Int. J. Quality and Innovation, Vol. 1, No.

3, 2011 259

Defect control analysis for improving quality and

productivity: an innovative Six Sigma case study

M. Shanmugaraja* and M. Nataraj

Department of Mechanical Engineering,
Government College of Technology,
Thadagam Road, Coimbatore 641013, Tamilnadu, India
E-mail: raja8011@yahoo.co.in
E-mail: m_natanuragct@yahoo.com
*Corresponding author

N. Gunasekaran
Angel Collge of Engineering and Technology,
Angel Nagar, Dharapuram Main,
Road, Tirupur 641 665, Tamilnadu, India
E-mail: guna_kct_cbe_tn_in@yahoo.com

Abstract: This research study proposes an innovative analysis for controlling

the defects in aluminium die casting industry. In this analysis, casting process
of a two-stroke engine oil pump body is concentrated. The component selected
has often rejected due to blow hole defects. Six Sigma, the zero defect
approach, is used in this study. Define measure, analyse, improve and control
(DMAIC) problem solving methodology is applied for problem analysis.
Taguchis experimental design is used for process validation and improvement.
The confirmation experiment showed that the rejection rate was reduced
to 4.8% from 17.22%. The application of Six Sigma program with Taguchi
technique has developed an innovative cost effective methodology for
controlling defects in die casting process in less experimental time.

Keywords: defects; aluminium die casting; blow holes; productivity; quality;

Six Sigma; define measure, analyse, improve and control; DMAIC; Taguchi
experimental design; analysis of variance; ANOVA; signal to noise ratio; S/N.

Reference to this paper should be made as follows: Shanmugaraja, M.,

Nataraj, M. and Gunasekaran, N. (2011) Defect control analysis for improving
quality and productivity: an innovative Six Sigma case study, Int. J. Quality
and Innovation, Vol. 1, No. 3, pp.259282.

Biographical notes: Muthuswamy Shanmugaraja obtained his BE in

Mechanical Engineering and ME in Industrial Engineering. He is doing
his research in quality management. He has been a Consultant Engineer
in automotive industry. His research interests include Six Sigma quality
control and implementation, quality function deployment, and combined
methodologies for quality assurance in any industries.

M. Nataraj obtained his BE (Mech.) in 1984 from Madurai Kamaraj University,

and ME (Engineering Design) in 1991 and PhD in Product and Process Design
Optimisation in 2006 from the Bharathiar University, India. He is currently a
Faculty in the Department of Mechanical Engineering at Government College

Copyright 2011 Inderscience Enterprises Ltd.

260 M. Shanmugaraja et al.

of Technology, Coimbatore, India. He is a Life Member of the Indian Society

of Mechanical Engineers (ISME). He has published five papers in international
journals and published three papers in Indian journals and presented 30 papers
at national conferences/international conference. He is awarded certificate of
merit in recognition of his research work by Indian Institute of Technology
Roorkee, India. His area of research is product and process design optimisation,
product development using design of experiments approach. He is supervising
eight research scholars in mechanical engineering area.

Nallasamy Gunasekaran obtained his BE and ME in Mechanical Engineering

Discipline and awarded Doctoral in year 2006. He has more than 20 years of
teaching experience and published a number of papers in various international
journals during the term. Presently, he is the Principal of Angel College of
Engineering and Technology, Tirupur, India.

1 Introduction

Quality and productivity are an integral component of organisations operational

strategies (Juran, 1988). In the globalisation of markets and operations, focus on quality
and productivity is of utmost importance (Feigenbaum, 1991; Elshennawy et al., 1991).
Quality improvement in operations and production has been one of the most significant
influences for organisation to be successful (Pande et al., 2000). Corporate consistently
strives to build quality into their products based on customer needs (Juran, 1988). For
manufacturing quality products, continuous improvement (CI) methodologies have been
developed to get better productivity of the operations (Hobbs, 2004; Nave, 2002). During
the past two decades, the quality progress has provided a broad collection of CI methods
to accelerate the process of improving quality and productivity that supports the business
growth (Cox et al., 2003).
Six Sigma is one of the recent CI approaches which are applied in the best-in-class
companies (Bessant and Francis, 1999). Six Sigma is a highly structured process
improvement framework that uses both statistical and non-statistical tools and techniques
to eliminate process variation and thereby improve process performance and capability
(Antony and Banuelas, 2002). Minimising defects to the level of 3.4 defects per Million
opportunities (DPMO) is at the heart of this methodology (Harry, 1998; McAdam and
Lafferty, 2004). To achieve target, this approach seeks to identify and eliminate
defects, mistakes or failures in business processes by focusing on process performance
characteristics (Snee, 2004).
Six Sigma approach have been found to be important profit drivers in a variety
of industries (Caulcutt, 2001), highlighting the economic dimension of quality
improvement. By employing DMAIC methodology, most of the Six Sigma efforts are
focused on taking variability out of the existing processes (Park et al, 2005; Bhote, 2002).
DMAIC is anagram of the major phases within the methodology namely, define measure,
analyse, improve and control (Breyfogle, 1999). The define phase entails the definition of
the problem and critical-to-quality (CTQ) characteristic. The measure phase selects most
appropriate quality characteristic to be improved and establishes metrics. In analysis
phase, the root causes of defect are analysed. In improve phase, simple but powerful
 Defect control analysis for improving quality and productivity 261

