Transcript Six Sigma Black Belt Training

LSSG Green Belt
Training
Measure: Finding and Measuring
Potential Root Causes
DMAIC Six Sigma - Measure
Objectives

Identify Inputs and Outputs

Control
Improve

Measure Process Capability

Analyze
Define
Measure

Determine key inputs and outputs
for the process and measures to be
analyzed
Collect data and compare customer
requirements to process variation
Revise Charter

Validate project opportunity and
perform charter revision
Agenda for Measure
1.
2.
3.
4.
Types of Measures/Setting Targets
Data Collection and Prioritization, MSA
SPC, Control Charts
Process Capability
Measures
Purpose of measurement:
Performance of a process
vs. Expectations
Objective
Lose 13 Pounds in
3 months
Secondary
Objective
Lose 1 Pound per
week
Select Measures
Driver(s)
Calories consumed
less Calories
burned
Critical Success
Factors (Drivers
Run 4 miles/day
and consume less
than 1500
calories/day


“SMART” Objectives
Clear operational
definitions
E.g. Losing Weight
Must measure both the result (Y) and the
drivers (Xs). Measure daily – to determine if
CSFs are met, and to make adjustments to
plan.
LSS Measurement
Measurement Plan
Data
What data
type?
Operational Definitions and Procedures
How
measured?
What
conditions?
How to ensure consistency of
measurement?
By who?
Where
measured?
What sample
size?
What is the data collection plan?
Measurement vs. Control
Causes/
Effects
Measurement
System
Control
System
Historical data
Current data
Measurement is not control! So, what is it?
Setting Targets
Set Targets

Objective/Meaningful

Management-employees collaboration

Team goal compatible with value stream objective
Balanced Score Card
Perspectives
Financial
Customer
Internal Process
Learning & Growth
Agenda for Measure
1.
2.
3.
4.
Types of Measures/Setting Targets
Data Collection and Prioritization, MSA
SPC, Control Charts
Process Capability
Data Collection and Prioritization
Some Collection Tools

Customer Survey

Work / Time Measurement

Check Sheet
Some Prioritization Tools

Pareto Analysis

Fishbone Diagram

Cause and Effect Matrix
Work Measurement
Goals of Work Measurement
 Scheduling work and allocating capacity
 Motivating workers / measuring performance
 Evaluating processes / creating a baseline
 Determining requirements of new processes
Time Studies





Typically using stop watches
For infrequent information - estimates OK
Measure person, machine, and delays independently
Medium Duration - not too short; not too long
Eliminate Bias - Compute Standard times from
observed times
Time Study: Calculations


Step 1: Collect Data (Observed Time)
Step 2: Calculate Normal Time from Observed Time,
where:
NormalT ime ObservedT imeper unit * (1  Performanc
e Rating)
use  when operator worksfaster then normal

