Professional Documents
Culture Documents
20090801JMPT
20090801JMPT
net/publication/4332563
CITATIONS READS
17 1,642
3 authors, including:
Soon-man Kwon
Changwon National University
76 PUBLICATIONS 663 CITATIONS
SEE PROFILE
All content following this page was uploaded by Soon-man Kwon on 22 December 2015.
a r t i c l e i n f o a b s t r a c t
Article history: Most rapid prototyping (RP) technologies apply a layered manufacturing (LM) process to efficiently fab-
Received 5 December 2008 ricate 3D physical models. However, a critical drawback that reduces the surface quality of the RP parts
Received in revised form 30 April 2009 occurs by utilizing LM. Hence, topics related to surface roughness have been a key issue in RP. In this paper,
Accepted 17 May 2009
a new approach to model surface roughness in fused deposition modeling (FDM) is proposed. Based on
actual surface roughness distributions of FDM parts, a theoretical model to express surface roughness
Keywords:
distribution according to changes in surface angle is presented by considering the main factors that cru-
Fused deposition modeling
cially affect surface quality. The proposed expression is verified by implementation and comparison with
Surface roughness
empirical data. Also, the effectiveness of the main factors is analyzed and discussed.
© 2009 Elsevier B.V. All rights reserved.
1. Introduction ing purposes. For this reason, surface roughness is a key issue in RP
(Ippolito et al., 1995; Reeves and Cobb, 1997; Mahesh et al., 2004).
In consumer products, the trend is gradually leading towards In the broad spectrum of RP, a number of research works related
preferring sensitivity factors over functional parts. Using only dig- to the topic of surfaces have been presented, but most of them
ital data, such as CAD models and virtual reality, is not sufficient have been concerned with the problem of the part itself for pro-
to reflect sensitivity effects in the design stage of the manufactur- cess optimization. That is, there have not been many approaches
ing process. As a person can touch and feel natural objects, the role focusing on investigating and modeling surface roughness. The fol-
of physical models is expanding. An advanced technology that can lowing approaches are directly associated with surface roughness
efficiently form 3D physical models has been introduced to meet representation. Reeves and Cobb (1998) proposed a mathematical
this demand (Jacobs, 1996). model to represent the surface roughness of SL parts. Attributes
Rapid prototyping (RP) is a manufacturing technology that fab- of layer thickness, surface angle, and layer profile were used for
ricates 3D physical models using a layered manufacturing (LM) modeling surface roughness. The calculated roughness values of the
process that stacks and bonds thin layers in one direction. In presented model were compared with those of a test part designed
comparison with the previous numerically controlled (NC) man- to measure and obtain average surface roughness accurately. Addi-
ufacturing technology, RP can rapidly fabricate high level models tionally, reduction of inherent surface deviation was carried out by
with complex shapes without geometric restriction under more analyzing and applying an effect called “print-through.” The rough-
comfortable work conditions. Hence, the manufacturing technol- ness model and reduction improvement seem to be reasonable for
ogy has been widely applied in various fields, from industrial a limited range of surface angles. Paul and Voorakarnam (2001)
products to medical appliances. SL (stereolithography), SLS (selec- investigated the source of surface roughness in LOM. An expres-
tive laser sintering), FDM (fused deposition modeling), and LOM sion for predicting the surface roughness of LOM prototypes was
(laminated object manufacturing) are representative RP technolo- deduced based on centerline average roughness and a fundamen-
gies (Lee, 1999; Chua et al., 2003). tal model that represents penetration depth of the SL part resin.
