12.

6 Statistical Process Control

Figure 12.5 Three Zones in Control Charts

CONTROL CHART FOR VARIABLE DATA

We will look at the following types of control charts using variable data.
∑ x bar and range chart (we can also plot standard deviation instead of range)
∑ x and moving range chart

X Bar and Range Chart

We plot this chart in the following manner. Samples from regular production are picked up at random. The
minimum number of samples or sample size should be greater than or equal to four to get consistent
results. Usually the samples are picked up at regular intervals in a shift or in a day or more. Usually 25
such samples are picked up. Measurements are to be carried out on a selected parameter and then listed
sub-group wise. For each sub-group, the average and range are to be calculated.
Example 12.1
A jewellery shop was making 10 grams gold coins for sale. At regular intervals, four coins each were
selected at random. The weights of the gold coins were measured. The table below indicates the measured
values of 25 sub-groups of gold coins (called x). Against each column, the x bar values were calculated,
which are nothing but the average of the four values. Just following that is the range of each sub-group,
which is nothing but the difference between the largest value and the lowest value.
Values of gold coin — x
Sub group 1 2 3 4 5 6 7 8 9 10 11 12 13
10.1 9.9 10.2 10 10.6 10 10 10.1 10.3 10.1 9.8 10.1 10
9.9 10 10.1 9.8 9.9 9.9 9.8 9.9 10 10 10.1 10.2 9.6
9.9 9.9 9.9 10.2 10.2 10.1 10.1 9.9 9.9 10 10 10 10.2
10 10.1 10 10 10 9.9 9.9 9.9 10 10.1 10.1 10 10.1
 Control Charts for Improving Process Capability 12.7

Calculated x bar
9.975 9.975 10.05 10 10.175 9.975 9.95 9.95 10.05 10.05 10 10.075 9.975
Calculated range
0.2 0.2 0.3 0.4 0.7 0.2 0.3 0.2 0.4 0.1 0.3 0.2 0.6
Values of gold coin — x
Sub group 14 15 16 17 18 19 20 21 22 23 24 25
10.2 9.9 10.3 9.9 10.6 10.2 9.8 10.1 10.3 10.1 9.8 10.1
9.8 10 10.1 9.8 9 9.9 9.8 9.9 10.1 9.9 10.1 10.2
9.7 10 9.2 10.2 10.2 10.1 10.1 9.9 9.9 10 9.7 9.6
10 10.1 10 10.1 9.8 9.9 9.9 10.1 10 10.1 10.1 10
Calculated x bar
9.925 10 9.9 10 9.9 10.025 9.9 10 10.075 10.025 9.925 9.975
Calculated range
0.5 0.2 1.1 0.4 1.6 0.3 0.3 0.2 0.4 0.2 0.4 0.6
Now, we have to calculate the average range —R bar. This is nothing but the average of the ranges of
the 25 sub-groups.
R bar = 0.412
Now, we have the calculate the Upper Control Limit (UCL) for the R chart.
UCL = D4 ¥ R bar
The Lower Control Limit (LCL) is calculated using the following formula:
LCL = D3 ¥ R bar
We have to get the values of D3 and D4 from Table C. Since D3 is zero for a sample size of 4, LCL will
also be equal to zero.
UCL = D4 ¥ Rbar = 0.9394, LCL=0
Now, we plot the R chart, which is given in Fig.12.6:

Figure 12.6

We now find that the range R in the chart exceeds the control limits in respect of two samples namely
sub-group number 16 and 18. This could be due to assignable causes or in some cases measurement errors.
Therefore, these two sub-groups are to be eliminated. Data after removing outliers are given below:
12.8 Statistical Process Control

