Lecture on Control Chart

Control Chart
Based on the theory of probability and sampling, Shewhart’s control charts
provide a powerful tool of discovering and correcting the assignable causes of
variation outside the “stable pattern” of chance causes, thus enabling us to
stabilize and control our processes at desired performances and thus bring the
process under statistical control.

In industry one is faced with two kinds of problems:

➢ To check whether the process is conforming to standards laid down, and
➢ To improve the level of standard and reduce variability consistent with
cost considerations.
Shewhart’s control charts provide an answer to both.
Therefore, control chart, as conceived and devised by Schewhart’s, is a simple
pictorial device for detecting unnatural patterns of variations in data resulting
from repetitive processes i.e., control charts provide criteria for detecting lack of
statistical control.

Control charts are simple to construct and easy to interpret and tell us at a
glance whether the sample point fall within 3 −  control limits (discussed
below) or not. Any sample point going outside the 3 −  control limits is an
indication of the lack of statistical control, i.e., presence of some assignable
causes of variation which must be traced, identified and eliminated.

A typical control chart consists of the following three horizontal lines:

▪ A central line (CL) to indicate the desired standard or the level of
the process.
▪ Upper control limit (UCL) and
▪ Lower control limit (LCL),

1
Together with a number of sample points as exhibited in the following diagram
which depicts the principle of Shewhart’s control chart.

In the control chart, Upper control Limit (UCL) and Lower Control Limit (LCL)
are usually plotted as dotted line and central line (CL) is plotted as a bold
(dark) line. If 𝑡 is the underlying statistic then these values depend on the
sampling distribution of 𝑡 and are given by

UCL = E(t) +3 S. E.(t)

LCL = E(t) – 3 S.E.(t)
CL = E(t)

3 −  Control Limits
3 −  Limits were proposed by Dr. Shewhart for his control charts from various
considerations, the main being probabilistic considerations.
Consider the statistic t = t ( x1 , x2 ,..., xn ) , a function of the sample observations
x1 , x2 ,..., xn . Let

E ( t ) = t and Var ( t ) =  t2

If the statistic𝑡 is normally distributed, then from the fundamental area property
of the normal distribution, we have

2
 P  t − 3 t  t  t + 3 t  = 0.9973
 P  t − t  3 t  = 0.9973
 P  t − t  3 t  = 0.0027

In other words, the probability that a random value of 𝑡 goes outside the 3 − 
limits, viz., t  3 t is 0.0027, which is very small. Hence, if 𝑡 is normally
distributed, the limits of variations should be between t + 3 t and t − 3 t which
are termed respectively the Upper Control Limit (UCL) and Lower Control
Limit (LCL).

Tools for Statistical Quality Control (SQC)

The following four, separate but related techniques, are the most important
statistical tools for data analysis in quality control of the manufactured products.

➢ Shewhart’s Control Chart for Variables i.e., for a characteristic which can
be measured quantitatively. Many quality characteristics of a product are
measurable and can be expressed in specific units of measurement such as
diameter of a screw, tensile strength of steel pipe, specific resistance of a
wire, life of an electric bulb, etc.
Such variables are of continuous type and are regarded to follow normal
probability law. For quality control of such data, two types of control charts
are used and technically these charts are known as:
(a) Charts for X (mean) and R (range),
(b) Charts for X (mean) and  (standard deviations).

➢ Shewhart’s Control Chart for fraction Defective or p-chart. This chart is

used if we are dealing with attributes in which case the quality
characteristics of the product are not amenable to measurement but can be
identified by their absence or presence form the product or by classifying
the product as defective or non-defective.
3
➢ Shewhart’s Control Chart for the ‘Number of Defects’ per unit or c-
Chart. This is usually used with advantage when the characteristics
representing the quality of a product is a discrete variable e.g., (i) the
number of defective rivets in an aircraft wing and (ii) the number of surface
defects observed in a roll of coated paper or a sheet of photographic film.

Control Charts for Variables

These charts may be applied to any quality characteristic that is measurable. In
order to control a measurable characteristic, we have to exercise control on the
measure of location as well as the measure of dispersion. Usually X and R
charts are employed to control the mean (location) and standard deviations
(dispersion) respectively of the characteristic.