statistical tools/techniques are used to reduce the defect or process variations. In control
phase, the way of sustaining the improvement is formulated and put in force
(Pyzdek, 2001; Montgomery, 1998). The DMAIC frame work utilises various tools and
techniques like control charts, quality function deployment (QFD), failure mode and
effect analysis (FMEA), design of experiments (DoE) and statistical process control
(SPC) for variation management to drive out defects in operations. Among the available
collection of tools and techniques, application of DoE is at the heart of DMAIC cycle
(Breyfogle, 1999a).
DoE technique helps to identify key process parameters and to subsequently
adjust them in order to achieve sustainable performance improvements (Pande et al.,
2002). This technique determines significant factors and factor interactions that
affect variability within a product at the improvement stage of Six Sigma application
(Antony, 2001). There may be situations when there is a trade-off in the selection
of process parameter levels to identify the optimum parameter level (Antony et al.,
1999). To tackle the conflicting situations, Japanese scientist Taguchi (1986) has
developed robust design approach (Robinson et al., 2003). Robust design is an
efficient and systematic methodology that applies statistical experimental design for
improving product and manufacturing process design (Tsui, 1992). Robust design
approach analyses the signal-to-noise (S/N) ratios as a means of finding a robust
solution (Maghsoodloo, 1990). The loss function is a contribution of robust design
approach to address the process variability. This quadratic loss function informs
about money value and product variability to all management. Classical DoE do not
directly emphasise the reduction of process variability and translate this need to
economical considerations for management (Breyfogle, 1999). Moreover, classical
DoE approach is deemed too costly and time consuming because of full factorial
designs (Ross, 1988). Orthogonal array (OA)-based Taguchi DoE has been recognised as
an economical tool to optimise process parameters in industry for quality and
productivity (Nataraj et al., 2007). Taguchi technique found an alternative
experimental design strategy to drive out the fit fall in classical approach. The basic
promise between classical and Taguchi approach to higher quality is that one proposes
that higher quality costs more and the other proposes that higher quality costs less
(Kackar, 1985).
There are currently few publications that have spoken the application of
Taguchi methodology in Six Sigma projects. This study is an initiative to practice
Six Sigma DMAIC concepts with Taguchis experimental design. Since the casting
process is prone to defects, this new approach is proposed in a die casting industry.
Today, most of the automobile components are manufactured by die casting process;
any reduction in defect level can benefit not only the manufacturer but also the end
user. This paper presents a case study performed to eliminate the casting defects in two
stroke engine oil pump body. The organisation was encountering blow holes defect in oil
pump body casting. In Six Sigma improve phase, Taguchi experimental design approach
is used to optimise the casting process parameters. L27 OA is used to make casting
process optimisation with minimum amount of data. The application of a Six Sigma
methodology (DMAIC) with Taguchis DoE has developed a structured defect control
framework to improve productivity and product quality of die casting processes with less
cost and time.
262 M. Shanmugaraja et al.

2 Relationship between quality and productivity

Organisations wants to increase profits, it should increase productivity as well as quality

(Park, 2003). Without improving quality, increasing the productivity may not be right.
But improving quality will results in improved productivity. Deming (1986), looking at
the relationship between quality and productivity, stresses improving quality in order to
increase productivity. To become an excellent company, the management should find
ways to improve quality as well as productivity simultaneously. Then, several benefits
result:
Productivity rises.
Quality improves.
Cost per good unit decreases. Price can be cut.
Workers morale improves because they are not seen as the problem.
Stressing productivity only may mean sacrificing quality and possibly decreasing output
as case 1 explained in Table 1. Increasing productivity by 10% may pressurise the
production resources and may results in 20% defective. Number of good items drops to
80% lesser than original 90%. Very often quality improvement results in productivity
improvement as illustrated in case 2 in Table 1. By continually trying to improve quality
of items produced, the defect rate can be brought to 1% from 10%.
This results in increased productivity by 9% as shown. Any focus on improving the
process quality results in a reduction of defective units, yielding an increase in good
units, quality, and eventually productivity.
Table 1 Productivity vs. quality focus to improvement

Case 1 Case 2
Focus: productivity Focus: quality
Before After Before After
Total items produces 1,000 1,100 1,000 1,000
Defective items 100 220 100 10
Good items 900 880 900 990
Improvement 90% 80% 90% 99%

Improving productivity through quality can produce all the desired results: better quality,
less rework, greater productivity, lower unit cost, price elasticity, improved customer
satisfaction, larger profits and more jobs. After all, customers get high quality at a low
price, vendors get predictable long-term sources of business, and investors get profits, a
win-win situation for everyone.

3 The Six Sigma programme

Six Sigma is a philosophy for company-wide quality improvement (McAdam and

Lafferty, 2004). Six Sigma is a powerful breakthrough business improvement strategy
 Defect control analysis for improving quality and productivity 263

that enables companies to use simple and powerful statistical methods for achieving
and sustaining operational excellence (Harry and Schroeder, 1999). The statistical
representation of Six Sigma describes quantitatively how a process is performing
(Banuelas et al, 2005).

Figure 1 Non-central normal distribution with 3.4 ppm sigma quality level (see online version for
colours)

The numerical goal of Six Sigma is reducing defects to less than 3.4 parts per
million (PPM), reducing cycle time and dramatically reducing costs which impact
the bottom line (Behara et al., 1995; Goh and Xie, 2004). As a process improvement
method, Six Sigma is largely mechanistic or process focused (McAdam and Lafferty,
2004). Figure 1 illustrates the basic measurement concept of Six Sigma (Park,
2003). Sigma quality level includes a 1.5 shift in mean value to account typical
shifts and drifts of the process mean. Table 2 shows the number of PPM that would be
outside the specification limits if the data were normally distributed with mean shifted to
1.5 .

Table 2 Sigma quality level with 1.5 mean shift

Specification limit Percent Defective PPM

1 30.23 697,700
2 69.13 308,700
3 93.32 66,810
4 99.3790 6,210
5 99.97670 233
6 99.999660 3.4
264 M. Shanmugaraja et al.

DMAIC is a disciplined and quantitative approach of Six Sigma for improving

the product or process quality (Hahn and Doganaksoy, 2000). DMAIC provide teams a
methodological framework to guide them in the conduct of improvement projects (Pande
et al., 2002; Pyzdek, 2003). The flowchart for DMAIC quality improvement process is
sketched in Figure 2 (Linderman et al., 2003).

