Chapter         23

Statistical Quality Control

23.1 Introduction
Control charting is one of the tools of statistical quality control (SQC). It is the most technically
sophisticated tool of SQC. Walter A. Shewhart (1920) of the Bell Telephone Labs developed it
as a statistical tool to study the manufacturing process variation for the purpose of improving
the economic effectiveness of the process. These methods are based on continuous monitoring
of process variation.
A typical control chart is a graphical display of a quality characteristic that has been measured
or computed from a sample versus the sample number or time. The chart contains a centre line
that represents the average value of the quality characteristic corresponding to the under control
state. Two other horizontal lines, called the upper control limit (UCL) and the lower control limit
(LCL) are also drawn. These control limits are chosen so that if the process is in control, nearly
all of the sample points will fall between them. As long as the points plot within the control
limits, the process is assumed to be under control, and no action is necessary.
However, a point that plots outside of the control limits is interpreted as evidence that the
process is out of control, and investigation and corrective action is required to find and eliminate the
assignable causes responsible for this behaviour. The control points are connected with straight-line
segments for easy visualization. Even if all the points plot inside the control limits, if they behave
in a systematic or non-random manner, then this is an indication that the process is out of control.
Assignable cause: It is the cause which can be controlled after inspection. Most of the causes for
variation are assignable cause. These causes can be avoided by proper control.
Chance cause: It is the cause which cannot be avoided. It may occur suddenly due to unpredictable
reasons.

23.2 Sources of Variations

There are five sources of variations: process, materials, environment, operators, and inspection.
Variation due to process includes tool wear, way of operation, tool vibration, jigs and fixtures,
etc. Variation due to materials includes quality characteristics of raw materials, such as tensile
strength, ductility, malleability, hardness, toughness, thickness, porosity, moisture contents, etc.
Variation due to the environment includes temperature, light, radiation, humidity, etc. Variation
due to operator includes skill, motivation, wages, etc. Finally, variation due to inspection includes
incorrect application of quality standards and inspection device.
 Statistical Quality Control  527

23.2.1 Types of Variations

There are three types of variations in production of piece parts:
   (i) Within piece variation: This type of variation in quality parameters varies with in
piece, for example, surface roughness of the different faces of the piece.
(ii) Piece-to-piece variation: This type of variation in quality parameters varies from piece
to piece, for example, change in surface finish due to continuous tool wear.
(iii) Time-to-time variation: This type of variation in quality parameters varies due to
different time of the day, for example, product produced in the morning will be better
than the product produced after launch due to laziness of the worker.

23.3 Chart Techniques

Following steps are followed to produce a control chart:
  (i) Select the quality characteristics: Selection of quality characteristics is very important
in control chart for variables/attributes. Only one-quality characteristics can be used at a
time in control chart for the variable. But, in case of control charts for attributes, more than
one quality characteristics can be considered to define that the product is defective or not.
Similarly, in C-chart a no. of defects can be considered as count of defects in the products.
(ii) Choose the rational subgroup: There are two ways to select the subgroup samples –
product produced at one instant of time or close to that instant as possible and products
produced over a period of time so that they can represent all the products. On the basis
of subgroup size, there are some empirical judgement as discussed below:
• As the subgroup size increases, the control limits become closer to central value,
which makes the control chart more sensitive.
• As the subgroup size increases, the inspection cost per subgroup increases.
• When destructive testing is used and the item is expensive, a small subgroup size of
2 or 3 is preferred.
• On the basis of statistical distribution, subgroup averages become normal when
subgroup size equals or exceeds 4 even the subgroups are taken from non-normal
population.
• When the subgroup size exceeds from 10 or 12, the range (R-chart) should be replaced
by standard deviation (S-chart).
(iii) Collect the data: After deciding the subgroup size, data should be collected to prepare
the control chart.
  (iv) Determine the trial central line and control limits: On the basis of the nature of
data collected, use suitable control chart and determine the trial central line and control
limits to check which subgroup is out of control.
  (v) Establish the revised central line and control limits: After finding the subgroup out of
control, establish the revised central line and control limits by discarding those subgroups.
(vi) Achieve the purpose: Finally, achieve the purpose of control chart for which it was
prepared.
528  Industrial Engineering and Management

23.4 Control Charts for Variables

Control charts for variables monitor characteristics that can be measured and have a continuous
scale, such as height, weight, volume, or width. When an item is inspected, the variable being
monitored is measured and recorded. For example, if we produce bolts, thread dimension might
be an important variable. We can take samples of bolts and measure their thread major and minor
diameters. Two of the most commonly used control charts for variables monitor both the central
tendency of the data (the mean) and the variability of the data (either the standard deviation or
the range). Note that each chart monitors a different type of information. When observed values
go outside the control limits, the process is assumed not to be in control. Production is stopped,
and employees attempt to identify the cause of the problem and correct it.

23.4.1 Objectives of Variable Control Charts

There are following objectives of variable control charts:
(i) To achieve quality improvement in products.
(ii) To determine the process capability.
(iii) To provide information for decision with regard to product specification.
(iv) To provide information about the current production process.
(v) To provide information about the decisions taken for currently produced items.

23.4.2 X-bar and R-chart

The X-bar (arithmetic mean) and R (range) control chart is used with variables data when
subgroup or sample size is between 2 and 15; however, this is not a fixed rule. The steps for
constructing this type of control chart are:
Step 1: Determine the data to be collected. Decide what questions about the process you plan to
answer. Refer to the data collection module for information on how this is done.
Step 2: Collect and enter the data by subgroup. A subgroup is made up of variable data that
represents a characteristic of a product produced by a process. The sample size relates
to how large the subgroups are. Enter the individual subgroup measurements in time
sequence in the portion of the data collection section of the control chart labelled
measurements (X1, X2, X3, … Xn).
Step 3: Calculate and enter the average for each subgroup. Use the formula below to calculate
the average (mean) for each subgroup and enter it on the line labelled Average in the
data collection section.
X1 + X 2 +  + X n
X =
n
X is the average of the measurements within each subgroup,
where
X i is the individual measurements within a subgroup,
i = 1, 2,…, n is the number of measurements within a subgroup or subgroup

size, and
 Statistical Quality Control  529

Step 4: Calculate and enter the range for each subgroup. Use the following formula to calculate
the range (R) for each subgroup. Enter the range for each subgroup on the line labelled
Range in the data collection section.
Range = Largest value in each subgroup − Smallest value in each subgroup
R = X max − X min 

Step 5: Calculate the grand mean of the subgroup’s average. The grand mean of the subgroup’s
average (X-bar) becomes the centre line for the control chart, i.e., X chart.
g

X1 + X 2 +  + X g ∑X j =1
j

X = =
g g
where X is the grand mean of subgroup’s average.
X j is the average of jth subgroup, and
g is the total no. of subgroups.
Step 6: Calculate the average of the subgroup ranges. The average of all subgroups becomes the
centre line for the R-chart.
g

