STATISTICAL PROCESS CONTROL

Statistical Process Control &

Process Capability

y
nl
O
By:

e
Quality Growth Services Pvt. Ltd.

os
H-13, Kirti Nagar, New Delhi 110015
www:qgspl.com E-Mail: [email protected]
Phone: +91 11 25431737/25438598 rp
Pu
g
in

Introduction
in

 Please take a minute to tell us Your

ra

understanding and application of SPC on a

scale of: 1-10 (10 Can Teach : 1 No
rT

Knowledge
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 2
 Ground Rules, Logistics, & Rewards

 Ground Rules:
- Respect all individuals in the room
- Class format requires participation
- Mind your cell phones and pagers

y
- Other: ______________________
______________________

nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 3
Pu
g
in

Agenda
in

You will learn:

ra

 Relationship in Voice of Process & SPC

rT

 Control Charts Concept – Variables &

Attribute
Fo

 Control Charts-Variable
 Control Chart – Attributes
 Process Capability & Performance

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 4
 Ultimate Goal Of An Organization

Profits, Survival and Growth: How?

- Products & Services must be preferable
to Customers over Competitors

y
- Organization must continually meet Needs

nl
& Expectations of Customers

O
i.e. Voice of Customers also called QUALITY

e
os
5
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 5
Pu
g
in

Statistical Process Control

in
ra

is one of the Mandatory Core Tools in

rT

IATF 16949:2016 Quality System

Fo

(9.1.1.1 & 9.1.1.2)

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 6
 Levels of Process Control

Error Proofing

Statistical Process
Control

y
nl
Traditional Control
(Inspection)

O
No Control

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 7
Pu
g
in

SPC History
in
ra

Developed By
rT

Dr. Walter A. Shewhart

During 1920’s in Bell Lab
Fo

Since then SPC has evolved to cover

different processes

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 8
 Variation : The Basic Principle of SPC

Everything is Different
• No two things are exactly alike…

y
• No two people are same…

nl
• No two products are exactly same…

O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 9
Pu
g
in
in

Common Special
ra

Causes Causes
(Natural (Assignable
rT

Variation) Variation)
Under Out of
Fo

Variation:
Statistical Two Control
Control Categories

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 10
 Common Causes

• Common to all individual readings in

time periods.

y
A process operating under common

nl
cause is called

O
Under Statistical Control

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 11
Pu
g
in

Special Causes
in

• Sudden in nature
ra

• Usually attract the attention of local

rT

people associated with the process.

Fo

They are not common to all time periods,

but they can cause process fluctuations
which are large in magnitude

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 12
 Process Behavior

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 13
Pu
g
in

How to estimate process behavior?

in
ra

Shape
rT
Fo

Location
Spread

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 14
 Processes can differ in:

Location Spread

y
nl
O
Shape

e
Or any combination

os
of these….

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 15
Pu
g
in

Statistics for location

in

Median
ra

• The mid
value of
rT

data String
Mean Mode
• Average of a • The value
Fo

set of values with highest

frequency

Location

Also known asQGS\SPC

Measure of Central Tendency
17
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 17
 Statistics for Spread

Range
• The difference Standard
of maximum
and minimum Deviation
values

y
nl
Spread

O
e
Also known as Measure of Dispersion

os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 18
Pu
g
in

Shape
in
ra
rT
Fo

Histogram

Also known as Measure of Distribution

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 19
 Mean

Average of a set of values

y
nl
O
e
os
Also known as Measure of Central Tendency

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 20
Pu
g
in

Is average good enough…

in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 21
 Is Central Tendency Enough to Represent Distribution ?

 Find the value of “n” and “X” for the following 2

distributions.

Are The 2
distributions

y
Same ?

nl
n= X=
1 2 3 4 5 6 7 8 9

O
n= X=
What is the
Difference ?

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 22
Pu
g
in

Range
in
ra

Range : The difference between the largest and the

rT

smallest of a set of numbers. It is designated

by a capital “R”
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 23
 Is Range a good Measure of Variability ?
Find the value of “n”, “X” and “R” for the following 2 distributions.

Range (R)= Max Value-Min. Value

Do the 2
n= X= R= distributions
have same

y
n= X= R= variability ?

nl
O
1 2 3 4 5 6 7 8 9
How do we
measure
average

e
variability from

os
the Center ?

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 24
Pu
g
in

Measure of Variation (Spread)

in

Calculate Standard Deviation for these 2 Distributions.

ra

n= X= R= s=
rT

1 2 3 4 5 6 7 8 9
n= X= R= s=

What is the
Fo

application of
Standard
Deviation ?