However, as the RP process is inherently performed by layered A full-factorial experiment was performed to analyze the effects
manufacturing, the surface finish of the RP part is inevitably exces- of layer thickness and orientation angle. The results show that the
sively rough. The defect is extreme on the inclined surfaces of the roughness model is restrictively available for prototypes with com-
parts and is known as the “stair stepping effect.” As a result, the paratively thin paper thickness and at small orientation angles.
surface finish of the parts is not satisfactory for general engineer- Luis Pérez et al. (2001) proposed a theoretical model to represent
a roughness average of general LM parts. The roughness model
was expressed considering layer thickness and horizontal distance
∗ Corresponding author. Tel.: +82 55 751 6694; fax: +82 55 757 5622. between layers as main factors. SL test parts were manufactured,
E-mail address: dkahn@gnu.ac.kr (D. Ahn). and an experimental analysis of the resulting surface roughness was
0924-0136/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2009.05.016
Author's personal copy
5594 D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600
D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600 5595
directly influenced by the interval. Hence, the air gap effect is not
considered in this paper.
Based upon the factors and the profile schematic, an expression
that can compute the surface roughness value according to sur-
face angle variation in FDM is derived by the following procedure.
The cross-sectional shape of the filament is defined as an ellipse.
Thus, the cross-section profile of the filament on the n-th layer is
expressed as:
x2 y2
+ 2 =1 (1)
a2 b
where a and b denote the half-lengths of the long and short axes of
the ellipse, respectively. Thus, the width (w) and height (h) of the
ellipse are given by:
w = 2a (2-a)
h = 2b (2-b)
If the overlap interval is equal to the value c, the actual interval (t)
of the layer thickness is obtained as:
t =h−c (3)
The value of the surface angle at one point on the surface of the
RP part is defined by the angle between the fabrication direction
and the normal vector of the spot, as illustrated in Fig. 4. Therefore,
when the fabrication process is performed at surface angle , the
Fig. 3. Magnified cross-sectional shape of the filament used in FDM. (a) Section view ellipse curve of the next layer (n + 1) is expressed as:
of deposited filaments (Sun et al., 2008) and (b) shape of deposited filaments (Ahn 2 2
et al., 2002). (x − k1 ) (y − k2 )
+ =1 (4)
a2 b2
where k1 and k2 denote translated lengths of the n-th filament
according to the x axis and y axis directions, respectively, repre-
Fig. 4. Schematic for modeling the surface profile of the FDM-processed part.
Author's personal copy
5596 D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600
sented as:
⎧
⎪
⎨
t
0 < < , =
/
tan 2
k1 = (5)
⎪
⎩0 =
2
k2 = t (6)
−(x − x1 )
y= + y1 (10)
tan( − )
Author's personal copy
D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600 5597
Fig. 6. Fabrication and measurement of an FDM test part. (a) Test part and surface
roughness measurement and (b) Empirical surface roughness data.
Fig. 7. Comparison of the measured data and computed values of surface roughness.
(a) Computed Ra for a layer thickness of 0.254 mm and (b) computed Ra for a layer
Thus, the coordinate value yt1 is obtained by inserting the value x1 thickness of 0.178 mm.
into Eq. (10). Therefore, a temporary centerline (lt ) is expressed by:
Additionally, two points (pt2 , pt3 ) are calculated by intersecting the At = Ap − Av (17)
line defined by Eq. (11) and the ellipse curve given by Eq. (1). Next, Thus, the final closed areas Afp and Afv are determined by moving
the other boundary line (lb2 ) is derived as: the temporary centerline until At is close to zero, as explained in Eq.
−(x − x2 ) (9). Hence, the final computed area required to obtain the average
y= + y2 (12) surface roughness is calculated as:
tan( − )
Next, point pt4 (xt4 , yt4 ) is calculated by intersecting the line defined Af = Afp + Af v (18)
by (12) and the line given by Eq. (11). The distance between the two boundary lines is given by:
After obtaining all four points, the area (Ap , Av ) is calculated by
1/2
Eqs. (13)–(16), as shown in Fig. 4. Each set (f(x), f(y)), (lt (x), lt (y)), l = [(x1 − x2 )2 + (y1 − y2 )2 ] (19)
and (lb (x), lb (y)) is an expanded form based on variables x and y of
the ellipse curve, the temporary centerline, and the boundary line, Therefore, the final roughness value at a surface angle is computed
respectively. by the following:
pt2 Af
Ap = [f (y) − lt (y)]dy (13) Ra = (20)
l
pt3
xt1 xt2
All roughness values at surface angles between 0◦ and 180◦ are
obtainable by the presented expressions and computation process.