Mean ( X ) and range (R) charts. No production process is perfect enough to

produce all the items exactly alike. Some amount of variation, in the produced
items, is inherent in any production scheme. This variation is the totality of
numerous characteristics of the production process viz., raw material, machine
setting and handling, operators, etc. As pointed out earlier, this variation is
the result of
(i) Chance causes and
(ii) Assignable causes.
The control limits in the X and R charts are so placed that they reveal the
presence or absence of assignable causes of variation.

(a) Average – mostly related to machine setting, and

(b) Range – mostly related to negligence on the part of the operator.
Steps for X and R charts

4
Measurement: Actually, the work of a control chart starts first with
measurements. Any method of measurement has its own inherent variability.
Errors in measurement can enter into the data by:
▪ the use of faulty instruments,
▪ lack of clear-cut definitions of quality characteristics and the method of
taking measurements, and
▪ lack of experience in the handling or use of the instrument, etc.

Since the conclusions drawn from control chart are broadly based on the
variability in the measurements as well as the variability in the quality being
measured, it is important that the mistakes in reading measurement instruments
or errors in recording data should be minimized so as to draw valid conclusions
form control charts.

Selection of Samples or Sub-groups: In order to make the control chart

analysis effective, it is essential to pay due regard to the rational selection of the
samples or sub-groups. The choice of the sample size 𝑛 and the frequency of
sampling i.e., the time between the selection of two groups, depend upon the
process and no hard and fast rules can be laid down for this purpose. Usually 𝑛
is taken to be 4 or 5 while the frequency of sampling depends on the state of the
control exercised. Initially more frequent samples will be required (15 to 30
minutes) and once a state of control is maintained, the frequency may be
relaxed. Normally 25 samples of size 4 each or 20 samples of size 5 each under
control will give good estimate of the process average and dispersion.

Remark: While collecting data it may not be necessary to go exactly at the

specified time; in fact, this should not be practiced. This is to avoid (i) the

5
operative being careful at the time of sampling, or (ii) any periodicities of the
process to coincide with sampling.

▪ Calculation of X and R for each sub-group.

Let X ij , j = 1, 2,..., n be the measurements on the ith sample (i = 1, 2, ..., k). The

mean 𝑋̅𝑖 , the range Ri and the standard deviation si for the ith sample are given
by
1
𝑋̅𝑖 = ∑ 𝑋𝑖𝑗
𝑛
𝑗
𝑅𝑖 = max 𝑋𝑖𝑗 − min 𝑋𝑖𝑗 (𝑖 = 1, 2, 3, … , 𝑘) (1.1)
𝑗 𝑗
1 2
𝑠𝑖2 = ∑(𝑋𝑖𝑗 − 𝑋̅𝑖 )
𝑛
𝑗 }
Next we find 𝑋̿(the averages of sample means), ̅𝑅(average of sample ranges)
and s (average of sample standard deviations) as follows:

1
𝑋̿ = ∑ 𝑋̅𝑖
𝑘
𝑖
1
𝑅̅ = ∑ 𝑅𝑖 … (1.2)
𝑘
𝑖
1
𝑠̅ = ∑ 𝑠𝑖
𝑘 }𝑖

▪ Setting of Control Limits

It is well known that if  is the process standard deviation (standard deviation
of the universe from which samples are taken), then the standard error of sample
mean
S .E. ( X i ) =  / n , (i = 1, 2,...k )

Also, from the sampling distribution of range we know that

6
 E ( R ) = d 2 .

Where d 2 is a constant depending on the sample size? Thus, an estimate of 

can be obtained from R by the relations:

R = d2 .   = R / d2 … (1.3)

Also X gives an unbiased estimate of the population mean  , since

𝑘
1
𝐸(𝑋̿) = ∑ 𝐸(𝑋̅𝑖 ) = 𝜇
𝑘
𝑖=1

Control Limits for X chart:

