Research Article

Advances in Mechanical Engineering

2019, Vol. 11(4) 1–12
Ó The Author(s) 2019
Optimization of process parameters to DOI: 10.1177/1687814019841460
journals.sagepub.com/home/ade
improve dimensional accuracy of
investment casting using Taguchi
approach

Sangita N Bansode , Vikas M Phalle and Shankar S Mantha

Abstract
In this study, Taguchi method is used to investigate the effect of critical process parameters of investment casting on
dimensional variations of thin walled, complex geometrical stainless steel component. A set of experiments have been
conducted to calculate shrinkages which is a measure of dimensional variation. The process parameters considered here
are number of ceramic coats, pouring time, pouring temperature, and casting cooling rate. The signal-to-noise ratio and
analysis of variance are used to study the influence of these parameters on shrinkage. Optimized condition for reduced
shrinkage deviation can be obtained by selecting lower number of ceramic coats, faster and laminar metal pouring, lower
superheated metal temperature, and faster cooling rate. Analysis of variance reports that the number of ceramic coats is
the most significant parameter with more than 80% contribution whereas other parameters are insignificant.
Investigation results also indicate that constraint dimension (width) has 25%–27% higher dimensional variability than
non-constraint dimensions (height) for the selected component.

Keywords
Investment casting, distortion, shrinkage, dimensional accuracy, Taguchi method

Date received: 18 November 2018; accepted: 12 March 2019

Handling Editor: Shun-Peng Zhu

Introduction are required to bring the casting dimension in specified

control limits. These additional correction operations
Investment casting (IC) is widely preferred process increase lead time as well as additional tooling and
among all the precision casting process to produce labor cost. The IC process consists of multiple stages
components, having complex geometry, stringent toler- and process variability at each stage contributes to
ances, and excellent surface finish. The IC process casting dimensions. Shrinkage at wax, mold, and cast-
offers extensive choice of cast metal and large range in ing stages plays major role in deciding the final casting
size and shape of the component. In foundry, a thin-
walled hollow rectangular component with varying
thickness, is considered as a complex casting.
Geometrical distortion is a major issue in such complex Department of Mechanical Engineering, Veermata Jijabai Technological
stainless steel casting due to large temperature gradient Institute (VJTI), Mumbai, India
and high melting point. Distortion is nonconformance
Corresponding author:
of measured casting dimensions from the specified Sangita N Bansode, Department of Mechanical Engineering, Veermata
drawing dimension. Major correction operations such Jijabai Technological Institute (VJTI), Matunga, Mumbai 400019, India.
as reworking of wax die and casting coining operations Email: sangeetabansode@somaiya.edu

