processing operations, including curing and sintering, are used to solidify the object. The
linear dimensional accuracy is one of the most important end-product qualities in manufacturing process. However, it is not easy
to control it since the whole process combines several steps and covers different disciplines. Sintering process can affect the
shrinkage rate significantly through using different sintering profiles. This project focuses on investigating the effect of sintering
parameters on linear dimensional accuracy with stainless steel 316L. The Taguchi method is used to design 9 groups of
orthogonal experiments. After that, the results are converted to signal-to-noise (S/N) ratio and using analysis of variance
(ANOVA) to analyze. There are 3 sets of recommended sintering parameters to achieve the best dimensional accuracy for each
axis and there is also 1 set of recommended sintering parameters which considers dimensional accuracies of all 3 axes.
Comparing with the shrinkage rates measured using default sintering profile, the dimensional accuracies are improved.
© 2017 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
© 2017 The Authors. Published by Elsevier B.V.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-reviewunder
Peer-review underresponsibility
responsibility
of of
thethe Scientific
organizing Committee
committee of 45th
of the NAMRI/SME.
SME North American Manufacturing Research Conference

Keywords: Binder Jetting (BJ) process; Taguchi Method; Sintering Parameters Optimization; Signal-to-noise (S/N) ratio; Analysis of Variance
(ANOVA)

* Corresponding author. Tel.: +1-514-398-2523; fax: 514-398-7365.


E-mail address: yaoyao.zhao@mcgill.ca

2351-9789 © 2017 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 organizing committee of the 45th SME North American Manufacturing Research Conference
doi:10.1016/j.promfg.2017.07.077
780 Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790

1. Introduction

Additive manufacturing (AM) technology which is used to be called as Rapid Prototyping (RP), was developed
since 1980s. In AM processes, an object is created by adding material layer-by-layer to form a thin cross-section of
selected pattern directly from a CAD model [1]. There are several advantages of AM technology. First, the
capability of fabricating complex geometric shapes is one of the most important advantages of AM process
comparing to conventional manufacturing processes. Secondly, the number of required processes and resources can
be significantly reduced using AM [2]. According to the American Society for Testing and Materials (ASTM),
there are 7 categories of AM technology: VAT photopolymerization, material jetting, binder jetting, material
extrusion, powder bed fusion, sheet lamination and direct energy deposition [3].
Several of the AM processes can work with metal. The most well-known two types are the powder bed fusion
method and binder jetting [4]. Binder Jetting (BJ) process is a prominent AM technology that was developed at MIT
in early 1990s [2]. The object is created by selectively depositing binder on the powder bed by the ink print head to
join the powders layer-by-layer. BJ process can be divided into 7 steps: printing, curing, de-powdering, sintering,
infiltration, annealing and finishing [5]. Figure 1 shows the printing system of ExOne M-lab machine. During the
build process, BJ does not employ any high power heat sources compared to other AM technologies which use laser
or electron beam to melt metal powders. Therefore, BJ process is more cost-effective and it is also capable to print
large parts. Moreover, BJ process can print a wide range of materials including metals, sands, polymers and
ceramics. Some common metal materials used in BJ process are stainless steel 316, stainless steel 420 and iron.
Since the green part is weak after printing, curing process is needed after printing to increase the strength of green
part. After that, the de-powdering process is needed before putting the part into sintering furnace. A soft brush and
air duster are needed to clean the surrounding powder in order to avoid excessive powders attachment on the part
that may lead to linear dimension inaccuracy. Sintering process is an important process to improve the part density
and strength through burning off the binder and solidifying the part. The binder decomposition and neck growth
between particles occur in post-processing operations during which dimensional shrinkage and porosity reduction
happen. This affects the dimensional accuracy of the final part, which is an important measuring index of quality
control. Therefore, sintering process should also be studied and controlled to achieve desired dimensional accuracy.
This research focuses on improving the linear dimensional accuracies by optimizing the sintering parameters. In the
following sections, the related previous research and literature review are summarized in section 2, methodologies
are introduced in section 3, experimental results, analysis and discussion are presented in section 4 and conclusions
are presented in section 5.

