7 Quality Improvement Tools

Objectives of Todays Program

To impart skills in using Quality Improvement tools for Problem-solving


; Develop QC way of thinking under the TQM Framework



To develop the habit of making annual improvements in Quality and annual reductions

in quality - related costs




7 Quality Improvement Tools
Quality is a never ending process.


I t is not a destination, but a journey.


- Dr. Edward Deming



Total Quality Definition
Total Quality is total in 3 senses :-

it covers every Process
every Job &
every Person


TQM is based on the principles of :

a) Customer Satisfaction

b) Continuous Improvement

c) Speaking with facts

d) Respect for people


Improvement Storyboard
Problem Solving Model

PDCA is one of the models evolved by the father of TQM Movement, Dr Edward Deming
for Problem solving / Process - improvements.


Plan
Select the process / problem
Describe current process
Describe all possible causes &
agree on root cause
Develop an effective and workable
solution
Do
Implement the solution or process
change
Plan
Do
Check
Review and evaluate the result
of the change
Check
Act
Reflect and act on learnings
Act
Typical Tools used : Brainstorming, Affinity Diagram, Check Sheet, Control chart,
Histogram, Interrelatinship Diagram, Pareto Chart, Process Capability, Radar chart
Run chart.
Improvement Storyboard
Indian Scenario


Plan
Do Check
Fight
Argue on who has to take the
blame !
Fight
7 Quality Improvement Tools
Frequency Sum
A
B
C
7
4
11
Check Sheet
100
0
50
10
20
Pareto Diagram
Stratifying Diagram
Histogram
Fishbone Diagram
10
5
0
2 4 6 8 10
Correlation Diagram
10
5
15
20
0
UCL
CL
LCL
Control Chart X, R
7 QC Tools
Frequency Sum
A
B
C
7
4
11
Check Sheet
100
0
50
10
20
Pareto Diagram
Stratifying Diagram Histogram
Fishbone Diagram
10
5
0
2 4 6 8 10
Correlation Diagram
10
5
15
20
0
UCL
CL
LCL
Control Chart X&R
DATA COLLECTION TOOL DATA ANALYSIS TOOLS
DATA DISPLAY & MONITORING TOOLS
Check Sheet
Frequency Sum
A
B
C
7
4
11
Check Sheet
(Counting & accumulating data)
Check sheet is used to systematically record and compile
data from historical sources, or observations as they
happen, so that patterns and trends can be clearly
detected and shown.
It makes the patterns in the data become obvious.
It forces agreement on the definition of each condition or
event (every person has to be looking for and recording the
same thing)
Check Sheet
Agree on the definition of the events or conditions being observed.
Eg. If you are looking for reasons for late payments, agree on the
definition of late

Decide who will collect the data; over what period; and from what
sources.
- Data collector should have time and knowledge
- Collect data over sufficient period to represent typical event
- Data collection by stratification is necessary to reflect important
differences within a population

How do I do it ?
Check Sheet
Examples :: Rejections of machined castings received from 3 suppliers
Defect
Supplier-A Supplier-B Supplier-C Total
No. of parts
Blow-holes
Total
Dimensional
deviations
Surface finish
Damages
Uncleared patch
27
12
10
24
19
19 35 38 92
Project : Rejections in Castings
Location : Incoming Inspection
Name : abc
Dates : 3/10/00 to 3/16/00
What type of data do you have ?
Words ?? Numbers ???

Attribute Data
Attribute data can be counted and plotted as discrete points
Eg. % of good or bad, Pass or fail, Yes or no.
No. Of people in this room, % defective trucks per week

