Process Capability Analysis

Continuous Improvement

Instructor File
Week Two – Module Six
 Black Belt/ Green Belt Training atEssex
Week One Week Two
5 Self-taught Modules Measure Phase
 1. Six Sigma Overview  Introduction
 2. Recognizing Improvement Opportunities  Clarify the Process/Product & Customer CTs
 3. Defining Opportunities in Sigma TRAC  Characterize the Defect
 4. Meet Minitab®  Determine Measure of Intended Function
 5. Data Collection and Analysis  Validate the Measurement System
 Evaluate Process Behavior
7. Evaluate Process Capability
PROJECT
8. Identify Potential Xs MANAGEMENT

Week Three Week Four

Analyze Phase Improve and Control Phases
1. Introduction 1. Introduction
2. Estimation & CI 2. Determine Counteractions
3. Hypothesis Testing 3. Polynomial and Multiple Regression
4. Correlation and SLR 4. Fractional Factorial Designs
5. ANOVA 5. Robust Design (Static)
6. Basic DOE 6. Control Phase
LEADERSHIP VSM
7. Full Factorial Designs 7. Error-proofing
8. Verifying Xs Project Presentations 8. Summary Project Presentations

Week Five: Review Projects through Improve Phase

Superior Essex, Feb 2014 2 W2-7 Process Behavior_Inst.ppt
Define
Define
1 Identify what’s important to the customer. Define project scope.

Measure
Measure
2 Determine what to measure (Y) and validate the measurement system.

3 Quantify current performance and estimate improvement target.

Analyze
Analyze
4 Identify causes (Xs) of variation and defects.

5 Provide statistical evidence that causes are real.

Improve
Improve
6 Determine solutions (ways to counteract causes) including operating
levels and tolerances.
7 Install solutions and provide statistical evidence that the solutions work.

Control
Control
8 Put controls in place to maintain improvement over time.

9 Provide statistical evidence that the improvement is sustained.

 About This Module
Measure
Measure
2 Determine what to measure (Y) and validate the measurement system.

3 Quantify current performance and estimate improvement target.

 The current process performance (baseline performance) is quantified

in terms of process capability:
 How capable is the process of consistently generating product that
meets specifications (what the customers want and need)?

Process Capability Analysis should be conducted on the

measure of the process’ intended function, the Process Y.
Superior Essex, Feb 2014 4 W2-7 Process Behavior_Inst.ppt
 Process Capability Analysis: Process Y

Process
Process Capability
Capability Analysis
Analysis IS
IS
NOT
NOT performed
performed onon the
the
Business
Business Y
Y (downtime).
(downtime). EXAMPLE:
“CUT HEADLINERS” PROJECT

Output Metrics:
 YB = Downtime minutes per day
Headliner cut
to shape  YP = Cut dimension (mm)

Process
Process Capability
Capability
Analysis
Analysis IS
IS performed
performed
on
on the
the Process
Process YY (cut
(cut GMT 360 STD PET HEADLINER

dimension).
dimension).

Superior Essex, Feb 2014 5 W2-7 Process Behavior_Inst.ppt

 Process Capability Improvement

Cut Dimension (mm)

Histogram of Dimension B

25
Initial
Initial State:
State: By improving the
20
POOR
POOR
CAPABILITY
CAPABILITY capability of the
Frequency

15

Process Y (cut
10

0
0.0 0.5

LSL
1.0 1.5 2.0 2.5

USL
dimension), the
team eliminated
Cut Dimension (mm)
20
Histogram of Dimension B (m.m.)

downtime!
15
Frequency

10

After
After Modified
Modified
0
0.0 0.5

LSL
1.0
A fter
1.5 2.0 2.5

USL
Collision
Collision Sensor
Sensor
Installed
Installed Cut Dimension (mm)
Histogram of Dimension B (m.m.)

BETTER
BETTER 16

14

CAPABILITY
CAPABILITY After
After Auto
Auto Clamp
Clamp 12

10

Installed
Installed Frequency 8

Note: These capability pictures HIGH

HIGH PROCESS
PROCESS 6

CAPABILITY!
4

were made using data from CAPABILITY! 2

corresponding control charts. 0

0.0 0.5 1.0 1.5 2.0 2.5

Superior Essex, Feb 2014 6 W2-7 Process Behavior_Inst.ppt

 Learning Objectives

After completing this module, you will be able to…

1. Determine if data distributions are Normal or non-Normal.

2. Fit a curve to data distributions, regardless of shape.
3. Determine areas under the Normal curve.
4. Quantify process capability from variable data in terms of
Cp, Cpk, Pp, Ppk & PPM.
5. Quantify process capability from attribute data in terms of
PPM or DPU.

Superior Essex, Feb 2014 7 W2-7 Process Behavior_Inst.ppt

 Evaluate Process Capability: Contents

Part One: What is Process Capability?

Part Two: Variable Data:
a) Describing variable data (shape, central tendency,
variation)
b) Process Capability from Variable Data
» Normal data
» Non-normal data
c) Estimating Percentage Beyond Specification Limits
d) Cp, Cpk, Pp Ppk Indices
Part Three: Attribute Data:
– Process Capability from Attribute Data
Part Four: What is Process Sigma Level?
Part Five: Set Improvement Target
Part Six: Example of Measure Phase

Superior Essex, Feb 2014 8 W2-7 Process Behavior_Inst.ppt

 PART ONE

What is Process Capability?

Superior Essex, Feb 2014 9 W2-7 Process Behavior_Inst.ppt

 Process Capability Defined

 Process capability is an estimate of the process’ ability to

meet specifications today and in the future - a form of
statistical inference.

LSL USL

Individual
Measurements

How capable is the process of consistently

producing parts within specification?
Superior Essex, Feb 2014 10 W2-7 Process Behavior_Inst.ppt
 How to Evaluate Process Capability

The approach used to evaluate process capability depends upon

the type of data used: Variable or Attribute.
 For variable data: is the data Normal or Non-normal?

? ?
An unstable
process is
unpredictable.

The process must be stable (in statistical control) in

order to obtain reliable estimates of process capability.

Superior Essex, Feb 2014 11 W2-7 Process Behavior_Inst.ppt

 Four Stable Processes

A B

What can be
said about the
LSL USL LSL USL
capability of
these
C D processes?

LSL USL LSL USL

Superior Essex, Feb 2014 12 W2-7 Process Behavior_Inst.ppt

 How Capable is It?
6 Standard Deviations

LSL USL
CAPABLE PROCESS

6 Standard Deviations 6 Standard Deviations

LSL USL
A VERY LSL A VERY, VERY USL
CAPABLE PROCESS CAPABLE PROCESS

Superior Essex, Feb 2014 13 W2-7 Process Behavior_Inst.ppt

 PART TWO

Evaluating Process Capability

using Variable Data

Superior Essex, Feb 2014 14 W2-7 Process Behavior_Inst.ppt

 Characterizing Variable Data

 Three characteristics are used to describe a set of

variable data:
– Shape (Histogram)
– Central Tendency (Mean, Median, Mode)
– Variation (Range, Standard Deviation, Variance)
 All three characteristics are used when estimating
process capability.

