See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259484891

SPC and Process Capability Analysis – Case Study

Conference Paper · June 2009

CITATIONS READS
4 47,725

2 authors:

Tatjana Sibalija Majstorovic Vidosav

Metropolitan University University of Belgrade
101 PUBLICATIONS 943 CITATIONS 183 PUBLICATIONS 1,398 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Tatjana Sibalija on 31 December 2013.

The user has requested enhancement of the downloaded file.

 International Journal ’’Total Quality Management & Excellence’’, Vol. 37, No. 1-2, 2009.

SPC AND PROCESS CAPABILITY ANALYSIS– CASE STUDY

UDC:
Tatjana V. ŠIBALIJA, Vidosav D. MAJSTOROVIĆ,
Faculty of Mechanical Engineering, University of Belgrade

Abstract: This paper presents one postulates of one of the most important quality engineering techniques
Statistical Process Control (SPC), embracing quality engineering tools: control charts and process
capability measurement. Their application is explained on a case study, which presents one part of Six
Sigma pilot project conducted in the observed manufacturing system.
Key Words: Statistical process control (SPC), control charts, process capability, Six Sigma.

1. INTRODUCTION are more eligible due to better quality of

information about the observed process. .
SPC implies application of statistical 2. Quality of measuring system. In order to
methods for identification and control of special provide reliable measurements of quality
causes of variation in the process. SPC presents a characteristics, it is necessary to perform
set of control charts based on statistical measuring system analysis and approve
principles, which monitor the process behaviour measuring system for measurement of the
(via observed significant factors) and indicate observed characteristics, prior to
process behaviour changes in order to eliminate implementation of SPC.
the potential problem on time, before it occurs. 3. Initial data collection for control chart
SPC program has two specific functions: (i) formation. Prior to formation of control chart,
as a monitor, to verify that process is under sampling plan must be determined at such
control, or to indicate that a process is not in way to adequately represent the observed
control, based on interpretation of control-chart characteristic of product and/or process and
abnormalities or other indicators; (ii) as a quality initial measuring data must be collected.
improvement technique, for the purpose of Training of operators who will measure
improving process capability [1]. characteristics and collect measuring data
Within SPC, control charts and capability must be properly performed.
studies are the main tools used to describe 4. Definition and documenting of OCAP (Out of
processes graphically. These two quality Control Plan Action). OCAP is the plan of
engineering tools are explained in this paper, as actions that must be carried out when the
well as their application on a real case study. observed characteristic is out of control. Each
control chart has its own OCAP. The purpose
2. SPC - BACKGROUND of OCAP is to instruct the operator who
collects the measuring data how to react on
Implementation of SPC for the observed each particular problem that indicates out of
process/system is carried out according to the control condition.
following steps [2]: 5. Calculation of control limits is performed with
1. Selection of quality characteristics / process respect to the type of control chart.
parameters (factors) relevant to the observed Calculation of control Limits for X , R and
product and/or process. Each control chart
X , σ charts is explained in the section 2.1.
follows the behaviour of the observed quality
characteristic with respect to process 6. Estimation of control. After control chart
parameters. Each point at the chart presents setup, it is possible to estimate weather the
characteristics of one sample, and it is process is statistically stable. The following
compared to prior points (measuring value situations indicate that process is not
and distribution) in order to estimate the statistically stable:
process trend. Table 1. shows the instructions • One or more points out of control limits;
for selection of the type of control chart • WER (Western Electric Rules) are special
according to the nature of the quality rules indicating that process is not stable or
characteristic (attribute or numerical data) and data are not correct (fig 1.), applicable only
sample size. Data collection for attribute for numerical data ( σ presents standard
control charts is faster and easier than for deviation of the process with respect to the
numerical charts, but numerical control charts observed characteristic):
 International Journal ’’Total Quality Management & Excellence’’, Vol. 37, No. 1-2, 2009.