Fig. 1. printing system of ExOne M-lab machine


Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790 781

2. Literature review

Three-dimensional printing (3DP) is a shaping process to form the part layer-by-layer. The ink print head passes
above the powders and prints the liquid binder [6]. Therefore, 3DP is similar to BJ process. The following section
reviews related research works on both 3DP and BJ process. The main objectives of these literatures are different.
Most of them focus on achieving full density [4, 6-10] and controlling the dimensional change [4, 8-11]. Table 1
summarizes the reviewed literature and factors they investigated.

Table 1. Summary of literature and factors they investigated.


Verlee et Johnston et Doyle et Do et Bai et al. Bai and Williams Sun et al.
al. [6] al. [11] al. [4] al. [7] [8] [9] [10]
Sintering
temperature 9 9 9 9 9 9
Sintering Sintering time 9 9 9 9
parameters
Heating rate 9
Atmosphere 9
Orientation 9
Printing
Layer thickness 9
parameters
Printing position 9
Particle shape 9 bimodal
powder
Particle Particle size 9 mixtures 9
Sintering aid 9 9

Johnston et al. focus on investigating dimensional changes of initial-stage sintering of 316L stainless steel in 3DP
[11]. By changing the isothermal peak temperature (1010 degree, 1100 degree, 1180 degree and 1263 degree) and
the heating rate (4 degrees/min, 7 degrees/min, 10 degrees/min and 20 degrees/min) to reach the peak isothermal
temperature, the in-sintering strains were measured by dilatometer. They found that besides the strain formation at
1263 degree, the total strains were almost the same regardless of the heating rate. The heating rate was found to
affect the strain formations significantly over time. Also, the amount of strain produced at isothermal temperature
increased with the increase of the isothermal peak temperature. Finally, during heating intervals, the strain formation
increased rapidly when the temperature was above 1120 degree. Comparison of experiment results with theoretical
model was done. However, the theoretical model can only apply to rapid heating rate (between 15-30 degree/min).
Using copper material in binder jetting process, Bai and Williams [9] found that both final part density and
volumetric shrinkage increased when sintering temperature increases. The densification improved while increasing
the sintering time; however, the highest density was not achieved at the longest sintering time. Sintering in reduced
atmosphere was compared with that in the air. The densification was not dramatically improved; however, the
sintered density did improve. After that, Bai et al. continuously investigated methods to improve the sintered density
by using bimodal powder mixture of copper in BJ process [8]. The results showed the bimodal powder mixtures had
effects on sintered density and powder bed density.
Sun et al. investigated the effect of adding ܵ݅ଷ ܰସ into stainless steel 420 powder on final density, dimensional
accuracy and mechanical properties [10]. It has been found that the sample would distort with high portion of ܵ݅ଷ ܰସ
and high sintering temperature. Do et al. demonstrated that adding the sintering aid in 420 stainless steel powder can
improve the final density of parts in 3DP [7]. They discovered, firstly, the final surface finish could be improved by
adding smaller additives. Secondly, the highest density found in the experiments was 97%. However, the sample
was distorted because extensive liquid phase was formed during the sintering by adding too much additives. Thirdly,
among the undistorted samples, the highest density was found at 90.22%. Finally, during the printing process, the
position of samples in the print bed was found to affect the shrinkage because of the compact factor of the powder.
782 Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790

Verlee et al. did the investigation about the effects of sintering temperature, sintering time, particle size and
particle shape on porous properties of the final part fabricated by 3DP [6]. The flowing results were gathered. First,
the sintering temperature and particle size affected the achieved density. Generally speaking, the higher the sintering
temperature, the higher the sintered density, and the lower the open porosity were achieved. The maximum sintering
temperature for 316L powders was found at 1430 degree. Collapse occurred above this temperature. Secondly, the
smaller the particles are, the higher surface energy and the higher densification can be achieved. Moreover,
according to mass transport, the finer the powders are, the smaller the pores between the particles will be which lead
to earlier densification during sintering. Therefore, the smaller the particles, the better density would be achieved at
lower sintering temperature. Finally, the particle shape was found to affect the final open porosity and density of the
part. The spherical powders had higher final density than non-spherical powders.
According to these literatures, few research focuses on sintering process after BJ process. Also, few of them
studies sintering shrinkage rate with stainless steel 316L powder in BJ process. Sintering parameters optimization in
BJ process has not been reported. Therefore, this research focuses on optimizing the sintering parameters to improve
the linear dimensional accuracy. In BJ process, both printing process and sintering process can affect shrinkage rate.
Thus, they need to be controlled properly in order to achieve desired dimensional accuracy.