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering

dimensions. Wax shrinkage is one of the largest shrin- Using Taguchi approach, optimized levels of the
kages in total shrinkage of component. selected parameter have been recommended to improve
Several researchers1–3 have contributed their work in the dimensional accuracy of casting. This reduces the
finding out influence of wax injection process para- requirement of additional correction operations thus
meters on dimensional stability of wax pattern. increasing foundry productivity significantly. The study
Morwood et al.4 reported that usage of filled wax, simultaneously compares the analysis of constrained
lower temperature gradient, and rigid restraint offered and non-constrained dimensions.
by metallic core assures minimum shrinkage deviation
of wax pattern. In mold stage, the base coat of ceramic
mold is formed by zircon slurry and stucco. High melt-
Methodology and experimentation
ing point and very low coefficient of thermal expansion The aim of this experimental research is to investigate
of zircon provides accuracy and consistency of mold the effect and identify the optimized condition for
dimensions as reported by Bates et al.5 Mold thickness selected casting process parameters to improve dimen-
is controlled by applied number of ceramic coats on sional accuracy of the IC component using Taguchi
wax pattern. Singh et al.6 and Kumar et al.7 considered technique. The overall methodology for research work
this factor along with other while studying rapid IC is illustrated in Figure 1 and described below.
and hybrid IC.
Casting stage is having the maximum contribution
and variability in total shrinkage which affects casting Details of the selected component
dimensions significantly.8–10 Okhuysen et al.11 reported In this research work, a thin-walled hollow rectangular
from survey-based study that dimensional discrepancies component with varying thickness has been selected as
will potentially result when same set of tooling allowan- shown in Figure 2(a). This component has applications
ces has been used for different cast alloys to produce in various industries. The weight of casting is approxi-
components. Harste and Schwerdtfeger12 measured mately 1.05 kg and the material used is stainless steel
shrinkage of steel during solidification and subsequent (CF8). The two critical dimensions of the component
cooling and confirmed the carbon content affects the are inside width (66 6 0.45 mm) and height
shrinkage of steel. Effect of pouring temperature on (96 6 0.55 mm) of a rectangular cavity. The thickness
dimensional accuracy has been studied by Farhangi of part (along Z direction) is 42 mm as shown in section
et al.13 and Seidu and Onigbajumo14 reported that B-B. The component has three isolated thick sections
increase in temperature causes increase in thermal P, Q, and R in which hot spots get developed during
strain. The selection of pouring technique affects the solidification of casting. To remove these hot spots,
metal temperature at pouring, viscosity, temperature three independent feeds are required to be attached to
distribution, and thermal stress in casting.15 Singh the component as shown in Figure 2(b). The feed pads
et al.16 examined the effect of metal composition, slurry at P and Q location influences and restricts the normal
layer combination, and pouring temperature on dimen- and free solidification of the ‘‘width’’ hence considered
sional deviation of casting and informed that all the as constrained dimensions while as ‘‘height’’ dimensions
factors are significant. Kang et al.17 attempted intensive are free to shrink referred as unconstrained dimensions.
cooling of riser by water spray which results in uniform Measurements were taken on both these dimensions
temperature distribution and reduced residual deforma- (width and height) throughout the experiment.
tion. Sabau18 reported that in IC, mold configuration
restricts air flow surrounding casting surfaces and
affects the uniform cooling of mold. Distortions create
Selection of process parameters
inefficiencies throughout the casting process that can IC process consists of multiple stages and each stage is
adversely impact lead times and generate considerable contributing to dimensional changes of component.
waste.19 Control on casting dimensions improves foun- Various process parameters at major stages of invest-
dry productivity significantly.20 This research work is ment casting which affect the dimensional accuracy
carried out by conducting a set of experimentation to have been listed in Ishikawa (cause and effect) diagram
study the effect of various process parameters at as shown in Figure 3. These parameters can be grouped
selected levels on dimensional accuracy of casting. To into four different stages wax, mold, casting, and post
simplify and reduce the experimental work, several casting stages. At wax stage, composition of wax and
researchers6,16,21,22 have applied Taguchi technique. wax injection process parameters such as pressure, tem-
This research work also uses Taguchi approach to peratures, and holding time are influential. Use of filled
determine the contribution of parameters in dimen- wax and wax chill prevents excessive shrinkage at thick
sional variability. Shrinkage of the component has been sections of the casting. Insertion of metallic spacer in
compared with shrinkage of standard test piece from ejected wax pattern helps to reduce deformation. Use of
the same mould to find out dimensional deviation. highly automated wax injection machine which
Bansode et al. 3

Figure 1. Steps of Taguchi technique.

Figure 2. (a) Schematic diagram of component and (b) feeders of component.

provides precise control and monitoring of process temperature number of coats, stucco size affects the
parameters and results in consistent and accurate wax mold dimensions. Number of ceramic coats signifi-
patterns. At mold stage, the mold composition, mold cantly affects the thickness of mold and thus strength of
4 Advances in Mechanical Engineering

Figure 3. Ishikawa cause and effect diagram for the dimensional accuracy of IC process.

Table 1. Selected parameters and their levels.

Notation Parameters Level 1 Level 2

A Number of ceramic coats 7 (low) 9 (high)

B Pouring technique Ladle (uncontrolled) Furnace (controlled)
C Casting cooling method Open-air (fast) cooling Covered (slow)
D Pouring temperature (°C) 1590 1640

mold thus selects as one of the parameters for Design of pouring or by furnace pouring. In the ladle
experiment (DoE). Major dimensional change occurs pouring, metal has been transferred from fur-
during solidification and cooling of casting at casting nace to ladle and then ladle to the hot mold by
stage. Shrinkage occurs in three phases: liquid shrink- tilting ladle manually which takes approxi-
age, solidification shrinkage, and solid shrinkage. Thus, mately 10–12 s. In the furnace pouring, mold
majority of the parameter has been taken from this capacity and furnace melting capacity are equal
stage except cast alloy composition as it is controlled by and furnace is having tilting arrangement. Hot
customer as per requirement of part. Thus, significant mold has been clamped on furnace top and
and controllable input parameters such as pouring tem- complete charge (molten metal) is poured by
perature, pouring rate, and casting cooling rate along controlled rotation of the furnace. This results
with number of ceramic coats are the four parameters laminar and steady flow with pouring time of
selected for experimentation to study their impact on 5–6 s. The pouring technique affects the mold
dimensional accuracy of casting. Two levels of each filling time, viscosity, temperature distribution,
parameter have been selected for the present work and and thermal stress in casting. Thus, selecting the
are given in Table 1 followed by brief description. ladle and furnace pouring as two levels.
3. Pouring temperature: In IC, the intricate shape
1. Number of ceramic coats: Thickness of mold is component has been poured with superheated
controlled by applied number of ceramic coats metal temperature which is generally kept above
on wax pattern which depends on size and the melting temperature of the alloy to avoid
weight of the casting. Use of thin ceramic shell the ‘‘not fill defect’’ in the casting. Metal pour-
utilizes material efficiently and reduces process ing operation needs high coordination between
time but may bulge or leak during pouring and the pouring operators. Time delay in pouring
solidification. On the other hand, thick mold drops the melt temperature significantly which
supports metallostatic pressure of hot molten further affects casting quality. The standard
metal during the casting process and avoids temperature set by method study technique is
probability of leakage and bulging. Nine coats 1590°C and by varying superheat degree to
which are standard are considered for level 2 1640°C, other level has been selected for cast
and seven coats for level 1. metal (stainless steel alloy).
2. Pouring technique: The molten metal has been 4. Casting cooling method: To avoid surface dec-
poured in the hot ceramic mold either by ladle arburizing of stainless steel casting, the top of
Bansode et al. 5