First revision after eliminating outliers

10.1 9.9 10.2 10 10.6 10 10 10.1 10.3 10.1 9.8 10.1
9.9 10 10.1 9.8 9.9 9.9 9.8 9.9 10 10 10.1 10.2
9.9 9.9 9.9 10.2 10.2 10.1 10.1 9.9 9.9 10 10 10
10 10.1 10 10 10 9.9 9.9 9.9 10 10.1 10.1 10
x : 9.975 9.975 10.05 10 10.175 9.975 9.95 9.95 10.05 10.05 10 10.075
R: 0.2 0.2 0.3 0.4 0.7 0.2 0.3 0.2 0.4 0.1 0.3 0.2
10 10.2 9.9 9.9 10.2 9.8 10.1 10.3 10.1 9.8 10.1
9.6 9.8 10 9.8 9.9 9.8 9.9 10.1 9.9 10.1 10.2
10.2 9.7 10 10.2 10.1 10.1 9.9 9.9 10 9.7 9.6
10.1 10 10.1 10.1 9.9 9.9 10.1 10 10.1 10.1 10
x : 9.975 9.925 10 10 10.025 9.9 10 10.075 10.025 9.925 9.975
R : 0.6 0.5 0.2 0.4 0.3 0.3 0.2 0.4 0.2 0.4 0.6
We will now calculate R bar and UCL and LCL.
R bar = 0.3304
UCL = D4 ¥ Rbar = 0.7534, LCL = 0
The R chart is plotted again and given in Fig.12.7:

Figure 12.7

Now, there are no outliers. Therefore, the R bar (central value) and UCL and LCL are noted and frozen.
It is now time to plot x bar chart. The x bar chart is nothing but a graph of x bar on the y-axis and sub-group
numbers in the x-axis. The calculation of x bar bar (grand average) and UCL and LCL for the x bar chart
are given below:
Grand average (x bar bar) = 10.002
UCL for x bar = x bar bar + A2 ¥ R bar = 10.243
LCL for x bar = x bar bar – A2 ¥ R bar = 9.761
The grand average is the average of the 23 sub-groups (since we have removed two outliers in the range
chart). The values for A2 is to be obtained from Table B.
The x bar chart is given in Fig.12.8.
Let us now summarize the steps involved for arriving at x and R charts.
1. Decide on the parameter to be controlled.
 Control Charts for Improving Process Capability 12.9

x bar chart
10.2
10.1

Range
10
9.9
9.8
9.7
1 3 5 7 9 11 13 15 17 19 21 23
Sub-group number

Figure 12.8

2. Pick up at least 4 samples at a time and measure the performance on the selected parameter as per
step 1; pick up the samples 25 times.
3. Find x and R for each subgroup.
4. Find R as an average of R of the 25 subgroups.
5. Calculate UCL and LCL for R chart, taking D4 and D3 from Table C.
UCL R = D4 R bar
LCL R = D3 R bar
6. Plot the R-values.
7. Check whether any R-values lie outside UCL or LCL.
8. If all R-values are within limits, go to step 11.
9. If upto two R-values are outside the limits, eliminate the respective subgroups and repeat steps 4 to
8.
10. If more than two R-values lie outside limits initially or if some values go out of limits after
recalculation in step 9 stop. Find out special causes and take action to improve the process. Start all
over again after the process is stable.
11. The grand average x bar bar is the average of the remaining sub-groups. Calculate x excluding the
subgroups, if any whose R were out of limits.
12. Calculate LCL, UCL, action limits as well as warning limits for x and R Charts.
UCL x bar = x bar bar + A2 R bar.
LCL x bar = x bar bar – A2 R bar.
Now, plot the x bar chart. Check that sample averages ( x ) of all the sub-groups lie within the
control limits.
Now, let us take one more example of x bar and R control chart.
Example 12.2
A manufacturer of 300 grams weight wants to carry out a process capability study. For this purpose, he has
picked up 25 sub-groups of 4 samples of weights manufacturing over the period of 2 days. Values of 300
grams weights are given:
1 2 3 4 5 6 7 8 9 10 11 12 13
301 299 302 300 306 300 300 301 303 301 295 301 300
299 300 301 298 299 299 298 299 300 294 301 302 296
299 299 299 302 302 301 301 299 295 310 300 300 302
300 301 300 302 307 300 301 302 300 302 312 301 304
12.10 Statistical Process Control

The corresponding x bar and the range of each group are given below:
x bar
299.75 299.75 300.5 300.5 303.5 300 300 300.25 299.5 301.75 302 301 300.5
range
2 2 3 4 8 2 3 3 8 16 17 2 8
Values of 300 grams weights – x
14 15 16 17 18 19 20 21 22 23 24 25
302 299 303 299 306 302 298 301 303 301 302 301
298 300 301 298 298 299 298 299 301 299 303 301
297 300 298 302 302 301 301 299 299 300 302 301
304 301 301 301 304 304 302 302 301 301 302 302
x bar
300.25 300 300.75 300 302.5 301.5 299.75 300.25 301 300.25 302.25 301.25
range
7 2 5 4 8 5 4 3 4 2 1 1
The R bar, UCL and LCL for the above data are calculated and given below:
R bar = 4.96
UCL = D4 ¥ Rbar = 11.309, LCL = 0