Variable data

Variable data can be measured and plotted on continuous scale

Eg. Length, time, volume, weight...
Check Sheet
Inference
All Suppliers have problem of blowholes. Probably due to
bad design of casting.
Only Supplier B has problem of unclear patches. Probably
due to pattern equipment.
Supplier-C has high rejections for dimensional deviations.
Machining process needs to be studied.
All Suppliers have problem of damages. Probably due to
wrog packing suggested by the Customer.
Check Sheet
Histograms
Histograms
A Histogram is a graphic summary of variation in a set of
data.
All activities have variability Eg. Values in a set of data
almost always shows variation
The patterns of variability reveal facts about the process
Patterns of variability can suggest and test theories
Interpretation
- Classify Pattern
- Develop explanation for the Pattern
Common Histogram Patterns
Bell shaped - Normal Distribution
Double - peaked
LSL USL LSL USL
Process stable and capable of meeting specifications Two process streams mixed. Ie data from two different
times, operators, machines, etc
Common Histogram Patterns
Edge - peaked
LSL USL
LSL
USL
Special assignable cause, measuring error or
data entry / recording error.
Suggests two distinct processes at work
Second peak indicates abnormality.
Isolated - peaked
Common Histogram Patterns
Plateau
Comb
LSL
USL LSL USL
Process variation too large for specification limits
Granularity - Inspectors rounding measurements or
measuring instrument resolution not fine enough
Common Histogram Patterns
Process shifted to left, or measurements
out of calibration
Skewed
LSL USL LSL USL
Truncated
Truncated data - Nonconformances not reported
Histograms and Limits of acceptability
LSL
USL
Acceptable, even if process varies slightly.
LSL USL
Process acceptable now, but slight change
will make it unacceptable. Should reduce
variability
Nominal
Nominal
How to construct Histograms
1 Decide on the process measure & Gather data
Obtain the table of raw variable data, measured on a continuous scale. Eg. Weight,
temperature, time, dimensions, speed etc.

2 Prepare a frequency table from the data collected.
Count the number of data points n in the sample
Determine the range R in the sample. Range = High value - Low value.
Determine the no. of class intervals k needed.(Use table for reference)
Determine the class width H. Class width = R / k

No. Of Data points n No. Of Classes k
< 50 5 - 7
50 - 100 6 - 10
100 - 250 7 - 12
> 250 10 - 20
How to construct Histograms
3 Determine the Class boundaries or end points
Use the smallest individual measurement in the sample or round off appropriately. This is
the lower end point.
Add the class width H to the lower end point. This gives lower end point for next class
interval. (Each class interval must be mutually exclusive)
Continue process till k class intervals are completed, and the table is completed.

4 Draw Histogram from frequency table
On Y axis (vertical) draw the frequency scale
On X axis (horizontal) draw the scale related to the variable under measure.
For each class interval, draw a vertical bar with height equal to frequency tally.
Complete the graph with title, labeling, targets and specification limits


How to construct Histograms
S D S D S D S D S D
A 9.95 C 9.86 B 9.64 B 9.85 A 9.93
A 10.04 C 9.81 B 10.13 B 10.15 A 10.04
A 9.96 C 9.98 C 9.87 B 10.07 A 10.04
A 10.00 C 9.91 C 9.83 B 9.87 A 9.95
A 9.88 C 9.84 C 9.96 B 10.11 B 9.82
A 10.03 C 9.87 C 9.88 B 9.92 B 9.95
A 10.01 A 9.98 C 9.94 B 10.12 B 9.88
A 9.97 A 9.99 C 10.03 B 9.86 B 10.08
A 9.96 A 9.97 C 9.86 C 9.86 B 10.14
A 9.91 A 9.91 C 9.93 C 9.80 B 9.68
B 10.12 A 10.01 C 9.82 C 9.97 B 9.91
B 9.83 A 10.04 C 9.93 C 9.84 B 9.89
B 10.07 A 9.96 A 9.97 C 9.87 B 10.08
B 9.86 A 10.02 A 10.04 C 10.07 B 9.81
B 10.14 A 9.98 A 10.13 C 9.81 C 9.84
B 10.07 A 10.08 A 10.01 C 9.85 C 9.95
B 9.88 B 9.86 A 10.03 C 9.94 C 9.82
B 10.13 B 9.80 A 9.96 C 9.81 C 9.95
B 9.84 B 10.13 A 10.14 A 10.05 C 9.88
B 10.19 B 9.94 A 10.02 A 9.99 C 9.87
C 9.90 B 9.67 A 10.00 A 10.06 C 9.92
C 9.88 B 10.16 A 9.94 A 10.01 C 9.81
C 9.91 B 10.10 B 10.12 A 9.96 C 10.02
C 9.97 B 9.85 B 10.04 A 10.01 C 9.83
Data on Pin Diameters (120 values)
Part No. 128B Pin
Specification :
10.00 +/1 0.2 mm