the mold cup is sprinkled with exothermic flux Table 2. Orthogonal array L8(27).
immediately after pouring. This flux burns and
generates large amount of heat and holds the Factor trial A B C D E F G
mold top cup and hence the molten metal hot 1 1 1 1 1 1 1 1
for longer duration. This happens in covered 2 1 1 1 2 2 2 2
environment which restrict the air circulation 3 1 2 2 1 1 2 2
and decrease the metal cooling rate to 10°C/ 4 1 2 2 2 2 1 1
min. The alternate method is allowing the mold 5 2 1 2 1 2 1 2
6 2 1 2 2 1 2 1
to cool in open air without cover which 7 2 2 1 1 2 2 1
increases the cooling rate to 20°C/min. To 8 2 2 1 2 1 1 2
understand the effect of variation in cooling rate
on casting dimensions, these two cooling meth-
ods (open and covered) have been selected for (55°C–65°C), and holding time (60 s). Critical
study. dimensions: width and height were measured on
wax patterns at AA#, BB#, CC# and at DD#,
EE#, FF#, respectively, as shown in Figure 1.
Orthogonal array Eight wax assembly trees have been formed by
In today’s competitive market, manufacturing indus- attaching eight wax patterns, four on each side
tries have to produce quality products at reduced cost of vertical wax riser plate which represents eight
with smaller lead time. Taguchi’s orthogonal array is casting trials as per L8 array. All wax patterns
the most efficient tool which helps in reducing total and tree assemblies have been numbered for
number of experiments. In this work, the four process identification as per designed trial.
parameters namely number of ceramic coats, pouring  The wax assemblies have been coated first with
time, pouring temperature, and casting cooling rate primary coat of zircon slurry at 120–140 s viscos-
each at two levels have been considered. Selection of ity and the secondary coat of aluminum silicate
appropriate orthogonal array for selected parameters is having slurry viscosity as 50–60 s. Slurry viscos-
the important step in Taguchi technique. Total degrees ity has been measured using ford cup of size B.
of freedom for each parameter when varied at two lev- Each coat was followed by stuccoing. Number of
els is 4. A two-level orthogonal array (L827) with eight secondary coats has been varied as per L8 array.
experimental runs (degrees of freedom = 8–1 = 7) is These molds were then allowed to dry. The
selected for this research. Orthogonal array is the smal- whole process has been carried out in humidity
lest possible design of combinations in which all the (45%–60%) and temperature (25°C) controlled
parameters are varied simultaneously and their effect environment and completed in 3–4 days. The
and performance interaction can be examined concur- dried mold had been taken to auto clave to
rently. The array name (L827) indicates that it can use 7 remove wax from mold using steam of tempera-
factors at two levels. The number 8 represents total ture 150°C at pressure of 10 bar. Mold was then
number of experiments to be conducted as shown in sintered in oil fired furnace at temperature
Table 2. In present case, each of the four factors has 1050°C–1100°C for 2 h which has been taken out
been varied at two levels and no consideration has given of furnace just before pouring. The interval time
to interaction effect of the parameters. Thus, four fac- between removal of mold from furnace and
tors have been assigned sequentially to first four col- actual metal pouring is allowed up to 40 s. If
umns of L8 orthogonal array while as remaining three delay is beyond 40 s, the mold is not poured and
kept unassigned as explained by Roy.23 left unfilled as ceramic shell is thin wall construc-
tion which cools rapidly. Delay in pouring
reduces the temperature of preheated mold and
IC procedure causes early solidification of metal before reach-
As per the experimental layout shown in Table 3, cast- ing to the intricate corners of each casting. This
ing trials were conducted. Filled wax is used for wax will result to cold shut and no fill defects in
pattern. As per conventional IC process, ceramic mold casting.
is formed using zircon slurry and flour for primary coat  Cast metal of low carbon, high alloy, austenitic
and aluminum silicate for other coats for stainless steel stainless steel of required composition has been
casting. Conventional IC process has been described as melted to pouring temperature and poured in
follows: preheated molds. Molds were filled by hot mol-
ten metal having different pouring temperatures
 Wax patterns have been injected using set wax by changing superheat degree in corresponding
injection pressure (30–40 kg/cm2), temperature
6 Advances in Mechanical Engineering

