CHAPTER TWO - THEORY OF

CONTROL CHARTS

Methods And Philosophy of Statistical

Process Control

Control Charts for Variables

Control Charts for
Attributes

Process And Measurement

System Capability Analysis

Introduction
In market economy, the maintenance of quality has a profound importance in
manufacturing and servicing environment. For doing so, one of the earliest tools is
Statistical Process Control(SPC).
The application of statistical techniques to control a process.
While its key aspect is to obtain predictable processes that produce
consistent results by quickly detecting the occurrence of assignable causes
of process shifts, be it above or below control limits or unnatural patterns,
so that investigation of the process and corrective action may be
undertaken before many nonconforming units are manufactured.
In a nutshell, the eventual goal of SPC is the elimination of variability in the process.
SPC can be applied to any process. Its seven major tools are:
1) Histogram
5)
Defect
concentration
2) Check sheet
diagram
3) Pareto Chart
6) Scatter diagram
4) Cause and effect diagram
7) Run chart
Dr.Gere K

 cont.
While these tools , often called the magnificent seven, are an important part of
SPC, they comprise only its technical aspects. SPC builds an environment in
which it is the desire of all individuals in an organization for continuous
improvement in quality and productivity.
Of the seven tools, the Shewhart control chart is probably the most
technically
sophisticated. It was developed in the 1920s by Walter A. Shewhart of
the Bell Telephone Laboratories.
To understand the statistical concepts that form the basis of SPC, we
must first describe Shewharts theory of variability.
In any production process, regardless of
how
well
designed
or
carefully
maintained it is, a certain amount of
inherent or natural variability will always
exist.

Dr.Gere K

Variation
There is no two natural items in any category are the same.
Variation may be quite large or very small.
If variation is very small, it may appear that items are identical,

but precision instruments will show differences.

Categories of variation
Within-piece variation
One portion of surface is rougher than another portion.
Apiece-to-piece variation
Variation among pieces produced at the same time.
Time-to-time variation
Service given early would be different from that given later
in the day.

Dr.Gere K

Sources of Variation
Equipment:-Tool wear, machine vibration,
Material:- Raw material quality
Environment:-Temperature, pressure, humidity,
lighting
Operator:-Operator performs- physical &
Types of Variation
emotional
1. Common or Chance Causes
Variation due to chance (common) causes is inevitable in any process or
product. They are difficult to trace and control even under best conditions of
production. Since these variations may be due to some inherent
characteristics of the process or machine which functions at random. W.E.
Deming contended that only management can address common cause
variation since it is inherent in the process as designed by management.
2. Assignable Causes or Special Causes

Assignable (special) cause variation posses greater magnitude as compared to

those due to chance causes and can be traced or detected. The power of
Shewhart control chart lies in its ability to separate out the assignable causes
of variations.
Dr.Gere K

Use of Control Charts

Separate common and special causes of variation
Determine whether a process is in a state of statistical control or out-ofcontrol
Estimate the process parameters (mean, variation) and assess the
performance of a process or its capability.
The control chart is an on-line process-monitoring technique widely used for this
purpose. Control charts may also be used to estimate the parameters of a production
process, and, through this information, to determine process capability. The control chart
may also provide information useful in improving the process. Finally, remember that the
eventual goal of statistical process control is the elimination of variability in the process.

Types of Data
Variable data
Product

characteristic that can be measured

Length, size, weight, height, time, velocity

Attribute data

Product characteristic evaluated with a discrete choice

Good/bad, yes/no

Dr.Gere K

How to develop a control chart

1.Select a quality characteristic to be measured
Identify a characteristic to study - for example, part length or any other variable
affecting performance.
2. Choose a subgroup size to be sampled
Choose homogeneous subgroups
Homogeneous subgroups are produced under the same conditions, by the
same machine, the same operator, the same mold, at approximately the same
time.
Try to maximize chance to detect differences between subgroups, while
minimizing chance for difference with a group.
3. Collect data
Generally, collect 20-25 subgroups (100 total samples) before calculating the control

limits.
Each time a subgroup of sample size n is taken, an average is calculated for the
subgroup and plotted on the control chart.
Dr.Gere K

Control Charts
The key instrument of SPC is the control chart invented by W. A. Shewhart in the
1920s. A control chart is a graphical comparison of performance data to
computed control limits drawn as limit line on the chart.
The primary function of control chart is to determine which type of variation is
present and whether adjustments need to be made to the process. It can be as
damaging to adjust a process which is operating in control (only common causes
variation present) as it is to fail to adjust a process which is operating out of control
(assignable causes of variation present). It is, therefore, important to be able to
determine what type of variation is present in a process.