SHAPE CENTRAL TENDENCY VARIATION

Superior Essex, Feb 2014 15 W2-7 Process Behavior_Inst.ppt

 A Histogram is Used to Evaluate Shape

Recall: A histogram is a bar graph in which data are grouped into classes.
The height of each bar shows how many data values fall in each class.

Does a histogram preserve the time

order in which the data was collected?
Superior Essex, Feb 2014 16 W2-7 Process Behavior_Inst.ppt
 Data Sets Produce Various Shapes

NORMAL DISTRIBUTION

BELL SHAPE

NON-NORMAL
DISTRIBUTION

RIGHT SKEWED LEFT SKEWED

(POSITIVELY (NEGATIVELY
SKEWED) SKEWED)

UNIFORM DISTRIBUTION BIMODAL DISTRIBUTION

What are some likely examples of

variable data for each of these shapes?
Superior Essex, Feb 2014 17 W2-7 Process Behavior_Inst.ppt
 The Proper Grouping of Data

 When creating a histogram, the data must be properly

grouped in order to understand the shape of the data
distribution.
 The following guidelines are useful in order to properly group
data sets to produce a useable histogram:

Number of data points Number of classes

Under 50 5-7
50 -100 6 -10
100 - 250 7-12
Over 250 10 - 20

For the given sample size, the right

number of classes should be used.

Superior Essex, Feb 2014 18 W2-7 Process Behavior_Inst.ppt

 Histogram Example

Minitab® Display
99
100
90
80
70
60
50
40
30
20
10 1
0
0 1000 2000 3000

What does this Histogram suggest?

Superior Essex, Feb 2014 19 W2-7 Process Behavior_Inst.ppt
 Exploring Data Using Histograms

 A value of 280 was incorrectly entered as 2800

Minitab® Display Minitab® display with

with incorrect data. correct data.
99
100 30 27
90 24
80
19
Frequency

70
20
60
14
50
40 9
30 10
20 4
3
10 1
0 0
0 1000 2000 3000 220 240 260 280 300 320 340 360

Do not simply view a histogram without questioning

the nature of the distribution – explore!
Superior Essex, Feb 2014 20 W2-7 Process Behavior_Inst.ppt
 The Normal Distribution

“Non-normal”
data will be
discussed later.

Many process outputs display a “bell” shape, or Normal

Distribution. The properties of the normal distribution are
used to make predictions about a process’ capability.

Superior Essex, Feb 2014 21 W2-7 Process Behavior_Inst.ppt

 Properties of the Normal Distribution

68% of parts will fall 95% of the parts will fall 99.73% of the parts will fall
between between between
+/- 1 standard deviations +/- 2 standard deviations +/- 3 standard deviations
from the mean from mean from mean
Superior Essex, Feb 2014 22 W2-7 Process Behavior_Inst.ppt
 Statistical Test for Normality

 While the histogram can be used to make a “judgment call”

about normality, a statistical test for normality is also
conducted.
 Minitab® uses the Anderson-Darling test to determine if a set of
data can be treated as normal data.

Interpreting the P-value (probability value):

 The P-value is the probability Example: Assembly Time

of getting the particular sample
if the population is normal.
 P-value < 0.05 means that the
chance of getting this sample
from a normal population is
very small (less than 5%). P-value = 0.001 Not Normal

A P-value < 0.05 suggests the population is not normal.

Superior Essex, Feb 2014 23 W2-7 Process Behavior_Inst.ppt
 Workbook: Exercise 7.1

Test for Normality

Superior Essex, Feb 2014 24 W2-7 Process Behavior_Inst.ppt

 Exercise 7.1: Test for Normality

Allotted Time
10 MINUTES

ASSIGNMENT:
ASSIGNMENT:
1.
1. Open
Open the
the data
data file
file “7.1
“7.1 Tensile.MTW”.
Tensile.MTW”.
2.
2. For
For each
each example
example (P (P Time,
Time, Tensile):
Tensile):
a)
a) Create
Create aa histogram
histogram using
using Minitab
Minitab®®..
b)
b) Run
Run an
an Anderson
Anderson Darling
Darling normality
normality test
test using
using Minitab
Minitab®®..
c)
c) Answer:
Answer: IsIs the
the data
data normally
normally distributed?
distributed?

To run Anderson-Darling normality test on P Time data:

Select: Stat > Basic Stat > Normality test.
Fill in: Variable = P Time; Percentile Lines = None;
Tests for Normality = Anderson Darling. Click OK.

Superior Essex, Feb 2014 25 W2-7 Process Behavior_Inst.ppt

 Workbook: Exercise 7.2

Describing Data

Superior Essex, Feb 2014 26 W2-7 Process Behavior_Inst.ppt

 Exercise 7.2: Describing Data

Allotted Time
20 MINUTES

ASSIGNMENT:
ASSIGNMENT:
 Open
Open Minitab
Minitab data
data file
file “7.2
“7.2 Describing
Describing Data.MTW”
Data.MTW”
 For
For the
the data
data set
set to
to which
which your
your team
team has
has been
been assigned:
assigned:
1.
1. Describe
Describe the
the data
data set
set in
in terms
terms of
of central
central tendency,
tendency,
variation
variation and
and shape.
shape. Is
Is the
the data
data normally
normally distributed?
distributed?
2.
2. Use
Use Minitab
Minitab®® to
to generate
generate graphs.
graphs.
3.
3. Present
Present the
the descriptive
descriptive statistics
statistics and
and graphs
graphs to
to the
the
class
class and
and discuss
discuss what
what your
your preliminary
preliminary analysis
analysis
shows
shows about
about the
the data.
data.

The five situations are summarized on the next slide.

Superior Essex, Feb 2014 27 W2-7 Process Behavior_Inst.ppt

 Exercise 7.2: Describing Data

 In five situations; data has been collected.

 All the data is time ordered.
A. COST: Daily costs collected over 15 weeks.
Subgroups of size n=4
B. CUTOFF: Felt strips cut by hand with a positive
stop at 110mm. Subgroups of size n=5
C. ID: Inner dimension on a seat track. Nominal is
8cm.
Subgroups of size n=5
D. THICKNESS: Foam sheet thickness.
Subgroups of size n=5
E. DELIVERY TIME: Time in days for delivery of
prototype parts. Subgroups of size n=4

Superior Essex, Feb 2014 28 W2-7 Process Behavior_Inst.ppt

 Fitting Distributions to Predict for the Population
Example #1: Histogram
Histogram ofofsample
of a sample n=100 of n=100 Normal Population Distribution
25
0.4

20

0.3

15
Frequency

Density
0.2
10

0.1
5

0.0
0 8 9 10 11 12 13
8 9 10 11 12 13

Example #2: Histogram

Histogram of sampleof sample of n=100
of n=100 Lognormal Population Distribution
50

0.25

40
0.20

30
Frequency

Density
0.15

20
0.10

10 0.05

0 0.00
0 2 4 6 8 0 2 4 6 8

We begin by looking at Normal Populations.