Table 3. Expt. layout after assigning the parameter values in orthogonal array.

Experiment No. of ceramic Pouring Casting cooling Pouring temperature,

no. coats (A) technique (B) method (C) °C (D)

1 7 Ladle Open 1590

2 7 Ladle Open 1640
3 7 Furnace Covered 1590
4 7 Furnace Covered 1640
5 9 Ladle Covered 1590
6 9 Ladle Covered 1640
7 9 Furnace Open 1590
8 9 Furnace Open 1640

Table 4. Percentage shrinkage deviation for width dimensions.

Trial A B C D R1 R2 R3 R = (Yi ) S/N ratio

1 1 1 1 1 9.46 8.08 8.92 8.82 –18.93

2 1 1 1 2 6.46 6.54 7.28 6.76 –16.61
3 1 2 2 1 2.64 2.45 3.43 2.84 –9.16
4 1 2 2 2 11.66 11.87 13.07 12.2 –21.74
5 2 1 2 1 26.88 25.88 27.31 26.69 –28.53
6 2 1 2 2 32.05 31.77 30.77 31.53 –29.98
7 2 2 1 1 24.94 25.39 26.11 25.48 –28.13
8 2 2 1 2 21.68 25.23 24.43 23.78 –27.54
T 17.26

S/N: signal-to-noise.

Table 5. Percentage shrinkage deviation for height dimensions.

Trial A B C D R1 R2 R3 R = (Yi ) S/N ratio

1 1 1 1 1 6.15 7.61 8.23 7.33 –17.36

2 1 1 1 2 6.63 6.73 7.1 6.82 –16.68
3 1 2 2 1 3.07 3.93 4.25 3.75 –11.56
4 1 2 2 2 6.22 6.12 6.86 6.4 –16.14
5 2 1 2 1 22.65 21.45 22.77 22.29 –26.97
6 2 1 2 2 23.32 22.59 24.32 23.41 –27.39
7 2 2 1 1 12.67 14.37 15.89 14.31 –23.15
8 2 2 1 2 13.55 16.7 16.94 15.73 –23.98
T 12.51

S/N: signal-to-noise.

mold using two different techniques and then and l2 is corresponding linear dimensions at
allowed to cool in open air or in covered condi- cast stage.
tion as per L8 array design.  Shrinkage deviation have been calculated by
 Critical dimensions of width and height were comparing shrinkage of the component with
again measured on 64 castings each dimension standard shrinkage of the metal at three loca-
at three specified locations as shown in Figure tions. These shrinkage deviations R1,R2,R3 rep-
1 from which average width and height of the resent the measure of dimensional accuracy and
component has been calculated. Linear shrink- considered as the responses for L8 orthogonal
age for both the dimensions of the selected array (Table 2). These values along with average
component has been calculated as per equation shrinkage deviation are shown as response val-
(1). Where l1 is linear dimensions at wax stage ues in Tables 4 and 5
Bansode et al. 7

  " #
ðl1  l2 Þ S 1X r
Scomponent = 100 ð1Þ =  10 log10 ðMSDÞ =  10 log10 Yi 2
ð3Þ
l2 NLB r i=0
 