3. Methodologies

3.1. Determination of influential sintering factors

The default sintering profile from ExOne is introduced first in this section. There are 4 isothermal holds in total.
The isothermal holds at 250 degrees and 420 degrees are to ensure complete binder elimination before sintering. To
control the temperature gradient across the part while it is heating, there is the third isothermal hold at 630 degrees.
At this temperature, the change from conduction to radiation heating occurs. The samples are held at this
temperature for 90 minutes. The fourth isothermal hold is at 1100 degrees for 90 minutes. The fourth isothermal
hold is the most important stage as the temperature is high enough for the neck growth. Then, the temperature
decreases at 10 degrees/min until it reaches to 50 degrees. The heating rate is at 10 degrees/ min for all the heating
stages except that the heating rate from 630 degrees to 1100 degrees which is 12 degrees/min. The three design
factors are chosen as isothermal sintering temperature, heating rate and sintering time which are introduced below.
Isothermal sintering temperature
Isothermal sintering temperature is one of the most important factors that would affect the shrinkage rate, has
been investigated by many researchers [6-11]. The temperature has complex effects on the sintering process due to
the continuously changing microstructure. Figure 2 shows the influence sintering temperature has on shrinkage rate.
It is obvious that isothermal sintering temperature is one of the parameters to be investigated.
Heating rate
Heating rate is also a changeable parameter in sintering profile. However, there is limited research about it.
Johnston et al. found that the theoretical model can only apply to rapid heating rate [11]. Therefore, a model based
on experimental data should be built to establish the relationship between heating rate and shrinkage rate.
Sintering time at isothermal sintering temperature
Same as the peak sintering temperature, the sintering time was also researched extensively [6, 8-10]. Figure 2
shows the influence of sintering temperature and sintering time on neck size, shrinkage and densification for the
sintering in powder metallurgy. It can be seen that sintering time is also one of the parameters to investigate.

(a) (b) (c)

Fig. 2. Influence of sintering time and sintering temperature on (a) neck size, (b) shrinkage and (c) densification. [12]
Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790 783

3.2. Design of experiment

The experiments of this research is designed following the Design of Experiment (DOE) theory. Experiments are
usually conducted to examine which factor has larger impact on the process output and at which level the factor
should be chosen to achieve the desired output. There are many different methods to design the experiments such as:
response surface design, mixture design, space filling design, full factorial design, Taguchi method, etc. [13].
Full factorial design is one of typical DOE methods that every combination should be run. As shown in table 2,
each factor has 3 levels. If using full factorial design method, all possible combinations are 27 meaning 27
experiments should be done. Since the cost of full factorial experiments is too high, Taguchi method is chosen in
this research. Taguchi method is a cost-effective method which uses orthogonal array to design the experiments.
Table 3 is the Lଽ orthogonal array table which can be operated by maximum four factors. The extra column can
be left for noise factor. Since small difference between two levels will not cause large difference of results, there are
only 3 levels of each factor. For the convenience, the combination of each run is denoted as the capital letter of each
factor and number of the level [14]. For example, the group 2 can be denoted as A1B2C2 which means the sintering
temperature is 1010 degree, heating rate is 12 degrees/min and the sintering time is 90 minutes.

Table 2. Experimental factors with each level.


Factors Range Level 1 Level 2 Level 3
Factor A (sintering temperature) 1010-1190 degrees Celsius 1010 1100 1190
Factor B (heating rate) 4-20 Celsius degrees/min 4 12 20
Factor C (sintering time) 60-120 minutes 60 90 120

Table 3. Lଽ (3ସ ) experimental orthogonal array table.