Figure 2 Six Sigma DMAIC information flow (see online version for colours)

Phase 0 (Definition): This phase is concerned with identification of the process or

product that needs improvement. It is also concerned with benchmarking of key
product or process characteristics of other world-class companies.
Phase 1 (Measurement): This phase entails selecting product characteristics; i.e.,
dependent variables, mapping the respective processes, making the necessary
measurement, recording the results and estimating the short- and long-term
process capabilities.
Phase 2 (Analysis): This phase is concerned with analysing and benchmarking the key
product/process performance metrics. Following this, a gap analysis is
undertaken to identify the common factors of successful performance.
Phase 3 (Improvement): This phase is related to selecting those product performance
characteristics which must be improved to achieve the goal. Once this is done,
the characteristics are diagnosed to reveal the major sources of variation. Next,
the key process variables are identified usually by way of statistically designed
experiments including Taguchi methods and other robust DoE. The improved
conditions of key process variables are verified.
 Defect control analysis for improving quality and productivity 265

Phase 4 (Control): This last phase is initiated by ensuring that the new process
conditions are documented and monitored via SPC methods. After the settling
in period, the process capability is reassessed. Depending upon the outcome of
such a follow-on analysis, it may become necessary to revisit one or more of
the preceding phases.

3.1 Taguchi method

In the 1980s interest in quality improvement among quality engineers and statisticians in
the West grew substantially. Most emblematic among the originators of this interest is the
Japanese Engineer Genichi Taguchi. He invented and promoted an offline quality control
concept that encompasses all stages of the product/process development (Nataraj and
Arunachalam, 2006). This concept represents a view on quality in which variation plays
a dominant role. This view on quality is generally accepted (Nair, 1992). Two prime
concerns of Taguchis approach are:
1 How to reduce economically the variation of a products function in the customers
environment?
2 How to ensure that decisions found optimum during laboratory experiments will
prove to be valid and reliable in manufacturing? (Nataraj and Arunachalam, 2006)
Furthermore, Taguchi introduced an alternative experimentation methodology using
OAs (Ross, 1988) to determine the near optimum settings of process parameters for
performance robustness (Nataraj et al., 2006b). The adequacy of this methodology has
been the subject of much debate among statisticians (Nair, 1992), though it is popular in
engineering (Jugulum and Dichter, 2001). Taguchi emphasises the importance of
variation reduction in quality improvement. Based on the results of an experiment,
settings for the control variables are chosen such that the process is made robust against
variation in the nuisance variables. The performance of a robust product is minimally
affected by environmental conditions in the field, by the extend of usage, or by
item-to-item variation during manufacturing (Nataraj et al., 2006a). The Taguchis
experimental approach is accompanied with analysis of variance (ANOVA), OAs, and
S/N ratio (Ross, 1988; Nataraj et al., 2006b).

3.2 Analysis of variance

ANOVA is a statistical method used to interpret experimental data and make necessary
decisions. The ANOVA breaks the total variations into accountable sources; total
variation is decomposed into appropriate components. The ANOVA analysis may be of:
1 No-way ANOVA in which the total variation (SST) is broken into variations due to
average/mean of all data point relative to zero (SSm)and variations of individual data
points around the mean (SSe).
SST = SS m + SSe

2 One-way ANOVA considers the effect of one controlled parameter upon the
performance of a product or process. Here the total variation (SST) is broken into:
266 M. Shanmugaraja et al.

variations due to average/mean of all data point relative to zero (SSm)

variations of individual data points under each factor level around the average of
all observations (SSA)
variations of the individual observations around the average of observations
under each factor level (SSe)

SST = SSm + SS A + SSe

3 Two-way ANOVA is the highest order of ANOVA analysis where two controlled
parameters are in the experimental situation. Here the total variation (SST) is
decomposed into:
variation due to factor A (SSA)
variation due to factor B (SSB)
variation due to interaction of factors A and B (SSAB)
variation due to error (SSe).

SST = SS A + SS B + SS A B + SSe

3.3 Orthogonal arrays

Based on DoE theory, Taguchis OA provide a method for selecting an intelligent subset
of the parameter space. OA is a matrix having rows and columns filled with factor levels
orthogonally. This orthogonal arrangement assures each and every level of each and
every factor comes in combination with each and every level of each and every other
factor. The number of columns in an OA indicates the maximum number of factors that
can be studied. OA is an efficient Taguchi test strategy to evaluate several factors in a
minimum number of tests as shown in Table 3 and obtains much information from few
trails.
Table 3 OA configurations

Orthogonal array Factors and levels No. of tests No. of tests in classical DoE
Two level array
L4 3 factors at 2 levels 4 8
L8 7 factors at 2 levels 8 128
L12 11 factors at 2 levels 12 2,048
L16 15 factors at 2 levels 16 32,768
L32 31 factors at 2 levels 32 2.147E + 09
Three level array
L9 4 factors at 3 levels 9 81
L18 8 factors at 3 levels 18 6,561
L27 13 factors at 3 levels 27 1.594E + 06
 Defect control analysis for improving quality and productivity 267

3.4 S/N ratio

Robust design approach uses a statistical measure of performance called S/N ratio
(Nataraj et al., 2006b). S/N ratio is used to choose the control levels that best cope with
noise while accounting for both mean and variability. It is a ratio of the mean (signal) to
the standard deviation (noise). Several S/N ratios are made available for choice based on
the type of quality characteristic to be considered as illustrated below:
1 lower is better
1

r
S / N LB = 10 log y2 (where r = number of tests)
i =1 i
r
2 nominal is best
S / N NB = 10 log Ve (for variance only)
Vm Ve
S / N NB = 10 log (for mean and variance)
rVe
3 higher is better
1 r
1
S / N HB = 10 log
r y
2
i =1 i

4 Case study

This case study deals with the reduction of casting defects in aluminium pressure die
casting industry. The problem was tackled using a Six Sigma DMAIC problem solving
methodology. The basic equation of Six Sigma, Y = f(x), defines the relationship between
a dependent variable Y or the outcome of a process and a set of independent variables or
possible causes which affect the outcome. In this case study, Y is the loss in
productivity due to an unacceptable number of defective castings.