R1 + R2 +  + Rg ∑R
j =1
j

R= =
g g
where Ri is the individual range for each subgroup
R is the average of the ranges for all subgroups
g is the number of subgroups
Step 7: Calculate the UCL and LCL for the averages of the subgroups. At this point, the chart
will look like a run chart. Now, the uniqueness of the control chart becomes evident as
the control limits are calculated. Control limits define the parameters for determining
whether a process is in statistical control. To find the X-bar control limits (Figure 23.1),
use the following formula:
UCL X = X + zσ X = X + 3σ X = X + A2 R
LCL X = X − zσ X = X − 3σ X = X − A2 R
3σ ′ 3
Here, 3σ X = = R = A2 R
.R 
n d2 n
To find the R control limits (Figure 23.2), use the following formula:
R ⎛ 3d3 ⎞
UCL R = R + 3σ R = R + 3d3σ = R + 3d3 = 1+ R = D4 R
d2 ⎜⎝ d2 ⎟⎠
R ⎛ 3d3 ⎞
LCL R = R − 3σ R = R − 3d3σ = R − 3d3 = 1− R = D3 R
d2 ⎜⎝ d2 ⎟⎠
Here, σ shows the standard deviations.
530  Industrial Engineering and Management

The value of the factor A2, D4, and D3 are given in the statistical tables in the ‘Appendix’
section.
3.54
Upper control limit
UCLX 3.5
Subgroup average

3.46
Central line
X 3.42

3.38

Lower control limit

LCLX 3.34

3.3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Subgroup number

Figure 23-1: Presentation of X-bar chart

UCLR 0.67 Upper control limit

Range

R 0.335
Central line

Lower control limit

LCLR 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Subgroup number

Figure 23-2: Presentation of R-chart

Establish the Revised Control Limits

If some data in X-bar and R-charts are out of control, establish the revised control limits by
discarding the subgroups out of control in the corresponding chart. To find the grand average and
control limits following formula can be used.
g g

∑X −X
j =1
d ∑R−R
j =1
d

X0 = ; R0 =
g − gd g − gd
 Statistical Quality Control  531

where X d are discarded subgroup averages

g d is number of discarded subgroups
Rd are discarded subgroup ranges
Using the standard value of the factors, the central lines and the 3σ control limits for actual
operations are obtained using the formula:
R
UCL X = X 0 + Aσ 0 = X 0 + A 0
d2
R0
LCL X = X 0 − Aσ 0 = X 0 − A
d2
and UCL R = D2σ 0 
LCL R = D1σ 0
where the value of A, D1, and D2 are the factors which will be taken from the ‘Statistical Table
Appendix’.

Example 23.1: A workshop produced the following data (in millimetre) regarding a product
manufactured on 5th August 2014 during two shifts of six hours each as shown in Table 23.1.
Find the control limits for X-bar and R-chart.

Table 23-1: Data regarding a product manufactured as on 5.08.2014

Subgroup no. X1 X2 X3 X4 X R
1 36 41 46 37 40 10
2 34 36 34 40 36 6
3 68 59 69 64 65 10
4 44 40 38 34 39 10
5 34 43 41 42 40 9
6 46 41 41 44 43 5
7 36 38 41 33 37 8
8 51 49 52 48 50 4
9 42 36 43 47 42 11
10 39 38 38 41 39 3
11 41 37 37 37 38 4
12 47 35 40 38 40 12
13 45 42 38 39 41 7
14 43 45 50 42 45 8
15 39 29 35 33 34 10
16 34 29 40 41 36 12
17 58 28 44 38 42 30
(Continued)
532  Industrial Engineering and Management

Table 23-1: Data regarding a product manufactured as on 5.08.2014 (Continued)

Subgroup no. X1 X2 X3 X4 X R
18 38 37 32 33 35 6
19 48 45 55 56 51 11
20 45 37 40 38 40 8
21 35 40 42 39 39 7
22 39 36 39 42 39 6
23 35 38 36 43 38 8
24 43 44 38 39 41 6

Solution:
g g

X1 + X 2 +  + X g ∑X
j =1
j
990 R1 + R2 +  + Rg ∑R
j =1
j
211
X = = = = 41.25; R = = = = 8.79
g g 24 g g 24
UCL X = X + zσ X = X + 3σ X = X + A2 R = 41.25 + 0.729 × 8.79 = 47.65
LCL X = X − zσ X = X − 3σ X = X − A2 R = 41.25 − 0.729 × 8.79 = 34.84
The value of A2 is taken for subgroup size, n = 4 from the statistical table given in Appendix 3.
The X -chart is shown in Figure 23.3. The 3rd, 8th, 15th, and 19th data are out of control.
We have to find the revised control limits for X .

70

60
Mean

50 UCL = 47.6556

Average = 41.2500
40
LCL = 34.8444

30
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00

Subgroup number

Figure 23-3: X-bar chart

UCL R = D4 R = 2.282 × 8.79 = 20.058

LCL R = D3 R = 0 × 8.79 = 0
 Statistical Quality Control  533

The value of D4 and D3 are taken for subgroup size, n = 4 from the statistical table given in
Appendix 3.
The seventeenth point is out of control in R-chart as shown in Figure 23.4; therefore, we
have to draw revised control limits for R-chart.

30

UCL = 20.0630
20
Range

Average = 8.7917
10

LCL = .0000
30
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00

Subgroup number

Figure 23-4: R-chart

Revised Control Limits

g

∑X −X
j =1
d
990 − 65 − 50 − 34 − 51
X0 = = = 39.5
g − gd 20
g

∑R−R
j =1
d
211 − 30
R0 = = = 7.86
g − gd 23
R0 7.86
UCL X = X 0 + Aσ 0 = X 0 + A = 39.5 + 1.5 × = 45.226
d2 2.059
R0 7.86
LCL X = X 0 − Aσ 0 = X 0 − A = 39.5 − 1.5 × = 33.773
d2 2.059
7.86
and UCL R = D2σ 0 = 4.698 × = 17.93 
2.059
7.86
LCL R = D1σ 0 = 0 × =0
2.059
The value of A, d2, D2, and D1 are taken for subgroup size, n = 4 from the statistical table given
in Appendix 3.
534  Industrial Engineering and Management

23.4.3 X-bar and S-chart

X-bar and R-charts are the most common charts for variables, but many companies use X-bar
and S-charts. R-chart uses only two values in a subgroup maximum and minimum values.
But, the S-chart uses all the data in subgroups. S-chart is therefore more accurate than R-chat.
For subgroup size less than 10, both R-chart and S-chart show the same variation, but for the
subgroup size more than 10, S-chart is preferred. Standard deviation for each subgroup can be
found using the following formula:
2
n
⎛ n ⎞
n∑ X i2 − ⎜ ∑ X i ⎟
i =1 ⎝ i =1 ⎠
Sj =
n ( n − 1)
where Sj is standard deviation of jth subgroup.
Centre line in X-bar chart and S-chart can be found as:
g

∑X
j =1
j
X1 + X 2 +  + X n
X = ; where X j = ; ∀j
g n
2
g n
⎛ n ⎞ n

∑S
j =1
j n∑ X i2 − ⎜ ∑ X i ⎟
i =1 ⎝ i =1 ⎠ ∑(X i − X)
S= ; where S j = = i =1
; ∀j
g n ( n − 1) (n − 1)
Control limits in X-bar and S-chart can be calculated as:
UCL X = X + A3 S ; LCL X = X − A3 S
UCL S = B4 S ; LCLC = B3 S
The value of A3, B4, and B3 can be taken for specific subgroup size from statistical table given in
Appendix 3.

