Kathmandu University

Course: MATH 208, ENVE II/II

Prepared by: Kiran Shrestha, Dr. Samir Shrestha

Introduction to Statistical Quality Control

 Statistical Quality Control is a process of maintaining quality of industrial products under some standard
specifications.
 The term „quality‟ in statistical quality control means an attribute of the product that determines its fitness for use.
 In fact, quality means level of standard which in turn depends on four M‟s and they are-
materials, manpower, machines and management.
 Two apparently identical parts made under carefully controlled conditions, from the same batch of raw material, and
only seconds apart by the same machine, can nevertheless be different in many respects.
 Any manufacturing process, however good, is characterized by a certain amount of variability, which is of random
nature and which cannot be completely eliminated.
 In fact, variation in the quality of manufactured product in the repetitive process in industry is inherent and inevitable.
 These variations are broadly classified as being due to the two causes, they are …
o Chance causes
o Assignable causes
 Chance causes of variations results from many minor causes that behave in a random manner. The variations due to
these causes are beyond the control of human hand and cannot be prevented/ eliminated under any circumstances.
 The range of such variation is known as „natural tolerance of the process‟.
 The second type of variation attributed to any production process is due to non-random causes, so called assignable
causes.
 These types of variation may occur at any stage of production process, i.e., right from the arrival of raw material to
final delivery of the manufactured products.
 Some of the important factors of assignable causes of variation are-
o defective raw materials
o new technology or operation
o negligence of operators
o improper handling of machines
o faulty equipments
o unexperienced technical staffs, etc.
 These causes can be identified and eliminated and are to be discovered in a production process before the production
becomes defective.
 Statistical Quality Control (SQC) is the method of identifying presence of assignable causes in any production
process.
 The main purpose of SQC is to devise statistical technique which would help us in separating the assignable causes
from the chance causes, thus enabling us to take immediate remedial action whenever assignable causes are present.
 A production process is said to be statistically under control, if it is governed by the random or chance causes alone,
in the absence of assignable causes of variations.
 SQC helps in the detection and correction of many production troubles and substantial improvement in the product
quality.
 It tells us when to leave a process alone and when to take action to correct troubles, thus preventing frequent and
unwanted adjustments.
 It provides better quality assurance at lower inspection cost.

1
 SQC reduce waste of time and material to the absolute minimum by giving an early warning of the cost of production
and hence may lead to more profit.
 One of the methods used to detect the presence of assignable cause of variation in a production process is „Cotrol
Chart‟.

Control Charts
Introduction
 Control chart is a simple pictorial device for detecting unnatural pattern of variation in data resulting from repetitive
processes.
 The concept of control chart was first discovered and developed by Walter A. Shewart, a physicist, in 1924.
 These charts provide criteria for detecting lack of statistical control.
 Control charts are based on the theory of probability and sampling.
 Control charts provide a powerful tool of discovering and correcting the assignable causes of variation, thus enabling
us to stabilize and control processes at desired performance and thus bring the process under statistical control.
 Control charts are simple to construct and easy to interpret.
 If fact, control charts tell us whether the sample points fall within 3 limits or not.
 Any sample point going outside the 3 limits is an indication of lack of control, i.e., presence of some assignable
causes of variation which must be identified and eliminated.

Basic Theory of Control 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 ant production scheme. This variation is the result of both chance causes and assignable causes.
While dealing with measurements, it is customary to exercise control over the average quality of a process (or mean) as
well as its variability. Considering that the measurements are sample statistics from a normal population X, we have
P(  - 3σ  X   + 3σ) = 0.99
These two limits  - 3σ and  + 3σ are used in control charts for detecting whether a process is in control or not. The limit
represented by  - 3σ is called lower control limit (LCL) and the limit represented by  + 3σ is called upper control limit
(UCL). So,
LCL =  - 3σ, and
UCL =  + 3σ
Since for normal population X,  = E(X) and σ = 𝑉(𝑋) , so it follows that
𝐿𝐶𝐿 = 𝐸 𝑋 − 3 𝑉(𝑋)
and
𝑈𝐶𝐿 = 𝐸 𝑋 + 3 𝑉(𝑋)
In control charts, values of different sample statistics such as sample mean, sample standard deviation, sample range, etc.
of a number of samples from production process are plotted for different samples. On identifying LCL and UCL in the
chart, one can observe whether the sample values lie between these two limits or not.
Types of Control Chart
Following two types of control charts are used as most important statistical tools for data analysis in quality control-
a) Shewart‟s control charts for measurements (variables)
b) Shewart‟s control charts for attributes