Case 1: When standards are given i.e., both  and  are known.
The 3 −  control limits for X chart are given by
3𝜎
𝐸(𝑋̄) ± 3𝑆. 𝐸. (𝑋̄) = 𝜇 ± = 𝜇 ± 𝐴𝜎, (𝐴 = 3/√𝑛)
√𝑛

If 𝜇′ and   are known or specified values of  and  respectively, then

UCLx =   + A 
(1.4)
LCLx =   − A  

Case 2: Standards not given, If both  and  are unknown, then using their
estimates X and  given in (1.2) and (1.3) respectively, we get the 3 − 
control limits on the X chart as:
R 1  3   3 
X 3 . = X    R = X  A2 R ,  A2 = 
d2 n  d2 n   d 2 n 

UCLx = X + A2 R 
  (1.4a)
LCLx = X − A2 R 

7
Since d 2 is a constant depending on n, A2 = 3/ ( d2 n ) also depends only on n and

its values have been computed and tabulated for different values for n form 2 to
25 are given in Table.

If on the other hand, the control limits are to be obtained in terms of s rather
than R , then an estimate of  can be obtained from the relation
s = C2   = s / C2 (1.4b)
n−2
2  2 
!
where C2 = .
n  n−3
 !
 2 

 3  
UCLx = X +   s = X + A1s 
 nC  
 2 
 (1.4c)
 3  
LCLx = X −   s = X − A1s 
 nC 
 2  

The factor 𝐴1 = 3/(√𝑛𝐶2 ) has been tabulated for different values of n from 2 to 25 in
Table.

Control Limits for R-chart.

3 −  control limits for R-chart are given by E ( R )  3 R . E ( R ) is estimated by R

and  R is estimated from the relation

 R = cE ( R ) = cR

where c is constant depending on n.

UCLR = R + 3cR = (1 + 3c ) R = D4 R 
  (1.5)
LCLR = R − 3cR = (1 − 3c ) R = D3 R 

The values of D4 = 1 + 3c and D8 = 1 − 3c have been tabulated for different values

of n from 2 to 25 in Table.

8
However, the control limits for R-chart can be obtained directly from the
assumed or known value of  as follows:
UCLR = D2 
 (1.5a)
UCLR = D1 

Where D1 and D2 are tabulated for values of n from 2 to 25 in Table.

Remarks:
It should be noted carefully that the control limits for X and R charts are based
upon the assumption that different samples or sub-group are of constant size n.

▪ Construction of Control Chart for X and R i.e., plotting of Central Line

and the Control Limits.
Control charts are plotted on a rectangular co-ordinate axis-vertical scale
(ordinate) representing the statistical measures X and R and horizontal scale
(abscissa) representing the sample number. Hours, date or lot numbers may also
be represented on the horizontal scale. Sample points mean or ranges are
indicated on the chart by points, which may or may not be joined.

For X chart the central line is drawn as a solid horizontal line at X and UCLx
and LCLx are drawn at the computed values as dotted horizontal lines.
For R chart, the central line is drawn as a solid horizontal line at R and UCLR
is drawn at the computed values as a dotted horizontal line. If the sample size
is seven or more (n ≥ 7), LCLR is drawn as dotted horizontal line at the
computed value otherwise (n < 7) LCLR is taken as zero.