Noise Sintering Sintering Noise
Group No. Factor A Factor B Factor C Heating rate
Factor temperature time factor
1 1 1 1 1010 4 60
2 1 2 2 1010 12 90
3 1 3 3 1010 20 120
4 2 1 2 1100 4 90
5 2 2 3 1100 12 120
6 2 3 1 1100 20 60
7 3 1 3 1190 4 120
8 3 2 1 1190 12 60
9 3 3 2 1190 20 90

Table 4 shows the experiment specification. Printing parameters used in this research are the optimal printing
parameters investigated by Chen and Zhao [15]. The layer thickness is 100µm, binder saturation is 75%, drying time
for each layer is 30s and heating rate is 70%. For each set of experiment, three samples are printed. The average
shrinkage rates are calculated. It is known that different printing position along x-axis will affect the shrinkage rate
[7], while printing orientation has less effect on shrinkage rate [4]. Thus the positions and orientation are kept the
same for all the experiment runs.
The default sintering profile has the combination of A2B2C2 which should be conducted first. The dimensions
after sintering are not exactly equal to 15mm which is the design dimension. The dimensions of x-axis, y-axis and z-
axis are 14.95mm, 14.83mm and 15.11mm respectively.
784 Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790

Table 4. Experiment specification.


Machine ExOne X1-Lab
Powder PM-R1-S3-30 stainless steel 316 powder
Binder BS004
Cleaner CL001
Curing oven Constant temperature drying oven, Model RCO-D
Curing time 5 hours
Sintering furnace Tube furnace GSL-1300-40X
Dimension measurement tester Digimatic standard caliper, Mastercraft 58-6800-4
sample shape Cubic
Number of samples in one set 3
Sample Dimension (mm) 15x15x15

4. Experimental Results, Analysis and Discussion

4.1. Calculations of signal to noise ratio

After the experiments, the signal to noise(S/N) ratio and analysis of variance(ANOVA) are used to analyze the
results. During the experiments, for each axis, the sample is measured 3 times using a digital caliper. Then, the
average dimension is calculated. Before calculating the S/N ratio, the shrinkage rates of each axis are calculated
using the equation 1. In this research, the design dimension is 15mm. The purpose is to let the real dimension as
close as possible to 15mm. Hence, the shrinkage rate is expected to equal to 0. Therefore, the smaller-the-better
formula is used to calculate the S/N ratio (equation 2) [16].
(real dimension of sintered sample - design dimension of sample)
shrinkage percentage = × 100%
(design dimension of sample)
(1)
S 1
10 log10 ( ¦ Yi 2 )
N n
(2)
where Y୧ represents multiple values of performance characteristic (measured shrinkage rates) and n represents the
total number of multiple values.
After calculating the S/N ratio, the level of response can be calculated using equation 3. For example, the level 1
responses of sintering temperature on dimensional accuracy of x-axis can be calculated as adding the S/N ratios of
group 1-3 (Level response = 55.66062 dB +51.42072 dB +45.00924 dB =152.0906 dB). Other level responses of
certain factor can be calculated in the same way.
n S
level _ response ¦ i 1
(
N
)i
(3)
Figure 3 shows the level response tendency chart of effect of factors on shrinkage rate of x-axis. The effect of
sintering temperature on shrinkage rate is monotonic decreasing. It means the higher the sintering temperature, the
lower the dimensional accuracy of x-axis is. The effect of sintering time on shrinkage rate is also monotonic
decreasing. For the effect of heating rate on shrinkage rate, the tendency is increasing at first and then decreasing.
The level response tendency of effect of design factors on shrinkage rate of y-axis is shown in Figure 4. The
effect of sintering temperature on shrinkage rate is monotonic decreasing. Both heating rate and sintering time
increase at first and then decrease.
Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790 785

The level response tendency of effect of factors on shrinkage rate of z-axis is shown in Figure 5. For the sintering
temperature, the trend is monotonic increasing which is different from the trend on x-axis and y-axis. The effect of
heating rate on dimension accuracy on z-axis is increasing at first and then decreasing. For the sintering time, the
trend is decreasing at first and then increasing.

Fig. 3. Level response tendency chart of effect of factors on shrinkage rate on x-axis.

Fig. 4. Level response tendency chart of effect of factors on shrinkage rate on y-axis.