Shewart’s control charts for measurements (variables)

Control charts for measurement includes
(a) Control Charts of process mean and
(b) Control charts for process variability
2
 Control Chart for process mean (or 𝒙-chart)
Control charts for process mean is also called 𝒙 chart.
Case I – When both process mean and standard deviation are known
If process mean and standard deviation are known to be  and  respectively, then the central line (CL) of 𝒙 chart, is
given by , i.e.,

CL = 

The lower control limit (LCL) is given by

𝜎
𝐿𝐶𝐿 =  − 3.
𝑛
The upper control limits (UCL) are given by
𝜎
𝑈𝐶𝐿 =  + 3.
𝑛
That is,
CL = 
𝜎
LCL =  − 3. 𝑛
𝜎
UCL =  + 3. 𝑛

Example 1: A plastics manufacturer extrudes blanks for use in the manufacture of eyeglass temples. Specifications
require that the thickness of these blanks have  = 0.150 and  = 0.002 inches. Use the specifications to calculate a
central line, lower control limit and upper control limit for 𝑥 chart with n = 5.
Solution: For 𝒙 chart
Here,  = 0.150 inch.  = 0.002 inchand 𝑛 = 5
The control lines for a 𝑥 chart are given by

𝐶𝐿 = 
3
𝐿𝐶𝐿 =  − 
𝑛
3
𝑈𝐶𝐿 =  + 
𝑛

𝐶𝐿 =  = 0.150
3
𝐿𝐶𝐿 =  −  = 0.150 – 1.342 × 0.002 = 0.147
𝑛

3
𝑈𝐶𝐿 =  +  = 0.150 + 1.342 × 0.002 = 0.153
𝑛

Case II - When process mean and standard deviation are unknown

If both mean and standard deviation that means population parameters are unknown, it is necessary to estimate them on
the basis of preliminary samples. For this purpose, it is usually desirable to obtain the results of 20 𝑜𝑟 25 consecutive
samples each of size 4 𝑜𝑟 5 are taken when the process is in control.
If 𝑘 samples each of size n are used, then mean of 𝑖 𝑡ℎ sample is

3
 𝑛
1
𝑥𝑖 = 𝑥𝑖𝑗
𝑛
𝑗 =1
and the grand mean of the 𝑘 sample means is
𝑘
1
𝑥= 𝑥𝑖
𝑘
𝑖=1
This statistic is used to estimate the process mean 𝜇. Note here that it is an unbiased estimate of the process mean 𝜇.
In the same way, the process standard deviation  can be estimated either from the standard deviation or from the range of
𝑘 samples.
Case II (a) - Using sample standard deviation to estimate control lines (when both process mean and standard
deviation are unknown)
If 𝑠𝑖 is the standard deviation of 𝑖 𝑡ℎ sample, then
𝑛
1 2
𝑠𝑖 = 𝑥𝑖𝑗 − 𝑥𝑖
𝑛−1
𝑗 =1
and the mean of sample standard deviation ‘𝑠’ is
𝑘
1
𝑠= 𝑠𝑖
𝑘
𝑖=1
Thus, the control lines, when both process mean and standard deviation are not known and range is used to estimate the
standard deviation are given by

𝐶𝐿 = 𝑥
𝐿𝐶𝐿 = 𝑥 − 𝐴1 . 𝑠
𝑈𝐶𝐿 = 𝑥 + 𝐴1 . 𝑠

The values of 𝐴1 for different sample sizes are listed in the Table XIII (Factors for Quality Control, Page 633)

Example 2: Construct an 𝒙 chart for 20 samples of size 3 which had sample mean 𝑥 equal to 21.2, 19.4, 20.4, 20.4,
20.4, 19.0, 20.3, 21.1, 21.6, 22.1, 24.4, 23.9, 24.9, 24.1, 21.8, 19.5, 20.3, 22.5, 23.4, 23.3 and sample standard deviation 𝑠
is equal to 2.0, 0.8, 1.1, 0.9, 1.0, 0.3, 1.3, 2.0, 0.8, 1.0, 1.5, 1.0, 1.5, 0.8, 1.3, 2.9, 4.3, 1.2, 0.3, 3.1

For 𝒙 chart,
21.2 + 19.4 … … … … … … … … … … … … … . +23.3
𝑥= = 21.7
20
2.0 + 0.8 + ⋯ … … … … … … … … … … … … . +3.1
𝑠= = 1.455
0