Scomponent  Sstandard
Sdeviation = 100 ð2Þ where Yi is the observed value of the response charac-
Sstandard
teristic for the corresponding trial, that is, shrinkage
deviation, and r is the number of repetitions. R1, R2,
and R3 for width and height represents the shrinkage
Results and discussion calculation done at three locations AA#, BB#, CC#, and
After conduction of experiment and measurement of DD#, EE#, FF#. R = (Yi ) represents the average. Using
shrinkage at three locations, the collected data have equation (3), S/N ratio for both the dimensions at each
been analyzed by means of calculating signal-to-noise experimental run is calculated and the values are listed
(S/N) ratio. The means analysis and S/N ratio analysis in Tables 4 and 5. These values are used as input to sta-
plot has been plotted using Minitab software to deter- tistical software Minitab 17 to find out the main effect
mine optimized condition of process parameter. The of parameters on shrinkage deviation.
analysis of variance (ANOVA) has been calculated
using statistical formulas to find out the contribution
of each parameter. The range within which the true
Main effects of selected parameters
value of estimated mean will lie has been calculated by Specially designed L8 orthogonal array facilitates to
finding confidence interval at 95% significance level. find out the main (average) effect of factors by combin-
The detail analysis of ceramic coats which is the most ing the similar levels of factors from each trial. The
significant factor has been done by varying numbers of main effect plot of means and S/N ratio for selected
coats in small steps to determine the effect on shrinkage casting process parameters: number of ceramic coats,
of critical dimensions of component. pouring technique, casting cooling method and pouring
temperature is shown in Figures 4(a), 4(b) for width
and 5(a), 5(b) for height dimensions. The plots for both
S/N ratio analysis the critical dimensions indicate that as the number of
For experimentation with repetitions, use of S/N ratio ceramic coats (A) increases from seven (7) to nine (9),
permits to determine the variance index. The greater shrinkage deviation increases significantly.
the value of this index, the smaller the product variance. Ceramic mold undergoes heating and cooling stages
The objective of the study is to reduce the shrinkage such as dewaxing, preheating, and pouring which may
variation of component from standard value. Thus, the affect the casting dimensions. Ceramic material used
S/N ratios for this work have been calculated on the for mould is thermally stable. Thus mould shrinkage/
basis of ‘‘lower is better’’ approach which is logarithmic expansion is considered as negligible. The shell is ther-
function based on mean square deviation (MSD) and mally stable in autoclave as temperature is 150°C, but
given by the following equation it may crack due to thermal expansion of wax in the
mold, if autoclave furnace takes longer time to build

Main Effects Plot for Means (Width) Main Effects Plot for SN ratios (Width)
Data Means Data Means
No. 0f ceramic coats Pouring technique Casting cooling method Pouring temperature No. 0f ceramic coats Pouring technique Casting cooling method Pouring temperature
-1 5.0

25
-1 7.5
Mean of SN ratios
Mean of Means

-20.0
20

-22.5

15
-25.0

10 -27.5

-30.0
5 1 2 1 2 1 2 1 2
1 2 1 2 1 2 1 2
Signal-to-noise: Smaller is better

(a) (b)

Figure 4. (a) Main effects plot for means of width shrinkage deviation. (b) Main effects plot for S/N ratio of width shrinkage
deviation.
8 Advances in Mechanical Engineering

Figure 5. (a) Main effects plot for means of height shrinkage deviation. (b) Main effects plot for S/N ratio of height shrinkage
deviation.

Table 6. Response table: The average of two levels for each parameter.

Level 
A B 
C 
D

Width average

1 7.65 18.450 16.210 15.958

2 26.87 16.075 18.315 18.567
Height average
1 6.075 14.963 11.047 11.920
2 18.935 10.047 13.963 13.090

the steam pressure. Thermal coefficient of expansion of parameter levels with lower shrinkage deviation have
ceramic is negligible as compared to the metal; hence, been identified which represent optimised condition of
the shell will constrain the free shrinkage of cast metal. parameters. However, higher levels from S/N ratio plot
As the shell thickness increases, the constraint to free also represent the same optimised condition for selected
shrinkage of metal increases, thus increasing the shrink- parameters. These optimum levels indicate that the
age variation. lowest shrinkage deviation (maximum dimensional
These results are in line with Singh et al.6 The fur- accuracy) can be achieved by selecting seven number of
nace pouring technique (B) provides minimum turbu- ceramic coats (A  1 ), furnace pouring method (B 2 ), open-
lence of metal flow as compared to ladle pouring 
air cooling of casting (C1 ), and lower pouring tempera-
technique thus resulting in reduced shrinkage deviation.  1 ). Predicted response at the identified optimum
ture (D
The cooling method of casting (C) under covered sheet conditions of selected parameters has been calculated
reduces the cooling rate, thus having high shrinkage using equation (4) and shown in Table 7
variability as compared to open-air cooling. Higher
pouring temperature results large temperature gradient  1  T Þ + ðB
Ropt = T + ðA  1  T Þ + ðD
 2  T Þ + ðC  1  T Þ
which causes thermal strain and leads to more shrink- ð4Þ
age deviation. This observation on temperature para- P
meter is in line with Farhangi et al.13 where T = Ri =n; grand average performance =
 1B
17.26 (width) and 12.51 (height) and A 1D
 2C  1 values
taken from response Table 6 for width as well as height
Estimation of optimized condition
In this experimental work, quality characteristic chosen
for response (output) is ‘‘shrinkage deviation’’ and for Analysis of variance
improved dimensional accuracy, consideration is given ANOVA is carried out on the experimental results to
as ‘‘smaller is the better.’’ From Figures 4(a) and 5(a), distinguish between the significant and insignificant
Bansode et al. 9

Table 7. Estimate of performance at the optimum condition of parameter for width and height dimensions.