4.1 The company

The subject company is a south Indian SME who produces many automobile aluminium
components for their various clienteles in India and abroad. Among the various
components produced in lots, two stroke engine oil pump body has selected for this
analysis. The reason for selecting this component is that the oil pump body has more
rejections than the rest. The castings are manufactured in pressure die casting method.
Around 90% of the companys output is being supplied within India. Since the operation
is mainly of made to order nature, occurrence of defects in large-scale affects the
production output which in turn makes hard to meet orders. Moreover, the company has
agreered to some orders under penalty basic for the shortage in supply to customers. Even
most of the casting defects are noticeable after the casting process; some may be
identified after a preliminary machining process. Normally, the machining process is
carried out at the customer end. In such cases, the company has to supply more number of
castings than ordered to compensate the rejections. The component selected for this study
268 M. Shanmugaraja et al.

is one such type being used in two wheelers engine as lubricating pump. Since it is the
heart of engine assembly, presence of any defect in the casing may lead to severe
consequences in operation. At the customer end, the castings are put in heat treatment
process for 24 hours before starting the machining operations. During machining, blow
holes are noticed particularly at the flange section of the component. Presence of such a
casting defect may result in oil seepage when the assembly is put in operation. At this
point of rejection, the casing costs considerable to the client company. Hence, it becomes
imperative to produce porous free castings not only to reduce rejections but also to
minimise the cost of rejection.

4.2 Define phase

This phase of the DMAIC methodology is aimed at defining the scope and goals of
implementing Six Sigma in terms of customer requirements and developing a process that
delivers these requirements. This defect control study was of the most important to top
management of the company as it was known that an effective solution to this problem
would have a significant impact on overall productivity as well as client satisfaction on
product quality. In this project, occurrence of blow holes in the oil pump body is the
prime concern. After performing a number of brainstorming exercises it is arrived at the
conclusion that the presence of blow holes was primarily influenced by the die casting
process. The solution to this problem was unknown and the impact of the problem was
very severe as it leads to a low quality and high rejection.

4.3 Measure phase

This is essentially a data-collection phase. At this phase, the following two important
aspects were addressed.

4.3.1 Data collection

Production and rejection statistics of oil pump body for the completed 20 batches are
collected from the companys record and it infers that the rejection was in increasing
trend as shown in Figure 3.
Figure 3 Rejection trend chart before study (see online version for colours)

4.3.2 Calculation of present sigma level

The existing process capability is expressed in terms of sigma quality level for the
purpose of comparing the improvement after the case study. Sigma quality level of each
batch is calculated through the following steps:
 Defect control analysis for improving quality and productivity 269

calculating defects per unit of the batch (DPU)

calculating defects per opportunities (DPO)

calculating defects per million opportunities (DPMO)

determining the sigma level corresponding to DPMO level.

The defects per unit (DPU) is:

Total number of defects observed in the batch

DPU =
Total number of units prodeuced in the batch

In this case, rejections due to blowholes are only concerned. Any other opportunities for
rejection are not accounted. Hence, the number of opportunities is one.
Hence, defects per opportunities (DPO) is:
DPU
DPO =
1
By the same token, defects per million opportunities (DPMO) is:
DPMO = DPO 1, 000, 000
The sigma quality level with 1.5 shift is determined (Breyfogle, 1999) by the
equation:

sigma quality level = 0.8406 + 29.37 2.221 ln( DPMO)

Table 4 summarises the sigma level calculation for all batches and it clearly indicates that
process performance is poor and it needs improvement. Existing process capability is
varying from 1.86 to 3.23 with an average of 2.51 and the defect level ranges
between 42,000 to 346,200 units with a mean of 17.22% out of a million outcomes.
Table 4 Sigma quality level of batches (see online version for colours)

Rej. Sigma Rej. Sigma

Batch DPU DPMO Batch DPU DPMO
(%) level (%) level
1 13.49 0.1349 134,900 2.61 11 8.16 0.0816 81,600 2.90
2 6.32 0.0632 63,200 3.04 12 13.07 0.1307 130,700 2.63
3 12.99 0.1299 129,900 2.63 13 13.35 0.1335 133,500 2.62
4 13.92 0.1392 139,200 2.59 14 21.82 0.2182 218,200 2.28
5 10.07 0.1007 100,700 2.79 15 19.38 0.1938 193,800 2.37
6 21.19 0.2119 211,900 2.30 16 32.81 0.3281 328,100 1.92
7 15.08 0.1508 150,800 2.54 17 34.62 0.3462 346,200 1.86
8 24.69 0.2469 246,900 2.18 18 24.04 0.2404 240,400 2.20
9 4.28 0.0428 42,800 3.22 19 29.2 0.292 292,000 2.03
10 4.2 0.042 42,000 3.23 20 21.69 0.2169 216,900 2.28
Average defect level = 17.22% Average sigma quality level = 2.51
270 M. Shanmugaraja et al.

4.4 Analyse phase

This phase is intended to analyse the data to determine the direction of process
improvement. In this case, it is important to identify the possible sources of variation
which lead to blowholes. A cause and effect study is conducted and parameters, thought
to contribute to more blowholes are listed as shown in Figure 4.
Figure 4 Cause and effect analysis (see online version for colours)

Among the possible parameters, five (metal temperature, intensifier pressure, metal
degassing frequency, 2nd phase turns, and metal mixing ratio) are selected for further
analysis. The rest were distinguished as constant parameters and kept steady throughout
the analysis. This selection of parameters of interest and their working range are based on
field expertise and field executives opinion.

4.5 Improve phase

In this phase, Taguchi DoE is conducted with the five process parameters identified from
the analysis phase.

