Soo King Lim

1.0 Attribute Statistical Process Control Chart .................................. 3

1.1 Monitoring Proportion of Defective Part in a Lot ........................ 3
Example 1 ......................................................................................................... 4
1.2 Monitoring Number of Defective Part in a Lot............................. 5
Example 2 ......................................................................................................... 5
1.3 Monitoring Number of Defective Part ........................................... 6
Example 3 ......................................................................................................... 7
1.4 Monitoring Average Number of Defective Part ............................ 8
Example 4 ......................................................................................................... 9
Example 5 ....................................................................................................... 10
1.5 Average Run Length of p-Control Chart .................................... 11
Example 6 ....................................................................................................... 12

-1-
 Soo King Lim

Figure 1: Data of example 1 ................................................................................... 4

Figure 2: Attribute statistical process control chart for example 2 ........................ 6
Figure 3: Data of example 3 ................................................................................... 7
Figure 4: c control chart of example 3 ................................................................... 8
Figure 5: Data of example 4 ................................................................................... 9
Figure 6: U control chart of example 4 ................................................................ 10
Figure 7: Data of example 5 ................................................................................. 10
Figure 8: Control limits of U chart for i = 1, 2, 3, 4, 5, 6, 7, and 8 .................... 11
Figure 9: U control chart of example 5 ................................................................ 11

-2-
 Soo King Lim

1.0 Attribute Statistical Process Control Chart

In the case where quality is measured as attribute such as the number of defect in a
component or a product or a batch of components or products, number or proportion
of defectives in a batch, etc., attribute control charts are used. The attribute control
charts that will be discussed are proportion of defective part in a lot, the number of
defective part in a lot, number of defective part, and average number defective part.

1.1 Monitoring Proportion of Defective Part in a Lot

The proportion of defective part in a lot is denoted by p. It is defined as the number
of defective part divided by the size in a batch. The test hypothesis is the null
hypothesis H0: p = p0 versus the alternative H1: p  p0. Note that for a manufacturer,
it wishes that p0 is the target whereby it should be as small as possible.

The test statistic is the sample proportion of defectives p or p̂ = x/n, where n

is the sample size of the batch and x is number of defective part in the sample batch.
The general formulae of control limit are Ep Z / 20 , where E p  p0 and standard
p0 (1  p0 )
deviation is 0  Stdev  p   . Hence the control limits are
n

p0 (1  p0 )
LCL or UCL = p0  Z  / 2 (1)
n

p0 (1  p0 )
For α = 0.0026, the control limits are p0  3 . The central line is p0 .
n

If p0 value is not specified, it must be estimated from the data. Let’s assume
that there are m sample batches, each with sample size n are collected and that the
total number of defective part in these m sample batches is d then the estimate of the
proportion of defectives per sample batch of sample size n is p = d/(mn), p̂0 and p .
Then the control limits are

p (1  p )
LCL  p  3 (2)
n

CL  p (3)

p (1  p )
UCL  p  3 (4)
n

-3-
 Soo King Lim

The control chart is called a p chart.

Example 1
The readings in Fig. 1 are the number of defective parts in 18 sample batches, each
containing a total of 50 items i.e. n = 50. Establish the control limits.

Solution
The total number of defective parts in the 18 sample batches collected is 229. Each
sample batch has 50 items. Thus, the average fraction of defective in these 18 sample
batches is p = 229/(18x50) = 0.254. The limits using average fraction of defective
p (1  p ) 0.254(1  0.254)
as the estimate of p0 are p  3 ; 0.254  3 , which are 0.0693
n 50
and 0.439 and the center line is 0.254.

Batch Number Number of Defect Proportion of Defect

i xi p  x i / 50
1 9 0.18
2 10 0.20
3 11 0.22
4 13 0.26
5 13 0.26
6 8 0.16
7 18 0.36
8 12 0.24
9 11 0.22
10 8 0.16
11 14 0.28
12 21 0.42
13 18 0.36
14 10 0.20
15 8 0.16
16 18 0.38
17 19 0.38
18 8 0.16
p 0.254
Figure 1: Data of example 1

In this example, all the p values are within the calculated control limits. Hence,
these limits can be used for monitoring p. If one or more p value falls outside the
limits, then these values have to be removed and a new p is calculated. This
procedure has to be repeated until all the p values used in the estimation of p are
within the control limits.

-4-
 Soo King Lim

1.2 Monitoring Number of Defective Part in a Lot

The number of defective part in a lot is denoted by np. The hypotheses tested are;
null hypothesis H0: mean number of defectives = np0 versus alternative hypothesis
H1: mean number of defectives  np0.

