Вы находитесь на странице: 1из 4

Journal of Materials Processing Technology 118 (2001) 385±388

Critical parameters in¯uencing the quality of prototypes


in fused deposition modelling
R. Anithaa,*, S. Arunachalamb, P. Radhakrishnana
a
PSG College of Technology, Coimbatore 641 004, Tamilnadu, India
b
Manufacturing Engineering Subject Group, School of Engineering, Coventry University, Coventry, UK

Abstract

Rapid prototyping (RP) meets the current needs in the industry to shorten design cycles and improve the design quality. Fused deposition
modelling (FDM) is one of the key technologies of RP. Various process parameters used in FDM affect the quality of the prototype. Work
was undertaken to assess the in¯uence of the parameters on the quality characteristics of the prototypes using Taguchi technique. This paper
discusses the results of the study and the conclusions arrived from it. # 2001 Elsevier Science B.V. All rights reserved.

Keywords: Rapid prototyping; FDM; Taguchi

1. Introduction evaluate the major factors affecting the dimensional varia-


bility of the castings produced.
Rapid prototyping (RP) technique is one of the most
promising techniques to reduce product development time
by way of realising the prototype or a prototype component 2. Taguchi technique
that can be directly used in assemblies, product testing, or
tooling for short or medium run production of the several RP In the recent years, Taguchi's techniques have been
techniques widely used today, fused deposition modelling immensely used to optimise both process design and product
(FDM) accounts for a signi®cant percentage of the machines design, based on comprehensive experimental investigation
in use and almost half of the machines which are introduced [4]. The technique has been employed to analyse super-
in the market belong to this category. Since this process abrasive reaming process for comparisons to be made with
could be used for a variety of applications and the cost of the existing hole ®nishing processes. Based on the experi-
prototype is generally high, there is a need for optimising the mental results, optimum process parameters have been
process parameters both from technological and economic reported in literature [5]. The primary advantages of the
point of views. This work attempts to obtain optimum design of experiments using Taguchi's technique include
process conditions using Taguchi techniques. simpli®cation of experimental plan, feasibility of study of
The quality of a prototype is manifested by several interaction effects among the different parameters. The
parameters. For many engineering applications, surface process variables in FDM are road width, build layer thick-
®nish is an important criterion. Several attempts have been ness, and speed of deposition, though there are other factors
made in the past to make a systematic analysis of errors and like temperature, humidity and wire diameter which are,
the quality of the prototypes. Susila et al. [1] have studied the however, kept constant in this study. Anitha et al. [6] studied
problem of optimisation by considering the direction of the applicability of the method to surface grinding problems
orientation of model build up. They found that the choice and found that the results obtained are quite satisfactory.
of correct orientation resulted in minimum build time. Male Therefore, the authors are convinced to adopt this technique
et al. [2] have made a time, cost and accuracy comparison in to identify an optimum process model for FDM.
the case of investment casting tooling produced using
stereolithography technique. The dimensional variability
of LOM models was studied by Wang et al. [3]. They 3. Selection of process parameters
adopted the ®sh bone diagram approach to systematically
The objective of the study is to analyse the effect of
*
Corresponding author. process variables on the surface roughness of the compo-

0924-0136/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 4 - 0 1 3 6 ( 0 1 ) 0 0 9 8 0 - 3
386 R. Anitha et al. / Journal of Materials Processing Technology 118 (2001) 385±388

Table 1 Table 3
List of process variables and their levels Measured Ra values

Factors Level 1 Level 2 Level 3 Trials Ra value

Layer thickness (mm) 0.1778 0.254 0.3556 1 7.34


Road width (mm) 0.537 0.622 0.706 2 8.32
Speed deposition (mm) 100 150 200 3 9.26
4 2.79
5 3.43
6 5.03
nents produced by the FDM process. The surface roughness
7 3.24
of the components produced is measured by the CLA value. 8 5.19
The objective is to minimise the surface roughness (Ra value). 9 2.89
Table 1 shows the variables and levels selected for the study. 10 9.85
The design matrix, an L18 orthogonal array (Table 2), was 11 10.73
12 3.82
chosen to account for the factors and their levels.
13 7.97
14 2.63
15 6.59
4. Results and analysis 16 3.04
17 2.88
18 6.21
The study involved a sample component and 18 models.
The roughness of these models was measured using Surtronic
surface roughness measurement tester and listed in Table 3.
4.2. ANOVA analysis
4.1. Signal to noise (S/N) ratio
ANOVA analysis provides signi®cance rating of the
The signal to noise ratio measures the sensitivity of the various factors analysed in the study. Based on the above
quality characteristic being investigated to those uncon- rating, factors which in¯uence the objective functions sig-
trollable external factors. To minimise the problem, the ni®cantly could be identi®ed and proper control measures
governing relationships for the S/N ratio in terms of the adopted. In a similar way, those factors with minimum
experimentally measured values of Ra, i.e., yi is calculated in¯uence could be suitably modi®ed to suit economic con-
as follows: siderations.
The ANOVA computations were carried out based on
S=N ratio ˆ 10 log 10 MSD
P procedure outlined in Ref. [7] and listed in Tables 8 and 9. A
where MSD ˆ …yi ytar †2=n, ytar the target value that is variable possessing the maximum value of variance is said
to be achieved, n the number of samples. The S/N ratio to have the most signi®cant effect on the process under
values obtained for the trials are listed in Tables 4±7. consideration.