Establish the Revised Control Limits

If some data in X-bar and S-charts are out of control, establish the revised control limit by
discarding the subgroups out of control in the corresponding chart. To find the grand average and
control limits, following formula can be used.
g g

∑X −X
j =1
d ∑S −S
j =1
d
S0
X0 = ; S0 = ; σ0 =
g − gd g − gd c4

UCL X = X 0 + Aσ 0 ; LCL X = X 0 − Aσ 0

UCL S = B6σ 0 ; LCL S = B5σ 0

The value of A, B6 and B5 can be taken for specific subgroup size from statistical table given in
Appendix 3.
 Statistical Quality Control  535

Example 23.2: Find the control limits for X-bar and S-chart from the data given in Example 23.1.

Solution:
Subgroup no. X1 X2 X3 X4 X s
1 36 41 46 37 40 4.546061
2 34 36 34 40 36 2.828427
3 68 59 69 64 65 4.546061
4 44 40 38 34 39 4.163332
5 34 43 41 42 40 4.082483
6 46 41 41 44 43 2.44949
7 36 38 41 33 37 3.366502
8 51 49 52 48 50 1.825742
9 42 36 43 47 42 4.546061
10 39 38 38 41 39 1.414214
11 41 37 37 37 38 2
12 47 35 40 38 40 5.09902
13 45 42 38 39 41 3.162278
14 43 45 50 42 45 3.559026
15 39 29 35 33 34 4.163332
16 34 29 40 41 36 5.597619
17 58 28 44 38 42 12.54326
18 38 37 32 33 35 2.94392
19 48 45 55 56 51 5.354126
20 45 37 40 38 40 3.559026
21 35 40 42 39 39 2.94392
22 39 36 39 42 39 2.44949
23 35 38 36 43 38 3.559026
24 43 44 38 39 41 2.94392

X1 + X 2 +  + X g ∑X
j =1
j
990
X = = g = = 41.25
g 24
g

∑S
j =1 93.624
S= g = = 3.901
24
UCL X = X + A3 S = 41.25 + 1.628 × 3.901 = 47.6
LCL X = X − A3 S = 41.25 − 1.628 × 3.901 = 34.899
536  Industrial Engineering and Management

UCL S = B4 S = 2.266 × 3.901 = 8.84

LCL S = B3 S = 0 × 3.901 = 0
The value of A3, B4 and B3 are taken for subgroup size, n = 4 from statistical table given in
Appendix 3.
The X-bar chart considering the standard deviation in place of range is shown in Figure 23.5
and 23.6. It is observed that the data in subgroups 3rd, 8th, 15th, and 19th are out of control in X-bar
chart and 17th subgroup id out of control in s-chart. Therefore, we have to find the revised control
limits.
g

∑X −X
j =1
d
990 − 65 − 50 − 34 − 51
X0 = = = 39.5
g − gd 20
g

∑S − S
j =1
d
93.624 − 12.543
S0 = = = 3.525
g − gd 23
S0 3.525
σ0 = = = 3.82
c4 0.9213
UCL X = X 0 + Aσ 0 = 39.5 + 1.5 × 3.82 = 45.23
LCL X = X 0 − Aσ 0 = 39.5 − 1.5 × 3.82 = 33.77
UCL S = B6σ 0 = 2.088 × 3.82 = 7.97
LCL S = B5σ 0 = 0 × 3.75 = 0
The value of A, B6 and B5 are taken for subgroup size, n = 4 from statistical table given in
Appendix 3.
70

60

50
UCL = 47.6027
40
Mean

Average = 41.2500

30
LCL = 34.8973
20

10
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00

Subgroup number

Figure 23-5: X-bar chart

 Statistical Quality Control  537

15

UCL = 8.8420
10
Standard deviation

5 Average = 3.9019

LCL = .0000
0
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00

Subgroup number

Figure 23-6: S-chart

Note: It has been observed that for the subgroups of size less than 12, there is no significant
difference between X-bar and R-chart and X-bar and S-chart. But, for the subgroups
larger than 12 X-bar and S-chart gives better result than that of X-bar and R-chart.

23.5 Control Chart for Trends

Due to die wear, the measurement gradually increases until it reaches the upper reject limit. Then
die is replaced or reworked. Since the central line is on a slope, its equation must be determined.
This is estimated using least-square method of plotting a line to a set of points. The equation for
the trend line using the slope-intercept form is X = a + bW as shown in Figure 23.7.
where W = S  ubgroup numbers represent the horizontal line
a = Intercepts of vertical axis

a=
∑ X ∑ W − ∑ W ∑ (W . X )
2

g ∑ W − (∑ W )
2
2

g ∑ (W . X ) − ∑ W ∑ X
b = Slope of line =
g ∑ W − (∑ W )
2
2
538  Industrial Engineering and Management

URL

UCL

X
LCL

LRL

No. of subgroups

Figure 23-7: Trends chart

23.6 Process under control

The natural pattern of the variations of data on a control chart for the process under control can
be given below as:
   (i) About 2/3 (67 per cent) of the points lie near the centre line.
(ii) A few points (5 per cent) lie closer to control limits.
(iii) The points are located back and forth across the centre line.
  (iv) Points are balanced on both sides of the centre line.
(v) No point lies beyond the control limits.
When the process is under control, the following advantages can be observed by producer as well
as consumer:
   (i) There will be fewer variations in product specifications.
(ii) Due to the uniformity in products, only few samples are required for inspection or
quality testing.
(iii) The process capability (6s) can be easily attained. With the knowledge of process
capability, a no. of decisions related to product specifications can be taken, such
as amount of rework in case of insufficient tolerance, tight specifications/loose
specifications, degree of quality assurance, etc.

23.7 Process out of control

The process out of control can be shown in Figure 23.8.