The test statistic used is the number of defective part x in a sample batch of size
n, which is denoted as np. The control limits of this chart are obtained by multiplying
the lower and upper control limits of the p chart by n, which are

 p (1  p0 ) 
LCL or UCL n  p0  3 0  for α = 0.0026 (5)
 n 

also are equal to np0  3 np0 (1  p0 ) .

In equation (5), p0 can be replaced by its estimate, which is p . The center line
CL is CL = np0. This is called np chart.

Example 2
Using data shown in Fig. 1, establish the control limits for number of defective part
in the lot and plot the control chart.

Solution
The control limits are calculated using p = 0.254 and n = 50, and equation
 p (1  p0 )   0.254(1  0.254) 
LCL or UCL n  p0  3 0 , which is 500.254  3  after
 n   50 
substituting p0 value and n value.

It yields lower control limit LCL equal to 3.45 and upper control limit UCL equal
to 21.95.

The statistical control chart is shown in Fig. 2. It may sound ridiculous to have
fractional control limits since the data is non parametric type. Nevertheless, it is
acceptable for academic purpose.

-5-
 Soo King Lim

Figure 2: Attribute statistical process control chart for example 2

1.3 Monitoring Number of Defective Part

If the quality of a component or product is measured in terms of the number of defect
per component or product or batch then a c control chart is used. The letter c here
denotes the number of defect per component or product or per some appropriate
units of the product. Examples are the number of defect per 10 yards of a cable or
per 1.0 square yard of an aluminum sheet, etc. or the number of defect in a sample
of size n.

The test hypothesis is; null hypothesis H0: mean number of defect per piece or
some units or per batch = c0 versus alternative hypothesis H1: mean number of defect
per piece or some units or product c0.

The test statistic is the number of defect per the appropriate unit, which is c.
The expected value and the standard deviation with assumption that c obeys a
Poisson distribution are E(c) = c0; standard deviation of c = c0. Poisson distribution
are E(c) = c0 and standard deviation of c = c0. Hence, the control limits c0 ± Zα/2 c0
becomes equation (6) taking α is α = 0.0026.

-6-
 Soo King Lim

c0 ± 3 c0 (6)

The center line is c0. If c0 cannot be specified, it can be estimated by the mean
number of defect from one or more sample batches, which is denoted by c then, the
control limits become c  3 c .The center line is located at c .

Example 3
Data shown in Fig. 3 are the number of defect in 10 sample batches, where each
batch contains 15 items. Establish the control limits of number of defect in batch
and plot the control chart.

Number of
Batch number i
Defect in Batch c
1 18
2 12
3 14
4 9
5 16
6 10
7 14
8 15
9 19
10 17
Figure 3: Data of example 3

Solution
Assuming that the process is in control when these observations are collected, the
calculated limits of the c chart are c = (18 + 12 + 14 + 9 + 16 + 10 + 14 + 15 + 19
+ 17)/10 = 12.4.

The control limits are LCL = 12.4 - 3 12.4 = 1.84 and UCL = 12.4+ 3 12.4 = 22.96.
The center line CL is 12.4. All the c values used in the estimation are within the
control limits as is shown in Fig. 4.

-7-
 Soo King Lim

Figure 4: c control chart of example 3

1.4 Monitoring Average Number of Defective Part

If the quality is measured in terms of the average number of defect per unit and it is
not the actual number of defects per unit or the actual number of defect per sample
batch then a U control chart is used. Here, U denotes the average number of defects
per unit and c denotes the actual number of defect per sample batch containing n
items.

If U0 denotes the in-control or targeted mean of the average number of defect

per unit then the hypothesis being tested is; null hypothesis H0: mean number of
average defect per unit equal to U0 versus alternative hypothesis H1: mean number
of average defect per unit is not equal U0.

The test statistic used is the average number of defect per unit U. The expected
value E(U) and standard deviation of U are

E(U) = U0 (7)

U = U0 / n (8)

where U = c/n; E(c) = c0; Var(c) = c0; E(U) = c0/n = U0; Var (U) = Var(c)/n2 = c0/n2
= nU0/n2 = U0/n. n is the sample size. Hence, the lower and upper control limits of
the U chart are

LCL or UCL  U0  Z / 2 U0 / n (9)

-8-
 Soo King Lim

or U0  3 U0 / n for α = 0.0026. The center line is at U0. If U0 cannot be specified, it

is estimated by the sample mean number of defects per unit from observations
collected when the process is in control. Let this estimate be U then the control
limits are

U3 U/n (10)

The center line CL is U .

Example 4
The data set shown in Fig. 5 contains the number of defect in eight sample batches,
where each batch contains 15 items. Assuming that the process is in control when
these observations are collected, calculate the control limits of the U statistical
process control chart and plot the control chart.