Fig. 5. Level response tendency chart of effect of factors on shrinkage rate on z-axis.
786 Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790

4.2. Investigation of optimal sintering parameters

From Figures 3-5, the optimal sintering parameters of dimensional accuracy of each axis can be determined. The
higher the level response, the closer sintered dimension is to designed dimension. The optimal sintering parameters
are shown in Table 5. For dimensional accuracy of x-axis, the highest levels for each factor are A1B2C1.
Considering the dimensional accuracy of y-axis, the optimal sintering parameters are A1B2C2. By the same token,
the optimal sintering parameters for dimensional accuracy of z-axis are A3B2C1. There is also one set of optimal
sintering parameters A2B2C1, which considers the dimensional accuracies of all three axes. For the sintering
temperature, the trends of dimensional accuracy of x-axis and y-axis are similar. The best status appears at the
lowest sintering temperature. However, the opposite trend occurs at z-axis. The best dimensional accuracy occurs at
highest sintering temperature. Considering this situation, the second level of sintering temperature is regarded as
suitable parameter for combined sintering profile. For the heating rate, the second level becomes undisputed optimal
level. The first level of sintering time can be treated as suitable parameter in combined sintering profile.

Table 5: optimal parameters


Optimal parameters Sintering temperature Heating rate Sintering time
For dimensional accuracy of x-axis
1010 degree 12 degrees/min 60 minutes
A1B2C1
For dimensional accuracy of y-axis
1010 degree 12 degrees/min 90 minutes
A1B2C2 (group 2)
For dimensional accuracy of z-axis
1190 degree 12 degrees/min 60 minutes
A3B2C1 (group 8)
For considering dimensional accuracies of all 3 axes
1100 degree 12 degrees/min 60 minutes
A2B2C1

4.3. Analysis of variance

Analysis of variance (ANOVA) is also a performance measurement method which statistically indicates at what
percentage the factor influences the performance characteristic. Tables 6-8 show the ANOVA of x-axis, y-axis and
z-axis respectively. For ANOVA of x-axis, the sintering temperature has significant influence on shrinkage rate
among these three design factors which has 72.75% of contribution. Sintering time has secondary effect on
shrinkage rate which has 15.62% of contribution. For the dimensional accuracy on y-axis, the sintering temperature
has overwhelming effect which occupies 90.08%. Other two design factors (heating rate and sintering time) have
less effect and only occupy 3.44% and 5.37% respectively. For dimensional accuracy on z-axis, the sintering
temperature has maximum effect which occupies 63.98%. Heating rate has secondary effect which occupies
23.56%. A more intuitive form is the percentage contribution which the pie charts (figure 6) are drown for each axis.

Table 6: ANOVA of effects of factors on shrinkage rate of x-axis


FACTOR SS DOF VAR F-ratio CONFIDENCE PERCENTAGE CONTRIBUTION
A (sintering temperature) 262.32 2 131.16 10.47 91.28% 72.75%
B (heating rate) 16.90 2 8.45 0.67 40.28% 4.69%
C (sintering time) 56.31 2 28.15 2.25 69.20% 15.62%
ERROR 25.06 2 12.53 6.95%
TOTAL 360.58 8
Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790 787

Table 7: ANOVA of effects of factors on shrinkage rate of y-axis


FACTOR SS DOF VAR F-ratio CONFIDENCE PERCENTAGE CONTRIBUTION
A (sintering temperature) 85.96 2 42.98 81.41 98.79% 90.08%
B (heating rate) 3.29 2 1.64 3.11 75.69% 3.44%
C (sintering time) 5.12 2 2.56 4.85 82.91% 5.37%
ERROR 1.06 2 0.53 1.11%
TOTAL 95.43 8

Table 8: ANOVA of effects of factors on shrinkage rate of z-axis


FACTOR SS DOF VAR F-ratio CONFIDENCE PERCENTAGE CONTRIBUTION
A (sintering temperature) 182.25 2 91.13 9.52 90.49% 63.98%
B (heating rate) 67.12 2 33.56 3.51 77.81% 23.56%
C (sintering time) 16.32 2 8.16 0.85 46.02% 5.73%
ERROR 19.14 2 9.57 6.72%
TOTAL 284.84 8

Fig. 6. Percentage contributions of factors on shrinkage rate of x-axis, y-axis and z-axis.