Notation Factor description Level description Level Contribution

Width Height

A No. of ceramic coats 7 1 –9.605 –6.44

B Pouring technique Furnace pouring 2 –1.19 –2.46
C Casting cooling method Open cooling 1 –1.05 –1.46
D Pouring temperature (°C) 1590 1 –1.3 –0.59
Contribution from all significant factors (total) –13.15 –10.95
Current grand average performance (T)  17.26 12.51
Expected result at optimum condition 4.11 1.56

Table 8. Analysis of variance for width.

Source DOF Sum of squares (S) Variance (V) Variance ratio (F) Pure sum of squares (S#) Percentage contribution

A 1 2215.3 2215.3 282.38 2207.452 89.53

B 1 33.843 33.843 4.31 25.981 1.05
C 1 26.59 26.59 3.39 18.74 0.76
D 1 40.87 40.87 5.210 33.03 1.33
Error (e) 19 149.068 7.845 1 – 7.33
Total 23 2465.66 – – – 100

DOF: degree of freedom.

parameters and to determine the relative percentage SA0 = Pure sum of squares = SA  ðVe XfA Þ ð11Þ
contribution of each parameter which influences varia-
S0A
bility of results. Study of the derived ANOVA table PA = 3 100 ð12Þ
aids to determine the parameters to be controlled for ST
reduction in variability. Derived table has important Percentage contribution (%) of each factor in result
terms such as degree of freedom, sum of squares, per- variation (shrinkage deviation) is derived from
centage contribution, F value, and P value and is ANOVA analysis of experimental data using equations
obtained using following formulas (5)–(12) and as shown in Tables 8 and 9.
ANOVA works on the variation from average val-
ST = Sum of squares of all experimental resultsC:F:
ues to predict the percentage contribution of the input
ð5Þ parameters. The information about the contribution of
where the sum of squares provides a measure of the parameter helps to differentiate between the significant
total variations present and not significant parameters which controls the
response function. Percentage contribution of each pro-
X
n cess parameter defines the power of the parameter to
ST = ðRi Þ2  C:F: ð6Þ control the results. Observation of both Tables 8 and 9
i=1 indicates that the number of ceramic coats (A) is the
P most influencing factor as having the highest percent-
where C.F. = T2/n; T is grand total of results = Ri ;
n is the number of experimental trials age contribution as 89.53% for width and 81.06% for
height shrinkage deviation. The percentage contribu-
  tion of other factors such as pouring technique (1.05,
A1  A2 2
Contribution of each factor = SA = ð7Þ 11.75), casting cooling method (0.76, 4.06) as well as
4
pouring temperature (1.33, 0.56) is less significant in
SA Se width as well as height shrinkage deviation. Percentage
VA = Varaince = , Ve = error varaince = ,
fA fe ð8Þ contribution of all selected parameters is different for
f = degree of freedom width and height. These differences are observed as
width is constrained dimensions whereas height is non-
Se = error contribution = ST  ðSA + SB + SC + SD Þ constrained dimensions due to attached feeder on
ð9Þ casting.
VA Error contribution for both the critical dimensions is
FA = varaince ratio = ð10Þ 7.33% and 2.57% for width and height respectively.
Ve
These are less than 10% which indicate that all the
10 Advances in Mechanical Engineering

Table 9. Analysis of variance for height.

Source DOF Sum of squares (S) Variance (V) Variance ratio (F) Pure sum of squares (S#) Percentage contribution

A 1 992.28 992.28 724.29 990 81.06

B 1 144.94 144.94 105.80 143.57 11.75
C 1 50.98 50.98 37.18 49.61 4.06
D 1 8.21 8.21 5.99 6.844 0.56
Error (e) 19 26.00 1.37 1 2.57
Total 23 1222.426 – – – 100

DOF: degree of freedom.