Superior Essex, Feb 2014 29 W2-7 Process Behavior_Inst.ppt
 Area under the Normal Curve

 When the population is normal, areas under the normal curve

are used to estimate the percentage of product that exceeds
specification limits.

LSL USL

3% 12%

Superior Essex, Feb 2014 30 W2-7 Process Behavior_Inst.ppt

 Predicting for the Population: Normal Distributions

BASED ON THIS SAMPLE WE PREDICT FOR THE POPULATION

LSL = 10 USL = 40
10 LSL = 10 USL = 40
9
8
7 m = 27.1
6 s = 8.3
5
4
3
2 2% 6%
1
0
10 15 20 25 30 35 40
5 10 15 20 25 30 35 40 45 50

 In total, 8% of the product will

exceed specification limits.
 2% below lower spec, 6%
above higher spec

How are these estimates are obtained?

Superior Essex, Feb 2014 31 W2-7 Process Behavior_Inst.ppt
 Standard Normal Distribution

The Standard Normal

0 y
Distribution, often called z
the z distribution, has:  1 

Where z is the
number of
standard
deviations from
the mean. -3s -2s -1s m 1s 2s 3s
z
-3 -2 -1 0 1 2 3 Scale

Superior Essex, Feb 2014 32 W2-7 Process Behavior_Inst.ppt

 z is the Number of Standard Deviations from the Mean

-3s -2s -1s m 1s 2s 3s

What is the value of z

at the 2 point?
Superior Essex, Feb 2014 33 W2-7 Process Behavior_Inst.ppt
 Translate Actual Units to z Units

PLACEMENT TIME FOR CLERICAL POSITIONS

y = 10 . 0 days

s = 2 . 5 days

2.5 5.0 7.5 10.0 12.5 15.0 17.5 Days

-3 -2 -1 0 1 2 3 z Scale

What percentage of the placements

require from 5 to 15 days?

Superior Essex, Feb 2014 34 W2-7 Process Behavior_Inst.ppt

 Determining Corresponding z Values

PLACEMENT TIME FOR CLERICAL POSITIONS

%?

12.0
2.5 5.0 7.5 10.0 12.5 15.0 17.5 Days

-3 -2 -1 0 1 2 3 z Scale

y -m -
z= = 12.0 10.0 = 2.0 = 0.80
s 2. 5 2. 5

What percentage of the placements require 12 or more days?

Superior Essex, Feb 2014 35 W2-7 Process Behavior_Inst.ppt
 Using the z Table

PLACEMENT TIME FOR CLERICAL POSITIONS

z = 0.80

21%
P = 0.2119
From z Table
12.0
2.5 5.0 7.5 10.0 12.5 15.0 17.5 Days

-3 -2 -1 0 1 2 3 z Scale

y -m -
z= = 12.0 10.0 = 2.0 = 0.80
s 2.5 2. 5

Superior Essex, Feb 2014 36 W2-7 Process Behavior_Inst.ppt

 z-table

z Table in Decimal beyond a given value (for z values of 0.5 to 2.09)

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.5 0.308537533 0.305025719 0.301531771 0.298055944 0.294598489 0.291159655 0.287739682 0.284338808 0.280957264 0.277595276
0.6 0.274253065 0.270930848 0.267628834 0.264347230 0.261086235 0.257846044 0.254626846 0.251428824 0.248252158 0.245097021
0.7 0.241963578 0.238851994 0.235762424 0.232695018 0.229649924 0.226627280 0.223627221 0.220649876 0.217695369 0.214763817
0.8 0.211855334 0.208970026 0.206107994 0.203269335 0.200454139 0.197662492 0.194894473 0.192150158 0.189429614 0.186732906
0.9 0.184060092 0.181411225 0.178786354 0.176185520 0.173608762 0.171056112 0.168527597 0.166023240 0.163543057 0.161087061
1.0 0.158655260 0.156247655 0.153864244 0.151505020 0.149169971 0.146859081 0.144572328 0.142309686 0.140071125 0.137856610
1.1 0.135666102 0.133499557 0.131356927 0.129238161 0.127143201 0.125071989 0.123024458 0.121000541 0.119000166 0.117023256
1.2 0.115069732 0.113139509 0.111232501 0.109348617 0.107487762 0.105649839 0.103834747 0.102042381 0.100272634 0.098525394
1.3 0.096800549 0.095097982 0.093417573 0.091759198 0.090122734 0.088508052 0.086915021 0.085343508 0.083793378 0.082264493
1.4 0.080756711 0.079269891 0.077803888 0.076358555 0.074933743 0.073529300 0.072145075 0.070780913 0.069436656 0.068112148
1.5 0.066807229 0.065521737 0.064255510 0.063008383 0.061780193 0.060570771 0.059379950 0.058207562 0.057053437 0.055917403
1.6 0.054799289 0.053698923 0.052616130 0.051550737 0.050502569 0.049471451 0.048457206 0.047459659 0.046478632 0.045513949
1.7 0.044565432 0.043632903 0.042716185 0.041815099 0.040929468 0.040059114 0.039203858 0.038363523 0.037537931 0.036726904
1.8 0.035930266 0.035147838 0.034379445 0.033624911 0.032884058 0.032156713 0.031442700 0.030741845 0.030053974 0.029378914
1.9 0.028716493 0.028066539 0.027428881 0.026803350 0.026189776 0.025587990 0.024997825 0.024419115 0.023851694 0.023295398
2.0 0.022750062 0.022215525 0.021691624 0.021178201 0.020675095 0.020182148 0.019699203 0.019226106 0.018762701 0.018308836

Superior Essex, Feb 2014 37 W2-7 Process Behavior_Inst.ppt

 Workbook: Exercise 7.3

Determine Area Using z Table

Superior Essex, Feb 2014 38 W2-7 Process Behavior_Inst.ppt

 Exercise 7.3: Determine Area Using z Table

Allotted Time
5 MINUTES

ASSIGNMENT:
ASSIGNMENT:
Using
Using the
the previous
previous example:
example: “Placement
“Placement
Time
Time for
for Clerical
Clerical Positions”:
Positions”:
 Follow
Follow in
in your
your workbook
workbook toto determine
determine
what
what percentage
percentage of
of the
the placements
placements
require
require 14
14 or
or more
more days.
days.

Superior Essex, Feb 2014 39 W2-7 Process Behavior_Inst.ppt

 Measures of Process Capability

 In addition to estimating percentage of product that falls outside

specifications, we also quantify process capability using
capability indices.