S - Supplier A,B,C
D - Measured Pin Dia
How to construct Histograms
Data on Pin Diameters (120 values)
Part No. 128B Pin
Interpretation of Histogram
Data on Pin Diameters (120 values)
Part No. 128B Pin
Supplier A
9.60-9.64
9.65-9.69
9.70-9.74
9.75-9.79
9.80-9.84
9.85-9.89
9.90-9.94
9.95-9.99
10.00-10.04
10.05-10.09
10.10-10.14
10.15-10.19
Supplier B
9.60-9.64
9.65-9.69
9.70-9.74
9.75-9.79
9.80-9.84
9.85-9.89
9.90-9.94
9.95-9.99
10.00-10.04
10.05-10.09
10.10-10.14
10.15-10.19
Supplier C
9.60-9.64
9.65-9.69
9.70-9.74
9.75-9.79
9.80-9.84
9.85-9.89
9.90-9.94
9.95-9.99
10.00-10.04
10.05-10.09
10.10-10.14
10.15-10.19
Upper Spec Limit 10.2
Lower Spec Limit 9.8
Nominal Value 10.0
Understanding frequency polygons
Understanding frequency polygons
Central Tendency :

The most useful measure of Central tendency is the Mean or Average. It is simply the sum of all
measurements on individual units, divided by the number of units.

Mathematically X bar = x / n

Other measures of Central tendency are Mode, Median, HM & GM

Dispersion :
Measures of Dispersion are Range, Standard Deviation. Range is the difference between the largest
and smallest values in the data. It is not a very efficient measure.

Standard Deviation ( ) is the most important measure of dispersion.

Mathematically, = (x - xbar)
2

---------------
(n-1)

Significance of Normal Curves
New Scale
Old Scale
The second step : analysing & reengineering the process
so that the value of sigma drops. Thus, if the upper and
lower specification limits originally stood at +/- 3 sigma, they
will now automatically stand at +/- 6 sigma now
1 -1 -2 2 3 -3
2 4 6
-2 -4
-6
A Six sigma process should actually generate
no flaws at all. But since, over the long run,
every process shifts by +/- 1.5 sigma, 3.4 parts
out of every million will go even beyond the 6
level, generating a few defects.
6 -6 4 -4 2 -2
X
X
Significance of Normal Curves
New Design Width
99.73 percent
68.26 percent
95.46 percent
3 -3 2 1 -1 -2
-3
X
Most of the output of a process with meet the
specification at the Mean (X). But some will
deviate, to varying extent measured by the
Standard Deviation ().

Original Design width
Original Defects
2 3 -2 -1 1
X
The Six sigma approach tackles this problem
in 2 ways :

First, it widens the design width, stretching
the upper and lower specification limits so
that even if the product ranges between X+3
or X - 3 , it will function perfectly.
6 benchmark
Defect Rate
(PPM)
Sigma Competitive
Level
3.4 6
233 5
World Class
6,210 4
66,807 3
Industry
average
308,537 2
690,000 1
Non-
competitive
Statistical Definitions
2
8
Definition of Mean :

Average Ht of Water =( 2 + 8)
----------- = 5 feet
2
Water is shallow & safe !


2
8
6
Definition of Standard Deviation :

Average Ht of Water =( 2 + 8)
----------- = 5 feet
2
Deviation from Mean = 8 - 5 => 3 feet

Conclusion : Water is too deep !
Standard deviation is the right approach for
analysing variations.

Pareto analysis
100
0
50
20
Pareto analysis
Pareto analysis is a ranked comparison of factors related to a quality problem. It helps a
quality improvement project team to identify and focus on the vital few factors.

It is named after an Italian born economist, Wildredo Pareto who observed that a few
people held the majority of the Wealth.

Pareto Principle

Pareto Principle is an universal phenomenon. We see the concept of Vital few &
Useful Many in every day business situations, and holds without exception in problems
of quality.
The top 15 % of our customers account for > 70 % of our total revenues.
In a 25 step manufacturing process, 5 operations acount for over 65 % of scrap.
In a typical meeting, a few people tend to make majority of the comments, while most
people are relatively quiet.