23.7.1 Pattern of Data Out of Control

The patterns shown in Figure 23.9 of the data on control chart are assumed to be out of control.
 Statistical Quality Control  539

Observation out of control

Variation due to
assignable causes
UCL = (16.2)
Volume in ounces

Variation due
CL = (16.0)
to normal causes

LCL = (15.8)
#1 #2 #3 #4 #5 #6 Variation due
Sample number to assignable causes

Figure 23-8: Process out of control

One or more Seven or more consecutive Six points in a row

points outside the points on one side of the steadily increasing or
control limits central line decreasing

Fourteen points Two out of three Feefteen points in a row

alternating up and down consecutive points in within the center third of
the outer third of the the control region
control region

Eight points on both

sides of the center third
of the control region

Figure 23-9: Pattern of data out of control

23.7.2 Reasons for Process Out of Control

  (i) Change or jump in level: Jump in level in the X-bar chart may occur due to intentional
or unintentional change in process setting, deploying inexperienced operator, using
different raw materials, and a minor failure of machine parts. Similarly, Change in level
in R-chart may occur due to inexperienced operators, sudden increase in gear play, and
greater variations in incoming raw materials.
(ii) Trends or steady change in level: Steady change in level in X-bar chart may occur due
to continuous wear of tool and die, gradual deterioration of equipment, viscosity in a
chemical process, build-up of chips in jigs and fixtures, etc. Similarly, steady change in
level in an R-chart may occur due to an improvement in worker skill, a decrease in worker’s
skill due to fatigue, and gradual improvement in homogeneity of incoming materials.
540  Industrial Engineering and Management

(iii) Recurring cycles: Recurring cycles of data in the X-bar chart may occur due to seasonal
effect of incoming materials, recurring effect of temperature and humidity, periodic
chemical/mechanical/psychological treatment, and periodic rotation of operators.
Similarly, recurring cycles of data on R-chart may occur due to operator’s fatigue and
rejuvenation resulting from morning, noon and afternoon breaks and lubrication cycles.
  (iv) Two populations: Two populations in the X-bar chart may occur due to large differences
in material quality, two or more machines working on the same chart, large differences in
testing methods or equipments, etc. Similarly, two populations in an R-chart may occur due
to different workers using the same chart and materials purchased from different suppliers.
(v) Mistakes: There is a no. of sources of mistakes which leads in the process out of control.
These sources include measuring equipment out of calibration, errors in calculation,
error in using test equipment, and taking samples from different populations.

23.8 Process Capability and Specification Limits

Central Limit Theorem
If the population from which samples are taken is not normal, the distribution of sample average
will tend towards normality provided that the sample size n is at least 4. This tendency increases
as the sample size increases.
X −μ X −μ
z= =
σX σ
n

23.8.1 Control Limits and Specification Limits

Control limits are established from the averages of subgroups of produced items, whereas
the specification limits are the permissible variation of tolerance limits determined by the
manufacturer. The difference between UCL and LCL is known as process spread. If we compare
the process spread and the difference between specification limits, three situations may occur:
(i) when the process spread (UCL – LCL = 6s) is less than the difference between the specification
limits (U – L), (ii) when the process spread (UCL – LCL = 6s) is equal to the difference between
the specification limits (U – L), and (iii) when the process spread (UCL – LCL = 6s) is greater
than the difference between the specification limits (U – L) (Figure 23.10).
U UCL
UCL = U
UCL U

Center line Center line Center line

LCL L
LCL = L
L
6σ < U – L 6σ = U – L 6σ > U – L LCL
(a) (b) (c)

Figure 23-10: Control limits and specification limits

 Statistical Quality Control  541

Case I: 6s < U – L
Process spread is less than the difference between the specification limits is desirable since, in
this situation defective items are not produced.
Case II: 6s = U – L
Process spread is equal to the difference between the specification limits is alarming situation
however, defective items are not produced. In this case, the process may go out of control,
therefore assignable causes of variation must be corrected as soon as they occur.
Case III: 6s > U – L
Process spread is greater than the difference between the specification limits is undesirable since,
in this situation some defective items may be produced.

23.9 Process Capability

To determine the process capability (6s0) X-bar and R-chart should achieve the optimal quality
improvement without investment in new equipment or modification. The quick method to
determine the process capability is mentioned below as:
   (i) Take 20 subgroups of size 4 for a total 80 measurements. These subgroups should be
selected randomly to avoid any biasing.
(ii) Calculate the sample standard deviation S for each subgroup.
(iii) Calculate the average sample standard deviation.

S=
∑S = ∑S
g 20
S
  (iv) Estimate the population standard deviation, σ 0 =
c4
  (v) Process capability is equal to 6s0
U−L
  (vi) Capability Index, C p = ;
6σ 0
where C p is capability index, U-L is tolerance,
and 6σ 0 is process capability.

Example 23.3: Using the data provided in Example 23.1, find the process capability index.
The upper and lower specification index are given as 50 mm and 30 mm, respectively.

Solution:
Given U = 50 mm and L = 30 mm. From Example 23.1, we find that the mean of standard
deviation is 3.90 mm (as shown in Example 23.2) and for subgroup size of 4, c4 = 0.9213.
S 3.90
σ0 = = = 4.23
c4 0.9213

542  Industrial Engineering and Management

U − L 50 − 30
Cp = = = 0.788
6σ 0 6 × 4.23
i.e., U - L < 6s0
Since process spread is greater than the difference between specification limits, it is undesirable
since, in this situation, there is a chance to be produced some defective items.

23.10 Limitations of Chart for Variables

There are following limitations of control charts for variables:
(a) It cannot be used for the attributes type quality characteristics.
(b) If a product has a large number of variables, it will require individual X-bar and R-chart
for each variable which is not possible. In this case, attributes chart is preferred.

23.11 Control Charts for Attributes

Attributes are the characteristics of a product or services which can be measured in terms of binary
value such as pass–fail, accepted–rejected, good–bad, etc., but it cannot be measured in terms of
numerical values such as weight, height, length, width, etc. There are two types of attributes:
(a) Where measurement are not possible, for example, visually inspected items such as
colour, missing parts, scratches and damage.
(b) Where measurements are possible, but not made because of involvement of cost and
time. For example, diameter of a wire can be measured using micrometer, but for
easiness go–no go gauge is used to confirm the accuracy of the diameter.
Control charts for attributes can be used for both defects and defective analysis. Defects represent
the individual quality characteristics of the product, but defective represents the entire product.
A defective product may have a number of defects.
There are two different groups of control charts for attributes: one is for defective, i.e. p-chart
which is based on binomial distribution and other one is for defects, i.e. c-chart and u-chart based
on the Poisson distribution.