LSL USL

Recall:
The 6s process
spread contains
99.73% of the
total area.
Individual
Measurements

6s

Superior Essex, Feb 2014 40 W2-7 Process Behavior_Inst.ppt

 Cp Index Demonstrates Potential Capability

USL
LSL

Cp = (USL – LSL)/ 6s
Cp = 1  Cp is the ratio of Total
Tolerance to the 6
Process Spread.
3 3
USL  It shows how capable
LSL

the process would be if it

Cp =1.33 were perfectly centered.
 Cp = Potential Capability
4 4
USL
LSL

Cp = 2 This is a
“Six Sigma”
process.

6 6

Superior Essex, Feb 2014 41 W2-7 Process Behavior_Inst.ppt

 Cp Index: Tolerance Band in Terms of Standard Deviations

USL
LSL

Cp = (USL – LSL)/ 6s
Cp = 1
Cp = 6/6
=1
3 3
USL
LSL

Cp =1.33
Cp = 8/6
= 1.33
4 4
USL
LSL

Cp = 2 Cp = 12/6
=2

6 6

Superior Essex, Feb 2014 42 W2-7 Process Behavior_Inst.ppt

 Cpk Index Demonstrates Actual Capability

USL – Avg.
Cpk =

USL
3s
LSL

Cp = 2 Cpk takes into

Cpk = 1.33 account any off-
centering that
actual occurs.

Cp = 2
When process is
Cpk = 2 centered, Cp = Cpk.

Superior Essex, Feb 2014 43 W2-7 Process Behavior_Inst.ppt

 Cp, Cpk Vs. Pp, Ppk

PROCESS
PROCESS IN
IN PROCESS
PROCESS NOT
NOT IN
IN
STATISTICAL
STATISTICAL CONTROL:
CONTROL: STATISTICAL
STATISTICAL CONTROL:
CONTROL:
Cp Pp
Cpk Ppk

KEY POINT:
Includes the
 The process must be in statistical control case when
(stable over time) in order for Capability insufficient data
Indices to represent the general state of is available to
a process. test for stability.
 Otherwise, the indices (Pp, Ppk) only
reflect the snapshot in time in which the
data was taken.

Superior Essex, Feb 2014 44 W2-7 Process Behavior_Inst.ppt

 Computations

Cp, Cpk Pp, Ppk

PROCESS IN PROCESS NOT IN
STATISTICAL CONTROL: STATISTICAL CONTROL:

USL – LSL USL – LSL

Cp = 6 Pp = 6

Cpk = min CPU, CPL Ppk = min CPU, CPL

CPU = (USL - X-bar)/3s CPU = (USL - X-bar)/3s
CPL = (X-bar - LSL)/3s CPL = (X-bar - LSL)/3s

s = R-bar/ d2
Or some other estimate of “within s =
å ( x - x )2
subgroup” standard deviation n -1

The only difference between Cp and Pp (Cpk and Ppk)

is the way that the standard deviation is estimated.
Superior Essex, Feb 2014 45 W2-7 Process Behavior_Inst.ppt
 Steps for Evaluating Process Capability
Variable Data: Normally Distributed

Assumptions: Process is stable. Gage R&R passed.

1. Assure the data is normally distributed. (Histogram, Anderson-
Darling test).
2. Estimate the average and standard deviation of the process
(often by using information from an X bar or R chart).
3. Determine the process variation (the 6 standard deviation
spread) and compare to specification limits. Make the capability
picture.
4. Estimate the percentage of product that falls outside the
specification limits. Be sure to use the specification limits that
represent customer needs!
5. Determine the process’ potential capability by calculating the
Cp index.
6. Quantify the process’ actual performance by calculating the
Cpk index.

Superior Essex, Feb 2014 46 W2-7 Process Behavior_Inst.ppt

 1. Assure Data is Normally Distributed

Data collected from a short-term study:

 Run Anderson-Darling test.
 Data is coming from a normal population.

Histogram of Individual Measurements

X Chart

X = 178.6

R Chart

R= 8.4

The Anderson-Darling test was run. p – value = .852

The data can be treated as coming from a normal population.
Superior Essex, Feb 2014 47 W2-7 Process Behavior_Inst.ppt
 2. Estimate Process Average and Standard Deviation

x is directly used as
the estimate of the
X Chart
Process Average, m.
X = 178.6

R will be used in a
R Chart
calculation to estimate
R = 8.4 Process Standard
Deviation, .

R
PROCESS STANDARD DEVIATION: s»
d2
Where: d2 is a constant that depends on subgroup size
(see Table of Control Chart Constants).

Superior Essex, Feb 2014 48 W2-7 Process Behavior_Inst.ppt

 Process Average and Standard Deviation
Process average = 178.6
Average range = 8.4
Sample size = 5

R 8.4
POPULATION s= = = 3.6
d2 2.326

178.6
X PROCESS
PROCESS AVERAGE STANDARD DEVIATION

The target = 171.

Is the process centered on target?
Superior Essex, Feb 2014 49 W2-7 Process Behavior_Inst.ppt
 3. Determine Process Variation and Compare to
Specification Limits

According to

USL = 182
LSL = 160
specifications,
all product is to
fall between 160
and 182.

TARGET

171
167.8 178.6 189.4
x  3 x x  3

Process Variation:
We expect that 99.73% of the time,
we will produce product that falls
between 167.8 and 189.4.

Superior Essex, Feb 2014 50 W2-7 Process Behavior_Inst.ppt

 4. Estimate the Percentage Beyond Specifications

 To estimate the percentage of product that falls outside the

specification limits, we must first compute z upper and z lower.

USL = 182
LSL = 160

167.8 178.6 189.4

z lower is the number of standard z upper is the number of standard

deviations between the Process deviations between the Process
Average and the Lower Average and the Upper
Specification Limit. Specification Limit.

Superior Essex, Feb 2014 51 W2-7 Process Behavior_Inst.ppt

 Calculating z upper and z lower

z upper

USL = 182
LSL = 160

USL - X
zupper =
s
182.0 - 178.6
zupper = =0.94
3.6

167.8 178.6 189.4 z lower

X - LSL
zlower =
s
The lower spec is The upper spec is
5.17 standard 0.94 standard 178.6 - 160.0
deviations away deviations away
zlower = = 5.17
from the process from the process
3 .6
average. average.

Superior Essex, Feb 2014 52 W2-7 Process Behavior_Inst.ppt

 Estimating Percentage Beyond Specifications

 From a z table, we find

USL = 182
LSL = 160
that z = 0.94
corresponds to
proportion = 0.1736 17.4%

 This converts to
17.36% Defective
or: 167.8 178.6 189.4
 173,600 PPM
z upper = 0.94

0.00 0.01 0.02 0.030.04

0.8 0.211855334 0.208970026 0.206107994 0.203269335 0.200454139
0.9 0.184060092 0.181411225 0.178786354 0.176185520 0.173608762
1.0 0.158655260 0.156247655 0.153864244 0.151505020 0.149169971

For z = 0.94, proportion = 0.1736

Superior Essex, Feb 2014 53 W2-7 Process Behavior_Inst.ppt
 5. Determine the Process’ Potential Capability (Cp)
LSL USL  The Cp index reflects the
potential of the process if
the average were perfectly
Cp = 1 centered between the
specification limits.
 The larger the Cp index, the
better!
Cp > 1  For a Six Sigma process,
Cp = 2

USL – LSL
Cp = 6
Cp < 1

Superior Essex, Feb 2014 54 W2-7 Process Behavior_Inst.ppt

 Cp Index Exercise

USL = 182
LSL = 160

TARGET

171
167.8 178.6 189.4 Recall: s = 3.6

CALCULATE USL - LSL ?