9
Criterion for detecting lack of control in X and R charts:
As pointed out earlier, the main object of the control charts is to indicate when a
process is not in control. the criteria for detecting lack of control are, therefore,
of fundamental and crucial importance. The pattern of the sample points in a
control chart is the key to the proper interpretation of the working of the
process. The following situations depict lack of control:
▪ A point outside the control limits. The probabilistic considerations provide a
basis for hunting for lack of control in such a situation a point going outside
control limits is a clear indication of the presence of assignable causes of
variation which must be searched and corrected. A point outside the control
limit may result from an increased dispersion or change in level. Lack of
uniformity may due to the variation in the quality or raw materials,
deficiency in the skill of the operators, loss of alignment among machines,
change of working conditions, etc. It may be indicated by a point (or points)
above the upper limits for ranges. It may also result in points outside the
controls limits for means.
▪ A run of seven or more points. Although all the sample points are within
control limits usually the pattern of points in the chart indicates assignable
causes. One such situation is a run of 7 or more points above or below the
central line in the control chart. Such runs indicate shift in the process level.
On R-chart a run of points above the central line is indicative of increase in
process spread and therefore represents an undesirable situation, while a run
below the central line indicates an improvement in the sense that the
variability has been reduced i.e., the process could hold to closer tolerance.
▪ One or more points in the vicinity of control limited or a run of points
beyond some secondary limits e.g., a run of 2, 3 points beyond 2-  limits or
a run of 4, 5 points beyond 1-  limits.
▪ The sample points on 𝑋̅ and R charts, too close to central line, exhibit
another form of assignable-cause. This situation represents systematic
10
 differences within samples or sub-groups and results from improper
selection of samples and biases in measurements.
▪ Presence of trends. The trends exhibited by sample points on the control
charts are also an indication of assignable causes. Trend pattern a
phenomenon usually observed in engineering industry, indicates the gradual
shift in the process level. Trend may be upward or downward. Tools wear
and the need for resetting machines often accounts for such a shift, and it is
essential to determine when machine resetting becomes desirable bearing in
mind that too frequent adjustments are a serious setback production output.
▪ Presence of cycles. In some cases, the cyclic pattern of points in the control
charts (fig1.4 on page 1.16) indicates the presence of assignable causes of
variation. Such patterns are due to material or/and any mechanical reasons.

Interpretation of X and R charts

In order to judge if a process is in control, X and R charts should be examined
together and the process should be deemed in statistical control if both the
charts show a state of control. Situations exist where R-chart is in a state of
control but X charts are not. We summaries below, in a tabular form such
different situations and interpretation to be accorded to each.
Sl. Situation in
Interpretation
No. R-chart X -chart
1 In control Points beyond limits only on Level of process has shifted
one side
2 In Control Points beyond limits on both Level of process is changing
the sides in erratic manner-frequent
adjustments
3 Out of control Points beyond limits on both Variability has increased
sides
4 Out of control Out of control on one side Both level and variability
have changed
5 In control Run of 7 or more points on Shift in process level
none side of central line
6 In control Trend of 7 or more points. No Process level is gradually
point out side control limits changing
7 Runs of 7 or more _ _ _ _ _ _ Variability has increased

11
 points above central
line
8 Points too close to _ _ _ _ _ _ Systematic differences
the central line within sub-groups
9 ______ Points too close to the central Systematic differences
line within sub-group

Control chart for standard deviation or  chart:

In sampling from a population with standard deviation  , we have
n −1 2
E ( s2 ) =  (*)
n
 E ( s ) = C2 (**)

n−2
2  2 
!
Where C2 = .
n  n−3
 !
 2 

(*) is true for sampling from any distribution whereas (**) is true only for
normal population. The values of C2 have been tabulated for different values of
n from 2 to 25 in Table.

Hence in sampling from normal population, we have

Var ( s ) = E ( s 2 ) − E ( s )
2

 n −1 
= − C22   2
 n 
n −1
S .E.( s) = C3 .  where C3 = − C22
n
UCLs = E ( s ) + 3S .E. ( s ) = ( C2 + 3C3 )  = B2 . 
 (1.6)
LCLs = E ( s ) − 3S .E. ( s ) = ( C2 − 3C3 )  = B1. 
Central line = CLs = C2 .

12
Values of B1 and B2 have been tabulated for different values of n. If the value of
 is not specified or not known, then we use its estimate, based on s defined in
(1.2) as given in (1.4b) viz.,  = s / C2 . In this case
𝐶3
𝑼𝑪𝑳𝒔 = 𝐸(𝑠) + 3𝑆. 𝐸. (𝑠) = 𝑠̄ + 3 . 𝑠̄
𝐶2
3𝐶3
= (1 + ) = 𝑠̄ = 𝑩𝟒 . 𝒔̄
𝐶2
Similarly, we shall get
3𝐶3
𝑳𝑪𝑳𝒔 = (1 − ) 𝑠̄ = 𝑩𝟑 . 𝒔̄ (1.6a)
𝐶2

𝐶𝐿𝑠 = 𝑠̄
Where 𝐵3 and 𝐵4 have been tabulated for different values of 𝑛 in Table. Since 𝑛
can never be negative, if LCL given by (1.6a) comes out to be negative, as will
be the case for 𝑛 from 2 to 5, it will be taken to zero.