Understanding Pareto diagrams
Annual cost of poor quality
Automobile Plant (GM, Canada)
0
20
40
60
80
100
a b c d e f g h i j k
D
o
l
l
a
r
s

p
e
r

u
n
i
t

p
r
o
d
u
c
e
d
USD % Cum freq
a Final Inspection 39.56 35.26 35.26
b Repairs 30.88 27.52 62.79
c Scrap 15.77 14.06 76.84
d Warranty 10.32 9.20 86.04
e Assembly Division Repairs 5.55 4.95 90.99
f Material 3.09 2.75 93.74
g Maintenance 2.08 1.85 95.60
h Reliability 1.45 1.29 96.89
i Miscellaneous 1.25 1.11 98.00
j Vendor Scrap 1.15 1.03 99.03
k Indirect Inspection 1.09 0.97 100.00
Analysis by Category
Vital Few
Useful Many
%
How to interpret Pareto Diagram
Goal of Pareto analysis, is to separate the numerous problems or causes into two
categories : Vital Few , and the Useful Many.


0
10
20
30
40
50
60
70
80
90
100
1
Pareto Diagram & the Awkward Zone
Awkward
Zone
Useful Many
Vital Few
Categories
How to construct Pareto Diagram
1. Total the raw data and determine the grand total





2. Re-order the contributors from the largest to the smallest


3. Determine the cumulative percent of total for each contributor
on the ranked list.




Cause No of queries
Address Wrong 8
Time Incorrect 20
Rate factor Incorrect 4
Service code Wrong 32
Format confusing 6
Payment not credited 3
73
Cause No of queries
Service code Wrong 32
Time Incorrect 20
Address Wrong 8
Format confusing 6
Rate factor Incorrect 4
Payment not credited 3
73
Cause No of queries Cum %
Service code Wrong 32 44
Time Incorrect 20 71
Address Wrong 8 82
Format confusing 6 90
Rate factor Incorrect 4 96
Payment not credited 3 100
73
How to construct Pareto Diagram
4. Draw and label the vertical and Horizontal axes
5. Draw the bars to represent the effect of each contribution
6. Draw line graph to represent the cumulative percent of total
7. Analyse the Diagram
Customer queries on billing
0
10
20
30
40
50
60
70
80
90
100
Service
code Wrong
Time
Incorrect
Address
Wrong
Format
confusing
Rate factor
Incorrect
Payment not
credited
Reason for customer query
N
o
.

o
f

c
u
s
t
o
m
e
r

q
u
e
r
i
e
s
Useful Many
Vital Few
80
0
When to use Pareto analysis
1. Prioritizing problems
2. Analysing symptoms
3. Identifying Root causes
4. Checking Performance of Remedy
Cause & Effect Diagram
Cause & Effect Diagram
Cause & Effect Diagram was developed by Kaoru Ishikawa
in the year 1943. Hence it is termed as Ishikawa Diagram

Since its shape appears like a fishbone, it is also known as
Fishbone Diagram

Cause & Effect Diagram is used to identify, explore and
graphically display all the possible causes related to a
problem or condition
Cause & Effect Diagram
Cause & Effect Diagram constructed involving a group of
concerned people is generally found to be more fruitful in
solving problems compared to the one made by a single
person
Normally, Brainstorming session precedes the construction
of a Cause - Effect Diagram.
Cause & Effect Diagram
Steps involved in constructing a Cause & effect Diagram
1. Define clearly the effect or problem to be analysed. Eg:
high rejections, low profits, high absenteeism
2. Draw a horizontal line with an arrow at the right hand end
and a box in the front of it. Write the problem (effect) in this
box.

Effect
Cause & Effect Diagram
Steps involved in constructing a Cause & effect Diagram
3. Draw boxes parallel to the main arrow and connect the
boxes to the main arrow by small arrows.
4. Write the major categories of causes in each of the boxes


Effect
Machine Man
Material Method
Cause & Effect Diagram
Steps involved in constructing a Cause & effect Diagram
5. Take up one major category at a time and have a brain
storming on the possible causes of the problem
6. Write down each of the causes around the respective
category


Effect
Man
Cause-2
Cause-1
Cause & Effect Diagram
Steps involved in constructing a Cause & effect Diagram
7. Add subsidiary causes for each cause already identified
8. Continue adding possible causes to the diagram until each
branch reaches a root cause