- 7 consecutive points above or below average characteristic. If the same process is

value of characteristic ( X ), performed on several machines, it is possible
- 7 consecutive points in ascending or to detect the machines (charts) with lower
descending sequence (trend), capability with respect to the observed
- 2 from 3 points in zones [–3σ ; –2σ] and [+ characteristic and to react on time.
2σ ;o +3σ] (zone A), It is important to clarify two key terms in
- 4 from 5 points in zones [–2σ ; –σ] and [+ σ SPC: (i) Variability: Each process contains
; +2σ] (zone B), certain level of variability. Usual sources of
- 15 points in zone [–σ ; + σ] (zone C), variation cause the variation within the process
- 8 points in zones [ –2σ ; –σ] and [+ σ ; +2σ] but do not affect its stability, so the response is
(zone B) or bellow, predictable. Such process is called statistically
- 14 consecutive points alternating up-down. controlled (stable). Special sources of variation
• One or more individual values within the affect the variation distribution, and in such case,
observed sample out of specification limits. process response is not predictable until the
7. Identification of the cause of instability by special causes of variation are detected and
data analysis. After detection of data that eliminated. (ii) Sampling: SPC uses samples of
show process instability, the cause of data in order to measure behaviour of the process
instability could be found and eliminated by for the given moment of time. The goal is to
following the responding OCAP. For the obtain maximum quality information with
considered point showing instability or out of minimum quantity of data. Appropriate sample
control condition, the cause and appropriate size and frequency of sampling is of the special
action should be entered, from the list of importance for proper implementation of SPC.
causes and actions in the OCAP. For the effective application of SPC,
8. Calculation of process capability indices (Cp following points should be thoroughly
and Cpk), in order to perform benchmarking considered: optimal sample size; control of
and to compare with prior process behaviour process characteristics rather than control of
(aiming to continuously improve the process). product characteristics; control of numerical
characteristics rather than attribute
9. Process control and focus on chart
characteristics; accurate measurements; regular
(characteristics) with low Cpk value. Control
process control and stopping on time (before it
charts with low Cpk value indicate possible
produces defective products).
instability or out of control condition of
process, with respect to the observed
Table 1: Selection of the type of control charts [3].
Data type Definition Sample size Type of control chart
Number of defects in unit- constant "c" chart (number of defects)
Attribute data product variable "u" chart (defects per unit-product)
(continual and
Number of defective products constant sample size > 50 "np" chart
discrete values)
Data constant sample size > 50 "p" chart (defective units portion)
sample size =1 X , Rm chart with moving range
Numerical data
sample size <10 X , R chart
(continual values)
sample size ≥10 X , σ chart

+3σ zone A
avergae characteristic

+2σ
value in the sample

zone B
+σ
X zone C
-σ
-2σ zone B

-3σ zone A
sample number
Figure 1. Allocation of zones at the control chart for numerical data, for the application of WER [3].