5 Taguchis DoE methodology

Step 1 Determination of factor levels

It is decided to have 3 levels for each factor to accommodate the non-linear relationship
between factors as shown in Table 5.
 Defect control analysis for improving quality and productivity 271

Table 5 Process parameters with levels

Factor notation Controllable factors Metric Level 1 Level 2 Level 3

P Metal temperature Centigrade 665 690 715
Q Intensifier pressure Kg/cm2 220 240 260
R Metal degassing Shots per 320 240 160
frequency degassing
S II phase turns Nos. 3 3.25 3.5
T Metal mixing ratio Ratio 80:20 70:30 60:40

Step 2 Selection of OA
It is interested to study the two-factor interaction effects on the component in respect to
formation of blowholes. The selected interactions are:
metal temperature [P] and intensifier pressure [Q]
intensifier pressure [Q] and metal degassing frequency [R]
metal temperature [P] and metal degassing frequency [R].
To select the appropriate OA matrix to fit the case study, it is required to count the
degrees of freedom (DOF). To reach near optimum parameter level combination, the
minimum number of tests that must be performed will be identifiable with DOF. Total
DOF is counted by adding DOF of the number of control parameters which is equal to
one less than the number of level. This case study required minimum of 22 tests to get
optimum parameter settings as shown in Table 6. To cater this situation, an OA with at
least 22 DOF needed to be selected. Hence, L27, a three-level OA with DOF = 27 which
is greater than DOF of the factors and interactions, is selected.
Table 6 Degrees of freedom

Factors/interactions DOF
P 2
Q 2
R 2
S 2
T 2
PQ 22
QR 22
PR 22
Total DOF 22

Step 3 Arranging factors and interactions L27 OA columns

Main factors P, Q, R, S, and T are assigned to columns 1, 2, 5, 9 and 10, respectively by
using the linear graphs shown in Figure 5. The columns subsequent to main factor
columns evaluate the interaction of those factors. By using the triangular table shown in
272 M. Shanmugaraja et al.

Table 7, the interactions PQ, QR and PR are assigned to columns 3 and 4, 8 and 11, and
6 and 7, respectively and resultant L27 matrix is shown in Table 8.

Figure 5 L27 linear graph

Table 7 L27 triangular table

Column Column no.

no. 2 3 4 5 6 7 8 9 10 11 12 13
1 3* 2 2 6 5 5 9 8 8 12 11 11
4* 4 3 7 7 6 10 10 9 13 13 12
2 - 1 1 8 9 10 5 6 7 5 6 7
- 4 3 11 12 13 11 12 13 8 9 10
3 - - 1 9 10 8 7 5 6 6 7 5
- - 2 13 11 12 12 13 11 10 8 9
4 - - - 10 8 9 6 7 5 7 5 6
- - - 12 13 11 13 11 12 9 10 8
5 - - - - 1 1 2 3 4 2 4 3
- - - - 7 6 11 13 12 8 10 9
6 - - - - - 1 4 2 3 3 2 4
- - - - - 5 3 12 11 10 9 8
7 - - - - - - 3 4 2 4 3 2
- - - - - - 12 11 13 9 8 10
8 - - - - - - - 1 1 2 3 4
- - - - - - - 10 9 5 7 6
9 - - - - - - - - 1 4 2 3
- - - - - - - - 8 7 6 5
10 - - - - - - - - - 3 4 2
- - - - - - - - - 6 7 7
11 - - - - - - - - - - 1 1
- - - - - - - - - - 13 12
12 - - - - - - - - - - - 1
- - - - - - - - - - - 11
 Defect control analysis for improving quality and productivity 273

Table 8 Basic structure of L27 OA with factor assigned

Objective
Columns
function
Test S/N
no. P Q PQ PQ R PR PR QR S T QR * * ratio
Run 1 Run 2
1 2 3 4 5 6 7 8 9 10 11 12 13
1 1 1 1 1 1 1 1 1 1 1 1 1 1 y1,1 y2,1 S/N1
2 1 1 1 1 2 2 2 2 2 2 2 2 2 y1,2 y2,2 S/N2
3 1 1 1 1 3 3 3 3 3 3 3 3 3 y1,3 y2,3 S/N3
4 1 2 2 2 1 1 1 2 2 2 3 3 3 y1,4 y2,4 S/N4
5 1 2 2 2 2 2 2 3 3 3 1 1 1 y1,5 y2,5 S/N5
6 1 2 2 2 3 3 3 1 1 1 2 2 2 y1,6 y2,6 S/N6
7 1 3 3 3 1 1 1 3 3 3 2 2 2 y1,7 y2,7 S/N7
8 1 3 3 3 2 2 2 1 1 1 3 3 3 y1,8 y2,8 S/N8
9 1 3 3 3 3 3 3 2 2 2 1 1 1 y1,9 y2,9 S/N9
10 2 1 2 3 1 2 3 1 2 3 1 2 3 y1,10 y2,10 S/N10
11 2 1 2 3 2 3 1 2 3 1 2 3 1 y1,11 y2,11 S/N11
12 2 1 2 3 3 1 2 3 1 2 3 1 2 y1,12 y2,12 S/N12
13 2 2 3 1 1 2 3 2 3 1 3 1 2 y1,13 y2,13 S/N13
14 2 2 3 1 2 3 1 3 1 2 1 2 3 y1,14 y2,14 S/N14
15 2 2 3 1 3 1 2 1 2 3 2 3 1 y1,15 y2,15 S/N15
16 2 3 1 2 1 2 3 3 1 2 2 3 1 y1,16 y2,16 S/N16
17 2 3 1 2 2 3 1 1 2 3 3 1 2 y1,17 y2,17 S/N17
18 2 3 1 2 3 1 2 2 3 1 1 2 3 y1,18 y2,18 S/N18
19 3 1 3 2 1 3 2 1 3 2 1 3 2 y1,19 y2,19 S/N19
20 3 1 3 2 2 1 3 2 1 3 2 1 3 y1,20 y2,20 S/N20
21 3 1 3 2 3 2 1 3 2 1 3 2 1 y1,21 y2,21 S/N21
22 3 2 1 3 1 3 2 2 1 3 3 2 1 y1,22 y2,22 S/N22
23 3 2 1 3 2 1 3 3 2 1 1 3 2 y1,23 y2,23 S/N23
24 3 2 1 3 3 2 1 1 3 2 2 1 3 y1,24 y2,24 S/N24
25 3 3 2 1 1 3 2 3 2 1 2 1 3 y1,25 y2,25 S/N25
26 3 3 2 1 2 1 3 1 3 2 3 2 1 y1,26 y2,26 S/N26
27 3 3 2 1 3 2 1 2 1 3 1 3 2 y1,27 y2,27 S/N27

Step 4 Execution of experiments

The casting process is carried out as per the factor settings in each test condition and
1,000 components are produced in two replications. The number of good components is
recorded as response for each test. Since the experiment response is number of good
components, higher is best S/N ratio characteristic selected and calculated using the
below equation and recorded as shown in Table 9.
274 M. Shanmugaraja et al.