Man
Cause-2
Cause-1
Sub cause-1
Root cause
Cause & Effect Diagram
Steps involved in constructing a Cause & effect Diagram

9. Check the logical validity of each causal chain
10. Check for completeness
4W s - What, Why, When & Where !
5Ms - Men, Materials, Methods, Machines, Measurements
5Ps - People, Provisions, Procedures, Place, Patrons (Customers)
Cause & Effect Diagram
Group exercise

Poor quality
of photo-copy
Cause & Effect Diagram
Group exercise


Lost control
of car

Scatter Diagram
10
5
0
2 4 6 8 10
Scatter Diagram
Scatter Diagram is used to identify the possible relationship
between the changes observed in two different set of
variables.

What is does?
Supplies data to confirm a hypothesis that two variables
are related.
Provides both a visual and statistical means to test the
strength of a potential relationship.
Provides a good follow up to a Cause & Effect diagram to
find out if there is more that just a consensus connection
between causes and the effect.
Scatter Diagram
How to do it?

Collect 50-100 paired samples of data that you think may
be related and construct a data sheet.
Draw the horizontal (x axis) and vertical (y axis) lines of the
diagram.
Plot the data on the diagram.
Interpret the data.

Scatter Diagram
Types of correlation:
Perfect Positive Correlation:
An increase in y may depend on an increase in x.
Possible positive correlation:
An increase in x may increase y to some extent. In such a
case there may be other variables which contribute to the
relation.
No correlation:
There is no demonstrated connection between x and y axis
Scatter Diagram
Possible Negative Correlation:
An increase in x value, y may be decreased to some
extent. Other variables may influence in the relationship
between x and y variables.

Perfect Negative Correlation:
A decrease in y may depend on an increase in x.
Scatter Diagram
SENSOR EXPERIMENT - RH VS O/P VOLTAGE
(SCATTER DIAGRAM - POSITIVE CORRELATION)
35.00
40.00
45.00
50.00
55.00
60.00
0 20 40 60 80 100 120
RELATIVE HUMIDITY %
O
U
T
P
U
T

V
O
L
T
A
G
E

V
Scatter Diagram - Sensor RH vs o/p Voltage
RH Voltage RH Voltage RH Voltage RH Voltage
10 43.10 20 45.20 30 46.00 40 49.10
10 39.90 20 42.90 30 43.20 40 45.00
10 41.30 20 42.60 30 45.50 40 48.40
10 42.20 20 44.30 30 45.80 40 48.90
10 40.40 20 43.10 30 44.10 40 46.30
50 51.20 60 53.60 70 54.00 80 57.10
50 47.00 60 50.50 70 52.00 80 55.10
50 50.10 60 51.40 70 51.30 80 54.30
50 51.20 60 52.10 70 54.60 80 55.80
50 49.70 60 52.90 70 53.20 80 55.80
90 58.90 100 60.50
90 56.50 100 58.40
90 56.10 100 57.00
90 58.20 100 59.30
90 56.70 100 57.50
Stratifying Diagram
Stratification is the separation of data into categories.
It is used to diagnose which category contributes to the
problem being solved.
The characteristic used to separate the data is called a
Stratification variable
Each variable will have two or more values
A category is defined by specific combinations or ranges of
variables

Stratifying Diagram
1. Select the phenomenon to be studied
eg: Errors in invoices
2. Select the Stratification variables and collect data.
eg: Day of week, Week of month, Accountant
3. Establish the categories for each Stratification variable. (A
category is the value or range of values of the stratification
variable)
eg: Week of month has 5 categories - 1st , 2nd , 3rd, 4th and
5th week.
Stratifying Diagram
4. Sort the observations into the categories of one of the
stratification variables.
5. Calculate the phenomenon measured for each category.
Eg: No. of errors (%) in 1st wk, 2nd wk, 3rd wk, 4th wk, 5th
wk.
6. Display the results in graphical form.
7. Prepare and display the results for additional stratification
variables.
Stratifying Diagram
Example
A small company does silver plating on watch cases for a
large watch manufacturer. The parts have to be free of
blemishes after painting.
The plating Vendor has two plating lines. An old line and a
newly commissioned line. The parts have a mark to
indicate the line in which they were plated.
The plating is done in three shifts. The shift is marked on
the parts.
The plating line is handled (loading/unloading) by one
operator per shift per line. The operators id is also marked
on the parts.
Stratifying Diagram
Example

The watch bases are supplied by four sheet metal
Suppliers. The parts are identified with the Supplier mark
In view of the stringent Q-requirement and the low process
capability, 100% visual inspection was introduced and the
data obtained was stratified.