Types of control
charts:
1. Control charts for variables
2. Control charts for attributes

Dr.Gere K

Control Charts for Variables

Control charts for variables enable the monitoring of the natural variability
occurring in a process where the data is provided in measurable units rather than
counted ones. The charts will then be used to reduce this variability around the
nominal value. Charts are based on variability due to common causes and are
used to determine the presence of special causes.
Variable data are plotted on a combination of two charts. Using X bar chart and
range (R) chart. However, S chart (standard deviation) chart should be used in
place of a range chart for larger sample size (n>10).This is because; the range
method loses efficiency relative to S2 as sample size increases. For a sample size of
2, the two methods are equivalent. Where as, for a sample size of 10, the range
method efficiency is only 0.85% relative to S2.
The X chart plots sample means. It is a measure of between sample
variations and is used to assess the centering and long-term variation of the
process. The range chart and S chart measure the with in-sample variation
and assess the short-term variation of the process.

Dr.Gere K

X Bar and R Control Charts

The constructing of an
chart begins with the collection of a serious of samples
from a process. The samples consist of two or more observations (sample size of 3
to 10 are best) each. The individual observations are averaged for each sample to
determine the sample mean ( ). The averages of at least 25 to 30 sample means are
calculated that is X-double bar ( ). The underlying distribution for the
chart is
the normal distribution. The centerline (CL) of the chart is . The upper control limit
(UCL) is set at + 3sigma; the lower control limit (LCL) is set at - 3sigma.
Appropriate placement of the upper and lower control limits is an economic issue.
The intent would be to fix the limits in such a way as to balance the economic
consequences of failing to detect a special cause when it does occur and wrongly
identifying the presence of a special cause when it has really not occurred.
Experience has led to the use of 3 limits as a good balance of these risks.
The 25 to 30 sample means are then plotted on the control chart. If none of the
points fall outside the control limits and there are no discernible patterns in the plot,
then process is said to be in control.

Dr.Gere K

10

X
i 1

(1)

i 1

( 2)

R X k ( L arg est ) X k ( smallest ) (3)

i 1

( 4)

The centerline for the X-bar chart is X-double bar. The upper control limit and lower
control limit are calculated using

UCL X A2 R (5)
CL X A2 R (6)

UCL X A2 R (7)

X 3 x ,

R
,
n
d2

3
A2
n d2

The constant A2 is tabulated for various sample sizes in

Appendix Table VI.
Dr.Gere K

11

The centerline for the range chart is

UCL D4 R (8)

R 3 R ,

CL R (9)

LCL D3 R (10)

. The UCL and LCL are calculated by:

R 3d 3

R d 3

, 1 3 d 3 D 4 , 1 3 d 3 D3

The constant d3 , and d3 is tabulated for various sample sizes in

Appendix Table VI.
The range chart is evaluated first, because the limits of the X chart depends on the
magnitude of the common causes variation of the process measured by R , if it is in
control then the X chart is evaluated. An out of control signal on either chart is an
indication that the process is out of control.

Dr.Gere K

12

UCL X A2 R 5.009 (0.577)(.115) 5.075

Centerline X A2 R 5.009
UCL X A2 R 5.009 (0.577)(.115) 4.943

Dr.Gere K

13

Dr.Gere K

14

C
o
n
t
r
o
l
C
h
a
r
t
:
X
B
a
r
O
b
s
1
.5
5
0
7
5
U
C
L
=
5
.
0
7
6
9
A
v
e
r
a
g
e
1
0
6
44
3
.0

M
ean

.5
5
0
2
5
.4
0
.4
9
7
5
.9
5
01234S
5
6
7
8
9
1
0
ig
m
a
le
v:3
Dr.Gere K

15

UCL

D R 2.115 * 0.115 0.243

4

CL

LCL

R 0.115

D R 0 * 0.115 0
3

Dr.Gere K

16

C
o
n
t
r
l
C
h
a
r
t
:
R
C
h
a
r
t
bA
sveLr1a=g.0243150
00..225 O
U
C

R
ange

00..115
00..512345678910
S
igm
alev:3
Dr.Gere K

17

and R charts when standards are given

If the mean and standard deviation of a process are known, or if they are specified
by management, perhaps as goals to be achieved X and R control charts can be
constructed without analyzing past data. For 3-sigma limits, the control limits for
X are given by:
UC L x
C Lx