CP INDEX: Cp = = ____
6s

Superior Essex, Feb 2014 55 W2-7 Process Behavior_Inst.ppt

 6. Quantify Actual Process Performance (Cpk)

 Unlike the Cp index, the Cpk index takes into account off-
centering of the process.
 The larger the Cpk index, the better.

LSL USL LSL USL

6 6

CP = 1 CP = 1
CPK = 1 Cpk < 1

Superior Essex, Feb 2014 56 W2-7 Process Behavior_Inst.ppt

 Calculating the Cpk Index

zupper = 0.94 & zlower = 5.17

zmin =[Smaller of zupper & zlower]
zmin = zupper = 0.94

zmin 0.94
Cpk = = = 0.31
3 3
When does
Recall: Cp =1.02 Cpk equal Cp?

Superior Essex, Feb 2014 57 W2-7 Process Behavior_Inst.ppt

 Workbook: Exercise 7.4

Process Capability

Superior Essex, Feb 2014 58 W2-7 Process Behavior_Inst.ppt

 Exercise 7.4: Process Capability

Allotted Time
30 MINUTES

ASSIGNMENT:
ASSIGNMENT:
1.
1. For
For the
the example
example toto which
which your
your team
team is
is
assigned,
assigned, determine
determine the
the process
process
capability
capability by
by following
following the
the 66 steps.
steps.
 “Number
“Number ofof Calls
Calls ““ data
data
 “Decal
“Decal Position”
Position” data
data
2.
2. For
For each
each situation,
situation, Step
Step #1#1 has
has been
been
completed
completed for
for you.
you.

The 6 steps are shown on the next slide.

Superior Essex, Feb 2014 59 W2-7 Process Behavior_Inst.ppt

 Exercise 7.4: Process Capability

Assumptions: Process is stable. Gage R&R passed.

1. Assure the data is normally distributed. (Histogram, Anderson-
darling test).
2. Estimate the average and standard deviation of the process
(often by using information from an X bar or R chart).
3. Determine the process variation (the 6 standard deviation
spread) and compare to specification limits. Make the capability
picture.
4. Estimate the percentage of product that falls outside the
specification limits. Be sure to use the specification limits that
represent customer needs!
5. Determine the process’ potential capability by calculating the
Cp index.
6. Quantify the process’ actual performance by calculating the
Cpk index.

Superior Essex, Feb 2014 60 W2-7 Process Behavior_Inst.ppt

 Workbook: Exercise 7.5

Process Capability Using Minitab®

Superior Essex, Feb 2014 61 W2-7 Process Behavior_Inst.ppt

 Exercise 7.5: Process Capability Using Minitab®

Part One:
With Instructor
Use Minitab®

ASSIGNMENT:
ASSIGNMENT:
1.
1. Open
Open the
the Minitab
Minitab worksheet:
worksheet:
“7.5
“7.5 Process
Process Capability.MTW”
Capability.MTW”
2.
2. Work
Work along
along with
with the
the Instructor
Instructor to
to conduct
conduct aa Process
Process
Capability
Capability Study
Study using
using the
the “Decal
“Decal Position”
Position” data.
data.

Superior Essex, Feb 2014 62 W2-7 Process Behavior_Inst.ppt

 Exercise 7.5: “Decal Position” Situation
Xbar-R Chart of Position

0.2520
U C L=0.252272
Assumptions:
Sample M ean

0.2505
_
_
X=0.249278
0.2490

0.2475

LC L=0.246283
0.2460
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample
The process output is stable.
U C L=0.01098
0.0100
Sample Range

0.0075
_
0.0050 R=0.00519

0.0025

0.0000 LC L=0
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

Probability Plot of Position

99.9
Normal

Mean 0.2493
P-value = 0.518
StDev 0.002216
99
N 125
AD 0.326

The population distribution is

95 P-Value 0.518
90
80
70
Normal.
Percent

60
50
40
30
20
10
5

0.1
0.242 0.244 0.246 0.248 0.250 0.252 0.254 0.256
Position

Superior Essex, Feb 2014 63 W2-7 Process Behavior_Inst.ppt

 Exercise 7.5: Capability with Minitab® “Decal Position”
Capability Analysis (Normal Distribution)
Select Stat > Quality Tools > Capability Analysis (Normal).
Select “Single column” > Enter “Position”.
Select “Subgroup size” > Enter “5”.
Lower spec = .245; Upper spec = .255.
Select “OK”.

Superior Essex, Feb 2014 64 W2-7 Process Behavior_Inst.ppt

 Exercise 7.5: Capability with Minitab® “Decal Position”

Under what
conditions are the
within and overall
PPM estimates
approximately
equal?

Superior Essex, Feb 2014 65 W2-7 Process Behavior_Inst.ppt

 Exercise 7.5: Capability with Minitab® “Decal Position”

Process Capability of Position

LSL USL
Process Data Within
LSL
Target
0.245
*
Overall Cp = .75
U SL 0.255 Potential (Within) C apability
Sample Mean 0.249278 Cp 0.75
Sample N
StDev(Within)
125
0.00223202
C PL 0.64
C P U 0.85 Cpk = .64
StDev(O v erall) 0.00221614 Cpk 0.64
O v erall C apability
apability
Pp 0.75
P PL 0.64
PPU 0.86
Ppk 0.64

Under what
Cpm *

0.244 0.246 0.248 0.250 0.252 0.254

conditions are the
Observ ed Performance
P P M < LSL 40000.00
Exp. Within Performance
PP M < LSL 27652.80
Exp. O v erall Performance
P P M < LSL 26790.7
26790.75
within and overall
PPM estimates
P P M > U SL 0.00 PP M > U SL 5177.07 P P M > U SL 4909.35
P P M Total 40000.00 PPM Total 32829.86 PPM Total 31700.10

approximately
“Within” PPM = “Overall” PPM =
32,830 31,700 equal?

When statistical control is present.

The R-bar/d2 used for the “within” calculation of sigma and the s used for the
“overall” calculation will be approximately equal. The resulting within and overall
bell curves will be almost identical thus producing similar PPM estimates.
No method of estimating capability is reliable without statistical control.
Superior Essex, Feb 2014 66 W2-7 Process Behavior_Inst.ppt
 Exercise 7.5: Process Capability
Using Minitab®

Allotted Time
7 MINUTES

Part Two: You Try It

ASSIGNMENT:
ASSIGNMENT:
1.
1. Open
Open Minitab
Minitab®® worksheet:
worksheet:
“7.5
“7.5 Process
Process Capability.MTW”
Capability.MTW”
2.
2. Use
Use Minitab
Minitab®® to
to conduct
conduct aa Process
Process
Capability
Capability Study
Study using
using the
the “Calls”
“Calls” data.
data.
3.
3. What
What is
is the
the capability
capability of
of the
the process?
process?