We Can Make Prediction:

If there is no abnormal change in the process,
99.73% of the population is likely to fall with in Mean ± 3 Std. Dev.

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 25
 Understanding Process Behavior

To understand a process you need its:

 Location: Average
 Spread: Range/Standard Deviation
 Change in Process Location/Period over Period

y
nl
O
Control Charts enable us learn change in

e
location/Spread of the process over time

os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 26
Pu
g
in

The Normal Curve

in
ra
rT
Fo

-3 -2 -1 + +2 +3

68.26

95.46

99.73

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 27
 Control Chart
Transformation of a normal distribution curve in the form of 3 parallel lines, where
• The middle line indicates mean and called central line (C.L.)
• The upper line indicates Mean +3 Sigma and called upper control limit (UCL)
• The lower line indicates Mean -3 Sigma and called lower control limit (LCL)

U.C.L.

y
+3

nl
C.L.
Mean

O
e
-3

os
L.C.L.

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 28
Pu
g
in

Types of Data
in

 Attribute Data (Qualitative)

- Categories like type 1, type 2, type 3
ra

- Yes, No
Attribute data is purely binary
rT

- Go, No Go or Pass/Fail
- Good/Defective
- On-Time/Late
Fo

- Discrete (Count) Data

- # of Maintenance Equipment Failures, # of claims

 Variable or Continuous Data (Quantitative)

- Decimal subdivisions are meaningful
- Viscosity (Sec), Pressure, Wt./10Litre, Acid Value

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 29
 SPC – Charts For All Occasions
No Defects
Variable Attribute
Data? Data?

Yes % Defective

No Constant Constant
Rational
I & MR chart Sample Sample
Subgroups?
Size? Size?

y
No No

nl
Yes Yes Yes

O
p-chart u-chart
X-bar & R chart

np- or p-chart c- or u-chart

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 30
Pu
g
in

Steps for Control Charts

in

1. Fulfill precondition
ra

2. Complete preparatory steps

rT

3. Data Collection

4. Making Trial Control Limits

Fo

8 Steps
5 Validation of Control limits

6 Process capability Study

7. On going control

8. Improvement
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 31
 Average-Range ( X-R )Chart

 Precondition:
- Measurement must be variable
- Situation must be practically feasible to have
at least 2 measurements in short span.

y
Preparatory Steps

nl

- Create a suitable (conducive) environment

O
- Understand the process
- Ensure Level 1 control -Minimize unnecessary

e
variations .

os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 32
Pu
g
in

Average-Range ( X-R )Chart

in

 Preparatory steps
ra

- Select the Subgroup Size

Rational subgroup: Variability within subgroup should be
rT

small
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 33
 Average-Range ( X-R )Chart

 Preparatory steps
- Select Subgroup Frequency
 Detect change in the Process over span
of time.

y
 For initial study, may be consecutive or a

nl
very short interval.

O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 34
Pu
g
in

Average-Range ( X-R )Chart

in
ra

 Preparatory steps
- Select No. of Subgroups
rT

(for initial study: to define the control limits)

To incorporate Major source of variation
Fo

(Generally 25 subgroups or more containing

about 100 individual measurements)
- Decide Measurement system: Should
qualify MSA

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 35
 Average-Range ( X-R )Chart
 Data Collection
- On a data collection sheet, called control chart sheet

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 36
Pu
g
in

Average-Range ( X-R )Chart

in
ra
rT
Fo

Above data is of Dia. Of a shaft where specs are 11.7± 0.2 (Expressed
as 0.50 to 0.90)
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 37
 Average-Range ( X-R )Chart

 Calculate Average of each Subgroup

X= ( X1 + X2 + … + Xn )/ n

y
 Calculate Range of each Subgroup
R= Xmax. - Xmin.

nl
O
X1, X2,…., Xn are individual values within the subgroup
n is the Subgroup Sample Size, k = No. of Subgroups

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 38
Pu
g
in

Average-Range ( X-R )Chart

in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 39
 Average-Range ( X-R )Chart

 Calculate Process average

X = (X1 + X2 + … + Xk)/ k
 Calculate Average Range

y
R= (R1 + R2 + … + Rk )/ k

nl
O
X1, X2,…., Xn are individual values within the subgroup

e
n is the Subgroup Sample Size, k = No. of Subgroups

os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 40
Pu
g
in

Average-Range ( X-R )Chart

in

 Calculate Trial Control Limits for Range Chart

ra

UCLR = D4 R
rT

LCLR = D3 R.
 Calculate Trial Control Limits for Average Chart
Fo

UCLX = X + A2 R
LCLX = X - A2 R
D4, D3 and A2 are constant varying as per sample size
(n).