Step 3: Calculate Standard Time from Normal Time,
where:
StandardT ime NormalT imeper unit * (1  Allowances)
Time Study: Numerical Example
A worker was observed and produced 40
units of product in 8 hours. The supervisor
estimated the employee worked about 15
percent faster than normal during the
observation. Allowances for the job
represent 20 percent of the normal time for
breaks, lunch and 5S.
Determine the Standard Time per unit.
Data Analysis Tools
Run Chart
0.58
12
10
8
6
4
2
0
Diameter
Defects
Scatter Diagram
0.54
0.5
0.46
0
10
20
Hours of Training
30
1
2
3
4
5
6
7
8
9
10
11
12
Time
Can be used to illustrate the relationships
between factors such us quality and training
Can be used to identify when equipment or
processes means are drifting away from specs
Histogram
Control Chart
Frequency
500
UCL
480
460
440
LCL
420
Data Ranges
Can be used to display the shape of variation
in a set of data
400
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15
Use to identify if the process is predictable (in
control)
Cause and Effect Diagram
Machine
Man
Effect
Environmental
Method
Material
Pareto Charts
Root Cause Analysis
80% of the
problems may be
attributed to 20%
of the causes
Design
Assy.
Instruct.
Purch.
Training
Other
Continuous Improvement Process
Orlando Remanufacturing
And
Distribution Center
Phase 1: Internal Kickbacks
Equipment To Be
Remanufactured
Tear Down
And Wash
Remanufacture
Reassembly
Final
Clean-up
Unit
Not OK
To Customer
QA
Five Most Common Reasons For Returns
From QA
Missing/
Wrong
Part
Dirt/Rust
Defective
Part
Leaks
Poor
Insulation
Impact of Reasons for Returns from QA Weighted Average
Leaks
Dirt/Rust Stainless Missing/ Defective
Poor
Steel Wrong Part Part
Insulation
Weighted Avg. = % Occurring X Defect Cost (0-10,
Based on Time to Repair)
Why Dirt? (Fishbone)
Environment
Machinery
Materials
Cleansing Compounds
Tools for $$
Dust/Humidity
Space Limitations
Larger Wire Brushes
Poor Lighting
Dirt
Rework
Attention to Detail
Lack of Communication
QA to IT
Training
Rinse
Measurement
Methods
Man
Environment
Dust/Humidity
Poor Lighting
Space Limitations
Materials
Cleansing Compounds
Need Larger WireBrushes
Machines
Best Tools for $$?
Methods
Reworking Steel after
Valves are Installed
Need to Rinse Parts off
after Sandblasting
People
Need More Training
More Attention to Detail –
Do it Right First Time
Measurement
QA Manager Fixes
Some Things
Without Informing
the Technicians
Why Leaks? (Fishbone)
Environment
Machinery
Materials
“O” Rings Old
Reengineer Rims
Poor Lighting
Bad Tubing
High Temperature
Leaks
No Leak Testing Prior to QA
Quality Check
Identify Most Occurrences
Measurement
Use Wrong Clamps
Forget to Connect
Methods
Mishandle Units
Don’t Crimp Properly
Man
Environment
High Temperatures
Poor Lighting
Materials
Bad Tubing
“O” Rings Too Old (Dry)
Machines
Need Rims That Make it
Easier to Install Tubing
Methods
Check Units for Ways
They Could Leak
Does Testing Create
Leaks?
People
Use Wrong Clamps
Don’t Crimp Properly
Forget to Connect
Measurement
No Testing for Leaks
Prior to QA
Which Mfr./Model Leak
the Most?
Variation Analysis
Most variation without “special” causes will be normally distributed
Variation is typically
classifiable into the 6 M’s
Variation is additive
Variation in the process inputs will generate more variation in the
process output
Variation is Present in All Processes!
Output
Measurement System Analysis (MSA)
Goal - To identify if the measurement system can distinguish
between product variation and measurement variation
2
2
 obseerved
  2product   gage
Some key dimensions

Accuracy

Precision

Bias
Tools: Gage R&R, DOE, Control Charts
Agenda for Measure
1.
2.
3.
4.
Types of Measures/Setting Targets
Data Collection and Prioritization, MSA
SPC, Control Charts
Process Capability
SPC vs. Acceptance Sampling
Acceptance Sampling: Used to inspect a batch prior to, or after the
process
Send to
Accept
Take
Receive
Lot
Customer
Meet
Sample
Criteria?
Rework
Reject
/Waste
Statistical Process Control (SPC): Used to determine if process is
within process control limits during the process and to take
corrective action when out of control
500
UCL
480
460
440
LCL
420
400
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Statistical Process Control
Process in Statistical Control
Statistical process control is
the use of statistics to
measure the quality of an
ongoing process
UCL
LCL
A Process
is in control when all
points are inside the
control limits
Process not in Statistical Control
UCL
LCL
A Process
is not in control when one
or more points is/are
outside the control limits
Process not in Statistical Control
UCL
LCL
Special Causes
When to Investigate
In Control
UCL
Even if in control the process should
be investigated if any non random
patterns are observed OVER TIME
LCL
1
Trend - Constant Increase/Decrease
2
3
4
5
6
UCL
Close to Control Limit
UCL
LCL
1
2
3
4
5
6
LCL
1
2
3
4
5
Cycles
UC
L
UCL
Consecutive Points Below/Above Mean
LCL
5
10
15
20
LCL
1
2
3
4
5
6
Control Chart Development Steps
1
2
Identify Measurement
INPUTS
OUTPUT
X’s
3
Y’s
Start
0.1
Sample
Sample
Size
Defective
p
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Total
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
1500
4
3
5
6
2
1
6
7
3
8
1
2
1
9
1
59
0.04
0.03
0.05
0.06
0.02
0.01
0.06
0.07
0.03
0.08
0.01
0.02
0.01
0.09
0.01
Improve Process
4
Determine Control Limits
Collect Data
Eliminate
Special
Causes
0.08
Reduce
Common
Cause
Variation
0.06
Improve
Average
0.04
0.02
Defects
0
0
2
4
6
8
10
12
14
16
18
A
B
C
D
Frequently Used Control Charts
Attribute: Go/no-go Information, sample size of 50 to 100

Defectives


p-chart, np-chart
Defects

c-chart, u-chart
Variable: Continuous data, usually measured by the mean
and standard deviation, sample size of 2 to 10