Superior Essex, Feb 2014 67 W2-7 Process Behavior_Inst.ppt

 Exercise 7.5: “Calls Data” Situation
Xbar-R Chart of Calls

154
U C L=154.716 Assumptions:
Sample M ean

152 _
_
X=151.150

150

148

The process output is stable.

LC L=147.584
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

15
U C L=13.07
Sample Range

10

_
R=6.18
5

0 LC L=0
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

Probability Plot of Calls

99.9
Normal
P-value = 0.215
Mean 151.2
StDev 2.740
99

95
N
AD
P-Value
125
0.492
0.215
The distribution is Normal.
90
80
70
Percent

60
50
40
30
20
10
5

0.1
145 150 155 160
Calls

Superior Essex, Feb 2014 68 W2-7 Process Behavior_Inst.ppt

 Cautions: Using Cpk Vs. Ppk

1. Using the wrong capability index can lead to

estimates that grossly misrepresent reality!
2. Capability metrics are estimates that depend
on sample size.

Caution!

Superior Essex, Feb 2014 69 W2-7 Process Behavior_Inst.ppt

 Caution #1: Using Wrong Index is Misleading
 200 Observations were sampled from a Normal distribution with a m = 50 and a s = 1.
 Forty sub-groups of size 5 are used to compute Cpk while all 200 observations are
pooled together to compute Ppk.
 For this example, USL=55 and LSL=45.

Process Capability Sixpack of C1, ..., C5 Cpk and Ppk

Xbar C har t C apability H istogr am are
51
UCL=51.372 approximately
Sample Mean

50
_
_
X=50.098
equal.
49
LCL=48.825
1 5 9 13 17 21 25 29 33 37 48.00 48.75 49.50 50.25 51.00 51.75 52.50

R C har t Nor mal P r ob P lot

UCL=4.668
A D: 0.297, P : 0.588
4
Cpk=1.72
Sample Range

_
2 R=2.208

0 LCL=0
1 5 9 13 17 21 25 29 33 37 48 50 52 54

Last 4 0 Subgr oups C apability P lot

52
Within Within O v erall Ppk=1.76
S tDev 0.949181 S tDev 0.928056
Values

Cp 1.76 Pp 1.8
50 O v erall
C pk 1.72 P pk 1.76
48
C C pk 1.76 C pm *
S pecs
10 20 30 40
Sample

Superior Essex, Feb 2014 70 W2-7 Process Behavior_Inst.ppt

 Caution #1: Using Wrong Index is Misleading
 200 Observations were sampled from a distribution that was not
stable.
Process Capability Sixpack of C1, ..., C5
Xbar C har t C apability H istogr am Cpk and Ppk 1
1 1 1 1

are NOT equal.

52 1 1 1 1 1 1 1 1
UCL=51.504
Sample Mean

2 _
_
50 X=49.971

LCL=48.438
48 1
1 1 1
1
1 1
1
1 5 9 13 17 21 25 29 33 37 45.0 46.5 48.0 49.5 51.0 52.5 54.0

R C har t Nor mal P r ob P lot

UCL=5.619
A D: 0.990, P : 0.013

Cpk=1.45
Sample Range

4
_
R=2.657
2

0 LCL=0
1 5 9 13 17 21 25 29 33 37 45 50 55

Last 4 0 Subgr oups C apability P lot

55 Within Within O v erall
Ppk=0.85
S tDev 1.14246 S tDev 1.94792
Values

50 Cp 1.46 Pp 0.86
O v erall
C pk 1.45 P pk 0.85
C C pk 1.46 C pm *
45
S pecs
10 20 30 40
Sample

Cpk (and Cp) should not be used for non-stable data

because they ignore the “between subgroup” variation.
Superior Essex, Feb 2014 71 W2-7 Process Behavior_Inst.ppt
 Caution #2: Capability Metrics are
Estimates that Depend on Sample Size.

How reliable is a 30 piece sample?

Superior Essex, Feb 2014 72 W2-7 Process Behavior_Inst.ppt

 How reliable is a 30 piece Capability Study?
 5000 different 30 piece capability studies have been generated from an infinite
population of data. The data was generated in Excel.
 The population is known to have a m = 30 and a s = 1.
 The USL=35 and the LSL=25. As such, the “true” Cp and “true” Cpk are 1.67.
Replication x-bar s Pp Ppk
1 29.99 1.06 1.58 1.57
2 29.83 0.75 2.24 2.16
3 29.87 0.80 2.07 2.02
4 30.21 1.08 1.55 1.48
12 29.98 0.89 1.87 1.86
13 29.87 1.00 1.67 1.62
14 30.27 1.22 1.37 1.30
15 29.88 1.06 1.58 1.54
. . . . .
. . . . .
4997 30.34 0.96 1.73 1.61
4998 30.18 1.11 1.50 1.45
4999 29.94 0.99 1.68 1.66 Minimum
5000 30.11 0.89 1.88 1.84 Ppk= 1.06

Average 30.00 0.99 1.72 1.67

Minimum 29.32 0.53 1.08 1.06 Maximum
Maximum 30.59 1.54 3.12 2.99 Ppk= 2.99
Std. Deviation 0.18 0.13 0.24 0.24

Superior Essex, Feb 2014 73 W2-7 Process Behavior_Inst.ppt

 Estimating Capability, n = 30

Process Capability of n=3 0

( usi ng 95. 0 % confi de nce)

Process D ata
LSL USL
O verall C apability
Ppk=1.66
LSL 25 Pp 1.66
T arget * Lo w er  C L 1.24
U SL 35 U pper  C L 2.09
S a mple M ean 30 P PL 1.66
S a mple  N 30 PPU 1.66
St D e v ( O v erall) 1.00245 Ppk 1.66
Lower  CL 1.22
Upper  CL 2.11
C pm *
Lo w er  C L *

Ppk could
25 .5 27.0 2 8.5 3 0 .0 31.5 3 3 .0 3 4 .5 be as small
O bser v ed P erform ance
P P M  < LSL 0.00
Exp. O v erall P erform ance
P P M  < L SL 0.31
as 1.22
P P M  >  U SL 0.00
P P M  T otal 0.00
P P M  >  U SL 0.31
P P M  T otal 0.61
or as large
as 2.11

Minitab gives a Confidence Interval around the estimate.