2
 International Journal ’’Total Quality Management & Excellence’’, Vol. 37, No. 1-2, 2009.

2.1.1. Control charts • for the standard deviation ( σ )chart:

- upper control limit: URL = B4 σ
Control charts are quality engineering tool,
- lower control limit: LRL = B3 σ .
used separately or within SPC to measure and
where A1, B4 and B3 are coefficients which value
control process variation and detection of special
depends of the sample size, and σ is average
causes of variation.
value of standard deviations of all observed
Control charts are composed of sampling
samples.
results taken over time that are plotted as points
on charts. Different kinds of control charts
2.1.2. Process Capability
accommodate variable (continuous distributions)
or attribute (continual and discrete distributions) Capability studies are used to predict the
sampling. Variable control charts are composed overall ability of a continuous distribution
of two graphs. The top graph monitors process (variable type) process to make products within
statistical location (the average sample measuring the required specifications. Capability indices
value). It measures whether the process is used to express process capability and
adjusted properly, comparing the calculated performance are [5]:
process average to the nominal or target value. • Cp = Process Capability (simple and
The bottom graph monitors process variation straightforward indicator of process
(range or standard deviation of the sample). capability);
Attribute control charts are composed of only one • Cpk = Process Capability Index (adjustment
graph that monitors lot-to-lot variations in terms of Cp for the effect of non-centred
of percent or number of nonconforming distribution);
(defective products or defects). • Pp = Process Performance (a simple and
Depending of the sample size, X , R and X , σ straightforward indicator of process
are the most commonly used control charts in performance);
industrial practice. The first step in chart • Ppk = Process Performance Index
formation is to determine the sample size and (adjustment of Pp for the effect of non-
frequency of sampling. Usual practice is to centred distribution).
consider smaller sample size and regular Mathematical formulations of the above
sampling, since the purpose is to detect process indices are:
changes over time and not to disturb the regular
process realization. The second step implies Cp = Tolerance
6σ
collecting of initial sampling data (approximately
20 samples). Based on initial data, the average Cpk = minimum { USL − X ; X − LSL }
value and the range or standard deviation of 3σ 3σ
characteristic should be calculated. where σ is standard deviation of process with
If sample size is n ≤10 than X , R control chart respect to the observed characteristic:
is formed and control limits are calculated σ = R / d2
according to following [4]: d2 is coefficient which depends on sample size,
• for the average ( X ) chart: 6σ is natural process tolerance, Tolerance is
- upper control limit: UCL = X + A2 R specification tolerance of the process (Tolerance
= USL – LSL, where USL is upper and LSL is
- lower control limit: LCL = X - A2 R lower specification limit, with respect to the
• for the range ( R ) chart: observed characteristic (fig. 2.)).
- upper control limit: URL = D4 R Process with Cp = 1.0 or greater provides
response (products) that meet customer
- lower control limit: LRL = D3 R
specification, if process is centred in the middle
where A2, D4 and D3 are coefficients which value of the specification range. But, if average value
depends of the sample size, and R is average of the observed characteristic is changed than
value of ranges of all observed samples. distribution could be shifted so the process could
For sample size n >10 X , σ chart is be shifted with respect to specifications. Cp does
not consider the location of the process, in
considered and control limits are calculated as
contrast to Cpk. Cpk presents the specification
follows [4]:
width (USL – LSL) with respect to how well the
• for the average ( X ) chart:
process spread is located about the target and the
- upper control limit: UCL = X + A1 σ specification limits. A Cpk=1.0 or greater implies
- lower control limit: LCL = X - A1 σ that a process which is in control is predictable
 International Journal ’’Total Quality Management & Excellence’’, Vol. 37, No. 1-2, 2009.

over time, and it is capable to meet customer found that vital defects are mainly related to
specifications with respect to the observed product characteristic - pot enamel thickness.
characteristic. For processes that are expected to Ishikawa diagrams were used to analyse vital
meet "6σ" requirements, minimal required value defect and their main causes. They revealed that
is Cpk=2 and Cp=2 [4]. the majority of the defects are related mainly to
LSL USL sub-processes A5.2. – Base enamelling and
A.5.4. – Cover enamelling ([6]-[9]). Measuring
system analysis was performed in the Measure
phase: the observed measuring system used to
measure the most important product quality
characteristic - pot enamel thickness was found as
6σ adequate for the observed measurements [10].
Tolerance The analysis of process A5 - Automatic
enamelling was performed using SPC. Based on
Figure 2. Statistical presentation of process two-weeks sample data for production performed
capability. on Automat 2, X , R control charts were created
In contrast to Cp and Cpk that are used for for base enamel thickness (fig. 3.) and for total
short term, performance indices Pp and Ppk are (base plus cover) enamel thickness (fig. 4.).
used for long term analysis. Cp and Cpk compute Sample size was 5 measurements at one part
the index with respect to the samples of data (pot) and sampling frequency was 2 hours ([6]-
(average values of the samples), while Pp and [9]). Specification limits for base enamel
Ppk take into account whole process (all thickness are: LSL÷USL = 80÷120 µm ([8]-[9]),
individual data within all samples). Standard and for total enamel thickness: LSL÷USL =
deviation for Ppk is calcuated as: 180÷300 µm ([7]). Control limits for both charts
∑ (X − X i )2 were set according to the "6σ" requirements:
σ = i LCL/UCL = Average -/+ 3 σ ([4]).
N −1 From fig. 3 and fig. 4, following conclusions
where X is average value of characteristic, X i could be drawn ([8]-[10]):
is individual measuring value, N is total number - data for base enamel thickness characteristic
of measurements (i=1, .., N). are normally distributed (P<0.005),
For both Ppk and Cpk the 'k' stands for - X , R chart for base enamel thickness is in
'centralizing facteur'- it assumes the index takes control (there are no points out of control
into consideration the fact that data is maybe not limits),
centred (hence, index shall be smaller). It is more - process capability indices (Cp=Cpk=1.41,
realistic to use Pp & Ppk than Cp & Cpk as the Pp=1.5, Ppk=1.22) do not satisfy "6σ"
process variation cannot be tempered with by requirements: Cp, Cpk, Pp, Ppk > 2 ([4]);
inappropriate subgrouping (samples). However, - from capability histogram it is visible that
Cp and Cpk can be very useful in order to know process was off-centre, with respect to base
if, under the best conditions, the process is enamel thickness;
capable of fitting into the specifications or not. - data for total enamel thickness are normally
The values for Cpk and Ppk will converge to distributed (P<0.005),
almost the same value when the process is in - X , R chart for total enamel thickness is out of
statistical control, because the real standard control; data at the chart are gathering into
deviation and the sample standard deviation will two groups, with no visible criteria for their
be identical [5]. distinguishing; this indicates dispersion
problem;
3. CASE STUDY - process capability indices (Cp=5.58,
Cpk=4.43, Pp=3.24, Ppk=2.57) are very good
The pilot-project Six Sigma, for the observed and meet "6σ" requirements.
manufacturing system, was conducted according This indicated that process needs optimisation
to DMAIC (Define-Measure-Analyse-Improve- with respect to base enamel thickness ([8]-[9])
Control) methodology. In the Define phase, and with respect to cover enamel thickness ([7]).
manufacturing system was mapped using IDEFO Since it was not possible to measure cover
method. In order to rank and analyse defect in the enamel thickness directly, chart presented at fig.
manufacturing process A5 – Automatic 4 shows data for total thickness, which includes
enamelling, Pareto analysis was performed. It base and cover thickness.