X-charts for individuals (X-MR or I-MR)
X-bar and R-charts
X-bar and s-charts
SPC Attribute Measurements
Normal Distribution: Z-Value
p-Chart Control Limits
T otal N um b er of D efectives
p=
T otal N um b er of O bservations
m
-3
Sp =
p (1 - p )
n
UCL = p + Z sp
LCL = p - Z sp
-2
-1
0
1
2
3
Z
Z- VALUE is the number of Standard
Deviations from the mean of the Normal Curve
p
Sp
percentage defects (mean)
Standard deviation of p
Z
Number of standard
deviations
n
Number of observation per
sample (i.e., sample size)
UCL
Upper control limit
LCL
Lower control limit
p-Chart Example
1.
2.
Calculate the sample proportion, p, for each sample
Calculate the average of the sample proportions
Sample
Sample
Size
Defective
p
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Total
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
1500
4
3
5
6
2
1
6
7
3
8
1
2
1
9
1
59
0.04
0.03
0.05
0.06
0.02
0.01
0.06
0.07
0.03
0.08
0.01
0.02
0.01
0.09
0.01
59
p=
= 0.0393
1500
3.
Calculate the sample standard deviation
sp =
4.
p (1- p)
=
n
.0393(1- .0393)
= .0194
100
Calculate the control limits (where Z=3)
UCL = p + Z s p = .0393 3(.0194)= .0976
LCL = p - Z s p = .0393- 3(.0194)= - 0.0189 0
5.
Plot the individual sample
proportions, the average
of the proportions, and the
control limits
0.1
0.08
0.06
0.04
0.02
0
0
2
4
6
8
10
12
14
16
18
SPC Continuous Measurements
n
2
3
4
5
6
7
8
9
10
A2
1.88
1.02
0.73
0.58
0.48
0.42
0.37
0.34
0.31
D3
0
0
0
0
0
0.08
0.14
0.18
0.22
D4
3.27
2.57
2.28
2.11
2.00
1.92
1.86
1.82
1.78
X-bar, R Chart Control Limits
x Chart Limits
UCL = x + A 2 R
LCL = x - A 2 R
R Chart Limits
Shewhart Table of
Control Chart Constants
UCL = D 4 R
LCL = D 3 R
SPC Continuous Measurements
UCL = x + A 2 R  10.54 .58(0.58)= 10.87
LCL = x - A 2 R  10.54- .58(0.58)= 10.19
Sample
1
1
10.6
2
10.7
3
10.5
4
10.9
5
10.9
Sample Sample
Mean Range
10.7
Sample Mean
0.4
2
10.4
11.0
10.4
10.7
10.7
10.6
0.6
3
10.8
10.8
10.8
10.2
10.5
10.6
0.6
4
10.3
10.2
10.3
10.4
11.0
10.4
0.8
5
11.0
10.7
10.9
10.6
10.8
10.8
0.4
6
10.9
10.0
10.4
10.1
10.5
10.4
0.8
UCL
10.90
10.80
X-bar
Chart
10.70
Means
Observation
10.60
10.50
10.40
10.30
10.20
10.10
1
2
3
4
5
6
7
8
9
10
11
12
13
Sample
7
10.8
10.4
10.5
10.7
10.7
10.6
0.4
8
10.1
10.3
10.9
10.2
10.4
10.4
0.8
9
11.0
10.5
10.7
10.8
10.7
10.7
0.5
10
10.8
10.9
10.4
10.3
10.4
10.6
0.6
11
10.5
11.0
10.5
10.8
10.8
10.7
0.5
12
10.2
10.1
10.7
10.8
10.2
10.4
0.7
13
10.8
10.6
10.3
10.4
11.0
10.6
0.7
14
15
LCL
UCL = D 4 R  (2.11)(0.58)  1.22
LCL = D 3 R  (0)(0.58)  0
Sample Range
R
UCL
1.25
1.05
14
10.1
10.3
10.3
10.3
10.8
10.3
0.7
15
10.1
10.1
10.3
10.2
10.1
10.2
0.2
10.54
0.58
Total Average
Chart
0.85
0.65
0.45
0.25
LCL
0.05
-0.15
1
2
3
4
5
6
7
8
Sample
9
10
11
12
13
14
15
Proper Assessment of Control Charts
Find special causes and eliminate

If special causes treated like common causes,
opportunity to eliminate specific cause of variation is
lost.
Leave common causes alone in the short term