Figure 12.9

It is found that the range of sub-groups 10 and 11 are beyond the UCL. Therefore, these two outliers are
eliminated and given below:
First revision after eliminating outliers
301 299 302 300 306 300 300 301 303 301 300 302 299
299 300 301 298 299 299 298 299 300 302 296 298 300
299 299 299 302 302 301 301 299 295 300 302 297 300
300 301 300 302 307 300 301 302 300 301 304 304 301
x : 299.75 299.75 300.5 300.5 303.5 300 300 300.25 299.5 301 300.5 300.25 300
R: 2 2 3 4 8 2 3 3 8 2 8 7 2
 Control Charts for Improving Process Capability 12.11

303 299 306 302 298 301 303 301 302 301
301 298 298 299 298 299 301 299 303 301
298 302 302 301 301 299 299 300 302 301
301 301 304 304 302 302 301 301 302 302
x : 300.75 300 302.5 301.5 299.75 300.25 301 300.25 302.25 301.25
R: 5 4 8 5 4 3 4 2 1 1
The R bar, UCL and LCL for the range chart calculated after revision are given below:
R bar = 3.9565
UCL = D4 ¥ Rbar = 9.0209 LCL = 0

Figure 12.10

Now, there are no outliers. Hence, we will now proceed to plot the x bar chart. But, before that we have
to calculate the grand average and the control limits. The calculation is given below:
x bar bar = 300.65
UCL = X bar bar + A2 ¥ R bar = 303.54
LCL = X bar bar – A2 ¥ R bar = 297.76

Figure 12.11

In the x bar chart also, there are no outliers.

12.12 Statistical Process Control

x bar and R charts are quite useful for study of process performance when we collect data as variables. The
control chart indicates whether the process is in control. It indicates clearly the variation due to assignable
causes. The points we were eliminating in the subsequent iterations in the worked examples can be due to
assignable causes or error in measurements.
Warning and Action Zones
The warning zone starts at x ± 2 s/ n or
x ± 2/3 A2 R bar
The action zone starts at x + s/ n or
x ± A2 R bar
Thus the action zone starts at UCL and LCL in the average chart. The warning line starts at 2/3 of UCL
and LCL.
Example 12.3
In our worked example 12.1, in the x chart action zone lies above 10.24 and below 9.76.
The warning zone starts at 10.18.
Similarly, the lower warning limit starts at 9.84.
Therefore, when the results fall above 10.24, but below 9.76, then we have to analyze and take action.
The warning and action zones for x chart are illustrated in Fig. 12.12:

Warning and Action Zones

Figure 12.12

Control limits for Range

Known R Known s
1
Upper action line D .001 R D 0.001s
1
Lower action line D .999 R D0.999 s
1
Upper warning line D .025 R D0.025 s
1
Lower warning line D .975 R D0.975 s
 Control Charts for Improving Process Capability 12.13

The values of constants can be obtained from Table C. For the worked example 12.1:
Upper action line = 2.57 ¥ 0.33 = 0.85
Lower action line = 0.1 ¥ 0.33 = 0.03
Upper warning line = 1.93 ¥ 0.33 = 0.64
Lower warning line = 0.29 ¥ 0.33 = 0.1
We know UCL = 0.75, LCL = 0.
The control lines for the worked example are given in Fig. 12.13:

R Chart
Figure 12.13

Let us look at another example to reconfirm our understanding.