4.4. Physical Interpretation of Experimental Data

For all three axes, the sintering temperature has significantly high percentage contributions on shrinkage rate.
Equations 4 and 5 are the formulas of shrinkage during initial-stage [12].
n
'L Bt
( )2
L0 2 Dmn

(4)
Q
B B0 exp( )
RT
(5)
ο୐
where is the shrinkage, t is isothermal time, D is the particle diameter, T is isothermal temperature, Q is an
୐బ
activation energy and n, m, R and B଴ are the constants which depend on the material. The sintering temperature is
the exponent of equation 5. It means the small change of sintering temperature will affect the shrinkage rate a lot.
788 Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790

It can be seen that the trends of three axes are totally difference, especially for the trend of z-axis compared to the
other two axes. This is due to the differences of dimension on each axis after curing. The average dimensions of x-
axis, y-axis and z-axis are 15.1284 mm, 15.0249 mm and 15.3316 mm respectively. They are slightly larger than the
design dimension due to the existence of the binder. However, the dimension of z-axis expands the most compared
to the dimensions of x-axis and y-axis. Similar trend was found by Farzadi et al. that dimensions on z-axis exceeds
the most than the design dimension while the dimensions of x-axis and y-axis do not have much difference [17]. The
possible reason for this phenomenon is that y-axis is parallel to the movement of print-head, so the binder is dropped
coincides with the longitudinal direction of the sample along y-axis. The binder slightly bleed on x-axis and the
excessive binder sticks on z-axis which increases the dimension of z-axis. Since the dimensions of x-axis and y-axis
after curing are close to design dimension, the small shrinkage rates are needed on both two axes. However, the
dimension on z-axis after curing expands too much. Therefore, the relative larger shrinkage rate is needed to reach
the design dimension.

4.5. Confirmation test

Confirmation test is one of the most important procedure in Taguchi method. Once the optimal sintering profiles
have been selected, equation 6 can be used to calculate the estimate S/N ratio which can calculate the predicted
shrinkage rate [14]. There are two objectives to do the verification experiment. The first objective is to get accurate
predicted results. The second objective is to verify the improvement of performance characteristics. As shown in
table 5, the optimal parameters for dimensional accuracy of y-axis and z-axis have already been conducted.
Therefore, there are only 2 sets of optimal parameters (optimal parameters of dimensional accuracy on x-axis and
optimal parameters of considering all three axes) need to be verified.
m
(S/N) predict (S/N) mean + ¦i 1[(S/N) i  (S/N) mean ]
(6)
where (ܵ/ܰ)௠௘௔௡ is the total mean S/N ratio, (ܵ/ܰ)௜ is the average S/N ratio at optimal level and m is equal to
the number of design factors. For example, the optimal parameters of x-axis are A1B2C1, the total mean S/N ratio
of x-axis is equal to 46.652 dB, the mean S/N ratios of A1, B2 and C1 are 50.697 dB, 48.525 dB and 50.147 dB
respectively. Therefore, the predicted S/N ratio of optimal parameters for dimensional accuracy of x-axis can be
calculated as: (S/N)௣௥௘ௗ௜௖௧௘ = 46.652 + [(50.697 െ 46.652) + (48.525 െ 46.652) + (50.147 െ 46.652)] =
56.065 ݀‫ܤ‬. From the S/N ratio, the shrinkage rate can be calculated using equation (2) which the corresponding
shrinkage rate is -0.001573.
The results of verification experiments are shown in Table 9. For the optimal parameters of dimensional accuracy
on x-axis, the predicted shrinkage rate is -0.157%. The shrinkage rate measured in the experiment is -0.141% which
is close to the predicted value. Moreover, the sintered dimension on x-axis of verify experiment is the closest to the
design dimension among all the experiments. For the combined optimal parameters that consider the dimensional
accuracies of all 3 axes, the experimental value on x-axis is -0.17% which is close to the predicted value -0.166%.
The experimental shrinkage rate on y-axis is -1.089% which is close to the predicted shrinkage rate -1.092%. The
experimental shrinkage rate on z-axis is 0.482% which is close to the predicted shrinkage rate 0.485%. Therefore,
all the experimental shrinkage rates are close to the predicted shrinkage rates. It means the experiment is repeatable.

