Dao Vu Truong Son, PhD

International University, VNU-HCM

Email: [email protected]
A Technical component of TQM:
SPC/SQC

Part 1
Stabilizing and Improving
a Process with Control Charts

2
Outlines
Introduction to SPC
Target of SPC
Process Variation
The Structure of Control Charts
Stabilizing a Process with Control Charts
Advantages of a Stable Process
Improving a Process with Control Charts
Two Possible Mistakes in Using Control Charts
Some out-of-control evidence
Quality Consciousness and Types of Control Charts
Three Uses of Control Charts
Exercises

3
Learning Objectives (You will be able to:)
To explain the need for continual reduction of variation, even when
the quality characteristic is within specifications.
To discuss the use of control charts to stabilize and improve a
process.
To discuss the consequences of over and under-adjustment of a
process.
To detect out-of-control behavior using the 7 Western Electric
rules.
To describe how attributes control charts can be used for defect
prevention.
To describe how variables control charts can be used for never-
ending improvement.
To discuss the purposes of studying control charts.
4
What it is What it is not
- A procedure structured to - No wondermeans to
tackle, recognize and solve problems
control problems

- Gives info concerning the - short-term approach

stability and capability of
the process

- Based on measurements - Business of the

and interpretation of the quality departments
operator alone
- Based on control charts
and standard procedures

5
6
Recall: process variation as the result of either common
causes or special causes.
Control Charts and Variation:
1. Control charts are used to identify and differentiate
between common and special causes of variation.
2. When a process no longer exhibits special variation, but
only common variation: process is stable.
3. When only common causes of variation are present in a
process: must take action to reduce the difference
between customer needs and process performance by
endeavoring to move the centerline of the process closer
to nominal and/or by reducing the magnitude of common
variation.
The Need for the Continual Reduction of Variation 7
 All control charts
P Chart f or Fraction have a common
1 structure:
1. a centerline (the
0.2 UCL=0.2055
process average)
2. upper and lower
Proportion

control limits (3-

0.1
P=0.1112 sigma limits) that
provide information
on the process
LCL=0.01689 variation.
0.0

0 5 10 15 20 25
Sample Number

8
Control charts are constructed by drawing samples
and taking measurements of a process characteristic.
Each set of measurements is called a subgroup.
Control limits are based on the variation that occurs
within the sampled subgroups.
In this way, variation between the subgroups is
intentionally excluded from the computation of the
control limits; the common process variation becomes
the variation on which we calculate the control
limits.
The control limit computations assume that there are
no special causes of variation affecting the process. If
a special cause of variation is present, the control
chart, based solely on common variation, will
highlight when and where the special cause occurred.

9
Stabilizing a Process with Control Charts
Raw data for construction of control chart
No of No of Fraction of
Day
entries defects defective entries
1 200 6 0.03
2 200 6 0.03
3 200 6 0.03
4 200 5 0.025
5 200 0 0
6 200 0 0
7 200 6 0.03
8 200 14 0.07
9 200 4 0.02
10 200 0 0
11 200 1 0.005
12 200 8 0.04
13 200 2 0.01
14 200 4 0.02
15 200 7 0.035
16 200 1 0.005
17 200 3 0.015
18 200 1 0.005
19 200 4 0.02
20 200 0 0
21 200 4 0.02
22 200 15 0.075
23 200 4 0.02
24 200 1 0.005

Total 4800 102 10

 When the data
consist of a series of
fractions that are
Chart Title
defective or possess
0.08
some other
Fraction of defective entries

0.07
0.06 characteristic of
0.05 interest, the
0.04
appropriate control
0.03
0.02
chart is a p-chart.
0.01 This is a depiction of
0 the process output in
0 4 8 12 16 20 24
Day terms of an attribute
of interest - in our
example, the fraction
defective.

11
The centerline for a p-chart is the mean of the fraction
defective.

Total number of defectives in all subgrops under investigation

p
Total number of units examined in all subgroups under investigation

Control limits are calculated as plus and minus three

times the standard error. The standard error for the
average proportion is:

p(1 p)
p
n

Where n is the subgroup size.