Av1 = [lb1 (x) − f (x)]dx + [lt (x) − f (x)]dx (14)
x1 xt1
Moreover, the total computed Ra at all surface angles from 0◦ to
360◦ is also easily calculated because the surface roughness forms
yt3 yt4
a symmetric distribution based on a surface angle of 180◦ . Conse-
Av2 = [lt (y) − f (y)]dy + [lb2 (y) − f (y)]dy (15)
yt4 y2
quently, if the two adjacent ellipse curves intersect, the value (Ra )
at all surface angles is accurately and efficiently computed by the
Av = Av1 + Av2 (16) proposed methodology.
Author's personal copy
5598 D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600
Test Independent value Dependent value Verification of the proposed roughness expression is essential.
LT (t) W/H (r) OI (c) RW(w) RH (h) Thus, FDM test parts were fabricated, and empirical roughness
data was obtained by measurement. In parallel, computation of the
1 90 516 344
2 120 561 374
introduced equation was carried out by coding with an application
254 1.5
3 150 606 404 tool. A comparison of the results from the measured data and the
computed values of surface roughness is presented. Additionally,
4 50 342 228
5 178 1.5 80 387 258
the effects of the major factors that influence surface roughness are
6 110 432 288 analyzed and discussed.
Values t, c, w and h are in micrometers; r is dimensionless value.
3.1. Verification
2.3. Indeterminate region In order to verify the proposed roughness, two test parts were
fabricated by the layer thickness as shown in Fig. 6a. The RP machine
The indeterminate region is required for investigating the fea- and fabrication material used in manufacturing the test parts were
sible region of the proposed expression. If the surface angle () FDM Maxum and ABS, respectively. The applied layer thicknesses
gradually decreases, the two ellipse curves can intersect at one were 0.254 and 0.178 mm. The test part model was designed to
point, as shown in Fig. 5a. The surface angle at this point is given obtain measured roughness values accurately and efficiently. That
as the critical angle ( c ). The critical angle depends on the surface is, several rotated planes were formed at the surface of the model.
angle, layer thickness, and ellipse shape. If the surface angle is less The part geometry was similar to that of Reeves and Cobb (1997),
than the critical angle value, the two adjacent ellipses do not inter- and was designed to measure the average surface roughness of an
sect, as described in Fig. 5b. In this case, to prevent deformation and SL part. The total number of the measured roughness data was 60
to intensify the rigidity of the FDM part, the fabrication process at surface angles 0–180◦ by rotating the planes at 3◦ angle inter-
is generally performed by raster angle such as cross (0◦ /90◦ ) and vals. Surface roughness was measured by a Surftest Formtracer
crisscross (+45◦ /−45◦ ) deposition style at the inner region, except (Mitutoyo Corp., 1997). From almost 50 measurements, two rough-
for the outer surface area (Mukesh et al., 1996; Ahn et al., 2004). ness data sets (MSR1 and MSR2) were obtained according to the
Fig. 5c shows an example of the FDM part by the crisscross build two applied layer thicknesses of 0.254 and 0.178 mm, respectively,
style. Hence, the tendency of surface roughness distribution can be as shown in Fig. 6b. Each series of roughness data was taken by
extremely variable according to each measurement point (a–d), as considering the set of data within one standard deviation of the
shown in Fig. 5d. That is, the surface roughness profile cannot be mean surface roughness from 100 trial measurements. MSR2 data is
defined as a functional relation. Therefore, the region under the crit- smaller than that of MSR1 owing to the applied layer thickness dif-
ical angle is named the “indeterminate region” and is not treated in ference. Moreover, for both MSR1 and MSR2, the surface roughness
this paper. values (Ra ) measured near a surface angle of 90◦ are comparatively
Table 2
Measured data and computed values of surface roughness.