Control chart for attributes

In spite of wide application of X and R (or  ) charts as a powerful tool of
diagnosis of sources of trouble in a production process, their use is restricted
because of the following limitations:

1. They are charts for variables only i.e for quality characteristics which can be
measured and expressed in numbers.
2. In certain situation they are impracticable and un-economical e.g., if the
number of measurable characteristics, each of which could be a possible
candidate for X and R chart, is too large, say 30,000 or so then obviously
there can’t be 30000 control charts.

As an alternative to X and R charts, we have the control charts for attributes

which can be used for quality characteristics:

13
 i. which can be observed only as attributes by classifying an item as
defective or non-defective i.e., conforming to specifications or not and
ii. Which are actually observed as attributes even though they could be
measured as variables e.g., go and no-go gauge test results.

There are two control charts for attributes:

➢ Control chart for fraction defective (p-chart) or the number of
defectives (𝑛𝑝 or 𝑑 chart)
➢ Control chart for then number of defects per unit (𝑐 -chart).

P-chart or control chart for fraction defective:

If ‘𝑑’ is the number of defectives in a sample of size 𝑛 then the sample
proportion defective is p = d/n. Hence 𝑑 is a binomial variate with parameters 𝑛
and 𝑝.
∴ 𝐸(𝑑) = 𝑛𝑝 𝑎𝑛𝑑 𝑉𝑎𝑟(𝑑) = 𝑛𝑝𝑞, 𝑞 = 1 − 𝑝
𝑑 1
∴ 𝐸(𝑝) = 𝐸 ( ) = 𝐸(𝑑) = 𝑝
𝑛 𝑛 } (1.7)
𝑑 1 𝑝𝑞
𝑉𝑎𝑟(𝑝) = 𝑉𝑎𝑟 ( ) = 2 𝑣𝑎𝑟(𝑑) =
𝑛 𝑛 𝑛

Thus, the 3-  control limits for p-chart are given by

𝐸 (𝑝) ± 3. 𝑆. 𝐸. (𝑝) = 𝑝 ± 3√𝑝𝑞/𝑛 = 𝑝 ± 𝐴√𝑝𝑞 (1.8)

Where A = 3 n has been tabulated for different values of 𝑛 in Table.

Case (i) standards given

If 𝑝′ is the given or known value of p then
𝑈𝐶𝐿𝑝 = 𝑝′ + 𝐴√𝑝′ (1 − 𝑝′ )
𝐿𝐶𝐿𝑝 = 𝑝′ − 𝐴√𝑝′ (1 − 𝑝′ ) }
𝐶𝐿𝑝 = 𝑝′
14
Case (ii) standards not given
Let di be the number of defectives and pi the fraction defective for the ith
sample (i = 1, 2,...,k) of size ni . Then the population proportion p is estimated
by the statistic p given by
k

d n p i i
p= i
= i
(1.8b)
n i
n
k

i
i

It may be remarked here that p is an unbiased estimate of p, since

𝐸(𝑝̄ ) = ∑ 𝐸(𝑑𝑖 )/ ∑ 𝑛𝑖 = ∑(𝑛𝑖 𝑝)/ ∑ 𝑛𝑖 = 𝑝

𝑖

In this case
UCLp = p + A p (1 − p ) 


LCL p = p − A p (1 − p )  (1.8c)

CL p = p 

Control chart for number of defectives (d-chart)

If instead of p, the sample proportion defective, we used d, the number of
defectives in the sample, then the 3-  control limits for d-chart are given by

𝐸(𝑑) ± 3𝑆. 𝐸. (𝑑) = 𝑛𝑝 ± 3√𝑛𝑝(1 − 𝑝) (1.9)