The control lines for 𝑥 chart are given by

𝐶𝐿 = 𝑥
𝐿𝐶𝐿 = 𝑥 − 𝐴1 . 𝑠
𝑈𝐶𝐿 = 𝑥 + 𝐴1 . 𝑠

From table (See table XIII, Page 633) for 𝑛 = 3, 𝐴1 = 2.394

𝐶𝐿 = 𝑥 = 21.7

4
𝐿𝐶𝐿 = 𝑥 − 𝐴1 . 𝑠 = 21.7 − 2.394 × 1.455 = 18.22
𝑈𝐶𝐿 = 𝑥 + 𝐴1 . 𝑠 = 21.7 + 2.394 × 1.455 = 25.18

30

25 UCL

CL
20
LCL

15

10

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

Figure: 𝑥 chart

Case II (b) - Using sample range to estimate control lines (when both process mean and standard deviation are
unknown)
Since the sample size commonly used in connection with the control charts for measurements is small, there is
usually very little loss of efficiency in estimating  from sample ranges.
Let 𝑅𝑖 be the range of 𝑖 𝑡ℎ sample, and the mean of ranges of 𝑘 samples be 𝑅, i.e.,
𝑘
1
𝑅= 𝑅𝑖
𝑘
𝑖=1
Thus, the control lines, when both process mean and standard deviation are not known and range is used to estimate the
standard deviation are given by

𝐶𝐿 = 𝑥
𝐿𝐶𝐿 = 𝑥 − 𝐴2 . 𝑅
𝑈𝐶𝐿 = 𝑥 + 𝐴2 . 𝑅

The values of 𝐴2 for different sample sizes are listed in the Table XIII (Factors for Quality Control, Page 633)

Control Charts for process variability

In controlling a process, it may not be enough to monitor the process mean, but it is also required to monitor process
variability. Although an increase in process variability may become more apparent from increased fluctuations on the 𝑥 ‟s,
a more sensitive test of shifts in process variability is provided by separate control charts namely R-chart.

5
R-Chart
R-chart measures process variability based on sample ranges. In this chart, it is assumed that the distribution of range of
samples is approximately normal. We consider 𝑘 samples each of size n from a process. Let 𝑅𝑖 be range of 𝑖 𝑡ℎ sample. By
sampling distribution of sample range, we can see that
𝐸 𝑅𝑖 = 𝑅
𝑘
1
𝑤ℎ𝑒𝑟𝑒, 𝑅= 𝑅𝑖
𝑘
𝑖=1
In this case since the sample size commonly used in connection with the control charts for measurements is small, there is
usually very little loss of efficiency in estimating  from sample ranges.
As such following relations of sampling distribution of sample range are used to devise control limits.
So, the control lines for R-chart are given by

𝐶𝐿 = 𝑅
𝐿𝐶𝐿 = 𝐷3 𝑅
𝑈𝐶𝐿 = 𝐷4 𝑅

The values of 𝐷3 𝑎𝑛𝑑 𝐷4 for different sample sizes are listed in the Table XIII (Factors for Quality Control, Page 633)

Example 3: In 20 successive random samples of size 5, following values of means and ranges are obtained
Sample Mean Sample No.
No. Range Mean Range
1 0.152 0.004 11 0.149 0.003
2 0.147 0.006 12 0.153 0.004
3 0.153 0.004 13 0.15 0.005
4 0.153 0.002 14 0.152 0.001
5 0.151 0.003 15 0.149 0.003
6 0.148 0.002 16 0.146 0.002
7 0.149 0.006 17 0.154 0.004
8 0.144 0.001 18 0.152 0.005
9 0.149 0.003 19 0.151 0.002
10 0.152 0.005 20 0.149 0.004
Calculate 𝑥 𝑎𝑛𝑑 𝑅 of these data and use them to construct 𝑥 chart and R chart.
 Here,
0.152 + 0.147 + ⋯ … … … … … … … … … … … … . +0.149
𝑥= = 0.15015
20
0.004 + 0.006 + ⋯ … … … … … … … … … … … … . +0.0.004
𝑅= = 0.00345
20

For 𝒙 chart
The control lines are given by
𝐶𝐿 = 𝑥
𝐿𝐶𝐿 = 𝑥 − 𝐴2 . 𝑅
𝑈𝐶𝐿 = 𝑥 + 𝐴2 . 𝑅
From table (See table XIII, Page 633) for 𝑛 = 5, 𝐴1 = 1.596 𝑎𝑛𝑑 𝐴2 = 0.577