Table 9: results of verify experiments


S/N ratio Shrinkage rate
Sintering Profiles Parameters
Predicted Experimental Predicted Experimental (Dismatch)
Optimal parameters for x-axis A1B2C1 56.065 56.663 -0.157% -0.141% (-0.016%)
x-axis 55.606 55.355 -0.166% -0.170% (0.004%)
Optimal parameters which
y-axis A2B2C1 39.238 39.259 -1.092% -1.089% (-0.003%)
considering all 3 axes
z-axis 46.283 46.346 0.485% 0.482% (0.003%)
Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790 789

Table 10 shows the comparison between the results under the optimal parameters on each axis and the results
under default setting. It can be seen that the dimensional accuracy of x-axis after optimizing is improved 54.66%.
The dimensional accuracy of y-axis after optimization is improved 32.86%. There is significant improvement of
dimensional accuracy of z-axis which is 73.98%.

Table 10: percentage improvement for all three sets of optimal parameters of each axis
Experimental
Sintering Profile Parameters Initial shrinkage rate Percentage improved
shrinkage rate
Optimal parameters of shrinkage rate on
A1B2C1 -0.141% -0.311% 54.66%
x-axis
Optimal parameters of shrinkage rate on
A1B2C2 (group 2) -0.756% -1.126% 32.86%
y-axis
Optimal parameters of shrinkage rate on
A3B2C1 (group 8) -0.185% 0.711% 73.98%
z-axis
The results of percentage improvement for combined optimal parameters are shown in Table 11. For x-axis, the
dimensional accuracy improved 45.34%. For y-axis, the dimensional accuracy slightly improved 3.29%. For z-axis,
the dimensional accuracy improved 32.29%. The percentage improvements are less compared to the results only
considering the dimensional accuracy of one axis. This is due to the trends of level responses on three axes are
different.

Table 11: percentage improvement for combined optimal parameters on each axis
Sintering Profile Parameters Experimental shrinkage rate Initial shrinkage rate Percentage improved
Combined optimal
-0.170% -0.311% 45.34%
parameters on x-axis
Combined optimal
A2B2C1 -1.089% -1.126% 3.29%
parameters on y-axis
Combined optimal
0.482% 0.711% 32.29%
parameters on z-axis

5. Conclusions

In conclusion, the relation between the sintering parameters and linear shrinkages of BJ AM process has been
investigated in this research. In order to improve the dimensional accuracy of each axis, the isothermal sintering
temperature, heating rate and sintering time have been chosen as design factors. Taguchi method is used to design
the experiment to reduce the cost of experiment. Lଽ orthogonal array has been used in this research. After the
experiments, the dimensions of each axis are measured and the S/N ratio has been calculated. Then, ANOVA has
been used based on the S/N ratio to analyze the experiment results. The three sets of optimal sintering parameters of
dimensional accuracy of each axis have been found and there is also one set of optimal sintering parameters which
considers dimensional accuracies of three axes.
The confirmation tests have also been conducted. The optimal parameters of x-axis are the sintering temperature
at 1010 degree, heating rate at 12 degrees/min and sintering time for 60 minutes. The result is close to the predicted
result. Comparing to the shrinkage rate occurred under default sintering profile, the dimensional accuracy improved
54.66%. The optimal sintering profile of y-axis is sintering temperature at 1010 degree, heating rate at 12
degrees/min and sintering time for 90 minutes. The dimensional accuracy has been improved 32.86%. Also, for the
shrinkage rate of z-axis, the optimal sintering profile is the sintering temperature is at 1190 degree, heating rate is at
12 degrees/min and sintering time is 60 minutes. The dimensional accuracy has been improved 73.98%. When all
three axes are considered simultaneously, one set of optimal combined sintering parameters have been chosen with
the sintering temperature at 1100 degree, heating rate at 12 degrees/min and sintering time for 60 minutes. The
dimensional accuracies have been improved 45.34%, 3.29% and 32.29% for x-axis, y-axis and z-axis respectively.
790 Yujia Wang and Yaoyao Fiona Zhao / Procedia Manufacturing 10 (2017) 779 – 790