3
n

LC L x

(9)
(10)

3
n

(11)

The control limits for the

chart can then be written as

UC L x A (12)
C L x (13)
LC L x A (14)

Dr.Gere K

18

UC L R R 3

C LR R
LC L R R 3

d , we can

Since = d and

UC L R d 2 3d 3

write

(15)

C LR d 2

(16)

LC L R d 2 3d 3

(17)

Defining the constants that depend only on the sample size, n, as

D d 3d
2

UC L R D2
C LR d 2
LC L R D1

D d 3d

and

We obtain

(18)
(19)
(20)
Dr.Gere K

19

X and S control charts

While X-bar and R charts are widely used, it is occasionally desirable to

estimate the process standard deviation directly instead of indirectly
through the use of the range R. This leads to control charts for X-bar and
S, where S is the sample standard deviation.
The 3-sigma limits for the S chart with the standards given are as follows
2

UC Ls c4 3 1 c4
C Ls c4

(21)
(22)

LC Ls c4 3 1 c4

(23)

Where
B c
6

( xi x )
i 1

n 1

S C4

3 1 c4
4

SD of S 1 c4

B5 c 4 3 1 c 4

The control limits can be rewritten using B6 and B5 as

UC Ls B6

( 24)

UCL A

C Ls c4

(25)

LC Ls B5

(26)

UCL A
Dr.Gere K

20

UC Ls S

3S

1 C4

(27)

C Ls S

(28)

LC Ls S

3S

1 C4

(29)

Where
B
B

1
1

1 C4

1 C4

C
The control limits can then be written as
3

UC Ls B4 S

(30)

C Ls S

(31)

LC Ls B3 S

(32)

And for the X-bar chart

UCL X A3 S
UCL X A3 S
Dr.Gere K

21

Standard is
given

Standard is given

Dr.Gere K

22

Detection of Patterns on X-bar and R

charts
Even
though, no points are above and below the control limits the process can be
said out-of-control by looking on the unnatural patterns produced by special causes.
Some of the patterns that are frequently seen are:

Cycles
Cycles are short trends in the data that occur in repeated patterns. Causes of cycles
on the X-bar chart include temperature and humidity changes, operator fatigue,
rotation of operators and electrical fluctuations. While, operators fatigue, shift (day
or night) and worn tools or dies are for the R chart. An example of a cycle is shown
in Fig. A

Mixture
In a mixture pattern, the points tend to fall near the UCL and LCL with an absence
of fluctuations near the middle. On the X-bar chart, mixtures can occur with over
control. Where as, difference in materials and measuring equipment can cause a
mixture on the R chart.

Dr.Gere K

23

Stratification
Stratification is characterized by artificial constancy. Instead of fluctuating naturally
inside limits, the points are very close to the CL. On the X-bar chart, this can be
caused by incorrect calculation of the control limits. The pattern may occur on the R
chart when the sampling process collects one unit from each of several underlying
distributions. If the largest and smallest unit in each sample are similar, unnaturally
small fluctuations will result.

Sudden Shift
A sudden shift in level is shown by an instantaneous change in one direction or the
other. On X-bar chart. On X-bar chart, a sudden shift could be caused by change to a
new type of material, new operator, new inspector, new machines and so on. On the
R chart, change in motivation of the operators, new operators and new equipment
are few of the many causes of a sudden shift.

Dr.Gere K

24

Trend
A trend is shown by a continuous movement in one direction.
On the X-bar chart, a trend is caused due to gradual
deterioration of equipment, worker fatigue and
accumulation of waste products.
On the R chart, improvement or deterioration of operators
skill work fatigue and gradual change in homogeneity of
incoming quality are come of the causes.
Interpreting unnatural patterns is a challenge for quality control personnel.
Some of the major problems associated with the analysis of control chart
patterns can be summarized as follows.
the random noise might contaminate the present pattern, the effect may
change with the magnitude of the unnatural pattern.
a pattern may sometimes resemble other patterns. For instance, a short trend
may be a subset of other patterns.
the problem is much more complicated if there is more than one pattern of
interest, or if the signal-to-noise ratio is low.
Dr.Gere K

25

 This shows pattern recognition is a crucial problem in statistical

process control. Over the years, the zone test or run tests have
been the major tool for interpreting control charts.

Although the zone tests or run tests have been proven to be effective in
detecting out-of-control situations the interpretation of process data is
still a very difficult task.
The major difficulty lies in the fact that there is no one-to-one mapping
between a supplementary rule and an unnatural pattern. In practice, the
types of unnatural patterns that a process may experience are not
known in advance.