Example 12.4
The peanut manufacturer has observed the following at the end of the day.
No. of sub-groups = 20
Size of sub-group = 5
S x = 6000
S R = 40
Draw the control limits for x and Range.
Step 1 : Find out x
x = 6000/20 = 300
Step 2 : Find out R
R = 40/20 = 2
x Chart
UCL = x + A2 R
= 300 + 0.58 ¥ 2
= 301.16
LCL = 300 – (0.58 ¥ 2)
= 298.84
12.14 Statistical Process Control

Upper warning line = x + 2/3 A2 R

= 300 + 0.39 ¥ 2 = 300.78
Lower warning line = 300 – 0.78
= 299.22
Now we can draw the control limits for x

Action
UCL
301.16
Warning Zone
300.78

300 x

299.22
Warning Zone
LCL
298.84
Action

x Chart
Figure 12.14

R Chart
UCL = D4 R = 2.11 ¥ 2 = 4.22
Upper Action Line = 2.34 ¥ 2 = 4.68
Upper Warning Line = 1.81 ¥ 2 = 3.62
Lower Warning Line = 0.37 ¥ 2 = 0.74
Lower Action Line = 0.6 ¥ 2 = 0.32
LCL = 0, since D3 = 0
The control limits for R Chart are shown in Fig. 12.15:
Indications of Abnormality in the Process The process is said to be out of control, when the performance
lies outside ± 3 s limits (UCL or LCL) in the x bar chart. The performance of the variable should preferably
lie between ± 2 s. However, the performance of some patterns as given below calls for immediate action
to analyze the causes and taking corrective actions:
Performance within ± 2 s
1. Seven consecutive points showing a continuously increasing or decreasing pattern.
2. Seven points remaining close to the central line.
3. Seven consecutive values showing either a continuous upward trend or downward trend.
4. Values showing a cyclic pattern.
 Control Charts for Improving Process Capability 12.15

Action

4.68

Warning Zone UCL

4.22

R 3.62
2 R
0.74 Lower Warning Line

0.32 Lower Action Line

Action Zone LCL

0

R Chart
Figure 12.15

Performance in the warning zone (outside ± 2 s , but within ± 3 s )

5. No incidence of two consecutive points in the warning zone.
6. Not more than four points in warning zone.

PROCESS CAPABILITY INDICES

The fundamental requirement of any process is that it should be stable first. Stability is indicated by
consistent performance of the process within the limits set. Only variations allowed are the common cause
variations. Therefore, statistical control implies performance of the process within the set limits. Only
thereafter, the process will be predictable. Therefore, capability can be assessed only on a process, which
is consistently stable over a period of time.
The process is abnormal when there are variations due to special or assignable causes. The abnormal
process is out of control. When there are special causes, the process should be stopped to find out the
special causes and eliminating them.
We will study the capability of the process, which has variations only due to common causes. The
process capability indices are deduced to find a relationship between variations due to common causes and
tolerance.
Recall the areas under the normal curve. We know that the area under the curve between ± 2 s is
95.45%. This means 4.55% of the curve is outside this total 4 s limits. If the tolerance of the process is
equal to ± 2 s, then 4.55% of the products will be defective. If the tolerance is equal to ± 3 s, then 0.27%
of the products will be defective. The tolerance limits are actually set by customer and what is under the
manufacturer’s control is the process performance and statistical control limits.
The process capability has two indices namely Cp and Cpk.

Cp
Cp = USL – LSL/6 s
12.16 Statistical Process Control

Where USL : Upper Specification Limit

LSL : Lower Specification Limit
s = R /d2
Let us consider various possibilities with regard to Cp.
Cp < 1: This means that the normal curve extends beyond ± 3 s limits (UCL and LCL). Therefore, the
process is incapable. When Cp > 1, it becomes increasingly capable, provided the grand average is
maintained at (USL + LSL)/2.

Cpk
Cp can be applied only when the process is correctly centered about the mid specification since it takes
into account the precision with the total tolerance. To know how well a process is performing both accuracy
and precision should be compared. Cpk takes into account both the degree of random variation and accuracy.
This is the most popular process capability index.
Cpk = minimum of (USL – x /3 s, x – LSL/3 s)
When Cpk < = 1, the variations and centering may cause infringing into one of the tolerance limits and
hence process not capable. If the process means coincides with mid specifications, then both Cpk and Cp
will give the same value. Cpk gives additional information about the centering. Therefore, it is also called
process performance index. Hence, increasing value of Cpk means that the process is increasingly becoming
capable.
Example 12.5
Look at worked example 12.1 after eliminating outliers.
Let USL = 10.5 and LSL = 9.5
x = 10.002
R = 0.3304
s = R /d2 (take value of d2 from Table B)
= 0.3304/ 2.059 = 0.13
Cp = 10.5 – 9.5/6 (0.13) = 1.28
USL – x = 10.5 – 10.002 = 0.498
x – LSL = 10.002 – 9.5 = 0.502
Therefore, Cpk = 0.498/3 (0.13) = 1.27
Since Cpk is almost equal to Cp, we can conclude that the process is nearly centered. Since they are greater
than 1, it is capable provided we maintain grand average at mid-specification limits.
Example 12.6
Let the specification limit be 300 ± 10 for data is Example 12.2
x = 300.65
s = R /d2 = 3.9565 / 2.059 = 1.9
Cp = 20 / 6 ¥ 1.9 = 1.8
x – LCL /3 ¥ 1.9 = 300.65 – 290 / 5.7 = 1.9
 Control Charts for Improving Process Capability 12.17