Based on ANOVA result, the sintering temperature is the most important factor that affects the dimensional
accuracies of three axes.
This research concludes the relation between the sintering parameters and linear shrinkages of BJ AM process
which can be used for both the manufacturing and design stages. It also found 4 sets of optimal sintering parameters
which can be used to improve the dimensional accuracy in different demand. Based on that, the further work can
link the printing process and sintering process together that can better control the dimensional accuracies of each
axis.

References

[1] Y. Zhang and A. Bernard, "AM Feature and Knowledge Based Process Planning for Additive Manufacturing in Multiple Parts Production
Context."
[2] I. Gibson, D. W. Rosen, and B. Stucker, Additive manufacturing technologies vol. 238: Springer, 2010.
[3] B. Stucker, "Additive manufacturing technologies: technology introduction and business implications," in Frontiers of Engineering: Reports
on Leading-Edge Engineering From the 2011 Symposium, National Academies Press, Washington, DC, Sept, 2012, pp. 19-21.
[4] M. Doyle, K. Agarwal, W. Sealy, and K. Schull, "Effect of Layer Thickness and Orientation on Mechanical Behavior of Binder Jet Stainless
Steel 420 + Bronze Parts," Procedia Manufacturing, vol. 1, pp. 251-262, // 2015.
[5] S. Meteyer, X. Xu, N. Perry, and Y. F. Zhao, "Energy and material flow analysis of binder-jetting additive manufacturing processes,"
Procedia CIRP, vol. 15, pp. 19-25, 2014.
[6] B. Verlee, T. Dormal, and J. Lecomte-Beckers, "Density and porosity control of sintered 316L stainless steel parts produced by additive
manufacturing," Powder Metallurgy, vol. 55, pp. 260-267, Sep 2012.
[7] T. Do, C. S. Shin, D. Stetsko, G. VanConant, A. Vartanian, S. Pei, et al., "Improving Structural Integrity with Boron-based Additives for 3D
Printed 420 Stainless Steel," Procedia Manufacturing, vol. 1, pp. 263-272, // 2015.
[8] Y. Bai, G. Wagner, and C. B. Williams, "Effect of Bimodal Powder Mixture on Powder Packing Density and Sintered Density in Binder
Jetting of Metals."
[9] Y. Bai and C. B. Williams, "An exploration of binder jetting of copper," Rapid Prototyping Journal, vol. 21, pp. 177-185, 2015.
[10]L. Sun, Y. H. Kim, D. Kim, and P. Kwon, "Densification and Properties of 420 Stainless Steel Produced by Three-Dimensional Printing
With Addition of Si3N4 Powder," Journal of Manufacturing Science and Engineering-Transactions of the Asme, vol. 131, p. 7, Dec 2009.
[11]S. Johnston, D. Frame, R. Anderson, and D. Storti, "Strain Analysis of Initial Stage Sintering of 316L SS Three Dimensionally Printed
(3DP™) Components."
[12]R. M. German, Powder metallurgy and particulate materials processing : the processes, materials, products, properties and applications,
2005.
[13]A. JMP and M. Proust, "Design of Experiments Guide," 2010.
[14]H. Chen, "A process modelling and parameters optimization and recommendation system for binder jetting additive manufacturing process,"
McGill University, 2016.
[15]H. Chen and Y. F. Zhao, "Process parameters optimization for improving surface quality and manufacturing accuracy of binder jetting
additive manufacturing process," Rapid Prototyping Journal, vol. 22, pp. 527-538, 2016.
[16]K. Dehnad, Quality control, robust design, and the Taguchi method: Springer Science & Business Media, 2012.
[17]A. Farzadi, M. Solati-Hashjin, M. Asadi-Eydivand, and N. A. A. Osman, "Effect of layer thickness and printing orientation on mechanical
properties and dimensional accuracy of 3D printed porous samples for bone tissue engineering," PloS one, vol. 9, p. e108252, 2014.

You might also like