0 8.85 16.85 – – – – – –
3 46.31 26.31 – – – – – –
6 23.22 33.22 – – – – – –
9 44.18 34.18 – – – – – –
12 21.24 31.24 – – – – – –
15 49.36 39.36 – – – – – –
18 39.28 34.28 – – – – – –
21 45.82 40.82 – – – – – –
24 37.66 35.66 – – – – – –
27 38.11 32.11 – – – – – –
30 36.83 26.83 – – 36.51 – – 24.71
33 33.37 24.37 – 36.04 31.02 – 26.23 21.13
36 31.60 21.16 – 31.26 27.46 – 22.57 18.75
39 30.06 20.06 32.75 28.12 24.95 – 20.23 17.06
42 28.77 19.07 29.73 25.88 23.10 23.28 18.59 15.82
45 26.26 19.06 27.60 24.22 21.71 21.38 17.38 14.87
48 21.86 18.12 26.01 22.95 20.64 20.01 16.46 14.14
51 22.81 18.28 24.80 21.97 19.80 18.99 15.74 13.59
54 21.13 17.13 23.86 21.20 19.15 18.22 15.18 13.14
57 21.14 16.94 23.13 20.59 18.63 17.62 14.75 12.77
60 20.71 16.71 22.54 20.10 18.20 17.14 14.39 12.49
63 20.13 15.43 22.07 19.72 17.87 16.76 14.11 12.27
66 19.86 15.86 21.69 19.40 17.60 16.47 13.88 12.08
69 18.56 14.56 21.40 19.15 17.38 16.22 13.71 11.94
72 18.39 14.79 21.17 18.96 17.22 16.04 13.57 11.82
75 18.75 13.05 20.98 18.80 17.09 15.90 13.45 11.73
78 19.67 13.27 20.84 18.69 16.98 15.77 13.37 11.67
81 19.44 13.44 20.74 18.60 16.91 15.70 13.30 11.61
84 20.08 13.08 20.67 18.54 16.86 15.63 13.25 11.58
87 19.59 12.59 20.62 18.51 16.83 15.60 13.23 11.56
90 17.35 12.51 20.61 18.50 16.81 15.58 13.22 11.55
Author's personal copy
D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600 5599
smaller than those of the other surface angles. A common char- Table 3
Test condition for analyzing the effect of the ratio of width to height of the section
acteristic in most RP-processed parts, this is due to using the LM
area of the filament.
process, as mentioned above. However, in spite of the same mea-
surement conditions and layer thicknesses, the roughness of the Test Independent value Dependent value
two layer thicknesses are extremely different between the angle LT (t) OI (c) W/H (r) RH (h) RW(w)
ranges 0–30◦ and 150–180◦ from the horizontal planes. This is due 1 1.2 449
to the “indeterminate region,” as explained in Section 2.3. Hence, 2 254 120 1.5 374 561
that region is not considered in verification. 3 1.8 673
Computation of the proposed roughness expression was carried 4 1.2 310
out by coding in the Mathematica (Wolfram Research Inc., 2003) 5 178 80 1.5 258 387
application tool. The same layer thicknesses (t) were used as in fab- 6 1.8 464
ricating the test parts, 0.254 and 0.178 mm. The applied ratio (r) of Values t, c, w and h are in micrometers; r is dimensionless value.
the road width to the road height was 1.5. Other input values were
the same as those used in fabricating the test parts. The specific
input overlap interval was determined based on the verification
test conditions for verification are shown in Table 1. The variables
results (Fig. 7a and b), and the interval between ratios was set to 0.3
OI, RH, RW, and W/H denote overlap interval (c), road height (h),
to adequately distinguish each surface roughness distribution. The
road width (w), and the ratio (r), respectively; these factors are also
dependent values h and w were given by the independent values (t,
illustrated in Fig. 4. Actually, it is difficult to know the exact values
c, and r) from expressions (2) and (3).