Or
USL – x /5.7 = 310 – 300.65 / 5.7 = 1.7
\ Cpk = 1.7
The process is capable but not centered as the data also reveals.
Example 12.7
Given sample size = 4 and 25 samples were considered. The process is under control.
x = 105
USL = 110
LSL = 90
s=2
Cp = 20 /6 ¥ 2 = 1.66
USL – x is the lowest.
Therefore, Cpk = 110 – 105/3 ¥ 2 = 5/6 = 0.83
Since, Cp is 1.66, the process could be highly capable. But Cp = 0.82 indicates that the capability has
been reduced due to shifting of the mean to the right. There may be processes, which are under statistical
control, meaning that the variation is due to random causes. But the process may not be capable as inferred
from the Cpk value. This may be overcome by widening the tolerance limits, provided the customer accepts
it. If not, the rejects will be high which have to be weeded out before supply. Irrespective of sample size,
Cpk < 1 indicates an incapable process. If Cpk greater than 1, but upto 2, we have to confirm the process
capability through more samples. Cpk > 3 is definitely a capable process.

X and MR Chart
x and R chart requires at least four items in each sub-group. In some industries such as chemical industry
or when the cost of samples is high, we can have only one item per subgroup. In such cases, we cannot find
the range R. In such situations, we can use moving range (MR) chart. Moving range is the difference
between the value and the one immediately preceding it. In this case, we use the X and MR chart. Thus,
there are two sets of data, one is the measurement called ‘X’ and the other is the moving range called
‘MR’.
The parameters of X and MR charts are given below:

Parameter X MR
Centre x MR
UCL x + 3 ( MR /1.13) MR ¥ 3.27
LCL x – 3 ( MR /1.13) 0

Example 12.8
The 11 subgroups of 1 sample each had values as indicated below:
100 101 100 102 100 99 100 98
99 100 101
Calculate control limits for MR chart.
12.18 Statistical Process Control

Solution
Let us first calculate the average of the values x = 100
We will get 10 moving ranges for 11 subgroups or values. The absolute value of the difference between
one value and the next is to be calculated. Thus MR will be:
1 1 2 2 1 1 2 1 1 1
\ MR = 13/10 = 1.3
UCL MR = MR ¥ 3.27
= 1.3 ¥ 3.27 = 4.251
LCL MR = 0
Now, we can calculate control limits for the X chart.
UCL = x + 3 ( MR /1.13)
= 100 + 3 (1.3)/1.13 = 103.4
LCL = x – 3( MR /1.13)
= 96.6
Now we can plot the values in X and MR Chart.
Example 12.9
The temperatures of a steam bath were recorded and shown below (x values). Plot x and MR chart,
x values
100 101 100 102 103 101 100 99 101 102 103 102 101 100
101 102 102 103 101 99 98 100 101 102
The moving range is the difference between the adjacent values (MR). The moving ranges are calculated
and given below:
MR values
1 1 2 1 2 1 1 2 1 1 1 1 1 1
1 0 1 2 2 1 2 1 1
Then, we calculate X bar, which is the average of the x values. The average of MR values which is MR
bar. The control limits for MR charts are given below:
UCL = MR bar x D4
Note:
Since the sample size is one always, in the moving range chart, we take D4 as 3.27 corresponding to a
sample size of 2.
LCL = 0
For x chart,
UCL = x bar + 3 (MR bar/1.13)
LCL = x bar – 3 (MR bar/1.13)
The charts are given in Fig.12.16:
UCL MR 3.9809
LCL 0
 Control Charts for Improving Process Capability 12.19