Dr.Gere K

26

There might be several patterns associated with a

particular rule. For instance, the possible patterns
associated with the rule eight points in row on
both sides of the centerline with none in zone C
suggested by Nelson (1985), could be a mixture
or a systematic variation. These tests might
indicate that an unnatural pattern is present,
but do not explicitly indicate which pattern
really occurs. In addition, some of the supplementary
rule results in more false alarms without significantly
improving the performance of control charts. A typical
example is the commonly used trend rule. Davis and
Woodall (1988) evaluated the trend rule and concluded
that the trend rule is not effective in detecting a
linear trend. The limitations of supplementary tests
have motivated interests in developing algorithms
K
27
based neural Dr.Gere
networks
for process data analysis.

Dr.Gere K

28

Working Rules: Pattern Analysis Tests (PAT)

PAT 1: One point plots beyond zone A on either side of the mean.
PAT 2: Nine points in a row plot on the same side of the mean.
PAT 3: Six consecutive points are strictly increasing or strictly decreasing.
PAT 4: Fourteen consecutive points which alternate up and down.
PAT 5: Two out of three consecutive points plot in zone A or beyond, and all three
points plot on the same side of the mean.
PAT 6: Four out of five consecutive points plot in zone B or beyond, and all five
points plot on the same side of the mean.
PAT 7: Fifteen consecutive points plot in zones C, spanning both sides of the
mean.
PAT 8: Eight consecutive points plot at more than one standard deviation away
from the mean with some smaller than the mean and
Dr.Gere K

29

Dr.Gere K

30

Process capability
Process capability compares the output of an in-control process to the specification
limits by using capability indices. The comparison is made by forming the ratio of
the spread between the process specifications (the specification "width") to the
spread of the process values, as measured by 6 process standard deviation units (the
process "width").
Process Capability Indices
We are often required to compare the output of a stable process with the process
specifications and make a statement about how well the process meets
specification. To do this we compare the natural variability of a stable process with
the process specification limits.
A capable process is one where almost all the measurements fall inside the
specification limits. This can be represented pictorially by the plot below:

Dr.Gere K

31

There are several statistics that can be used to measure the capability of a process:
Cp, Cpk, Cpm. Most capability indices estimates are valid only if the sample size used
is 'large enough'. Large enough is generally thought to be about 50 independent data
values.
The Cp, Cpk, and Cpm statistics assume that the population of data values is normally
distributed. Assuming a two-sided specification, if and are the mean and standard
deviation, respectively, of the normal data and USL, LSL, and T are the upper and
lower specification limits and the target value, respectively, then the population
capability indices are defined as follows:

Dr.Gere K

32

Definitions of various process

capability indices:

To get an idea of the value of the Cp statistic for varying

process widths, consider the following plot

Dr.Gere K

33

Example
The design specifications for component are 100
0.5.
Whereas the process report shows that process average is
99.9mm and standard deviation is 0.18. Do these figures
call for any action by any one?
Solution
USL = 100.5 mm
LSL = 99.5mm
X-bar = 99.9mm
= 0.18
Cp = 1/6
= 0.925
Defective products will always be there. It is
therefore necessary to take action to
reduce the number of defectives produced.
Dr.Gere K

34

Control Charts for Moving Averages and

Ranges
Many organizations are involved in continuous processes, such as manufacturing
steel, aluminum, paint, oil or chemicals. In such cases it is recognized that the ( , R)
charts is inappropriate. So, in continuous processes extensive use is made of moving
average charts.
UCL X 3

LCL X 3

UCL D4 MR

MR

MR

d
U=34.8
CL=33.52
L=32.24
2

LCL D3 MR

UCL = 1.57
CL= 0.48
LCL = 0

Example
The viscosity of an aircraft primer paint is an important quality characteristic. The product is
produced in batches and as each batch takes several hours to produce, the production rate is too
slow to allow sample sizes greater than one. The viscosity of the previous 15 batches is given
below

Dr.Gere K

35

Viscosity of Aircraft Primer

Paint
Moving Range
Batch # Viscosity (x) (MR)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

33.75
33.05
34.00
33.81
33.46
33.02
33.68
33.27
33.49
33.20
33.62
33.00
33.54
33.12
33.84

0.70
0.95
0.19
0.35
0.56
0.34
0.41
0.22
0.29
0.42
0.62
0.54
0.42
0.72

33.46

0.48

Dr.Gere K

36