of the overlap interval in advance, because the value is a kind of
The computed surface roughness values vs. the ratio change
correction factor to adjust layer thickness. Fig. 3 shows an overlap
were also plotted, shown in Fig. 8. The results for layer thicknesses
interval between adjacent layers. In the experiment, the base value
of 0.254 and 0.178 mm are shown in Fig. 8a and b, respectively. In
of the overlap interval was determined based on the interval pre-
each layer thickness, as the ratio increases, the computed Ra values
sented in Fig. 3. And, two additional interval values were used to
also increase together, and the range that has the higher roughness
know the effect of the interval variation by 30 plus and minus inter-
values is expanded. This is caused by increasing the areas enclosed
vals in micrometers from the base value. That is, the interval values
by the assessed profile to the basic evaluation length as the ratio
are input by classifying three cases for each layer thickness as pre-
increases. Also, it is shown that the indeterminate region is reduced
sented in Table 1. Afterward, the values t, c, and r are independently
determined as input values, and values h and w are dependently
calculated by Eqs. (2) and (3).
Table 2 shows the measured data and the computed values
according to the test conditions given in Table 1 for surface angles
from 0◦ to 90◦ . From all the data and computed values, distribution
graphs were obtained with respect to surface angle from 0◦ to 180◦
at 3◦ intervals, as shown in Fig. 7.
Comparisons of the layer thicknesses 0.254 and 0.178 mm are
presented in Fig. 7a and b, respectively. It is shown that the
computed roughness distribution profiles are shifted down by
increasing the overlap interval. That is to say, the roughness val-
ues are diminished in proportion to the interval. And, the range of
the computed roughness depends on the critical angle ( c ) in all
test conditions. That is, the indeterminate region decreases as the
overlap interval increases from comparing the two roughness data
sets. Additionally, the profile curves of Test2 and Test5 are closest to
those of MSR1 and MSR2. Deviations between the measured data
and computed values were within 3–5 m over a limited surface
angle range by comparison between MSR1 and Test2; MSR2 and
Test5. However, by considering the distributed roughness values, it
can be estimated that even this deviation is due to measurement
error. Moreover, the roughness profile variations with surface angle
change are similar to those of the measured surface roughness val-
ues under all test conditions. Therefore, it can be concluded that the
validity of the proposed surface roughness expression is reasonable.
5600 D. Ahn et al. / Journal of Materials Processing Technology 209 (2009) 5593–5600
according to the increase in the ratio. This is due to expansion of the with the presented surface roughness expression. Therefore, more
overlap area between the adjacent filaments, as the length of the advanced process planning can be achieved in RP.
long axis comparatively increases. From the results, prediction of
surface roughness for a given nozzle size and layer thickness can be Acknowledgment
more accurately accomplished.
This work was supported by a Korea Research Foundation Grant
3.3. Overlap interval funded by the Korean Government (MOEHRD) (KRF-2008-005-
J01001).
The overlap interval is directly related to the layer thickness,
as explained in Section 2.2. Thus, the interval can also influence References
surface roughness. In this section, the effect of the interval is ana-
lyzed. Table 1 also shows the test conditions for the interval effect Ahn, D., Kim, H., Lee, S., 2009. Surface roughness prediction using measured data and
interpolation in layered manufacturing. J. Mater. Process. Technol. 209, 664–671.
because the overlap interval is constant according to each applied Ahn, D.K., Lee, S.H., 2007. Improving the surface roughness of SL parts using a coating
layer thickness. Hence, from the verification results presented in and grinding process. Int. J. Precision Eng. 8, 14–19.