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 41
 Table of Constants for Control Charts

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 42
Pu
g
in

Draw Average - Range Chart

in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 43
 Interpretation for Control Charts

 Interpretation should be started with R-Chart

first

 For each indication of special cause, analysis

y
should be done to identify the root cause and

nl
action to be taken.

O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 44
Pu
g
in

Setting Control Chart Limits for on going use

in

 For initial control chart,

ra

- Discard all the subgroup showing out of control situation (starting

from R-Chart)
- Re-Calculate Control Limits, plot and analyse for any out of
rT

control situation.
- Re discard if any out of control situation again found.
Fo

- Continue the above cycle until all plot indicate a control situation.
- Repeat same exercise with Average chart.
- If more than 50% data are required to be discarded, reject all data
and re-collect.
- Once initial control chart indicates control situation,
Calculate Initial Capability
Extend control limits for ongoing control.
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 45
 Voice of the Process

-3 

y
+3 

nl
O
Process Width
Design Width

e
os
LSL T USL

Voice of the Customer

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 46
Pu
g
in

Process Capability
in

 Calculate Process Standard

ra

Deviation
 = R/d2
rT

d2 is a constant varying as per

sample size (n)
 Calculate Process Capability (Cp)
Fo

Cp = (USL - LSL) / 6 USL = Upper Specification Limit

= Tolerance/ 6 LSL = Lower Specification Limit
 Calculate Process Capability (Cpk)
CpkU = (USL -X) / 3
or CpkL = (X - LSL) / 3
Whichever is minimum will be Cpk

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 47
 Process Performance

Standard Deviation
s =  ( Xi-X)2 i=1
for n=80
USL = 0.900
n-1 LSL = 0.500
Process Performance
Pp = (USL - LSL) / 6s = (0.900 - 0.500) / 6 x 0.0759

y
= 0.880
PpkU = (USL - X) / 3s = (0.900 - 0.738) / 3 x 0.0759

nl
= 0.710
PpkL = (X - LSL) / 3s = (0.738 - 0.900) / 3 x 0.0759

O
= 1.045
Ppk = 0.710

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 48
Pu
g
in

Process Capability:
in
ra

Cp Cpk Remarks
 Process Capable
rT

 Continue Charting
 Bring Cpk closer to Cp
 Process has potential Capability
Fo

X  Improve Cpk by Local action

 Process lacks basic Capability

X X  Improve process by Management

action

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 49
 Interpreting Control Charts
 The Interpretation of control charts is based on the statistical
probability of a particular pattern occurring by complete chance (or
being caused by random variation).
 All of the tests identify events that have a less than 0.3 % chance of
occurring by random chance (outside of 3s probability of being
caused by random variation).

y
 Control charts are divided into sigma zones above and below the

nl
average line.
- Zone C is < 1s from the mean.

O
- Zone B is between 1s and 2s.
- Zone A is between 2s and 3s.
- Beyond Zone A is > 3s.

e
 The following rules are from Journal of Quality Technology, October,

os
1984 and have been included in AIAG SPC manual-edition2
Control charts are instituted to be interpreted, not just for wallpaper.
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 50
Pu
g
in

Test 1 & 2
in

Test1. One Point Beyond Zone A

 Test 2
- Caused by a process
ra

x
UCL mean shift
A
B
rT

C
C
B
A
Fo

LCL Test 2. Nine Points in a Row

x on One Side of the Center Line
 Test 1 – the basic test. UCL
A
- Caused by a large change B
in the process C
- Requires immediate C
action B x
A
LCL

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 51
 Test 3 & 4
Test 3. Six Points in a Row Steadily
Increasing or Decreasing  Test 4 – Alternating Patterns
UCL - Over adjustment
A - Shift-to-Shift variation
B x - Machine-to-machine
C variation
C
B

y
A x

nl
LCL

Test 4. Fourteen Points in a Row

Alternating Up and Down

O
 Test 3 – Trends up or down
UCL
- Caused by A
- Mechanical wear B x
- Chemical depletion

e
C
- Increasing contamination C

os
- Etc. B
A
LCL

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 52
Pu
g
in

Test 5 & 6
in

Test 5. Two Out of Three Points

in a Row in the same Zone A or  Test 6
Beyond
ra

x x - Another test for shifts

UCL
A - Test 1, 5, 6 are related and
B show conditions of high
rT

C special cause variability.