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
significant parameters have been considered in this 8:185X 1:368
study and no factor has been left out. It also represents C:I:ðmÞ = EðmÞ 6
4:8
that there exists negligible variation due to, uncontrol- = 6 3:65ðWÞ and 1:52ðHÞ
lable factor, interaction of factors as well as variation
in condition. The column F value indicates significance
of the corresponding parameter by comparing with tab-
Confirmation trial
ular F values. The smaller experimental F value com-
pared to table values decides the factor to be pulled to The experimental trial for predicted optimized condi-
get realistic effect of significant factors. Larger F value tion of the parameter is essential if the optimized condi-
of the parameter represents that the variation in this tion is not from the experimental design suggested by
parameter results large variation on experimental Taguchi technique. In this case, it is 1-2-1-1, that is,
results. 1B
A  1D
 2C  1 , which is corresponding to highest S/N ratio
and lowest mean value from Figures 4 and 5. 1-2-1-1
level combination is not one of the experiments
Confidence interval designed for L8 array (refer Table 2). To validate the
optimized condition, confirmatory experiment was con-
The true value of estimated mean will lie within a range,
ducted using set of optimum value of parameters. Two
which has been calculated by finding confidence inter-
ceramic molds were formed by applying lower (7) num-
val (C.I.) at specific significance level. If the estimate of
ber of coats on wax pattern. Dried mold then preheated
the mean value of a set of observations is donated by
in furnace. Molten metal is heated in furnace at tem-
E(m), then C.I. is determined using following equations
perature of 1590°C and poured in preheated mold by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi furnace tilting method. The poured mold is then
F} ðf1 , f2 ÞXVe allowed to cool in open air by keeping it on sand bed
C:I:ðmÞ = EðmÞ 6 ð13Þ
ne without covering by sheet metal box. The dimensions
have been measured on these castings and compare
where F} ðf1 , f2 Þ = Varaince ratio for DOF f1 and f2 at the with wax dimensions to calculate shrinkage for the
level of significance }. optimized condition. The average value of percentage
Confidence level is (1  }) and E(m) = mean shrinkage deviation at optimal setting of parameter is
value of a set of observation 4.63 for width and 2.07 for height, and they are found
to be falling with in the predicted range of 0.46%–
f1 = DOF of meanðwhich is always equals 1Þ 7.76% for width and 0.04%–3.08% for height.
and f2 = DOF of error term

ðTotal number of results ½or number of S=N ratiosÞ

ne = Number of effective replications = ð14Þ
ðDOF of mean + DOF of all factors used in the estimate of the meanÞ

24
ne = = 4:8, EðmÞ = 4:11ðWÞ and
1+4
Shrinkage analysis of most contributing factor
1:56ðHÞ, Ve = 7:845ðWÞ and 1:37ðHÞ
The number of ceramic coats is the most significant fac-
for } = 0:01 that is for 99% confidence level, tor contributing in dimensional accuracy of the width
F0:01 ð1, 19Þ = 8:185 as well as height dimensions. The impact of this factor
has been studied by carrying out the few more trials at
Bansode et al. 11

Width Height Mould thickness

16
3.2

Mould thickness in mm
14
3 12

% Shrinkage
2.8 10

2.6 8
6
2.4
4
2.2 2
2 0
4 5 6 7 8 9 10
Number of ceramic coats

Figure 6. Influence of number of ceramic coats on percentage shrinkage and mold thickness.

different coat levels as discussed below. In this work, the critical dimensions, width and height.
seven ceramic molds were coated with different num- Attached feeders on width constrain the free
bers of coats (4 to 10) for the selected component along shrinkage of casting; thus, width results less
with test piece. Figure 6 shows graph of number of shrinkage compared to height, thus more varia-
coats versus mold thickness and percentage shrinkage. bility in width dimensions.
This indicates that as number of coat increases, mold
thickness increases while as percentage shrinkage of test
Declaration of conflicting Interests
piece, width, and height of component decreases but
gets stabilize after 10 coats. Increase in mold thickness The author(s) declared no potential conflicts of interest with
constraints the free shrinkage of alloy during solidifica- respect to the research, authorship, and/or publication of this
tion and cooling as compared to thin mold. The height article.
dimensions of the component follow the shrinkage path
of standard test piece as is free from feeder constraint Funding
whereas width dimension has lower shrinkage value The author(s) disclosed receipt of the following financial sup-
due to constraint offered by feeder. port for the research, authorship, and/or publication of this
article: The project has been funded under Technical
Education Quality Improvement Program (TEQIP II) of
Conclusion Government of India, World Bank Assisted Project for
Research and Development under Veermata Jijabai
In this study, the impact of selected casting process Technological Institute (VJTI), Matunga, Mumbai and
parameters on dimensional accuracy of complex stain- experimentation support from Uni Deritend Ltd, Nasik,
less steel casting has been studied using Taguchi India.
method.
ORCID iD
1. The optimum condition of casting process para-
Sangita N Bansode https://orcid.org/0000-0001-9690-4685
meters results in high dimensional accuracy by
reducing average shrinkage deviation of critical
dimensions more than 50%. The selection of References
lower number of coats for ceramic mold, con- 1. Wang D, He B, Li F, et al. Cavity pressure and
trolled metal pouring technique with sufficient dimensional accuracy analysis of wax patterns for
superheat temperature and open-air cooling investment casting. Mater Manuf Process 2013; 28:
reduces shrinkage variation in casting. 637–642.
2. The results report that the number of ceramic 2. Yarlagadda PK and Hock TS. Statistical analysis on
coats is the major factor toward achieving accuracy of wax patterns used in investment casting pro-
reduced shrinkage deviation as having contribu- cess. J Mater Process Tech 2003; 138: 75–81.
tion above 80% for both the critical dimensions. 3. Rahmati S, Akbari J and Barati E. Dimensional accu-
Other process parameters contribution observed racy analysis of wax patterns created by RTV silicone
rubber molding using the Taguchi approach. Rapid Pro-
as negligible.
totyping J 2007; 13: 115–122.
3. Increase in number of coats on ceramic mold
results the steep decrease in shrinkage of both
12 Advances in Mechanical Engineering