Superior Essex, Feb 2014 74 W2-7 Process Behavior_Inst.ppt

 Estimating Capability, n = 100

Process Capability of n=100

(using 95.0% confidence)

P rocess D ata
LSL USL
O v erall C apability
Ppk=1.66
LS L 25 Pp 1.66
Target * Low er C L 1.43
U SL 35 U pper C L 1.89
S ample M ean 30 PPL 1.66
S ample N 100 PPU 1.66
S tD ev (O v erall) 1.00252 P pk 1.66
Low er C L 1.42
U pper C L 1.90
C pm *
Low er C L *

Ppk could
25.5 27.0 28.5 30.0 31.5 33.0 34.5 be as small
O bserv ed P erformance E xp. O v erall P erformance as 1.42
P P M  < LS L 0.00
P P M  > U S L 0.00
P P M  < LS L 0.31
P P M  > U S L 0.31 or as large
P P M  Total 0.00 P P M  Total 0.61
as 1.90

Recall: Confidence Interval when n=30: from 1.22 to 2.11

What happens to the Confidence Interval when more data is available?

Superior Essex, Feb 2014 75 W2-7 Process Behavior_Inst.ppt

 Process Capability: One-sided Specifications

1. Assure that the data is normally distributed. (Process must be

stable).
2. Run Minitab® Capability Analysis.

LSL Only USL Only

Cpk =
z lower Cpk =
z upper
3 3

Superior Essex, Feb 2014 76 W2-7 Process Behavior_Inst.ppt

 One-sided Specifications: Weld Strength Example
Xbar-R Chart of Strength

110
U C L=111.66  Weld Strength is checked on four
units twice a shift.
Sample M ean

105

 Data on 25 Subgroups is available.

_
_
100 X=100.32

95

90
LC L=88.99
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

40

30
U C L=35.49 1. Assure data is normally
distributed:
Sample Range

20
_
R=15.56
10

0 LC L=0
Both charts show
1 3 5 7 9 11 13
Sample
15 17 19 21 23 25
statistical control. The
process is stable.
Probability Plot of Strength
Normal  Normality is accepted with
a p-value = .538.
99.9
Mean 100.3
StDev 7.011
99 N 100
AD 0.315
95 P-Value 0.538
90
80
70
Percent

60
50
40
30
20
10
5

0.1
80 90 100 110 120
Strength

Superior Essex, Feb 2014 77 W2-7 Process Behavior_Inst.ppt

 One-sided Specifications: Weld Strength Example
2. Run Minitab® Capability Analysis.
Process Capability of Strength

LSL
P rocess Data Within
LS L 75 Ov erall
Target *
USL * P otential (Within) C apability
S ample M ean 100.324 Cp *
S ample N 100 C P L 1.12
S tDev (Within) 7.55473 C PU *
S tDev (O v erall) 7.01057 C pk 1.12
O v erall C apability
Pp *
Cpk = 1.12
PPL 1.20
PPU *
P pk 1.20
C pm *

75.0 82.5 90.0 97.5 105.0 112.5

O bserv ed P erformance E xp. Within P erformance E xp. O v erall P erformance
P PM < LS L 0.00 P P M < LS L 401.05 P P M < LS L 151.77
P PM > U S L * PPM > USL * PPM > USL *
P PM Total 0.00 P P M Total 401.05 P P M Total 151.77

PPM = 401

Note: for one sided specifications the Cp index is not sensible.

Superior Essex, Feb 2014 78 W2-7 Process Behavior_Inst.ppt
 Process Capability Analysis: Non-normal Data
Follow these steps to determine the reason the data is non-normal:

1. If the process is not stable, work to stabilize it

before proceeding with capability analysis.
2. If the process is stable and the data is non-
normal, check to see if there are multiple
RIGHT SKEWED
(POSITIVELY SKEWED)
sources pooled together (e.g., different lines,
multiple cavities, multiple tools, etc.). If so,
perform separate analysis on each data source.
3. If the process is stable, data is non-normal and
LEFT SKEWED the data come from only one population
(NEGATIVELY SKEWED)
source, attempt to understand the “physics”
behind the shape of the distribution.

If it makes sense that the data is non-

normal, then find a function that fits
the data, and use that function to
estimate the capability of the process.

Superior Essex, Feb 2014 79 W2-7 Process Behavior_Inst.ppt

 Invalid Capability Analysis: Non-Normal Data
The Normal Distribution was used, when not appropriate: Normal
distribution
Process Capability of Normal
does not fit the
LSL data well.
P rocess Data Within
LS L 27 Ov erall
Target *
USL * P otential (Within) C apability
S ample M ean 32.8993 Cp *
C P L 0.92
S ample N 50
S tDev (Within) 2.13429 C PU * Reported Cpk
C pk 0.92
S tDev (O v erall) 2.19118
O v erall C apability
= 0.92 is
Pp
PPL
*
0.90
WRONG!
PPU *
P pk 0.90
C pm *

This analysis
28 30 32 34 36 38 40
is NOT sound!
O bserv ed P erformance E xp. Within P erformance Exp. O v erall P erformance
The curve
P PM < LS L 0.00
P PM > U S L *
P P M < LSL
P P M > U SL
2854.30
*
P P M < LS L 3547.94
PPM > USL *
does NOT FIT
P PM Total 0.00 P P M Total 2854.30 P P M Total 3547.94 the data!

Since the above curve does not fit the data, this capability
analysis is NOT VALID.
Superior Essex, Feb 2014 80 W2-7 Process Behavior_Inst.ppt
 Valid Capability Analysis: Non-Normal Data

 Finding a function that fits the data: Weibull

distribution is a
Process Capability of Non-Normal function that fits
Calculations Based on Weibull Distribution Model
the data well.
LSL
P rocess D ata O v erall C apability
LS L 27 Pp *
Target * PPL 2.55
USL *
S ample M ean 32.8993
PPU
P pk
*
2.55
Reported Ppk
S ample N
S hape
50
1.24112
E xp. O v erall P erformance = 2.55 is valid
P P M < LS L 0.00
S cale 2.84197
Threshold 30.2471 P P M > U SL *
P P M Total 0.00
O bserv ed P erformance
P P M < LSL 0.00
PPM > USL *
P P M Total 0.00

This analysis
is sound!
The curve fits
28 30 32 34 36 38 40
the data!

Since the above curve fits the data, this capability analysis is valid,
providing:
 The process is stable, and
 The
Superior Essex, data comes from a single population
Feb 2014 81
source. W2-7 Process Behavior_Inst.ppt
 Workbook: Exercise 7.6

Fitting a Curve to the Data

Superior Essex, Feb 2014 82 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Fitting a Curve to the Data

Part One:
With Instructor

ASSIGNMENT:
ASSIGNMENT:
 Open
Open the
the Minitab
Minitab®® worksheet:
worksheet:
“7.6
“7.6 Non-Normal
Non-Normal Data.MTW”
Data.MTW”
 Work
Work along
along with
with the
the Instructor
Instructor to
to analyze
analyze the
the
“Diameter”
“Diameter” data:
data:
1.
1. Check
Check for
for normality.
normality.
2.
2. Find
Find aa distribution
distribution that
that best
best fits
fits the
the data.
data.
3.
3. Run
Run aa capability
capability analysis.
analysis.
LSL
LSL == 49.7mm,
49.7mm, USLUSL == 50.3mm.
50.3mm.