12
Using this value, the upper and lower control limits for a
p-chart are given by:
p(1 p)
UCL(p) p 3
n

p(1 p)
LCL(p) p-3
n

We can now find the numerical values for constructing

our p-chart:
102
p 0.021
4,800
(0.021)(1 0.021)
UCL(p) 0.021 3 0.052
200
(0.021)(1 0.021)
LCL(p) 0.021 3 0.009
200

Notice that since a negative fraction defective is not

possible, the lower control limit is set at 0.00. 13
 New Terminal Malfunction
Untrained (Replaced with “Old”
Operator Standby Unit)

0.08

0.07
Fraction of defective entries

0.06

0.05 UCL

0.04

0.03

0.02

0.01

0 LCL
0 4 8 12 16 20 24
Day

Unknown

14
The action taken on the process stemming
from investigations of days 8 and 22 should
change the process so that the special causes
of variation will be eliminated. Consequently,
the data from days 8 and 22 may now be
deleted. After eliminating the data for the days
in which the special causes of variation are
found, the control chart statistics are
recomputed.
73
p 0.017
4,400
(0.017)(1 0.017)
UCL(p) 0.017 3 0.045
200
(0.017)(1 0.017)
LCL(p) 0.017 3 0.010
200
Hence, use LCL 0.00 15
 0.05

0.045 0.044

0.04

0.035
Fraction defectives

0.03

0.025

0.02
0.017
0.015

0.01

0.005

0
0 10 20 30 40 50
Day

16
A stable process is a process that exhibits only
common variation or variation resulting from inherent
system limitations. The advantages of achieving a
stable process are:
Management knows the process capability and can predict
performance, costs, and quality levels.
Productivity will be at a maximum, and costs will be minimized.
Management will be able to measure the effects of changes in
the system with greater speed and reliability.
If management wants to alter specification limits, it will have the
data to back up its decision.
A stable process is a basic requirement for process
improvement efforts.

17
Once a process is stable, it has a known capability. A
stable process may, nevertheless, produce an
unacceptable number of defects (threshold state)
There are two areas for action to reduce the difference
between customer needs and process performance.
➢ First, action may be taken to change the process
average.
➢ Second, management can act to reduce the level of
common variation with an eye toward never-ending
improvement of the process.
The workers can only suggest changes; they can not
effect changes to the system.

18
 0.03

0.026
0.025

0.02
Fraction defective

0.015

0.01
0.008

0.005

0
0 5 10 15 20 25
Day

19
There are two types of mistakes that the
user of a control chart may make: over-
adjustment and under-adjustment.

20
Avoiding both of these mistakes all of the time is an
impossible task.
Never adjusting the process - so that we never
make the mistake of over-adjusting - could result in
severe under-adjustment.
On the other hand, if we made very frequent
adjustments to avoid the problem of under-
adjustment, we would probably be over-adjusting.
Control charts provide an economical means to
minimize the total loss that results from these two
errors.
Consequently, control charts provide management
with information on when to take action on a
process and when to leave it alone.

21
The area between the control limits is divided into 6 bands.
1. C zones: centerline 1 standard error.
2. B Zones: between 1 and 2 standard errors from the centerline.
3. A zones: between 2 and 3 standard errors from the centerline.

Rules for identifying out-of-control points (7 rules).

- Backward looking
- Any out-of-control points found are marked with an X.
22
Rule 1. A process exhibits a lack of control if any
subgroup statistic falls outside of the control limits.

23
Rule 2. A process exhibits a lack of control if any two
out of three consecutive subgroup statistics fall in one
of the A zones or beyond on the same side of the
centerline.

24
Rule 3. A process exhibits a lack of control if four out
of five consecutive subgroup statistics fall in one of
the B zones or beyond on the same side of the
centerline.

25
Rule 4. A process exhibits a lack of control if eight or
more consecutive subgroup statistics lie on the same side
of the centerline.

26
Rule 5. A process exhibits a lack of control if eight or
more consecutive subgroup statistics move upward in
value or if eight or more consecutive subgroup statistics
move downward in value.

27
Rule 6. A process exhibits a lack of control if an unusually
small number of runs above and below the centerline are
present (a saw-tooth pattern).

28
Rule 7. A process exhibits a lack of control if 13
consecutive points fall within zone C on either side of the
centerline.