23.11.1 p-Chart
p-Chart is the most versatile control chart which can be used for single quality characteristics
(like X-bar and R-chart) as well as for multiple quality characteristics at a time to control the
entire product. It can be established for controlling the performance of operator, work centre,
department, plant, etc.
p-Chart Construction for Constant Subgroup Size
Trial centre line and control limits g

∑p
j =1
j

p=
g
 Statistical Quality Control  543

np j
where pj = , j indicates no. of subgroups, total no. of subgroups are g ,
n
n is number inspected in a subgroup,
np j is number defectives of j th subgroup,
p j is fraction defective in j th subroup, and
p is average fraction defective.
p(1 − p ) p(1 − p )
UCL = p + 3 ; LCL = p − 3
n n

Revised Control Limits and Centre Line

If some of the fraction defectives of subgroups go beyond the control limits, a revised control chart
is required. To establish the revised control chart, the subgroups having the fraction defectives out
of control should be discarded in the calculation of revised average fraction defective. A sample
of p-chart is shown in Figure 23.11.
g

∑p
j =1
j − pd
p0 =
g − gd

where pd is fraction defectives in discarded subgroups,

g d is subgroups.

p-Chart of defects
0.07
1
0.06
1
0.05 UCL = 0.04262
1
Proportion

0.04 0.034

0.03 0.025

0.02 P = 0.016

0.01 0.007

0.00 LCL = 0
3 6 9 12 15 18 21 24 27 30
Sample

Figure 23-11: p-Chart

p0 (1 − p0 ) p (1 − p0 )
UCL = p0 + 3 ; LCL = p0 − 3 0
n n
544  Industrial Engineering and Management

Example 23.4: Inspection results of electric bulb produced by a company in a month of

25 working days are shown in Table 23.2. Daily production rate is 500 bulbs. Find the control
limits using control charts for defectives.
Table 23-2: Daily production and defectives record of bulbs

Subgroup Number Number Fraction Subgroup Number Number Fraction

number inspected defective defective number inspected defective defective
(n) (np) (p) (n) (np) (p)
1 500 18 0.036 14 500 5 0.01
2 500 5 0.01 15 500 2 0.004
3 500 14 0.028 16 500 8 0.016
4 500 6 0.012 17 500 10 0.02
5 500 2 0.004 18 500 12 0.024
6 500 9 0.018 19 500 25 0.05
7 500 9 0.018 20 500 3 0.006
8 500 2 0.004 21 500 8 0.016
9 500 13 0.026 22 500 9 0.018
10 500 16 0.032 23 500 2 0.004
11 500 3 0.006 24 500 5 0.01
12 500 15 0.03 25 500 3 0.006
13 500 14 0.028

Solution: g

∑p
j =1
j
0.436
p= = = 0.01744
g 25
p(1 − p ) 0.01744(1 − 0.01744)
UCL = p + 3 = 0.01744 + 3 = 0.035
n 500
0.01744(1 − 0.01744)
LCL = 0.01744 − 3 =0
500
From p-chart shown in Figure 23.12, it is observed that first and nineteenth data are out of
control. We have to draw the revised control limits.
g

∑pj =1
j − pd
0.436 − 0.036 − 0.05
p0 = = = 0.0152
g − gd 23
p0 (1 − p0 ) 0.0152(1 − 0.0152)
UCL Revised = p0 + 3 = 0.0152 + 3 = 0.0316
n 500
p0 (1 − p0 ) 0.0152(1 − 0.0152)
LCL Revised = p0 − 3 = 0.0152 − 3 =0
n 500
 Statistical Quality Control  545

Note: Similarly, per cent defective chart and number defective chart can be found using
following formula for control limits.

0.05
Proportion nonconforming

0.04
UCL = .035003

0.03

0.02 Center = .017440

0.01

LCL = .000000
0.00
10.000
11.000
12.000
13.000
14.000
15.000
16.000
17.000
18.000
19.000
20.000
21.000
22.000
23.000
24.000
25.000
1.000
2.000
3.000
4.000
5.000
6.000
7.000
8.000
9.000

Subgroup number

Figure 23-12: p-Chart

Per cent defective chart:

Central line = 100 p0
⎡ p (1 − p0 ) ⎤ ⎡ p (1 − p0 ) ⎤
UCL = 100 ⎢ p0 + 3 0 ⎥; LCL = 100 ⎢ p0 − 3 0 ⎥
⎢⎣ n ⎥⎦ ⎢⎣ n ⎥⎦
Number defective chart:
Central line = np0
UCL = np0 + 3 np0 (1 − np0 ); LCL = np0 − 3 np0 (1 − np0 )

23.11.2 p-Chart Construction for Variable Subgroup Size

If the number inspected in subgroups varies from one subgroup to another, the control limits
should be drawn individually for each subgroup as shown in Figure 23.13. Revised control limits
can be drawn in similar way as in case of constant number inspected subgroups.
g

∑p
j =1
j

p=
g
546  Industrial Engineering and Management

np j
where pj = , j indicates no. of subgroups, total no. of subgroups are g ,
nj
n is number inspected in a subgroup,
np j is number defectives of j th subgroup,
p j is fraction defective in j th subroup, and
p is average fraction defective.

p(1 − p ) p(1 − p )
UCL j = p + 3 ; LCL j = p − 3
nj nj

0.16

0.14 UCL

0.12

0.10
Proportion

0.08
P

0.06

0.04

0.02
LCL
0.00
1 3 5 7 9 11 13 15 17 19
Sample

Figure 23-13: p-Chart for variable subgroup size

Minimizing the Effect of Variable Subgroup Size

When the subgroup size varies from subgroup to subgroup, it presents a complex chart that is
difficult to explain. By using average subgroup size, one limit can be calculated and placed on the
control chart. A case of control chart is explained using an example as given below:

Example 23.5: There are one month data as given in Table 23.3. The production rate varies
from day to day. Number of non-conforming components are also given with number inspected.
Draw a p-chart and find the control limits for the data produced.
 Statistical Quality Control  547

Table 23-3: Data for number inspected and number defectives

Subgroup Number Number Fraction Subgroup Number Number Fraction

no. inspected defective defective no. inspected defective defective
(ni) (np) ( p) (ni) (np) (p)
1 2385 47 0.02 14 1667 34 0.02
2 1451 18 0.012 15 2350 31 0.013
3 1935 74 0.038 16 2354 38 0.016
4 2450 42 0.017 17 1509 28 0.018
5 1997 39 0.02 18 2190 30 0.014
6 2168 52 0.024 19 2678 113 0.042
7 1941 47 0.024 20 2252 58 0.026
8 1962 34 0.017 21 1641 52 0.032
9 2244 29 0.013 22 1782 19 0.011
10 1238 39 0.032 23 1993 30 0.015
11 2289 45 0.02 24 2382 17 0.007
12 1464 26 0.018 25 2132 46 0.022
13 2061 47 0.023

Solution:

g g

∑p
j =1
j ∑ np
j =1
j
1035
p= = g
= = 0.020
g 50, 515
∑n j =1
j

p (1 − p ) 0.020 (1 − 0.020 ) 0.042

Control limits = p ± 3 = 0.020 ± 3 = 0.02 ±
nj nj nj

Putting the values of n for different subgroup, we can get the individual control limits as shown
in Figure 23.14.
We see that the data of 3rd, 19th, 21st and 24th are out of control. Therefore, we have to
calculate the revised control limits.
g

∑ p −∑ p
j =1
j d
0.514 − 0.038 − 0.042 − 0.032 − 0.007
p0 = = = 0.018
g − gd 21

Using the new value of p0, we can calculate the new control limits for the individual subgroup.
548  Industrial Engineering and Management