If common causes treated like special causes, you will
most likely end up increasing variation (called
“tampering”)
Taking the right action improves the situation
Quarterly Audit Scores
Did something unusual happen?
Score
0
1
2
3
Quarter
4
5
6
Quarterly Audit Scores
What do these lines represent?
Score
0
1
2
3
Quarter
4
5
6
Quarterly Audit Scores
Now what do you think?
Score
0
1
2
3
Quarter
4
5
6
Agenda for Measure
1.
2.
3.
4.
Types of Measures/Setting Targets
Data Collection and Prioritization
SPC, Control Charts
Process Capability
Process Capability Introduction
“Voice of the Process”
(The “Voice of the Data”)
Based on natural (common
cause) variation
Tolerance limits
(The “Voice of the Customer”)
Customer requirements/Specs
Process Capability
A measure of how “capable”
the process is to meet
customer requirements
Compares process limits to
tolerance limits
Process Capability Scenarios
A
C
specification
specification
natural variation
natural variation
B
D
specification
specification
natural variation
natural variation
Process Capability Index, Cpk
Capability Index shows if the process is capable of meeting
customer specifications
 X  LTL
UTL - X 

Cpk = m i n
or


3

3



Find the Cpk for the following:
Mean = 50.50
Stdev = 1.5
A process has a mean of 50.50 and a
variance of 2.25. The product has a
specification of 50.00 ± 4.00
50.00 ± 4.00
Interpreting the Cpk
Cpk > or = 0.33
Cpk > or = 0.67
Cpk > or = 1.00
Cpk > or = 1.33
Cpk > or = 1.67
Cpk > or = 2.00
Capable at 1 *
Capable at 2 *
Capable at 3
Capable at 4
Capable at 5
Capable at 6
* Processes with Cpk < 1 are traditionally called “not capable”.
However, improving from 1 to 2, for example, is extremely valuable.
Calculating Yield
100 units
Task
1
96 units
4 rwk
Traditional Yield (TY)
Task
2
98 units
2 rwk
TY 
0.9910  0.90
Task
3
95 units
5 rwk
Task
4
Total Output of the Final Task
Total Number of Units Started
Rolled Throughput Yield (RTY):
another way to get “Sigma” level
RTY  cum (
0.99100  0.37
90 units
10 rwk
96 units
Task
5
TY 
96
 0.96
100
Units Produced Without Rework
)
Total Number of Units Started
RTY  0.96 * 0.98 * 0.95 * 0.90 * 0.96  0.77
The Hidden Factory = TY - RTY
The Hidden Factory = 0.96-0.77 =0.19
Traditional Yield assessments ignore the hidden factory!
Six Sigma Quality Level
Six Sigma results in at most 3.4 DPMO - defects per
million opportunities (allowing for up to 1.5 sigma shift).
Is Six Sigma Quality
Possible?
IRS Tax Advice
DPMO
Doctor Prescription Writing
1,000,000
100,000
Restaurant Bills
93% good
Airline Baggage Handling
99.4% good
10,000
Payroll Processing
99.98% good
1,000
100
Domestic Airline Flight
Fatality Rate
(0.43PMM)
10
1
1
2
3
4
5
6
SIGMA
Source: Motorola Inc.
Six Sigma Quality
DPMO
Six Sigma Shift
T otalNumber of Defects
1,000,000opportunities

The drift away from target mean over time

3.4 defects/million assumes an average shift of 1.5 standard deviations

With the 1.5 sigma shift, DPMO is the sum of 3.39767313373152 and
0.00000003, or 3.4. Instead of plus or minus 6 standard deviations, you
must calculate defects based on 4.5 and 7.5 standard deviations from
the mean! Without the shift, the number of defects is .00099*2 = .002
DPMO.
Z
4.5
6.0
7.5
P(<Z)
0.99999660232687
0.99999999901341
0.99999999999997
1 - P(<Z)
0.00000339767313
0.00000000098659
0.00000000000003
* 1,000,000
3.39767313373152
0.00098658770042
0.00000003186340
Quality Levels and DPMO
Defects per million opportunities
Assumes 1.5 sigma shift of the mean
Sigma Level
DPMO (Defects per
million opportunities)
Reduction from previous
sigma level
1.0
697672
2.0
308770
55.74%
3.0
66811
78.36%
4.0
6210
90.71%
5.0
233
96.25%
6.0
3.4
98.54%
Regardless of the current process sigma level, a very significant
improvement in quality will be realized by a 1-sigma improvement!
Is Six Sigma Quality Desirable?
99% Quality means that


10,000 babies out of 1,000,000 will be given to the wrong
parents!
One out of 100 flights would result in fatalities. Would you fly?
What is the quality level for Andruw Jones?