4. Morwood G, Christodoulou P, Lanham B, et al. Con- base superalloys. J Mater Process Tech 2004; 153:
traction of investment cast H13 tool steel real time mea- 209–212.
surement. Int J Cast Metal Res 2000; 12: 457–467. 14. Seidu SO and Onigbajumo A. Effect of casting wall thick-
5. Bates CE, Griffin J and Jennings J. CAST IRON, Div. ness and pouring temperature on residual stress build up
5-Casting Technologies to Improve the Dimensional in aluminium 6063 casting. Leonardo J Sci 2016; 29:
Accuracy of Thin-Walled Iron Castings. Transactions of 148–160.
the American Foundry Society 2001; 109: 1163–1172. 15. Campbell J. Stop pouring, start casting. Int J Metalcast
6. Singh J, Singh R and Singh H. Dimensional accuracy 2012; 6: 7–18.
and surface finish of biomedical implant fabricated as 16. Singh R, Singh R, Dureja JS, et al. Investigations for
rapid IC for small to medium quantity production. J dimensional accuracy of Al alloy/Al-MMC developed by
Manuf Process 2017; 25: 201–211. combining stir casting and ABS replica based investment
7. Kumar P, Singh R and Ahuja IP. Investigations on casting. Compos Part B: Eng 2017; 115: 203–208.
dimensional accuracy of the components prepared by 17. Kang J, Hao X, Nie G, et al. Intensive riser cooling of
hybrid investment casting. J Manuf Process 2015; 20: castings after solidification. J Mater Process Tech 2015;
525–533. 215: 278–286.
8. Sabau AS and Porter WD. Alloy shrinkage factors for 18. Sabau AS. Numerical simulation of the IC process. Tran
the investment casting of 17-4PH stainless steel parts. Amer F 2005; 51: 273–276.
Metall Mater Trans B 2008; 39: 317–330. 19. Galles D and Beckermann C. Prediction of distortions
9. Horacek M. Accuracy of castings manufactured by the and pattern allowances during sand casting of a steel
lost wax process. Foundry Trade J 1997; 171: 423–424. bracket. International Journal of Cast Metals Research
10. Rosenthal HH. Shrink allowance for pattern dies in 2017; 30(3): 133–47.
investment casting. In: Proceedings of the 27th annual 20. Totten G, Howes M and Inoue T. Handbook of residual
meeting of the Investment Casting Institute, Rosemont, stress and deformation of steel. Materials Park, OH: ASM
IL, 1979, pp.1–11. International, 2002.
11. Okhuysen VF, Padmanabhan K and Voigt RC. Tooling 21. Cxolak M and Dısxpınar D. Taguchi approach for optimi-
allowance practices in the investment casting industry. In: zation of parameters that effect grain size of cast A357
Investment Casting Institute: 46th annual technical meet- alloy. Arch Foundry Eng 2017; 17: 35–42.
ing, Orlando, FL, 27–30 September 1998, pp.1–17. Mon- 22. Santhosh AJ and Lakshmanan AR. Investigation of duc-
tvale, NJ: Investment Casting Institute. tile iron casting process parameters using Taguchi
12. Harste K and Schwerdtfeger K. Shrinkage of round iron– approach and response surface methodology. China
carbon ingots during solidification and subsequent cool- Foundry 2016; 13: 352–360.
ing. ISIJ Int 2003; 43: 1011–1020. 23. Roy RK. A primer on the Taguchi method. Dearborn,
13. Farhangi H, Norouzi S and Nili-Ahmadabadi M. Effects MI: Society of Manufacturing Engineers, 2010.
of casting process variables on the residual stress in Ni-