C
B
A
Fo

Test 6. Four Out of Five Points in

LCL a row in the same Zone B and
x
Beyond
 Test 5 UCL
A x
- The second basic test B
- High variation w/o C
exceeding the 3s limit
C
- Major special-cause B x
variation A
LCL

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 53
 Test 7 & 8
Test 7. Fifteen Points in a Row in Zone C
(Above and Below Centerline)  Test 8 – Alternating Means
- Mixtures
UCL
A - Over control
- Two different processes on
B x the same chart
C
C
B

y
A Test 8. Eight Points in a Row on Both Sides

nl
LCL
of Centerline with None in Zone C

 Test 7 – The Whitespace UCL

O
A
Test x
B
- Occurs when within subgroup
variation is large compared to C

e
between group variation or, C
- Old or incorrectly calculated B

os
limits A
LCL

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 54
Pu
g
in

Exercise :1
in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 55
 Exercise : 2

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 56
Pu
g
in

Exercise : 3
in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 57
 Exercise : 4

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 58
Pu
g
in

Interpretation for Control Chart

in

X Bar Chart R CHART CONCLUSION

ra

Under Control Under Control Enjoy

rT

Under Control Out of Control Spread Changed

Fo

Out of Control Under Control Location Changed

Out of Control Out of Control Both Spread & Location

Changed

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 59
 What If Sub Grouping is Irrational?

Individuals (I & MR) Charts

 For use when there is no rational basis for sub grouping or when the
measurements are expensive and/or destructive
 MR stands for Moving Range. The moving range is calculated by

y
recording the differences between successive pairs of

nl
measurements
 When it comes to creating control charts, it's generally good to

O
collect data in subgroups, if possible. But sometimes gathering
subgroups of measurements isn't an option. Measurements may
be too expensive. Production volume may be too low. Products

e
may have a long cycle time.

os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 60
Pu
g
in

Moving Range Control Charts

in
ra

The Control Chart used for

rT

Slow production rate or

When sub-grouping is irrational
Fo

Destructive testing or
Homogenous process

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 61
 Moving Range Control Charts

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 62
Pu
g
in

Moving Range Control Charts

in

DATE READING RANGE

(X) (R)
ra

01.11.18 9.20
02.11.18 8.50 0.70
rT

03.11.18 9.40 0.90

04.11.18 10.50 1.10
05.11.18 9.30 1.20
06.11.18 11.10 1.80
Fo

07.11.18 10.40 0.70

08.11.18 10.40 0.00
09.11.18 9.00 1.40
10.11.18 10.00 1.00
11.11.18 11.70 1.70
12.11.18 10.30 1.40
13.11.18 10.20 0.10

Average 5.03 0.52

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 63
 Moving Range Control Charts

 Upper Control Limits UCLX = X+E2R

 Lower Control Limits LCLX = X-E2R
 Upper Control Limits UCLR = D4R

y
nl
 Lower Control Limits LCLR = D3R

O
Subgroup Size (n) d2 D3 D4 E2
2 1.128 - 3.267 2.660

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 64
Pu
g
in

Table of Constants for (I-MR) Charts

in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 65
 ATTRIBUTE CONTROL CHART

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 66
Pu
g
in

SPC – Charts For All Occasions

in

No Defects
Variable Attribute
ra

Data? Data?

Yes % Defective
rT

No Constant Constant
Rational
I & MR chart Sample Sample
Subgroups?
Size? Size?
Fo

No No
Yes Yes Yes

p-chart u-chart
X-bar & R chart

np- or p-chart c- or u-chart

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 67
 Control Charts for Attribute Data

 Defective Data
 Defect Data
- Package does/does not leak
- No of accident/Month
- Shade Close to Standard/Not
Close to Standard - Paint flaws on a casing
- Finish: Smooth Bits Free/Not - Errors on an invoice
Smooth Bits Free - No of Pin Holes/Sq. Metre

y
- Correct/Incorrect Shipment
 c- & u-charts

nl
 p- and np-charts - u-chart for the number of
- p-charts for proportion non- defects per inspection unit,

O
conforming units, sample size sample size not necessarily
not necessarily constant constant
- np-charts for number of non- - c-chart for the number of

e
conforming units, sample size is defects, sample size constant

os
constant

rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 68
Pu
g
in

Control Charts (Attributes)

in

Why Attribute Control Charts

ra

 For both technical as well as

rT

administrative Process
Control
Fo

 No additional data required

 Quick and inexpensive to
obtain
 Prioritize Problem Areas
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 69
 p- Chart
 When to use a p-chart
- When it is difficult or uneconomical to make a numerical
measurement
- When it is desired to combine different types of defects into an
overall proportion
- When the available data are for attributes

y
 Management summaries

nl
- Many management summaries are attribute forms and could
benefit from control chart analysis.

O
- Examples: scrap rates, quality audits, first-run yields, etc.