Stratifying Diagram
Example: Analysis of rejections for blemishes after silver plating of watch bases
Stratifying Diagram
Plating Line Shift Operator Watch base
Stratifying Diagram
Example

From the stratification, the following was concluded.
Although, the old plating line had higher rejections, the
major cause of the problem was one of the line operators.
The third shift showed a higher rejection rate. But it was
found that the problematic operator was on the third shift
during three of the weeks during the month.
Further investigation revealed that this operator was not
following the specified sequence in drying the parts after
plating.

Stratifying Diagram
Points to remember
Before a phenomenon can be analysed effectively with
stratification, the proper identifiers for each observation
must be selected.
Stratification alone will not point out the root cause of a
problem. It will however point out areas for further
investigation.
If the stratification does not yield useful results, additional
data should be collected and a second level stratification
should be used .

Control Charts
10
5
15
20
0
UCL
CL
LCL
Control Chart X, R
Control Charts
Basic definitions:
Variations
Non assignable causes
Assignable causes
Statistics
Process
Stable and Controlled process - Normal distribution
Curve
Machine capability study (Cm and Cmk)

Process capability study (Cp and Cpk)


Control Charts
Variation:

Variation is defined as Deviation from the standard.

e.g. If 20 +/- 2 mm is the spec and 19.00 mm is the observed
reading. Then variation is 1.00 mm.

Control Charts
Non assignable causes:
The non assignable causes are the causes which cannot
be totally eliminated from the process but can only be
reduced to a minimum level.
They are:
Inherent in the process
Sources are many
Follows a definite pattern
Unpredictable
Control Charts
Assignable causes:
The assignable causes are the forced causes by an
external agency and which are avoidable by following
certain precautions.

They are:
Economical to eliminate
Predictable
Forced into the process and few sources
Large in magnitude

Control Charts
Non assignable causes Assignable causes

Material property - r.m. variation Tool breakage/wear
Machine property - vibration M/c setting improper
Ambient conditions - Temp. Poor m/c maint.
Improper jigs/fixtures
Nonconducive
environment
Operator not setting
the dial properly -
process parameters

Control Charts
Statistics:
Statistics is the collection, organising, analysis,
interpretation and presentation of data so as to facilitate
action.
Process:
A process is a sequence of activities or operations to
convert inputs (raw material) into output (end products) by
using resources (machine / men / methods etc).
Stable process:
When a process has only non assignable causes then the
process follows a bell shaped curve and the process is
said to be stable and in statistical control.
Control Charts
Control Charts
Variable charts Attribute charts
Measured and plotted on Counted and plotted as discrete
a continuous scale. events, e.g. shipping errors,
eg. Time, Temp., Cost, % waste, absenteeism
figures
Defect data Defective data
Sample size = 1 Sample size bet. Sample size > 10 Constant sample Variable sample Constant sample Variable sample
3 & 5 size > 5 size size > 50 size >50
X and Rm X and R X and S
C chart U chart np chart p chart
Control Charts
Machine Capability Study: (Cm and Cmk)
To study the variation with respect to the machine alone by
observing consecutive readings (50 or 100) in a short time.

Conditions for study:
Machine to be set to the mean exactly at the start
Set up approval report to be made and approved
Least count of the measuring instrument should be at least
1/10th of the total tolerance given in the spec.
The measuring instrument should be a calibrated one
All assignable causes should be totally eliminated at the
start of study.
Control Charts
Process of study:
Collect consecutive 50 samples from the machine and
number them sequentially.
Record the readings in the chart.
Perform standard deviation calculations and arrive Cm and
Cmk values.
Use the capability test chart for graphical solution

Inference from study:
To know about capability of machine to meet the spec.
Based on the capability values to decide whether machine
needs servicing or maintenance.