1 r
1
S / N NB = log
r y
2
i =1 i

Table 9 Experiment response with S/N ratio

Test No. of good items out of 500 S/N ratio Test No. of good items out of 500 S/N ratio
no. Run 1 Run 2 (HB) no. Run 1 Run 2 (HB)
1 401 452 52.55 15 458 478 53.39
2 462 450 53.17 16 490 486 53.76
3 421 436 52.63 17 369 389 51.56
4 390 401 51.94 18 385 368 51.5
5 369 354 51.15 19 436 455 52.97
6 469 476 53.48 20 395 421 52.19
7 485 468 53.55 21 459 462 53.26
8 359 346 50.93 22 463 475 53.42
9 310 264 49.07 23 401 426 52.31
10 418 431 52.55 24 485 476 53.63
11 479 469 53.51 25 495 486 53.81
12 352 378 51.22 26 425 435 52.66
13 301 320 49.82 27 465 452 53.22
14 329 365 50.77

Step 5 Data analysis

The data collected in two replicates of 27 trials are analysed and mean values of response
and S/N ratio are arrived for main factors and interactions as shown in Table 10. Mean
response for factor P at level 1 is calculated by averaging the responses of tests in which
the factor P is kept at level 1.
(401 + 462 + 421 + 390 + 369 + 469 + 485 + 359 + 310)
=
29
(452 + 450 + 436 + 401 + 354 + 476 + 468 + 346 + 264)
+
29
= 406.3
Similarly, mean response and mean S/N ratio values for all the factors and interactions at
each level is calculated as shown in Table 10.

Step 6 Response curve analysis

Mean response and mean S/N ratio values are plotted against each level of factors to
draw the response curve. Response curves are graphical representation of change in
performance characteristics for the variation in factor levels. From Figure 6, optimum
factor levels are identified based on the higher is best S/N ratio characteristic and listed
in Table 11.
 Defect control analysis for improving quality and productivity 275

Table 10 Mean response and Mean S/N ratio

Factors/levels L1 L2 L3 Factors/levels L1 L2 L3
P 406.3 403.6 450.7 P 52.06 52.02 53.06
Q 432.1 413.1 415.4 Q 52.7 52.22 52.23
R 436.3 402.4 421.9 R 52.7 52.03 52.38
S 420.8 419.4 420.4 S 52.40 52.34 52.39
T 419.7 410.5 430.4 T 52.36 52.14 52.63
PQ 442.4 425.5 405.4 PQ 52.60 52.50 52.00
QR 412.2 436.1 412.3 QR 52.20 52.70 52.20
PR 425.4 420.9 414.2 PR 52.60 52.40 52.20

Table 11 Optimum factor settings

Notation Factor Optimum level Value

P Metal temperature L3 715C
Q Intensifier pressure L1 220 Kg/cm2
R Degassing frequency L1 320 shots/degas
S 2nd phase turns L1 3 no
T Metal mixing ratio L3 60:40

Figure 6 Response curves for process parameters (see online version for colours)
276 M. Shanmugaraja et al.

Step 7 ANOVA analysis

It is interested to identify the most influencing process parameter causes blow holes.
Pooled ANOVA analysis is done on the output response data at 95% confidence interval
as deployed in Table 12. The ANOVA summary showed that the factor Q, S and T has no
influence on the response where as the factors P (metal temperature) and R (degassing
frequency) have a substantial influence on the response. Also interactions of P with Q
and with R played a considerable role in blowholes formation.
Table 12 Pooled ANOVA analysis on response data (see online version for colours)

Source SS DOF V F ratio F-critical Percent contribution

P 25,150.26 2 12,575.13 6.5168 3.225 15.25
Q 3,851.148 [2] Pooled
R 13,790.48 2 6,895.241 3.5733 3.225 8.35
S 18.481 [2] Pooled
T 3,567.37 [2] Pooled
PQ 25,456.07 4 6,364.018 3.2980 2.6 15.40

PR 21,483.74 4 5,370.935 2.7834 2.6 13.00

QR 7,426.37 [4] Pooled

Error 79,115.65 41 1,929.65 48.00
Total 164,986.2 53 100.00
Note: Significant at 95% confident level.

Step 8 Mean response predicting

To validate the optimum factor level setting for future process, confident interval and the
mean response are estimated as below:

Good = P3 + Q1 + R1 + S1 + T 3 4 M good

where
P3 mean response at level 3 of factor P
Q1 mean response at level 1 of factor Q
R1 mean response at level 1 of factor R
S1 mean response at level 1 of factor S
T3 mean response at level 3 of factor T
Mgood overall mean response value.
 Defect control analysis for improving quality and productivity 277

From Table 10, the mean response values are taken and estimated means is calculated as:
Good = 450.7 + 432.1 + 436.3 + 420.8 + 430.4 4(420.18)
Good = 489.58

Confidence interval for the population is calculated using the following formula (Ross,
1988):

1
CI = F ;1;Ve Ve

e

where
F ;1;Ve F ratio required for (risk)

ve error degree of freedom

Ve error variance
ne experiment trails = 54.
In this study:
risk is taken as 0.10
Confidence = 1 risk
ve DOF for error variance is 41 from Table 4
Ve = 1,929.65 from ANOVA table
F (1, 41) = 2.84 (taken from f ratio table).
Hence,

CI = (2.84)(1,929.65)[1 / 54]
CI = 10.1
The estimated mean response is 489.58 and at 90% CI, the predicted optimum output
would be:

Good CI < Good < Good + CI

[489.58 10.1] < Good < [489.58 + 10.1]
479.5 < Good < 499.7

Step 9 Confirmation experimental result

A confirmation test is conducted with batch volume of 1,000 pieces at the optimum factor
settings. Out of 20,000 components produced in 20 batches, the client report showed that
1296 components are rejected due to blowholes after machining. The confirmation
experiment results inferred that the rejection rate is brought to 4.8% in average as shown
in Figure 7.
278 M. Shanmugaraja et al.