Figure 12.16

Since there are no outliers, we do not have to eliminate any points. Now, we can plot x chart and see
whether there are any outliers.
x bar = 101
MR bar = 1.2174
UCL X = 104.23
LCL X = 97.768
x Chart
104
102
100
Series 1
x

98

96

94
1 4 7 10 13 16 19 22
Values
Figure 12.17

This chart is useful in the following situations:

∑ Samples are more expensive
∑ The data is accumulated slowly over a period of time
∑ Each lot consists of only one sample

CONTROL CHARTS BY ATTRIBUTES

The control chart by variables is quite useful to study the performance against individual characteristics of
products. The control charts are to be plotted for each quality characteristics. An entity may contain
hundreds of quality characteristics. This means that we need to plot one chart for each characteristic. This
may neither be feasible nor desirable in the normal circumstances. Hence, this chart is more expensive. At
the same time, the usefulness of the chart should not be undermined because this gives micro level analysis
of the quality of the products manufactured. This can be used selectively.
12.20 Statistical Process Control

When a quality problem is found and if the organization wants to carry out further analysis, then
specific characteristics of the entity can be measured and controlled using X bar and R control charts.
Usually, the organizations collect data on the number of non-conforming products. This data is readily
available. Hence, it is rather easier to prepare control charts by attributes. The attribute charts are likely to
be less expensive. Furthermore, the top management will be interested in the overall results and hence
attribute charts fit the bill. Thus, the control chart by attributes could be the starting point in every
organization. Here, we count the occurrences of failures or defects in a product or number of defectives,
percentage of non-conforming products, etc. The x bar and R chart is based on normal frequency distribution.
However, the attributes charts are based on distributions as given below:
Binominal Distribution Poisson Distribution
np chart c chart
p chart u chart
Let us now look at the charts based on binominal distribution. Here, we count the number of non-
conforming products, i.e. the number of defectives. The np chart helps us to analyze the quality of the
process by looking at the number of non-conforming entities. The p chart on the other hand looks at the
fraction rejected due to non-conformances to specifications or fraction defectives.
Let us first look at the np chart.

np Chart
The np chart is used when the sample size is constant. Such conditions occur in manufacturing organizations.
The organization may carry out inspection of constant size of products and records the number of defective
units in each sample.
Example 12.10
The following table indicates the number of defectives in a toilet soap manufacturing organization. Sample
size is constant and equal to 100. The steps involved in finding out the control limits are given below:
∑ Collect data about the number of defective units in each sample. The sample size may be 100 or any
convenient number, which is called n. The number of defectives in each subgroup is called np.
∑ Collect these data periodically say 25 times. The number of samples, i.e. the number of sub-groups
= N.
∑ Calculate the average number of defectives – np bar, which is equal to Snp/N.
∑ Calculate p bar, which is the average fraction defective. This is equal to np bar/n.
∑ Now, the central line is np bar.
∑ The control limits are np bar +/– 3 ¥ ÷np bar (1–p bar)
np = defectives 0 1 5 6 4 2 8 2 1 5 3 4 6 1 0 5 7 2 6 1
np bar = 3.45
p bar = 0.0345
UCL = 8.925288
LCL = 0
The np chart plotted as per Fig. 12.18 is a graph of fraction defective vs. time or the sample number. We
can also plot percent defective chart using the same data. The differences lie in the following:
∑ Percent defective is 100 ¥ fraction defective
∑ In the control limits, we will have (100 – p bar) instead of (1 – p bar).
 Control Charts for Improving Process Capability 12.21