Fig. 7, the effect can be concluded as follows. As the overlap intervals Ahn, S.H., Lee, C.S., Jeong, W., 2004. Development of translucent FDM parts by post-
processing. Rapid Prototyping J. 10 (4), 218–224.
increase, the computed values (Ra ) of surface roughness decrease
Ahn, S.H., Montero, M., Odell, D., Roundy, S., Wright, P.K., 2002. Anisotropic materials
at all surface angles except for the intermittent region. The surface properties of fused deposition modeling ABS. Rapid Prototyping J. 8 (4), 248–257.
roughness profile at each interval variation seems to be approxi- Chua, C.K., Leong, K.F., Lim, C.S., 2003. Rapid Prototyping: Principles and Applications.
mately an offset curve of the other profile. This effect commonly World Scientific, Singapore, pp. 1–23.
Ippolito, R., Iuliano, L., Torino, P., 1995. Benchmarking of rapid prototyping techniques
occurs in the case of the two layer thickness condition. The result in terms of dimensional accuracy and surface finish. Ann. CIRP 44 (1), 157–160.
comes from reduction of the layer thickness by increasing the inter- Jacobs, P.F., 1996. Stereolithography and Other RP&M Technologies. Society of Man-
val. Further, as the interval increases, the intermittent region is ufacturing Engineers, New York, pp. 1–26.
Lee, K.W., 1999. Principles of CAD/CAM/CAE Systems. Addison Wesley, Mas-
reduced. This is also due to the overlap area expansion by the sachusetts, pp. 378–431.
interval increase. Consequently, the effects on surface roughness Luis Pérez, C.J., Calvet, J.V., Sebastián Pérez, M.A., 2001. Geometric roughness analysis
distribution of changes in overlap interval can be predicted. in solid free-form manufacturing process. J. Mater. Process. Technol. 119, 52–57.
Mahesh, M., Wong, Y., Fuh, J., Loh, H., 2004. Benchmarking for comparative evaluation
of RP systems and processes. Rapid Prototyping J. 10 (2), 123–135.
4. Conclusions Mitutoyo Corp., 1997. Mitutoyo Contracer CBH-400 User’s Guide.
Mukesh, K., Agarwala, V.R., Jamalabad, N.A., Langrana, A.S., Philip, J.W., Stephen, C.D.,
1996. Structural quality of parts processed by fused deposition. Rapid Prototyp-
In this paper, a new approach for formulating the surface rough- ing J. 2 (4), 4–19.
ness of FDM parts was presented. In order to express the surface Pandey, P.M., Reddy, N.V., Dhande, S.G., 2003. Improvement of surface finish by stair-
roughness in more detail, key variables determining the section case machining in fused deposition modeling. J. Mater. Process. Technol. 132,
323–331.
shape of the extruded filament were considered by investigating Paul, B.K., Voorakarnam, V., 2001. Effect of layer thickness and orientation angle on
the actual surface profile of the part. FDM test parts were fabricated surface roughness in laminated object manufacturing. J. Manuf. Process. 3 (2),
to verify the proposed surface roughness expression. Empirical and 94–101.
Reeves, P.E., Cobb, R.C., 1997. Reducing the surface deviation of stereolithography
computed roughness data were acquired by measuring the test
using in-process techniques. Rapid Prototyping J. 3 (1), 20–31.
part and implementing the proposed expression. By comparison Rodriquez, J.F., Thomas, J.P., Renaud, J.E., 2000. Characterization of the mesostructure
between the measured data and computed values, the validity of of fused-deposition acrylonitrile–butadienestyrene materials. Rapid Prototyp-
the proposed expression was proved. Additionally, the effects of ing J. 6 (3), 175–185.
Stratasys Inc., 2007. http://www.stratasys.com/.
surface angle, layer thickness, cross-sectional shape of the fila- Sun, Q., Rizvi, G.M., Bellehumeur, C.T., Gu, P., 2008. Effect of processing conditions
ment, and overlap interval on surface roughness were analyzed and on the bonding quality of FDM polymer filaments. Rapid Prototyping J. 14 (2),
evaluated. From the results, it was demonstrated that an elaborate 72–80.
Wolfram Research Inc., Copyright 1988–2003, MATHEMATICA, Ver. 5.0.
prediction of the surface roughness of FDM parts can be performed