6
𝐶𝐿 = 𝑥 = 0.15015
𝐿𝐶𝐿 = 𝑥 − 𝐴2 . 𝑅 = 0.15015 − 1.596 × 0.00345 = 0.144644
𝑈𝐶𝐿 = 𝑥 + 𝐴2 . 𝑅 = 0.15015 + 1.596 × 0.00345 = 0.155656

0.156 UC
0.154
0.152
0.15 CL
0.148
0.146
LCL
0.144
0.142
0.14
0.138
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Figure: 𝑥 chart
For 𝑹 chart
The control lines are given by
𝐶𝐿 = 𝑅
𝐿𝐶𝐿 = 𝐷3 𝑅
𝑈𝐶𝐿 = 𝐷4 𝑅

From table (See table XIII, Page 633) for 𝑛 = 5, 𝐷3 = 0 𝑎𝑛𝑑 𝐷4 = 2.115
𝐶𝐿 = 𝑅 = 0.00345
𝐿𝐶𝐿 = 𝐷3 𝑅 = 0 × 0.00345 = 0
𝑈𝐶𝐿 = 𝐷4 𝑅 = 2.115 × 0.00345 = 0.007297

Control Chart for Individual Measurements

When the production is very slow or costly it is impractical to allow the sample size to be greater than one. Other cases
include processes where every observation can be measured due to automated inspection. Then the Shewhart control chart
for individual measurement is appropriate in these situations. The Shewhart chart uses the moving range, MR of two
successive observations for estimating the process variability. The moving range for 𝑚 observation 𝑥1 , 𝑥2 , … . , 𝑥𝑚 is
defined by
𝑀𝑅𝑖 = 𝑥𝑖 − 𝑥𝑖−1 , 𝑖 = 2,3, … , 𝑚
The control limits for individual control chat are calculated as

𝐶𝐿 = 𝑥
𝑀𝑅
𝐿𝐶𝐿 = 𝑥 − 3
𝑑2
7 𝑀𝑅
𝑈𝐶𝐿 = 𝑥 + 3
𝑑2
 1 𝑚 1 𝑚
where 𝑥 = 𝑖=1 𝑥𝑖 sample is mean of the observation data 𝑥𝑖 and 𝑀𝑅 = 𝑖=2 𝑀𝑅𝑖 is the sample mean of the
𝑚 𝑚 −1
moving range 𝑀𝑅𝑖 .
The values of 𝑑2 for different sample sizes are listed in the Table XIII (Factors for Quality Control, Page 633)

The control limits for the moving range control chart are

𝐶𝐿 = 𝑀𝑅
𝐿𝐶𝐿 = 𝐷3 𝑀𝑅
𝑈𝐶𝐿 = 𝐷4 𝑀𝑅
The values of 𝐷3 𝑎𝑛𝑑 𝐷4 for different sample sizes are listed in the Table XIII (Factors for Quality Control, Page 633)

See Example 17-2 (Book: page 519)

Limitations of Control Charts for Variables

In spite of wide applications of 𝑥 and R charts for detecting the assignable causes of variation in a production process,
their use is restricted because of the following limitation-

a) They are charts for variables only, i.e., for quality characteristics which can be measured and expressed in
numbers.
b) In certain situations, they are impracticable to construct, e.g., if the number of measurable characteristics is too
large such as a radio, or a television, or any other electric appliances

(In such case it would be appropriate to construct the control charts for attributes.)

Control Charts for Attributes

Although more complete information can usually be gained from measurements on a finished product, it is often quicker
and cheaper to check the product against specifications on an “attribute” basis.

In spite of wide applications of 𝑥 and R charts for detecting the assignable causes of variations in a production process,
their use is restricted because firstly they are charts for variables only, i.e., for quality characteristics which can be
measured and expressed in number and secondly, in some cases they are impracticable and un-economic to construct.

In situations like this it would be more appropriate to construct the control charts for attributes.

There are following types of control charts of attributes-

a) Control charts for fraction defectives (p-chart)

b) Control chart for the number of defects (c-chart)
c) Control chart for number of defects per unit (u-chart)

For control limits of 𝑝 − 𝑐ℎ𝑎𝑟𝑡, 𝑐 − 𝑐ℎ𝑎𝑟𝑡 𝑎𝑛𝑑 𝑢 − 𝑐ℎ𝑎𝑟𝑡 and examples, see in the BOOK: PAGES 520-525