Figure 12.18

p chart
In the above example, if the sample size is not constant, then we can plot a p chart. The p chart is more
popular and is widely used. This chart is used to plot the fraction rejected. The fraction-rejected p is
defined as the ratio of number of non-conforming items to the total number of items inspected. As the name
suggests, it is a fraction less than 1.
The p chart can also be plotted as per cent rejected which will be hundred times p.
The steps involved in finding out the control limits are given below:
∑ Collect the data about the number of defective units in each sample of unequal size. The sample size
may be noted along with the number of defective units.
∑ Calculate p (per cent or fraction defective). When the number of defects are divided by the number
of samples and expressed as a fraction or a percentage, it is called p.
∑ The individual size of samples is equal to n. n is not a constant and it may vary from sample to
sample.
∑ Collect these data periodically say 25 times. The number of samples is equal to N.
∑ Average sample size is equal to Sn / N.
∑ Average fraction or per cent defective, p bar is equal to Sp/N.
∑ Now, the central line is p bar.
∑ The control limits for fraction defective chart are p bar ± 3 ¥ ÷p bar(1–p bar)/n bar.
∑ The control limits for per cent defective chart are p bar ± 3 ¥ ÷p bar(100–p bar)/n bar.
∑ If the lower control limit is less then zero, then LCL = 0.
Example 12.11
In a cell phone manufacturing plant samples were taken from each day’s production and tested and number
of defectives found each day was recorded. It is given below:
No of defectives 2 1 5 1 4 5 2 3 1 0 0 2 5 4 1 3 2 1 5 0
Sample size 50 55 80 70 90 60 72 80 90 50 81 92 55 63 70 59 58 62 70 75
percent defectives 4 1.8 6.3 1.4 4.4 8.3 2.8 3.8 1.1 0 0 2.2 9.1 6.3 1.4 5.1 3.4 1.6 7.1 0
Average sample size = 69
p bar = 3.51224
UCL = 10.15595
LCL = 0
The p chart follows.
12.22 Statistical Process Control

Percent defective chart

10

Percent defective
8
6
4
2
0
1 3 5 7 9 11 13 15 17 19
Sample

Figure 12.19

The selection of a p chart or np chart depends on the convenience. If the number of items inspected
varies from sample to sample, then p chart is ideal. However, if the sample size is constant, then np chart
will be useful.

c CHART
The control chart by variables namely X bar and R chart is plotted for chosen quality characteristics,
which is measured on the continuous scale. The p or np charts are useful for characterizing the number of
defectives. The control charts for non-conformities is called c chart. It is useful when a non-conforming
item contains one or more non-conformances, i.e. it contains one or more defects. When it is necessary to
study the total number of non-conformances in a product or a group of equal number of similar products,
we can use the control chart techniques based on Poisson distribution. The c and u charts are based on
Poisson distribution. The c chart is applicable when the number of products inspected is a constant. Usually
it is one item. The variable c is the number of defects found in the constant sample size. When the sample
size varies, i.e. when the opportunity for occurrence of non-conformities change from sample to sample
then a u chart is used.
c charts are used for understanding the number of defects in a specific portion of the population. For
instance, this can be used to analyze the number of defects in a specified area like 10 cm ¥ 10 cm in a
printed circuit board. The requirement is that the samples should have the same area or dimensions. For
instance, the contamination in one litre of water can be studied using a c chart by drawing samples
periodically.
The steps involved in finding out the control limits are:
∑ Collect data about the number of defects per unit of the same size. Collect 25 such samples. Call
them c.
∑ Calculate c bar, which is Sc/n.
∑ The control limits for c are c bar ± 3 ¥ ÷c bar.
Example 12.12
An automobile company has a painting section. The number of defects were counted in an area of 1 metre
by 1 metre of the chassis. The number of defects for the same area in different samples is listed in table
below. Plot a c chart.
C, no . of defectives 0 1 5 6 4 2 8 2 1 5 3 4 6 1 0 5 7 2 6 1
 Control Charts for Improving Process Capability 12.23

Solution
c bar = 3.45
UCL = 9.022253
LCL = –2.12225 (treat as zero)
C chart
10

6
c
Series 1
4

0
1 3 5 7 9 11 13 15 17 19
Sample number

Figure 12.20

u Chart
The u chart is quite useful in software industry. It is more flexible than the c chart. In a u chart, the unit
inspected is not a constant. Therefore, we normalize the defects for a standard unit.
Example 12.13
The number of defects found in the software program and the total number of lines of program are given in
the table below. Plot a u chart.
No. of defectives 2 1 5 1 4 5 2 3 1 1 1 2 5 4 1 3 2 15 2
Sample size (kilo
lines of code) 5 5 5 5 4 4 4 4 4 4 4 4 4 5 5 5 5 55 5
Solution
The number of defects and sample size are indicated in the above table. We have to normalize the above
by finding out defects per kilo line of code. The normalized defects per unit are given below. The central
line is u bar.
u bar = Su/N
where N = No. of samples
However, we have a difficulty in finding out the control limits. Here, we will have multiple control
limits depending upon each sample size. The general formula is:
Control Limit = u bar ± 3 ¥ ÷u bar /n
Therefore, while drawing the control limits corresponding to each sample size, we have to determine
both UCL and LCL for the size of the sample. In the above example, there are only two sample sizes.
Therefore, we have to calculate the control limits for the sample size of 4 and 5 separately. Once this is
done, we can plot the u as well as the control limits.
12.24 Statistical Process Control