29
No quality consciousness: Accept Everything without Question
Defect detection: Mass inspection. (yes-no)
Defect prevention: Attribute control charts. (goalpost view)
Never-ending improvement: Variable control charts. (Taguchi)
Innovations (quality creation)

30
Evaluating the Past
Evaluating the Present
Predicting the Near Future

31
Identify any points that indicate a
lack of control and explain why

32
Assoc. Prof. Ho Thanh Phong, International University
Assoc. Prof. Ho Thanh Phong, International University
Assoc. Prof. Ho Thanh Phong, International University
Quality Management

Dao Vu Truong Son, PhD

International University, VNU-HCM
Email: [email protected]
 Attribute Control Charts
Sections
Introduction
Types of Attribute Control Charts
The p Chart for Constant Subgroup Sizes
The p Chart for Variable Subgroup Sizes
The np Chart
Count Charts
➢ c Charts
➢ u Charts

2
 Introduction
Attribute data are data based on counts, or the number of times
we observe a particular event.

Attribute data may include such classifications as:

defective or conforming;
go or no-go;
acceptable or not acceptable;
or number of defects per unit.

3
 Types of Attribute Control Charts
There are two basic types of attribute control charts:

1. classification charts
2. count charts.

4
 Types of Attribute Control Charts
Classification Charts
Classification charts deal with either the fraction of items or the number of items in a series of
subgroups that have a particular characteristic.

1. p Chart.
➢ Used to control the fraction of items with the characteristic.
➢ Subgroup sizes may remain constant or may vary.

2. np Chart.
➢ Used to control the number of items with the characteristic.
➢ Used only with constant subgroup sizes.

5
 • Count Charts
Count charts deal with the number of times a particular characteristic appears in
some given area of opportunity.

1. c Chart.
➢ Used to control the number of times a particular characteristic appears in a
constant area of opportunity.
✓ For example:
❖ defects per air conditioner;
❖ accidents per workweek in a factory;
❖ and, deaths per week in a city.

2. u Chart.
➢ Used when the area of opportunity changes from subgroup to subgroup.

6
 Classification Charts
Conditions for Use
1. When each unit can be classified as either conforming or nonconforming: a
classification chart is appropriate.
2. Samples of n items are periodically selected from process output. For these n
distinct units comprising a subgroup:
➢ Each unit must be classifiable as either possessing or not possessing the characteristic of interest.
✓ For example, each unit in a subgroup might be classified as either defective or non-defective,
or conforming or nonconforming. The number of units possessing the characteristic of
interest is called the count, X.
➢ The probability that a unit possesses the characteristic of interest is assumed to be stable from
unit to unit.
➢ Within a given area of opportunity, the probability that a given unit possesses the characteristic
of interest is assumed to be independent of whether any other unit possesses the characteristic.

For data satisfying these conditions, we may use a p or np chart.

7
 Some definitions
Subgroup size (n): should be constant? How about the size of
n?
Subgroup frequency: to encompass all of the process’ cycles.
How to choose the frequency? – Expert!
Number of subgroup: >= 25 for p and np charts. Attribute
<> variable charts.

8
Assoc. Prof. Ho Thanh Phong, International University
 1. The p-Chart for Variable Subgroup Sizes
The chart with subgroup size is varies (Example in page 200 -203, text book)

9
Using Average Subgroup Size: An Example

10
11
 2. The np-Chart
The quantity np is the number of units in the subgroup with some
particular characteristic.

Traditionally, np-charts are used only when subgroup sizes are

constant.

The information used is the same as for p-charts with constant

subgroup sizes, these two charts are interchangeable.

12
Constructing the np-Chart

13
 Constructing the np-Chart

1. Centerline (np) =

14
 Constructing the np-Chart
1. Boundary between upper zones B and C =

2. Boundary between lower zones B and C =

3. The upper boundary between zones B and C for this example is given by

5. and the lower boundary between zones B and C is given by

6. Boundary between upper zones A and B =

7. Boundary between lower zones A and B =

8. The results for this example are:

15
16
 3. Count Charts
A defective item is a nonconforming unit.

A defect, on the other hand, is an imperfection of some type that

is undesirable, although it does not necessarily render the entire
good or service unusable.

One or more defects may not make an entire good or service

defective.

For example, we would not scrap or discard a computer because of a small

scratch in the paint.

17
 Conditions for Use

1. If we are to use the c-charts or u-charts,

➢ the events must be describable as discrete events;
➢ they must occur randomly within some well-defined area of opportunity;
➢ they should be relatively rare;
➢ they should be independent of each other.