Figure 7 Rejection run chart after study (see online version for colours)

6 Findings of the research study

Table 13 presents the consequential remarks of Six Sigma implementation with Taguchi
robust design methodology for controlling defects in aluminium die casting industry.
Table 13 Results of the research study

Process evaluation
Contribution
Process parameters SS Factor Optimum
SS DOF Cumulative
(SS/DOF) (SS/total) level
Highly critical factors
Metal temperature [P] 25,150.26 2 12,575.13 54.23 54.23
Critical factor
Degassing frequency 13,790.48 2 6,895.24 29.74 83.97 320 shots/
[R] degas
Non-critical factors
Intensifier pressure 3,851.15 2 1,925.575 8.30 92.27 220 kg/cm2
[Q]
Metal mixing ration 3,567.37 2 1,783.685 7.69 99.96 60:40:00
[T]
2nd phase turns [S] 18.48 2 9.24 0.04 100 3 no
Total 23,188.87 100.00
Performance evaluation
Rejections Productivity
Percent Percent
Min Max Avg savings Total items Defective Good
Productivity savings
(%) (%) (%) produced items items
Before 4.2 34.62 17.22 12.42 10,000 1722 8278 82.78 10.74
study
After 3.2 7.6 4.8 20,000 1296 18704 93.52
study
 Defect control analysis for improving quality and productivity 279

The cumulative contribution of process parameters up to 84% is found critical for casting
quality. Metal temperature and degassing frequency were found influencing parameters
for casting defect. Interactions of metal temperature with intensifier pressure and
degassing frequency also found critical for blowholes formation. 12.42% reduction in
defect level has resulted in productivity improvement by 10.74% with marginal cost of
quality control.

6.1 Calculation of sigma level after study

The defects per unit (DPU) is:
Total number of defects observed in the batch
DPU =
Total number of units produced in the batch
= 1, 296 / 20, 000
= 0.0648
DPU
DPO =
1
= 0.0648 / 1
= 0.0648
DPMO = DPO 1, 000, 000
= 0.0648 1, 000, 000
= 64,800
The sigma quality level with 1.5 shift is:

sigma quality level = 0.8406 + 29.37 2.221 ln( DPMO)

= 0.8406 + 29.37 2.221 ln(64,800)
= 3.03

7 Conclusions

This paper has explored how best casting industries can take steps to control or minimise
defects through Six Sigma programme. Six Sigma DMAIC approach has been extremely
utilised for statistical analysis of the problem. Taguchis robust design practice was used
as improvement strategy to discover optimum process parameter settings. L27 OA-based
experimental design was carried out to study the variations in casting quality at different
levels of the process parameters. From the response graph analysis, it could be observed
that the highest process yield could be attained at the combined settings of parameters
P(3), Q(1), R(1), S(1), and T(3), i.e., metal temperature at 715 degree centigrade,
intensifier pressure at 220 kg/cm2, degassing frequency of 320 shots/degas, 2nd phase
turns at 3 nos. and metal mixing ratio as 60:40. Pooled ANOVA analysis has performed
and found metal temperature and degassing frequency are vital process parameters which
contributed nearly 84% of variations in output casting quality. The experimental results
have been validated by confirmation tests and found the casting process capability
280 M. Shanmugaraja et al.

improved from 2.51 to 3.03 by reducing the defect rate by 12.42% (17.224.8) as
illustrated in Figure 8. The results of this project provided better motivation for the future
applications of Six Sigma methodology across all the components manufactured by the
company. Taguchi DoE has become a part of working culture of the company on the
realisation its worth by the management. The benefits of adopting the Six Sigma problem
solving methodology have been convinced of and it has been linked directly to the
strategic goals of the company.

Figure 8 Rejection percentage evaluations (see online version for colours)

8 Research implications

Much of this study is devoted to improve the product quality through the process
optimisation. But the costs associated with this improvement drive are not accounted
anywhere in this study. Investing much on quality improvement will have adverse impact
on productivity. This study will have substantial support where the cost of poor quality is
more. Before going to improve either quality or productivity, the cost of improvement
against savings should be analysed. Even this may not be the case for many, as an
improvement drive, it should have favourable results in all aspects of the business. Future
research may be evoked to incorporate cost components into the analysis for
synchronising the improvement on quality and productivity.

References
Antony, J., et al. (1999) Reducing manufacturing process variability using experimental design
technique: a case study, Integrated Manufacturing Systems, Vol. 10, No. 3, pp.162169.
Antony, J. and Banuelas, R. (2002) Key ingredients for the effective implementation of Six Sigma
program, Measuring Business Excellence, Vol. 6, No. 4, pp.2027.
Antony, J. (2001) Improving the manufacturing process quality using design of experiments:
a case study, International Journal of Operations and Production Management, Vol. 21,
Nos. 5/6, pp.812822.
Banuelas, R., Antony, J. and Brace, M. (2005) An application of Six Sigma to reduce waste,
Quality and Reliability Engineering International, Vol. 21, pp.553570.
 Defect control analysis for improving quality and productivity 281