Batch Number Number of Defect in Average Number of

i Batch c Defect per unit U = c/15
1 18 1.20
2 12 0.80
3 7 0.47
4 9 0.60
5 16 1.07
6 12 0.80
7 14 0.93
8 15 1.00
8

c i
U i 1
= 0.86
8x15
Figure 5: Data of example 4

Solution
The average number of defect U is U = [18 + 12 + 7 + 9 + 16 + 12 + 14 + 15]/(8x15)
= 0.86.

The control limits are U  3 U / n = 0.86  3 0.86 / 15 , which are 0.14 and 1.58. The
center line CL is 0.86. The statistical control chart is shown in Fig. 6.

-9-
 Soo King Lim

Figure 6: U control chart of example 4

Example 5
The data of the number of defect found in eight sample batches with unequal sample
batch size are given in Fig. 7. Calculate the control limits of the U chart and plot the
control chart.

Batch Sample Number of Defect Average Number of

Number i Size ni in Batch ci Defect per unit Ui = ci/ni
1 2 5 5/2 = 2.5
2 3 6 6/3 = 2.0
3 2 2 2/2 = 1.0
4 5 7 7/5 = 1.4
5 8 9 9/8 = 1.1
6 10 15 15/10 = 1.5
7 3 7 7/3 = 2.3
8 7 14 14/7 = 2.0
Total 40 65 U  1.63
Figure 7: Data of example 5

Solution
In this example, the sample sizes are not equal. Hence each sample batch will have
its own control limits. This is because the control limits U  3 U / n are functions of
the sample size. Grand mean U is calculated using all observations. Thus, U is U
= [5 + 6 + 2 + 7 + 9 + 15 + 7 +14]/[2 + 3 + 2 + 5 + 8 + + 3 + 7] = 65/40 = 1.63.

The control limits are equal to U  3 U / ni , where the subscript i denotes batch i.
The center line CL is 1.63.
For i = 1, the control limits are 1.63  3 1.63 / 2 , which are -1.08 and 4.34 respectively.
Since negative value is not permitted, the control limits for i = 1 are lower control
- 10 -
 Soo King Lim

limit LCL = 0, center line CL = 1.63, and upper control limit UCL = 4.34. The rest
of control limits for i value batches are shown in Fig. 8.

Average
Batch Sample Size
Number of LCLi CL UCLi
Number i ni
Defect Ui
1 2 2.5 0.00 1.63 4.34
2 3 2.0 0.00 1.63 3.84
3 2 1.0 0.00 1.63 4.34
4 5 1.4 0.00 1.63 3.34
8 8 1.1 0.28 1.63 2.98
6 10 1.5 0.42 1.63 2.84
7 3 2.3 0.00 1.63 3.84
8 7 2.0 0.18 1.63 3.08
Figure 8: Control limits of U chart for i = 1, 2, 3, 4, 5, 6, 7, and 8

The control limits of eight sample batches are given in Fig. 8. It can be seen that the
average numbers of defect Ui’s for all sample batches are within their respective
control limits. The U control chart is shown in Fig. 9.

Figure 9: U control chart of example 5

1.5 Average Run Length of p-Control Chart

The proportion of defective part in the lot is p or p̂ = x/n, where n is the sample
size of the batch and x is number of defective part in the sample batch. The formulae
p(1  p )
of control limits are p  Z  / 2 . For  = 0.0026, they are equal to
n

- 11 -
 Soo King Lim

p(1  p )
p3 . The probability q for m out of n samples from the lot is defective is
n
given by


q p
m
(11)

Thus, the average run length of p-control chart is

ARL = 1 / p 
m
(12)

Example 6
Let’s take the earlier illustration of Shewart control chart that sample size is n = 6,
the probability of the plotted point outside the control limit is p = 1 -
Pr 5.44  Z  0.55 = 1 - (0.7088 - 0) = 0.2912 and the ARL is 3.43. Calculate the
ARL of p-control chart using this data for 2 out of 6 and 6 out of 6 samples are
defective. Comment your results.

Solution
The average run length of the p-control chart for 2 out of 6 samples are defective is
ARL = 1 / p  = 1 / 0.29122 = 11.79.
m

The average run length of the p-control chart for 6 out of 6 samples are defective is
ARL = 1 / p  = 1 / 0.29126 = 1,640.
m

For 2 out of 6 samples are defective case, it would most probable occur once
between 11 and 12 sample batches.

For 6 out of 6 samples are defective case, it would most probable occur once every
1,640 sample batches.

- 12 -
 Soo King Lim

A P
Average run length ................................................. 10 p chart ....................................................................... 3
C U
c control chart ........................................................... 5 U control chart .......................................................... 7
N
np chart ..................................................................... 4

- 13 -