0.05

0.04
Proportion nonconforming

UCL
0.03

Center = .02
0.02

0.01
LCL

0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Subgroup number

Figure 23-14: p-Chart for variable subgroup size

Minimizing the Effect of Variable Subgroup Size

Step1: Find the average subgroup size
g

∑n
j =1
j
50515
nav = = = 2020.5 ≅ 2020
g 25
Step 2: Find the upper and LCLs using average subgroup size

p (1 − p ) 0.020 (1 − 0.020 )
UCL = p + 3 = 0.020 + 3 = 0.0293
nav 2020
p (1 − p ) 0.020 (1 − 0.020 )
LCL = p − 3 = 0.020 − 3 = 0.0106
nav 2020

When an average subgroup size is used, there are four situations which occur between the control
limits and the individual fraction defective values.
 Statistical Quality Control  549

Situation I: When a point (subgroup fraction defective) falls inside the limits and its subgroup
size is smaller than the average subgroup size, the individual control limits will be wider than
that for average subgroup size; therefore, individual control limits are not needed. The data will
be under control.
Situation II: When a point (subgroup fraction defective) falls inside the limits and its subgroup
size is larger than the average subgroup size, and the difference in subgroup size is substantial, the
individual control limits will be smaller than that for average subgroup size; therefore, individual
control limits are needed to check the data that is either under control or out of control.
Situation III: When a point (subgroup fraction defective) falls outside the limits and its subgroup
size is smaller than the average subgroup size and the difference in subgroup size is substantial,
the individual control limits will be wider than that for average subgroup size, but individual
control limits are needed to check the data that is either under control or out of control.
Situation IV: When a point (subgroup fraction defective) falls outside the limits and its subgroup
size is larger than the average subgroup size, the individual control limits will be smaller than
that for average subgroup size; therefore, individual control limits are not needed. The point will
be out of control.

23.12 Count of Defect chart

There are two types of count of defect charts: (i) Count of defect chart (c-chart) (ii) and Count
of defect per unit chart (u-chart)

23.12.1 c-Chart
In a c-chart, unit subgroup size is used and count of defects in each subgroup is inspected.
A defective part can have a number of defects. A defect shows the quality characteristics, whereas
defective represents the products having defects. There are following steps for constructing
c-chart.
Step 1: Collect the data regarding the count of defects.
Step 2: Establish trial central line and control limits using following formula.
g

∑c
j =1
c=
g
UCL = c + 3 c
LCL = c − 3 c

where c is average count of defect.

If some data are out of control, establish a revised control limits by discarding the data
out of control.
550  Industrial Engineering and Management

Step 3: Establish revised control limits

g

∑c −c
j =1
d

c0 =
g − gd
UCL = c0 + 3 c0
LCL = c0 − 3 c0

where c0 is average count of defect of revised data.

Example 23.6: There are 25 units of a product and the count of defects in each unit has been
given in Table 23.4. Establish a control limits using c-chart.

Table 23-4: Data for count of defects

Unit no. Count of defect Unit no. Count of defect

1 7 14 3
2 6 15 2
3 6 16 7
4 3 17 5
5 22 18 7
6 8 19 2
7 6 20 8
8 1 21 0
9 0 22 4
10 5 23 14
11 14 24 4
12 3 25 3
13 1

Solution:
g

∑c
j =1 141
c= = = 5.64
g 25
UCL = c + 3 c = 5.64 + 3 5.64 = 12.76
LCL = c − 3 c = 5.64 − 3 5.64 = −1.484  0

5th, 11th, and 23rd data are out of control as shown in Figure 23.15; therefore, a revised control
limit is required.
 Statistical Quality Control  551

25

20
Nonconformities

15
UCL = 12.7646

10

Center = 5.6400
5

LCL = .0000
0
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
25.00

Subgroup number

Figure 23-15: c-Chart

Revised Control Limits

g

∑c − c
j =1
d
141 − 22 − 14 − 14
c0 = = = 4.136
g − gd 22
UCL = c0 + 3 c0 = 4.136 + 3 4.136 = 10.23
LCL = c0 − 3 c0 = 4.136 − 3 4.136 = −1.965  0

23.12.2 u-Chart
c-Chart can be used only when subgroup size is one but when subgroup size is more than one,
u-chart is used. There are following steps to establish a u-chart (Figure 23.16);
552  Industrial Engineering and Management

UCL = 1.609
1.6
Fraction of nonconformance

1.4

1.2

0.8
LCL = 0.698

0.6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Subgroup number

Figure 23-16: u-Chart

Step 1: Collect the data regarding count of defect in each subgroup and find u.
cj
uj =
n

where j is serial number of subgroup.

Step 2: Find the average value of u.
g

∑u
j =1
j

u=
g
Step 3: Establish control limits.
u
UCL = u + 3
n
u
LCL = u − 3
n
If some data be out of control, establish revise control limits.
g

∑u − ∑u
j =1
j d

u0 =
g − gd
u0
UCL = u0 + 3
n
u0
LCL = u0 − 3
n
 Statistical Quality Control  553

Example 23.7: Establish a u-chart from the data given in Table 23.5.

Table 23-5: Data for count of defects per unit

Subgroup Number Count of Defect per Subgroup Number Count of Defect

no. inspected defect unit no. inspected defect per unit
(n) (c) (u = c/n) (n) (c) (u = c/n)
1 50 36 0.72 14 50 52 1.04
2 50 48 0.96 15 50 42 0.84
3 50 45 0.9 16 50 47 0.94
4 50 68 1.36 17 50 64 1.28
5 50 77 1.54 18 50 61 1.22
6 50 56 1.12 19 50 66 1.32
7 50 58 1.16 20 50 37 1.34
8 50 67 1.34 21 50 59 1.18
9 50 38 0.76 22 50 31 0.62
10 50 74 1.48 23 50 48 0.96
11 50 69 1.38 24 50 68 1.36
12 50 54 1.08 25 50 78 1.56
13 50 56 1.12

Solution:
g

∑u
j =1
j
28.85
u= = = 1.154
g 25

u 1.154
UCL = u + 3 = 1.154 + 3 = 1.609
n 50
u 1.154
LCL = u − 3 = 1.154 − 3 = 0.698
n 50
22nd point is out of control; therefore, revised control limits are required.
g

∑u
j =1
j − ud
u0 = = 1.17
g − gd
u0 1.17
UCL = u0 + 3 = 1.17 + 3 = 1.636
n 50
u0 1.17
LCL = u0 − 3 = 1.17 − 3 = 0.716
n 50
554  Industrial Engineering and Management

23.13 Demerits classification and control chart

On the basis of severity of defects, they can be classified as critical defects, major defects and
minor defects.
Critical defects: They are the defects in a product which result in hazardous or unsafe use. They
also prevent the performance of the product.
Major defects: These are the defects in a product which result in failure of working or performance.
Minor defects: These are not severe and do not affect the working performance of a product, but
affect the appearance or look of the product.
Control Chart
Control chart is established and plotted for a count of demerits per unit. The demerit per unit is
given by the formula:
D = wc uc + wma uma + wmi umi
where D = demerits per unit;
wc , wma , wmi = weights for critical, major, and minor defects, respectively;
uc , uma , umi = count of defects for critical, major, and minor defects, respectiively.
If wc, wma, and wmi are 9, 3, and 1, respectively, the formula is
D = 9uc + 3uma + 1umi
The value of D is calculated from the formula for each subgroup and central line and the 3s are
calculated using following formulas:
D0 = 9u0 c + 3u0 ma + 1u0 mi
92 u0 c + 32 u0 ma + 12 u0 mi
σ 0u =
n
UCL = D0 + 3σ 0 u ; LCL = D0 − 3σ 0 u

Example 23.8: Assuming that a 9:3:1 three-class weighing system is used, determine the central line
and control limits when u0c = 0.09, u0ma = 0.6, and u0mi = 3.0, and n = 50. Also calculate the demerits
per unit for 12th December when critical defects are 2, major defects are 24, and minor defects
150 for 50 units inspected on that day. Is the December 12 subgroup in control or out of control?