Behara, R.S., Fontenot, G.F. and Gresham, A. (1995) Customer satisfaction measurement and
analysis using Six Sigma, International Journal of Quality and Reliability Management,
Vol. 12, No. 3, pp.918.
Bessant, J. and Francis, D. (1999) Developing strategic continuous improvement capability,
International Journal of Operations & Production Management, Vol. 19, No. 11,
pp.11061119.
Breyfogle, F.W. (1999a) Implementing Six Sigma Smarter Solutions Using Statistical Methods,
Wiley, New York, NY.
Breyfogle, F.W. (1999b) Implementing Six Sigma, John Wiley & Sons, Inc., New York, NY.
Bhote, K.R. (2002) The Ultimate Six Sigma: Beyond Quality Excellence to Total Business
Excellence, 1st ed., AMACOM/American Management Association, New York, NY.
Caulcutt, R. (2001) Why is Six Sigma so successful?, Journal of Applied Statistics, Vol. 28,
Nos. 34, pp.301306.
Cox, J.F., Blackstone, J.H. and Schleier, J.G. (2003) Managing Operations: A Focus on Excellence,
Two Volume Set, North River Press, Great Barrington, MA.
Deming, W.E. (1986) Out of the Crisis, pp.2324, 9798, MIT Centre for Advanced Engineering
Study, Cambridge, MA.
Elshennawy, A.K., Maytubby, V.J. and Aly, N.A. (1991) Concepts and attributes of total quality
management, Total Quality Management, Vol. 2, No. 1, pp.7597.
Feigenbaum, A.V. (1991) Total Quality Control, 3rd ed., McGraw-Hill, London.
Goh, T.N. and Xie, M. (2004) Improving on the Six Sigma paradigm, The TQM Magazine,
Vol. 16, No. 4, pp.235240.
Hahn, G.J. and Doganaksoy, N. (2000) The evolution of Six Sigma, Quality Engineering,
Vol. 12, No. 3, pp.317326.
Harry, M.J. (1998) Six Sigma: a breakthrough strategy for profitability, Quality Progress,
Vol. 31, No. 5, pp.6064.
Harry, M.J. and Schroeder, R. (1999) Six Sigma: The Breakthrough Management Strategy
Revolutionizing the Worlds Top Corporations, Double Day, New York, NY.
Hobbs, D.P. (2004) Lean Manufacturing Implementation, 1st ed., Ross Publishing, Inc., Boca
Raton, FL.
Jugulum, R. and Dichter, A. (2001) Taguchi methods in American universities and corporations,
Quality Engineering, Vol. 13, No. 4, pp.607621
Juran, J.M. (1988) Juran on Planning for Quality, The free press, New York, NY.
Kackar, R. (1985) Off-line quality control, parameter design and the Taguchi method with
discussion, Journal of Quality Technology, Vol. 17, No. 4, pp.176209.
Linderman, K., Schroeder, R., Zaheer, S. and Choo, A. (2003) Six sigma: a goal theoretic
perspective, Journal of Operations Management, Vol. 21, No. 2, pp.193203.
Maghsoodloo, S. (1990) The exact relation of Taguchis signal-to-noise ratio to his quality loss
functions, Journal of Quality Technology, Vol. 22, pp.5767.
McAdam, R. and Lafferty, B. (2004) A multilevel case study critique of six sigma: statistical
control or strategic change?, International Journal of Operations & Production Management,
Vol. 24, Nos. 5/6, p.530.
Montgomery, D.C. (1998) Introduction to Statistical Quality Control, John Wiley & Sons, Inc.,
New York.
Nair, V. (1992) Taguchis parameter design: a panel discussion, Techno Metrics, Vol. 34, No. 2,
pp.127161.
Nataraj, M. and Arunachalam, V.P (2006) Optimizing impellor geometry for performance
enhancement of a centrifugal pump using the Taguchi quality concept, Part A: Proceedings of
Institute of Mechanical Engineers, Vol. 220, pp.765782.
282 M. Shanmugaraja et al.

Nataraj, M., Arunachalam, V.P. and Suresh, K.G., (2006a) Optimizing planer cam mechanism in
printing machine for quality improvement using Taguchi method: risk analysis with
concurrent engineering approach, International Journal of Computer Applications in
Technology, Vol. 26, No. 3, pp.164173.
Nataraj, M., Arunachalam, V.P. and Ranganathan, G., (2006b) Using risk analysis and Taguchis
method to find optimal conditions of design parameters: a case study, International Journal
of Advanced Manufacturing Technology, Vol. 27, pp.445454.
Nataraj, M., Arunachalam, V.P. and Balaji, B. (2007)A practical approach to optimize the coating
parameters to win customer confidence, Part B: Proceedings of Institute of Mechanical
Engineers, Vol. 222, pp.495506.
Nave, D. (2002) How to compare six sigma, lean and the theory of constraints, Quality Progress,
Vol. 35, No. 3, p.73.
Pande, P.S., Neuman, R.P. and Cavanagh, R.R. (2000) The Six Sigma Way, 1st ed., McGraw-Hill,
Inc., New York NY.
Pande, P.S., Neuman, R.P. and Cavanagh, R.R. (2002) The Six Sigma Way, Team Field book:
An Implementation Guide for Process Improvement Teams, McGraw-Hill, New York, NY.
Park, S. (2003) Six Sigma for Quality and Productivity Promotion, Asian Productivity
Organization, Japan.
Park, S.H., Hu, M., Barth, B. and Sears, R. (2005) Leveraging Six Sigma disciplines to drive
improvement, International Journal of Six Sigma and Competitive Advantage, Vol. 5,
Nos. 1 and 2, pp.121133.
Pyzdek, T. (2001) The Six Sigma Handbook, McGraw-Hill, London.
Pyzdek, T. (2003) The Six Sigma Handbook: A Complete Guide for Green Belts, Black Belts, and
Managers at all Levels, McGraw-Hill, New York, NY.
Robinson, T.J., Borror, C.M. and Myers, R.H. (2003) Robust parameter design: a review, Quality
and Reliability Engineering International, Vol. 20, pp.81101.
Ross, P. (1988) Taguchi Techniques for Quality Engineering, McGraw-Hill, London.
Snee, R.D. (2004) Six sigma: the evolution of 100 years of business improvement methodology,
International Journal of Six Sigma and Competitive Advantage, Vol. 1, No. 1, pp.420.
Taguchi, G. (1986) Introduction to Quality Engineering: Designing Quality into Products and
Processes, Asian Productivity Organization, Tokyo.
Tsui, K.L. (1992) An overview of Taguchi method and newly developed statistical methods for
robust design, IIE Transactions, Vol. 24, pp.4457.