Table 2
Design matrix used in the experiment

Trials Dummy Layer Road width Dummy Layer thickness Dummy Road width Speed
level thickness (mm)  speed (mm2) level  speed (mm2) level (mm) (mm)

1 0 1 1 0 1 0 1 1
2 0 1 2 0 2 0 2 2
3 0 1 3 0 3 0 3 3
4 0 2 1 0 2 0 3 3
5 0 2 2 0 3 0 1 1
6 0 2 3 0 1 0 2 2
7 0 3 1 0 1 0 2 3
8 0 3 2 0 2 0 3 1
9 0 3 3 0 3 0 1 2
10 0 1 1 0 3 0 2 1
11 0 1 2 0 1 0 3 2
12 0 1 3 0 2 0 1 3
13 0 2 1 0 3 0 3 2
14 0 2 2 0 1 0 1 3
15 0 2 3 0 2 0 2 1
16 0 3 1 0 2 0 1 2
17 0 3 2 0 3 0 2 3
18 0 3 3 0 1 0 3 1
R. Anitha et al. / Journal of Materials Processing Technology 118 (2001) 385±388 387

Table 4 When contribution of any factor is small, then the sum


S/N values of experimental study of squares, S, for that factor is combined with the error Se.
Trials Roughness values MSD S/N This process of disregarding the contribution of a selected
factor and subsequently adjusting the contributions of the
1 9.45 89.21 19.5
2 8.32 69.22 18.4 other factors is known as pooling [8]. In this experiment,
3 9.26 85.75 19.33 the contributions of the interactions between the layer
4 2.79 7.78 8.91 thickness and speed of deposition, and road width and
5 3.43 11.76 10.7 speed of deposition were found to be negligible. Hence
6 5.03 25.3 14.03
they are pooled and the contributions of other factors are
7 3.24 10.49 10.2
8 5.19 26.94 14.3 signi®cantly increased.
9 2.89 8.35 9.21
10 9.85 96.82 19.85 4.3. Correlation analysis
11 10.73 115.13 20.61
12 3.82 14.59 11.64
In process control, the aim is to control the characteristics
13 7.97 63.52 18.02
14 2.63 6.92 8.4 of the output of the process by controlling a process para-
15 6.59 43.43 16.37 meter. One succeeds if the parameters are chosen correctly.
16 3.04 9.21 9.64 The choice is usually based on judgement and knowledge of
17 2.88 8.29 9.18 the concerned technology. A correlation is assumed between
18 6.21 38.56 1.58
a variable product characteristic and a variable process
parameter.
In the present study, a relationship is assumed between the
Table 5 layer thickness (process parameter) and surface roughness
S/N analysis of layer thickness (product characteristic). Layer thickness is the property
Factor S/Navg which signi®cantly affects the quality of the prototypes in
RP. This is proved by the contribution at 99% level of
Layer thickness at 0.1778 mm 18.22
Layer thickness at 0.254 mm 12.74 signi®cance.
Layer thickness at 0.3556 mm 9.02 The results obtained are listed in Table 10. The correlation
(r) coef®cient obtained is 0.656. The range of values for r
lies between 1 and 1. The experimental value indicates a
Table 6 reasonably strong negative relation. Therefore, as layer
S/N analysis of road width thickness increases, the surface roughness decreases.
Factor S/Navg
4.4. Regression analysis
Road width at 0.537 mm 12.02
Road width at 0.622 mm 14.67
Road width at 0.706 mm 13.79 The regression analysis is attempted for layer thickness as
it is the dominant factor. The desired surface roughness is to
be as less as possible. The surface roughness value that is
Table 7
required is set at 3.15 Ra. Though surface roughness is to be
S/N analysis of speed of deposition as minimum as possible, the experimental value is set by the
range covered by the experiment. Statistical reasoning is that
Factor S/Navg
linearity of the relationship applies only to the range covered
Speed at 100 mm 13.72 by the data. Additional data would be required to check if the
Speed at 150 mm 14.985 surface roughness improves further in case the process
speed at 200 mm 11.28
parameter values are changed.