4
 International Journal ’’Total Quality Management & Excellence’’, Vol. 37, No. 1-2, 2009.

Xbar Chart Capability Histogram

110 UCL=110,09

Sample Mean
105 _
_
X=103,73

100
LCL=97,36
1 20 39 58 77 96 115 134 153 172 93 96 99 102 105 108 111 114

R Chart Normal Prob Plot

UCL=23,33
AD: 15,241, P: < 0,005
Sample Range

20

_
10 R=11,03

0 LCL=0
1 20 39 58 77 96 115 134 153 172 90 100 110 120

Last 25 Subgroups Capability Plot

112 Within Within Overall
StDev 4,74309 StDev 4,44462
Values

104 Cp 1,41 Pp 1,5

Overall
Cpk 1,14 Ppk 1,22
CCpk 1,41 Cpm *
96 Specs
170 175 180 185 190
Sample

Figure 3. X , R control chart for base enamel thickness [9].

Xbar Chart Capability Histogram

111111121111111112 2212 11 1 11 1
88 115222222122 115552221 21222222 122221222222 LSL USL
5 2 2 22 22 86 UCL=232.36
232 552 2 2 2 8 5 2 6 5 2 2 2 8
8 8 6 22 2 62 2 2 8 Specifications
Sample Mean

2 22 2 86 2 2 _
_
X=227.56 LSL 180
224 22 22
USL 300
2 58 8 2 LCL=222.75
1 1111111 111 1 11 1 1
1
1
11111111111111111111 11 1111 1 1 1
1 11111111
216 1 1
1
1 20 39 58 77 96 115 134 153 172 192 208 224 240 256 272 288

R Chart Normal Prob Plot

20 AD: 11.136, P: < 0.005
UCL=17.62
Sample Range

10 _
R=8.33
2
2222222222
3 2 2222
0 LCL=0
1 20 39 58 77 96 115 134 153 172 200 220 240

Last 100 Subgroups Capability Plot

240 Within Within Overall
StDev 3.58161 StDev 6.1671
Values

225 Cp 5.58 Pp 3.24

Overall
Cpk 4.43 Ppk 2.57
Cpm *
210 Specs
100 120 140 160 180
Sample

Figure 4. X , R control chart for total (base and cover) enamel thickness [7].
 International Journal ’’Total Quality Management & Excellence’’, Vol. 37, No. 1-2, 2009.