2. Exact conformance to these conditions is not always easy to verify.

18
 3.1 c-Charts
When all conditions for an area of opportunity chart are met, and when the
subgroup sizes remain constant, a c-chart is used.

Areas of opportunity that are constant in size are easier to manage than those
that vary.

19
 c: The number of events in an area of opportunity, the count for
each area of opportunity.

The sequence of successive c values, taken over time, is used to

construct the control chart.

The centerline for the chart is the average number of events

observed. It is calculated as:

20
 Counts, Control Limits, and Zones
1. Consider a firm that has decided to use a c-chart to help keep track of the
number of telephone requests received daily for information on a given
product. Each day represents an area of opportunity. Over a 30-day period,
1,206 requests are received, or an average:

➢ A count of 59 is within the control limits, while a count of 60 is beyond the

UCL.

21
 Counts, Control Limits, and Zones
1. Boundary between lower zones B and C =

2. Boundary between lower zones A and B =

3. Boundary between upper zones A and B =

4. Boundary between upper zones B and C =

5. Because the actual counts are whole numbers, the observation would fall into
zones as follows:

The zones each contain a reasonable number of whole numbers and are close
enough in size to be workable.

22
 Counts, Control Limits, and Zones
1. Consider the problem that would have been encountered if the
process average had been = 2.4 Here we would get:

➢ Boundary between lower zones B and C =

➢ Boundary between lower zones A and B =

➢ Boundary between upper zones A and B =

➢ Boundary between upper zones B and C =

23
 Zone Counts

Upper A 6, 7

Upper B 4, 5

Upper C 3

Lower C 1, 2

Lower B 0

Lower A 0

24
 Counts, Control Limits, and Zones
1. As a rule of thumb, the zone boundaries should not be used for
c charts with average counts of less than 20.0.
2. As the observable count shrinks, the use of variables control
charts must be instituted for continued process improvement.

25
Construction of a c-chart: An Example
1. Consider the output of a paper mill: the product appears at
the end of a web and is rolled onto a spool called a reel. Every
reel is examined for blemishes, which are imperfections. Each
reel is an area of opportunity.

26
 Construction of a c-chart: An Example

1. Centerline(c) =

27
28
 Small Average Counts
1. When average counts are small, data appearing as counts will
tend to be asymmetric. This may lead to over-adjustment (false
alarms) or under-adjustment (too little sensitivity).

29
Stabilizing a Process: An Example
1. An industrial washing machine manufacturer inspects
completed units for defects.

30
31
32
 The process is not in control. Special causes of variation are present.
Let us assume that the local operators responsible for the final
inspection act so that the special causes of variation for points 9,
10, 14, 19, 20, 22, and 23 are identified and the appropriate
corrections are made.
The data for points affected by known special causes that have been
eliminated are deleted from the data set, and the centerline and
control limits are recomputed:

33
The new limits are so close to the old limits that the old limits
are used for the next 24 machines produced.

34
 The first five data points for these next 24 machines are well
below the lower control limit.
Investigation by the local operators reveals that a substitute for
the regular inspector counted the defects on those five
machines. The substitute was not properly trained and did not
identify all the defects correctly.
The operators informed management, and management made
appropriate changes in policy so that this situation would not
recur.
These points can now be eliminated from the data. Beginning
with machine number 30, all counts are below the process
average.
Local operators decided that the process has been changed, so a
revised control chart is constructed beginning with point
number 30.

35
36
The process, as it stands, now appears to be in a state of control.
37
 3.2 u-Charts
In some applications the areas of opportunity vary in size

For example, samples taken from a roll of paper may need to

be manually torn from rolls, so that the areas sampled -- the
areas of opportunity -- will vary.

38
 The characteristic used for the control chart, u, is the ratio
of the number of events to the area of opportunity in
which the events may occur.

For observation i, we call

➢ ci : the number of events (such as imperfections) the observed,
➢ ai : the area of opportunity
➢ Thus, ui is the ratio for each point.

39
40
 Construction of a u-chart: An Example
➢ Consider the case of the manufacture of a certain grade of
plastic. The plastic is produced in rolls, with samples taken five
times daily. Because of the nature of the process, the square
footage of each sample varies from inspection lot to inspection
lot.

41
42
 An Example

➢ The control limits are different for each of the subgroups and
must be computed individually for each subgroup

43
44
45
 Limitations of Attribute Control Charts
As processes improve and defects/defectives become rarer, the
number of units that must be examined to find one or more of
these events increases.