Case (i) standards known

If 𝑝′ is the given value of p than
𝑈𝐶𝐿𝑑 = 𝑛𝑝′ + 3√𝑛𝑝′ (1 − 𝑝′ )
𝐿𝐶𝐿𝑑 = 𝑛𝑝′ − 3√𝑛𝑝′ (1 − 𝑝′ ) } (1.9a)
𝐶𝐿𝑑 = 𝑛𝑝′

15
Case (ii) standards not given
Using p as an estimate of p as in (1.8b), we get

UCLd = np + 3 np (1 − p ) 


LCLd = np − 3 np (1 − p )  (1.9b)

CLd = np 

Since p cannot be negative, if LCL as given by above formulae comes out to be
negative then it is taken to be zero.

*** p and d- charts for fixed sample size???

Interpretations of p -chart
1. From the p-chart a process is judged to be in statistical control in the same
way as is done for X and R charts.
If the entire sample points fall within the control limits without exhibiting
any specific pattern, the process is said to be in control. In such a case, the
observed variations in the fraction defective are attributed to the stable
pattern of chance causes and the average fraction defective p is taken as the
standard fraction defective p.
2. Points outside the UCL are termed as high spots. These suggest deterioration
in the quality and should be regularly reported to the production engineers.
The reasons for such deterioration could possibly be known and removed if
the details of conditions under which data were collected, were known. Of
particular interest and importance is if there was any change of inspection or
inspection standards.
3. Points below LCL are called low spots. Such points represent a situation
showing improvement in the product quality. However, before taking this
improvement for guaranteed, it should be investigated if there was any
slackness in inspection or not.

16
4. When a number of points fall outside the control limits, a revised estimate of
p should be obtained by eliminating all the points that fall above UCL (it is
assumed that points that fall below LCL are not due to faulty inspection).
The standard fraction defective p should be revised periodically in this way.

Control chart for number of defects per unit (C-chart)

The field of application of c-chart is much more restricted as compared to 𝑋̅ and
R charts or p-chart. Before we embark upon to discuss the theory behind c-
chart, it is imperative to distinguish between defect and detective.
An article which does not conform to one or more of the specifications, is
termed as defective while any instance of article’s lack of conformity to
specifications is a defect. Thus, every defective contains one or more of the
defects e.g., a defective casting may further be examined for blow holes, cold
shuts, rough surface, weak structure, etc.
Unlike d or np chart which applies to the number of defectives in a sample, c-
chart applies to the number of defects per unit. Sample size for c-chart may be a
single unit like a radio, or group of units or it may be a unit of fixed time,
length, area etc. For example, in case of surface defects, area of the surface is
the sample size; in case of casting defects a single part (such as base plate, side
cover) is the sample size. However, defined sample size should be constant in
the sense that different samples have essentially equal opportunity for the
occurrence of defects.

Limits for c-chart

In many manufacturing or inspection situations, the sample size n i.e., the area
of opportunity is very large (since the opportunities for defects to occur are
numerous) and the probability p of the occurrence of a defect in any one spot is
very small such that np is finite. In such situations from statistical theory we
know that the pattern of variations in data can be represented by Poisson
17
distribution, and consequently 3 − 𝜎 control limits based on Poisson
distribution are used.
Since for a Poisson distribution mean and variance are equal, if we assume that
c is Poisson variate with parameter, 𝜆, we get

𝐸(𝑐) = 𝜆 𝑎𝑛𝑑 𝑉𝑎𝑟(𝑐) = 𝜆

Thus 3 − 𝜎 control limits for c-chart are given by
𝑈𝐶𝐿𝑐 = 𝐸(𝑐) + 3√𝑉𝑎𝑟(𝑐) = 𝜆 + 3√𝜆
} … … … … . .1.11
𝑈𝐶𝐿𝑐 = 𝐸(𝑐) − 3√𝑉𝑎𝑟(𝑐) = 𝜆 − 3√𝜆
𝐶𝐿𝑐 = 𝜆