Thus, data at chart presented at fig. 4 contain significant factors, as identified from the
variation of base and of cover enamel thickness. experimental design, will be monitored and the
Due to his fact, it is necessary first to optimise whole process will be documented to ensure that
the process A5 with respect to base enamel improvements are maintained beyond the
thickness (optimisation of sub-process A5.2.) in completion of the pilot- project. The achieved
order to solve location problem (since previous process improvements will be monitored and
and current process is off-centre). Then, verified in everyday practice by using control
optimisation of the process with respect to cover charts and process capability analysis with
enamel thickness (optimisation of sub-process respect to base and total enamel thickness
A5.4.) should be performed [2], [11]. The characteristics.
purpose of such optimisation is to find optimal
process parameters setting (for both sub-
processes A5.2. and A5.4) that meet REFERENCES
specifications for the target base and total enamel
thickness and to reduce variability of process [1] Šibalija, T., Attaining Process Robustness through
with respect to both characteristics. Design of Experiment and Statistical Process Control,
Proceedings of the 11th CIRP International Conference
on Life Cycle Engineering – LCE 2004, pp.161–168,
6. CONCLUDING REMARKS Belgrade, 2004.
[2] Majstorović, V., Šibalija, T., Implementation of
SPC and process capability analysis present SPC, Report on researches conducted in Six Sigma
powerful means for the analysis of current and pilot-project No.01.01.B. 2007/1, Faculty of
previous process behaviour and they provide Mechanical Engineering, University of Belgrade,
information that serve as a basis for the process 2007 (on Serbian language).
improvement. Correct implementation of SPC [3] Majstorović V., Šibalija T, Soković M., Pavletić
assures possibility to detect special causes of D., Monograph: “Tempus ETIQUM – Six Sigma
process variation on time, in order to eliminate Model and Application” - in progress (on Serbian
language).
them before generating defective products.
[4] Pyzdek T., The Six Sigma Handbook, McGraw-
Process capability analysis entails comparing the Hill Companies, Inc, 2003.
performance of a process against its [5] www.isixsigma.com
specifications, thus enabling analysis of previous [6] Šibalija, T., Majstorović, V., Six Sigma
and current process performance, as well as Methodology – Case Study, Proceedings of the 8th
benchmarking. This is of special importance International Conference on The Modern Information
when comparing previous or current process Technology in the Innovation Processes of the
performance with the process performance after Industrial Enterprises–MITIP, pp.246-252, Budapest,
improvement. Hungary, 2006.
SPC and process capability analysis are [7] Šibalija, T., Majstorović, V. Six Sigma
Methodology Implementation in Serbian
inevitable steps in implementation of Six Sigma
Manufacturing enterprise, International Journal
methodology for the existing process and/or ’’Total Quality Management & Excellence’’, No.3,
system according to DMAIC cycle. Within the Vol. 35, pp. 17-22, 200 7.
scope of plot-project for the observed process [8] Majstorović, V., Šibalija, T., An Application of
(A5 - Automatic enamelling) improvement DMAIC Approach to Process Quality Improvement –
according to DMAIC methodology, SPC and Case Study, Proceedings of IFAC Workshop on
process capability analysis were used in Analyse Manufacturing, Modelling, Management and Control -
phase. Their application revealed the location MIM 2007, Budapest, Hungary, 2007.
problem in sub-process A5.2 and dispersion [9] Šibalija, T., Majstorović, V. An Application of
problem in sub-process A5.4. DMAIC Approach – Case Study from Serbia,
Proceedings of International Conference on Advances
In the Improve phase of DMAIC approach
in Production Management Systems–APMS 2008,
DoE was used to identify the optimal settings of pp.120-126, Espoo, Finland, 2008.
critical-to-quality factors (process and enamel [10] Šibalija, T., Majstorović, V., Measuring System
parameters), for automatic enamelling process Analysis in Six Sigma methodology application – Case
[11]. By implementing the optimum parameters Study, Proceedings of the 10th CIRP Seminar on
into practices, it is expected that the process Computer Aided Tolerancing - CAT 2007, Erlangen,
performance will be improved, therefore Germany, 2007.
improving process robustness and capability. In [11] Šibalija, T., Majstorović, V., (2007) An Example
order to ensure sustainability, achieved results of DoE Application For Automatic Enamelling
will be followed through Control phase of Process Improvement, International Journal ’’Total
Quality Management & Excellence’’, No.1-2, Vol.35,
DMAIC approach. The improved data on
pp.405-410, 2007.

View publication stats