If we consider a p-chart where the average fraction of

nonconforming items is 0.005, then on average we'd need to
examine 200 units to have an average count of just 1.00. In the
extreme, to maintain a reasonable average count as the area of
opportunity grows, 100 percent inspection becomes the
rule.

46
 • This implies inspecting all of the items and sorting
those that conform to some specification from those
that don't. Not only is this inspection costly, but it's
equivalent to accepting the fact that the process is
producing a constant fraction of its output as
defective and will continue to do so.

• Hence, attribute control charts are limited in terms of

the level of process improvement they make possible.
Additional process improvement is possible with
variables control charts.

47
 Exercises
A bank is studying the
proportion of transactions are
being made by ATMs. Plot the
control charts. Indicate
whether the process is
stable.ext book

48
Statistical Process Control
Part 3

Son Dao, PhD, ISE @ IU 1

Variable control charts
• Variables data consist of numerical measurement such as
weight, length, width…

• Variables data contain more information than attribute data.

Variables control charts (VC) deal with measurement
themselves, so they are more powerful than attribute charts.

• Using variables measurements and charts is the only way to

overcome the need for large and large subgroups.

• Three principal types of VC: (1) the x and R chart; (2) the x and
s chart; (3) the individuals and moving range chart.

Son Dao, PhD, ISE@IU 2

Subgroups size should be large enough …

Individuals and moving range charts: only one

variables measurement.
x and R charts: from 2-3, or 4-9 subgroup sizes.
x and s charts: 10 or more subgroups.

Subgroup frequency depends on the particular

application…

Son Dao, PhD, ISE@IU 3

Xbar and R chart
A pharmaceutical company performs SPC on
nominal specification of vial filled (52g).
Samples of 6 vials are selected every 5 min
during a 105-min period
Rbar = ∑R/k
UCL(R) = D4Rbar
LCL(R) = D3Rbar

Son Dao, PhD, ISE@IU 4

Xbar and R – Construct the charts
1 2 3 4 5 6
1 52.22 52.85 52.41 52.55 53.1 52.47
2 52.25 52.14 51.79 52.18 52.26 51.94
3 52.37 52.69 52.26 52.53 52.34 52.81
4 52.46 52.32 52.34 52.08 52.07 52.07
5 52.06 52.35 51.85 52.02 52.3 52.2
6 52.59 51.79 52.2 51.9 51.88 52.83
7 51.82 52.12 52.47 51.82 52.49 52.6
8 52.51 52.8 52 52.47 51.91 51.74
9 52.13 52.26 52 51.89 52.11 52.27
10 51.18 52.31 51.24 51.59 51.46 51.47
11 51.74 52.23 52.23 51.7 52.12 52.12
12 52.38 52.2 52.06 52.08 52.1 52.01
13 51.68 52.06 51.9 51.78 51.85 51.4
14 51.84 52.15 52.18 52.07 52.22 51.78
15 51.98 52.31 51.71 51.97 52.11 52.1
16 52.32 52.43 53 52.26 52.15 52.36
17 51.92 52.67 52.8 52.89 52.56 52.23
18 51.94 51.96 52.73 52.72 51.94 52.99
19 51.39 51.59 52.44 51.94 51.39 51.67
20 51.55 51.77 52.41 52.32 51.22 52.04
21 51.97 51.52 51.48 52.35 51.45 52.19
22 52.15 51.67 51.67 52.16 52.07 51.81
Son Dao, PhD, ISE@IU 5
Xbar and s chart
Centerline = sbar = ∑s/k

Use similar steps as in Xbar and R-chart

Son Dao, PhD, ISE@IU 6

Notes
Real data usually are multi-variate → grouping
data for analysis is important
Use technical department (with in-depth
knowledge about the process) to know the key
variables

Son Dao, PhD, ISE@IU 7

 Example: Paint filling data
Variations:
Head-to-head
Hour-to-hour
Measurement-to-
-measurement

Son Dao, PhD, ISE@IU 8

Arrangement 1

Son Dao, PhD, ISE@IU 9

Arrangement 2

Son Dao, PhD, ISE@IU

Useful constant

Son Dao, PhD, ISE@IU 11

Summary

Son Dao, PhD, ISE@IU 12

Son Dao, PhD, ISE@IU 13
Son Dao, PhD, ISE@IU 14