Solution:
D0 = 9u0 c + 3u0 ma + 1u0 mi = 9 × 0.09 + 3 × 0.6 + 1 × 3 = 5.61
92 u0 c + 32 u0 ma + 12 u0 mi 92 × 0.09 + 32 × 0.5 + 12 × 3
σ 0u = = = 0.54
n 50
UCL = D0 + 3σ 0 u = 5.61 + 3 × 0.54 = 7.23; LCL = D0 − 3σ 0 u = 5.61 − 3 × 0.54 = 3.99
⎛ 2⎞ ⎛ 24 ⎞ ⎛ 150 ⎞
DDecember12 = 9 ⎜ ⎟ + 3 ⎜ ⎟ + 1 ⎜ = 4.8(in control)
⎝ 50 ⎠ ⎝ 50 ⎠ ⎝ 50 ⎟⎠
 Statistical Quality Control  555

23.14 Lot-by-Lot Acceptance Sampling

23.14.1 Introduction
Lot-by-lot acceptance sampling is the basis to check the quality of products produced at different
time intervals, on different machines, in different plants. Different lots of product are prepared
based on the similarity in the production process. All the products in a lot cannot be inspected
due to costly nature of inspection, destructive testing, repetitive nature of job, lack of automation,
etc. Therefore, a sample(s) from each lot is selected for inspection; if number of defective is less
or equal to prescribed minimum number, then the lot is accepted and otherwise lot is rejected.
The acceptance or rejection of the lot is based on different sampling plans which have been
discussed in the next section of this chapter.

23.14.2 Application of Sampling Plan

There are following applications of acceptance sampling plan:
(a) If the testing process used for the product is destructive, sampling plan is used in place
of 100 per cent inspection.
(b) If the testing process is costly, sampling plan is used in place of 100 per cent inspection.
(c) Sampling plan may be used in a repetitive type of inspection to avoid fatigue or boredom.
(d) When control charts are not available, sampling plan may be used.
(e) When inspection process is not automated, sampling plan may be used.

23.14.3 Advantages of Sampling Plan

There are following advantages of sampling plan over control chart or 100 per cent inspection
process.
(a) It is more economical due to fewer inspections.
(b) There are less handling damages during inspection.
(c) It is most suitable for the destructive type of testing.
(d) There is upgradation of the inspection job from monotonous piece-by-piece to lot-by-
lot decisions.
(e) There is stronger motivation for quality improvement due to rejection of entire lot rather
than the return of defectives.
(f) Due to fewer inspectors, it is possible to simplify the recruiting, training and supervising
activities.

23.14.4 Disadvantages/Limitations of Sampling Plan

(a) There are risks of accepting defective lots and rejecting good lots.
(b) More time and efforts are devoted in the planning and documentation of the sampling
plan.
(c) Less information is provided about the product.
556  Industrial Engineering and Management

23.15 Sampling Plan

The initial step of the sampling plan is lot formation. Lot formation can influence the effectiveness
of the sampling plan. The guidelines for lot formation are given below as:
(a) Lot must be homogeneous, i.e. the all the products in the lot is produced by the same
machine, same operator, same input material, and so on.
(b) Lots should be as large as possible. Since the sample sizes do not increase as rapidly as
lot sizes, a lower inspection cost results with larger lot sizes.
The second step of the sampling plan is sample selection. The basic technique of sample selection
is to assign a series of numbers to each product in the lot and select the products for sampling
randomly. The random number can be generated manually or using software.
Type of Sampling Plans
There are three types of sampling plans used in the lot-by-lot sampling. These plans are as given
below:
(a) Single sampling (SS) plan
(b) Double sampling (DS) plan
(c) Multiple sampling plan

23.15.1 Single Sampling (SS) Plan

In SS plan, the decision about the lot is taken on the basis of quality of only one sample.
The following are the procedures for SS plan:
(a) A lot size (N) of the product is taken to check the quality.
(b) A sample size (n) is selected randomly from the lot.
(c) If the number of defectives in the sample exceeds the acceptance number (c), the entire
lot is rejected.
(d) If the number of defectives in the sample is equal to or less than the acceptance number,
the entire lot is accepted.

Flow Chart for Single Sampling Plan

Flow chart for SS plan is given in shown in Figure 23.17.
Inspect a sample of n from a lot of N

If the number of defective, d If the number of defective,

is less or equal to acceptance d is greater than acceptance
number, c; accept the lot. number, c; reject the lot.

Figure 23-17: Flow chart for single sampling plan

 Statistical Quality Control  557

Operating Characteristics Curve for SS Plan

There are following steps for plotting the operating characteristics (OCs) curve for single
sampling Plan:
(a) Assume process average, P0 value between 0–8 per cent. Take at least 7–8, P0 values to
plot smooth curve.
(b) Find the value of probability of acceptance, Pa using Poisson distribution table given in
Appendix III of this book (take cumulative value).
Pa = Pc or less
The Poisson distribution is used for sampling plans involving the number of defects or
defects per unit rather than the number of defective parts. It is also used to approximate
the binomial probabilities involving the number of defective parts when the sample (n)
is large and p0 is very small. When n is large and p0 is small, the Poisson distribution
formula may be used to approximate the binomial. Using the Poisson distribution
to calculate probabilities associated with various sampling plans is relatively simple
because the Poisson tables can be used.
nP
Pa = 0 e − nP0
c
(c) Plot the curve between 100P0 and Pa. The nature of the curve is shown in Figure 23.18.

Pa

0
100p0

Figure 23-18: OC curve for single sampling plan

Example 23.9: Draw the OC curve for single sampling plan; given N = 3000; n = 100; c = 2.