Table 8
ANOVA analysis (without pooling)

Factors Degrees of freedom, d Sum of squares, S Variance, V Variance ratio, F Contribution, P (%)

Layer thickness (mm) 2 74.24 37.12 12.73 49.37


Road width (mm) 2 24.34 12.17 4.175 13.36
Speed of deposition (mm) 2 24.72 12.36 4.240 13.63
Layer thickness  speed (mm2) 4 5.58 1.395 0.479 Negligible
Road width  speed (mm2) 4 0.936 0.234 0.080 Negligible
Error 3 8.744 2.915 ± ±
388 R. Anitha et al. / Journal of Materials Processing Technology 118 (2001) 385±388

Table 9
ANOVA analysis (with pooling)

Factors Degrees of freedom, d Sum of squares, S Variance, V Variance ratio, F Contribution, P (%)

Layer thickness (mm) 2 74.24 37.12 26.76 51.57


Road width (mm) 2 24.34 12.17 8.774 15.57
Speed of deposition (mm) 2 24.72 12.36 8.911 15.83
Layer thickness  speed (mm2) 4 5.58 Pooled ± ±
Road width  speed (mm2) 4 0.936 Pooled ± ±
Error 11 15.26 1.397 ± 5

Table 10 The other factors, road width and speed, contribute to 15.57
Results of correlation analysis and 15.83% at 99% level of signi®cance, respectively. The
S. No. Layer Surface x2 y2 xy signi®cance of layer thickness is further strengthened by the
thickness, x (mm) roughness (y) correlation analysis, which indicates a strong inverse rela-
1 0.1778 9.445 0.032 89.21 1.679 tionship with surface roughness.
2 0.1778 8.32 0.032 69.22 1.479 According to the S/N analysis, the layer thickness is most
3 0.1778 9.26 0.032 85.75 1.646 effective when it is at level 3 (0.3556 mm), the road width at
4 0.254 2.79 0.065 7.78 0.708 level 1 (0.537 mm) and the speed of deposition at level 3
5 0.254 3.43 0.065 11.76 0.871
(200 mm). According to the trials, sample 18 was found to
6 0.254 5.03 0.065 25.3 1.277
7 0.0356 3.24 0.126 10.49 1.152 give the best results.
8 0.0356 5.19 0.126 26.94 1.85
9 0.0356 2.89 0.126 8.35 1.028
10 0.1778 9.85 0.032 97.022 1.751 References
11 0.1778 10.73 0.032 115.13 1.908
12 0.1778 3.82 0.032 14.59 0.679
13 0.254 7.97 0.065 63.52 2.02 [1] B. Susila, et al., Studies on the influence of fabrication orientation on
14 0.254 2.63 0.065 6.92 0.668 the productivity of fused deposition modelling, in: Proceedings of 13th
15 0.254 6.59 0.065 43.43 1.674 CAR and FOF Conference, Colombia, Brazil, 1997.
16 0.3556 3.035 0.126 9.21 1.079 [2] J. Male, H. Tsang, G. Bennet, A time, cost and accuracy comparison of
17 0.3556 2.88 0.126 8.29 1.024 soft tooling for investment casting produced using stereolithography
18 0.3556 6.21 0.126 38.56 2.208 techniques, in: Proceedings of the Solid Freedom Symposium, The
University of Texas, Austin, TX, 1996.
[3] W. Wang, J.G. Conley, H.W. Stoll, Dimensional variability analysis in
post processing of rapid tooling, in: Proceedings of the Ninth Solid
Free Form Fabrication Symposium, Austin, TX, 1998.
For a surface roughness value of 3.15 Ra, the desired [4] S. Arunachalam, J. O'Sullivan, Role of design of experiments in
value for layer thickness would be 0.367 Ra. The other business manufacturing management Ð an application example, in:
factors can also be suitably set depending on the desired Proceedings of the International Conference on Business and
roughness value. Management, India, 1998.
[5] S. Arunachalam, et al., Process capability analysis of super abrasive
reaming compared with the existing hole finishing processes, in:
Proceedings of the International Conference on Quality Engineering
5. Conclusions and Management, 1997.
[6] R. Anitha, et al., Taguchi approach to improve surface topography of
The results revealed several interesting features of the how carbon steels finished by surface grinding, in: Proceedings of the
EUFIT'98 Conference, Aachen, Germany, 1998.
FDM processes. It is found that without pooling, only the
[7] W.L. Condra, Value Added Management with Design of Experiments,
layer thickness is effective to 49.37% at 95% level of Chapman & Hall, London, 1995.
signi®cance. But on pooling, it was found that the layer [8] R.K. Roy, A Primer on the Taguchi Method, Van Nostrand, New York,
thickness is effective to 51.57% at 99% level of signi®cance. 1990.

Вам также может понравиться