Case (i) Standard given. If 𝜆́ is the specified value of 𝜆, then

𝑈𝐶𝐿𝑐 = 𝜆́ + 3√𝜆́
𝐿𝐶𝐿𝑐 = 𝜆́ − 3√𝜆́ }……………………. (1.12a)
𝐶𝐿𝑐 = 𝜆́

Case (ii) Standards not given

If the value of 𝜆 is not known, it is estimated by the mean number of defects per
unit. Thus, if 𝑐𝑖 is the number of defects observed on the ith (i = 1, 2,..., k)
inspected unit, then an estimate of 𝜆 is given by
𝜆̂ = 𝑐̅ = ∑𝑘𝑖=1 𝑐𝑖 ⁄𝑘 ………………… (1.12)

It can be easily seen that 𝑐̅ is an unbiased estimate of 𝜆. The control limits, in

this case, are given by
𝑈𝐶𝐿𝑐 = 𝑐̅ + 3√𝑐̅
𝐿𝐶𝐿𝑐 = 𝑐̅ − 3√𝑐̅ }
𝐶𝐿𝑐 = 𝑐̅
Since c can’t be negative, if LCL given by above formulae comes out to be
negative, it is regarded as zero.

18
The central line is drawn at 𝑐̅ and UCL and LCL at the values given by (1.12a).
The observed number of defects on the inspected units is then plotted on the
control chart. The interpretations for c-chart are similar to those of p-chart.

C-chart for variable sample size or u-chart

In this case instead of plotting c, the statistic u = c/n is plotted, n being the
sample size which is varying. In ni is the sample size and ci the total number of
defects observed in the ith sample then
𝑢𝑖 = 𝑐𝑖 ⁄𝑛𝑖 (𝑖 = 1,2, … , 𝑘) …………….. (1.13)
gives the average number of defects per unit for the ith sample.

In this case an estimate of 𝜆, the mean number of defects per unit in the lot,
based on all the k-samples is given by:
1
𝜆̂ = 𝑢̅ = ∑𝑘𝑖=1 𝑢𝑖 …………………………. (1.13a)
𝑘

We know that if 𝑋̅ is the mean of a random sample of size n then S.E.(𝑋̅) =

𝜎𝑢 ⁄√𝑛. Hence the standard error of the average number of defects per unit is
given by
𝑆. 𝐸. (𝑢) = √𝜆⁄𝑛 = √𝑢̅⁄𝑛 (on using 1.13a); Where 𝜆 = 𝑢̅

Hence 𝟑 − 𝝈 control limits for u-chart are given by

𝑈𝐶𝐿𝑢 = 𝑢̅ + 3√𝑢̅⁄𝑛
𝐿𝐶𝐿𝑢 = 𝑢̅ − 3√𝑢̅⁄𝑛 }
𝐶𝐿𝑢 = 𝑢̅
As is obvious, control limits will vary for each sample. The central line,
however, will be same. The interpretation of these charts is similar to the p-chart
of d-chart.

19
Applications of c-chart
The universal nature of Poisson distribution as the law of small numbers makes
the c-chart technique quite useful. In spite of the limited field of application of
c-chart (as compared to 𝑋̅, R, p charts) there do exist situations in industry
where c-chart is definitely needed.
Some of the representative types of defects to which c-chart can be applied
with advantage are:
▪ c is number of imperfections observed in a bale of cloth
▪ c is the number of surface defects observed in (i) roll of coated paper or a
sheet of photographic film, and (ii) a galvanized sheet or a painted, plated
or enameled surface of given area
▪ c is the number of defects of all types observed in aircraft subassemblies
or final assembly
▪ c is the number of breakdowns at weak spots in insulation in a given
length of insulated wire subject to a specified test voltage
▪ c-chart has been applied to sampling acceptance procedures based on
number of defects per unit e.g., in case of inspection of fairly complex
assembled units such as T.V. sets, aircraft engines, tanks, machine-guns,
etc. in which there are very many opportunities for the occurrence of
defects of various types and the total number of defects of all types found
by inspection is recorded for each unit
▪ C-chart technique can be used with advantage in various fields other than
industrial quality control, e.g., it has been applied (i) to accident statistics
(both of industrial accidents and highway accidents), (ii) in chemical
laboratories, and (iii) in epidemiology

20