Solution:
To draw an OC curve, we have to take some arbitrary value of p0. The value of p0 can be taken
from 0.01 to 0.07 at close interval for smooth curve. The corresponding value of probability of
acceptance Pa (cumulative) can be taken from Poisson distribution table given in the ‘Appendix 4’.
The graph is plotted between 100p0 and Pa. The value of Pa = P2 or less corresponding to np0 is
collected in Table 23.6.
558  Industrial Engineering and Management

Table 23-6: Pa corresponding to different value of np0 at c = 2

p0 100p0 np0 Pa
(Process average) (Percentage process (Number defective) (Probability of
average) acceptance)
0 0 0 1
0.01 1 1 0.92
0.02 2 2 0.677
0.025 2.5 2.5 0.543
0.03 3 3 0.423
0.035 3.5 3.5 0.321
0.04 4 4 0.238
0.045 4.5 4.5 0.174
0.05 5 5 0.125
0.055 5.5 5.5 0.093
0.06 6 6 0.062
0.07 7 7 0.029

The graph shown in Figure 23.19 is slight deviated from the ideal graph due to lack of
exactness in values of Pa corresponding to different np0.

0.8

0.6
Pa
0.4

0.2

0
0 1 2 2.5 3 3.5 4 4.5 5 5.5 6 7
100p0

Figure 23-19: Single sampling plan for N = 3000, n = 90, and c = 1

23.15.2 Double Sampling (DS) Plan

The decision about the sample is based on one or two samples depend on the acceptance and
rejection number of first and second samples. The following are procedures for DS plan:
(a) A lot size (N) of the product is taken to check the quality.
(b) Two sample sizes (n1 and n2); two acceptance numbers (c1 and c2); and two rejection
numbers (r1, r2) are specified.
(c) Take the first sample of size n1.
 Statistical Quality Control  559

(d) If the number of defectives in the first sample does not exceed c1, the lot is accepted and
the second sample is not required.
(e) If the number of defectives (d1) in the first sample exceeds r1, the lot is rejected and the
second sample is not required.
(f) If the number defectives in the first sample are more than c1 but less than r1, a second
sample n2 is selected and inspected.
(g) If the number of defectives in combined first and second sample does not exceed c2, the
lot is accepted.
(h) If the number of defectives in combined first and second sample exceeds c2, the lot is
rejected.

Flow Chart for Double Sampling Plan

Flow chart for the DS plan is shown in Figure 23.20

Inspect a sample of n1 from a lot of N

If the number of defective, If c1 < d1 < r1, take a If the number of defective, d1
d1 is less or equal to second sample of size n2 is greater than the acceptance
the acceptance number, c1; and inspect them. number, c1; reject the lot.
accept the lot.

If the number of defectives in If the number of defectives

samples, d1 + d2 is less than or in samples, d1 + d2 is
equal to c2; accept the lot. greater than c2;
reject the lot.

Figure 23-20: Flow chart for double sampling plan

OC Curve for Double Sampling Plan

The following are the steps for plotting the OC curve for double sampling plan:
(a) Assume process average, p0 value between 0 and 8 per cent. Take at least 7–8 p0 values
to plot smooth curve.
(b) Find the value of probability of acceptance, Pa using Poisson distribution table given in
‘Appendix 4’ of this book (take cumulative value).
Pa = ( Pa )I + ( Pa )II

(
where   ( Pa ) = Pc or less I
I 1
)
r1 > di
and ( Pa )II = ∑ (P ) (P
c1 < di
di ( c2 − di ) or less )
I
560  Industrial Engineering and Management

(c) Plot the curve between 100p0 and Pa. The nature of the curve is shown in Figure 23.21.

Pa Pa II

Pa
Pa I

0 100p0

Figure 23-21: OC curve for double sampling plan

Example 23.10: Draw double sampling plan; given N = 3000; n1 = 200; c1 = 2; r1 = 5;

n2 = 250; c2 = 5.

Solution:
For DS plan, we have to find the probability of acceptance of the first sample for the defective less
than or equal to the acceptance number for the first sample, i.e., 2. If the number of defectives
in the first sample is in between acceptance number (2) and rejection number (5) of the first
sample, draw second sample of size 250. If the number of defective in the first sample is 3,
the maximum defective allowed in the second sample will be 2 or less, so that total defective
should not exceed the acceptance number of the second sample (i.e. 5). Similarly, if the number
of defective in the first sample is 4, the maximum defective allowed in the second sample will
be 1 or less. The probability of acceptance based on the second sample is found. Finally, the
summation of the probabilities of acceptance based on the first sample and the second sample are
combined as shown in Figure 23.22 using the following formula (Table 23.7).
1
0.9
0.8
0.7 Pa II
0.6
Pa 0.5
0.4
0.3 Pa I
0.2
0.1
0
1 2 3 4 5 6 7 8 9 10 11
100p0

Figure 23-22: Double sampling plan for N = 3000, n1 = 200; c1 = 2; r1 = 5; n2 = 250; c2 = 5; r2 = 6

 Statistical Quality Control  561

( Pa )I = ( P2 or less )I
( Pa )II = ( P3 )I ( P2 or less )II + ( P4 )I ( P1 or less )II
Pa = ( Pa )I + ( Pa )II

where
( Pa )I is probability of acceptance based on the basis of first sample;
( P2 or less )I is probability of acceptance based on the first sample for the defective less or equaal to 2;
( Pa )II is probability of acceptance based on second sample;
( P3 )I is probability of acceptance for the defective exactly equal to 3 in first sample;
( P2 or less )II is probability of acceptance of for the defective less or equal to 2 in second sample;
( P4 )I is probability of acceptance for the defective exactly equal to 4 in first samplee; and
( P1 or less )II is probability of acceptance of for the defective less or equal to 1 in second sample.

Table 23-7: Calculation of probability of acceptance for double sampling plan (Values are taken from
Poisson distribution table from the ‘Appendix 5’ of this book.)

p0 100p0 n1  p0 (Pa)I, (a) (P3)I (P4)I n2  p0 (P2 or less)II (P1 or less)II (Pa)II, (b) Pa = (a) + (b)

0 0 0 1 0 0 1
0.01 1 2 0.677 0.180 0.090 2.5 0.543 0.287 0.123 0.8
0.02 2 4 0.238 0.195 0.195 5 0.125 0.041 0.032 0.27
0.025 2.5 5 0.125 0.140 0.176 6.25 0.054 0.014 0.010 0.135
0.03 3 6 0.062 0.089 0.134 7.5 0.022 0.005 0.002 0.064
0.035 3.5 7 0.029 0.052 0.091 8.75 0.008 0.002 0.000 0.029
0.04 4 8 0.014 0.029 0.057 10 0.002 0.000 0.000 0.014
0.045 4.5 9 0.006 0.015 0.034 11.25 0.0007 0.000 0.000 0.006
0.05 5 10 0.002 0.007 0.019 12.5 0.000 0.000 0.000 0.002
0.055 5.5 11 0.001 0.004 0.010 13.75 0.000 0.000 0.000 0.001
0.06 6 12 0.000 0.002 0.005 15 0.000 0.000 0.000 0.000

23.15.3 Multiple Sampling Plan

Multiple sampling plan is very similar to the DS plane in which third sample, fourth sample and
so on are used for the decision of acceptance and rejection of the lot. It depends on the number
of defectives occurred in different samples.