Defects per unit u 0.4 0.2 1 0.2 1 1.3 0.5 0.8 0.3 0.3 0.3 0.5 1.3 0.8 0.2 0.6 0.4 0.2 1 0.4
U bar 0.57
UCL for 5 samples 1.6
LCL for 5 samples –1 treated as zero
UCL for 4 samples 1.7
LCL for 4 samples –1 treated as zero
The u chart is given below along with the control limits.

Figure 12.21

Comparison between control chart by variables and by attributes

Variable Charts Attribute Charts
Chart type X bar and R or s , x and MR np, p, c, u
Type of data Measured on a continuous scale Number of defectives or defects
Advantages Useful for improvement of product quality Data collection is easy and economical
characteristics
Detailed information about chosen characteristics Provides a holistic view
of a process or product
Focused on specific parameters Data readily available in every organization
Suitable for detailed analysis Easily understandable since it is pass or fail
criteria only
Disadvantages Cannot be used with go/no go type inspection Not suitable for control and improvement of
individual parameters of processes or products
Expensive Severity of defects not visible
Not detailed enough
Application Control and improvement of individual parameters Control of number of defects in a unit or
of processes such as temperature, dust, etc. and number of defectives—usually product
every characteristics of products such as length, oriented.
noise, gain, etc.
 Control Charts for Improving Process Capability 12.25

The need for control charts:

∑ Provides information about process capability
∑ Provides a basis for predicting future performance
∑ Provides a basis for measuring improvement
∑ Control charts help in getting the process under control and thereby prevent defects from occurring.
∑ In accordance with Juran’s chain reaction, the above leads to improving quality, productivity and
increased market share.
∑ Prevents unnecessary process adjustments and enables necessary process adjustments
∑ Provides a lot of diagnostic information
∑ Helps in identifying assignable causes immediately
∑ Gives feedback to the process owners
∑ Helps the management to keep the processes under control and effect savings
∑ Enables highest ROI.
The summary of control charts discussed in this chapter is:

Variables
Chart Name Measures plotted Centre line Control limits Remarks
x bar Average of individual x bar bar x bar bar x bar bar
samples—x bar ± A2 × R bar S xbar/N
(N = No. of samples)
R Range of each sample—R R bar UCL = D4 × R bar R bar = SR/N
LCL = D3 × R bar N = No. of samples
x Individual values—x x bar x bar ± 3 (MR bar/1.13) x bar = Sx/n
MR Moving range—MR MR bar UCL = MR bar ¥ 3.27 MR bar = SMR/n-1
LCL = 0 (n = no. of values)

Attributes
Chart Name Measures plotted Centre line Control limits Remarks
np np – number of np bar np bar ± 3 ÷np bar n – sample size
defectives in samples (1–p bar) N – No. of samples
of constant size of n np – defects in the
individual
sub-groups
np bar = Snp/N
p bar = np bar/n
p p – fraction defectives in p bar p bar ± 3 ÷p bar p bar = S p/N
samples of varying sizes (1–p bar)/n bar
N = No. of
samples in the data
n bar = average
sample size
(Contd.)
12.26 Statistical Process Control

(Contd.)
Chart Name Measures plotted Centre line Control limits Remarks
c c – No. of defects or c bar c bar ± 3 ÷c bar c bar - average
number of defects
deviations in samples in sample of
of constant size constant size
u u – No. of defects per u bar u bar ± 3 ÷u bar/n u – defects per
sample
unit in the samples of For each unique n we
variable sizes have to calculate the u bar = Su/N
limits n – sample size
N – No. of samples

GUIDANCE FOR SELECTION OF CHARTS

Guidance for selection of appropriate control chart is given in Fig.12.22:

Figure 12.22

This chapter gives the most important tools to assess the process capability. The aim of TQM is continuous
improvement of every process in the organization. The central limit theorem states that the sample averages
distribute more closely around m than the individual values of the population. Irrespective of the underlying
distribution of the population, if four or more samples are drawn at random from the lot, the sample