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 70
Pu
g
in

Subgroup Size and Sample Frequency

in

 Subgroup size
- Subgroups are usually quite large (50 to 200 or more!)
ra

- Ideally, each subgroup should have at least 5 non-conforming units.

- Minimum: 90% of the subgroups must have at least one non-
conforming unit.
rT

- Subgroup sizes need not be constant, but should be within ± 25% of

the mean subgroup size.
- The lower the number of non-conforming units, the larger the required
subgroup size.
Fo

- In general, attribute charts need much larger subgroup sizes than the
equivalent variable charts.
 Sample frequency
- Depends on the cost of operating in an out-of-control state.
- Short time intervals allow for faster feedback, but may not meet
subgroup size requirements.
 No of Sub Groups
- Major source of variation incorporate (25 or more)

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 71
 p-Control Chart

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 72
Pu
g
in

p-Control Chart
in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 73
 Calculating Control Limits For a p-Chart

 For each subgroup

- Record: n – the number of units inspected
np– the number of nonconforming units
- Calculate p – the proportion non-conforming:

y
nl
O
- Calculate the control limits:

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 74
Pu
g
in

p- Control Chart
in
ra
rT
Fo

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 75
 Interpreting p-Charts

 Attribute charts have fewer tests for special causes due to the
lower information content of attribute data. These are:
- One Point more than 3 Sigma from centre line
- Nine points in a row on same side of centre line
- Six points in row, all increasing or all decreasing
- Fourteen points in a row, alternating up and down.

y
nl
 A point outside the 3 limit is assumed to be from special cause
variation.

O
 When np<5, the likelihood of runs below p-bar increases.
Take care when interpreting runs in these cases.

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 76
Pu
g
in

Process Capability for Attribute

in

Capability (Attribute)= 1-p bar

or with the help of Z-table
ra

If defective rate is 2%, Capability shall be:

= 1-0.02= 0.98 OR 98%
rT

In terms of Z score it is 2.05 (Z Score value for 0.02 (i.e.2%)

Z score = (Specification – Average)/Std. Deviation
Fo

Recall Cpk= (Specification – Average)/(3*Std. Deviation)

Therefore Cpk= Z score/3
For Above example, Cpk= 2.05/3= 0.683

Exercise
Calculate Z Score & Cpk of Rework Batches Example done earlier for p-chart

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 77
 Process Capability for Attribute

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 78
Pu
g
in

np-Charts
in

 When to use an np-chart?

ra

- Use the same criteria as a p-chart:

 When it is difficult or uneconomical to make a numerical
rT

measurement
 When it is desired to combine different types of defectives into
a single value
Fo

 When the available data are for attributes

 When the data come from a binomial process
- And the subgroups are all the same size
 How is it different from a p-chart?
- The actual number of nonconforming parts are plotted,
rather than the proportion defective
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 79
 Control Limits for np-Charts

 For each subgroup

- Record: np– the number of nonconforming units
- Calculate np-bar–the mean # nonconforming units:

y
nl
- Calculate the control limits:

O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 80
Pu
g
in

Charts for Defects– c-Charts and u-Charts

in

 When to use c- or u-charts

ra

- When the data is attribute data of defects, not defectives

rT

- When the data comes from a Poisson distribution

- c-charts are charts constructed for number of occurrences
for a constant exposure
Fo

- u-charts are constructed for rate of occurrence for either

constant or varying exposure

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 81
 C Control Chart

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 82
Pu
g
in

Calculate c-Chart Control Limits

in

For each subgroup

ra


- Record: c – the number of defects found
rT

- Calculate c-bar – the mean number of defects:

Fo

- Calculate the control limits:

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 83
 C Control Chart

y
nl
O
e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 84
Pu
g
in

C Control Chart
in
ra
rT
Fo

 Interpretation rules for c- & u-charts are the same as those for p-and
np-charts.
 Because all of the points fall randomly within 3s process limits, you
can conclude the process is running normally.

© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 85
 u-Charts
 When
- When the data are from Poisson process
- When the data are rates of occurrence with different sample sizes

 Calculate Control Limits

- Calculate the process

y
average nonconformities:

nl
O
- Calculate the control limits:

e
os
rp
© Quality Growth Services Pvt. Ltd. Unpublished proprietary work available only under license. All rights reserved. 86
Pu
g
in
in

Keep in touch
ra
rT

www.twitter.com/Official_QGSPL www.facebook.com/QGSPL/ www.linkedin.com/company/qgspl/

Fo

E‐Mail Us Give us a call

[email protected] www.youtube.com 011‐25431737

For more information, visit our Website

www.qgspl.com