Superior Essex, Feb 2014 83 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Fitting a Curve to the Data: “Diameter”

Histogram of Diameter
35

30

25

Histogram shows a non-

Frequency

20

15 normal pattern.
10

0
49.65 49.73 49.81 49.89 49.97 50.05 50.13 50.21
Diameter

Probability Plot of Diameter

Normal
99.9
Mean 49.98

99
StDev
N
AD
0.1106
100
0.909
Normality is rejected:
95
90
P-Value 0.020
p-value = .020
80
70
Percent

60
50
40
30
20
10
5

0.1
49.6 49.7 49.8 49.9 50.0 50.1 50.2 50.3 50.4
Diameter

Superior Essex, Feb 2014 84 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Fitting a Curve to the Data: “Diameter”

Select Stat > Quality Tools > Individual Distribution Identification.

Select “Single column” > Enter “Diameter”.
Check “Use all distributions…”
Select “OK”.

Superior Essex, Feb 2014 85 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Curve Fitting: Session Output
 Each of the distribution types is fit to the data. A statistical test
(Anderson-Darling) is run to see if the distribution is an acceptable fit.

Goodness of Fit Test

Distribution: AD P LRT P
Normal 0.909 0.020
Lognormal 0.919 0.019 The 3-Parameter
3-Parameter Lognormal 0.918 * 0.732 Weibull
Exponential 45.672 <0.003 distribution gives
the best fit to the
2-Parameter Exponential 18.005 <0.010 0.000 data,
Weibull 0.288 >0.250 p-value > .500
3-Parameter Weibull 0.282 >0.500 0.340
Smallest Extreme Value 0.290 >0.250
Largest Extreme Value 3.013 <0.010
Gamma 0.925 0.020
3-Parameter Gamma 1.170 * 1.000
Logistic 0.846 0.016
Loglogistic 0.852 0.016
3-Parameter Loglogistic 0.846 * 0.778

Superior Essex, Feb 2014 86 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Curve Fitting: Graphical Output

Probability Plot for Diameter

G oodness of F it Test
2-P arameter Exponential - 95% C I Weibull - 95% C I
99.9 99.9 2-P arameter E xponential
90 90 A D = 18.005
P er cent 50 50 P -V alue < 0.010

P er cent
10 10 Weibull
A D = 0.288
P -V alue > 0.250
1 1
3-P arameter Weibull
0.1 0.1 A D = 0.282
0.0001 0.0010 0.0100 0.1000 1.0000 10.0000 49.2 49.5 49.8 50.1 P -V alue > 0.500
Diameter - T hr eshold Diameter
S mallest E xtreme V alue
3-P arameter Weibull - 95% C I S mallest Extreme V alue - 95% C I A D = 0.290
99.9 99.9 P -V alue > 0.250

90 90
50 50
P er cent

P er cent
10 10

3-Parameter 1 1

Weibull test
0.1 0.1
results 0.50 0.75 1.00 1.25 49.2 49.5 49.8 50.1
Diameter - T hr eshold Diameter

Superior Essex, Feb 2014 87 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Fitting a Curve to the Data: “Diameter”
Capability Analysis (Non-normal Distribution):
Select Stat > Quality Tools > Capability Analysis Non-normal Distribution
Select “Single column” > Enter “Diameter”.
Check “Distribution” > Enter “3-parameter Weilbull
Enter LSL: 49.7
Enter USL: 50.3
Select “OK”.

Note: If a
specification
limit is
actually a
boundary,
check the
appropriate
box.

Superior Essex, Feb 2014 88 W2-7 Process Behavior_Inst.ppt

 Exercise 7.6: Capability Analysis: Minitab® Output
Process Capability of Diameter
Calculations Based on Weibull Distribution Model

LSL USL

LS L
P rocess Data
49.7
O v erall C apability
Pp 0.89
Ppk = .65
Target
USL
*
50.3
PPL
PPU
0.65
1.37
(Minitab does not
S ample M ean 49.9768 P pk 0.65 compute Cpk
S ample N 100
S hape 12.5091
E xp. O v erall P erformance
P P M < LS L 17381.38
with non-normal
S cale
Threshold
1.17671
48.8483
PPM > USL 0.98 data)
P P M Total 17382.36
O bserv ed P erformance
P P M < LS L 10000.00
PPM > USL 0.00
P P M Total 10000.00

PPM = 17,382

49.7 49.8 49.9 50.0 50.1 50.2 50.3

Since the above curve fits the data, this capability analysis is valid if:
 The process is stable, and
 The data comes from a single population source.

Is the process capable?

Superior Essex, Feb 2014 89 W2-7 Process Behavior_Inst.ppt
 Exercise 7.6: Fitting a Curve to the Data

Allotted Time
15 MINUTES

Part Two: You Try It

ASSIGNMENT:
ASSIGNMENT:
 Open
Open the
the Minitab
Minitab®® worksheet:
worksheet:
“7.6
“7.6 Non-Normal
Non-Normal Data.MTW”
Data.MTW”
 Analyze
Analyze the
the “Flatness”
“Flatness” data:
data:
1.
1. Check
Check for
for stability
stability (n=5)
(n=5)
2.
2. Check
Check for
for normality.
normality.
3.
3. Find
Find aa distribution
distribution thatthat best
best fits
fits the
the data.
data.
4.
4. Run
Run aa capability
capability analysis.
analysis.
USL
USL == 10
10
Zero
Zero is
is aa lower
lower boundary.
boundary.

Superior Essex, Feb 2014 90 W2-7 Process Behavior_Inst.ppt

 PART THREE
Case Studies: Reviewing
Practical Application of Process
Capability Studies

Superior Essex, Feb 2014 91 W2-7 Process Behavior_Inst.ppt

 Case Study #1: Welding Process – Measure Phase

YP= Nugget Diameter (mm)

20 subgroups of n=5 welds, taken over 5 days

X-Bar and R charts - Nugget Diameter (mm)

5.5
U C L=5.353

Special
Sample M ean

5.0 _
_
X=4.884
Cause
4.5
LC L=4.416

4.0 1
Before the team
1
2 4 6 8 10 12 14 16 18 20 can estimate
Sample
process
1.6
U C L=1.717
capability, they
must find and
Sample Range

1.2
_
R=0.812
counteract the
0.8
“special cause”.
0.4

0.0 LC L=0
2 4 6 8 10 12 14 16 18 20
Sample

Superior Essex, Feb 2014 92 W2-7 Process Behavior_Inst.ppt

 Case Study #1: Welding Process – Measure Phase

YP= Nugget Diameter (mm)

20 NEW subgroups of n=5 welds, taken over 5 days

X-Bar and R charts - Nugget Diameter

 Process is
5.50
now stable.
U C L=5.389  Normality
5.25
test passed.
Sample M ean

_
5.00 _
X=4.955  Capability
4.75 can be
4.50 LC L=4.520
estimated.
2 4 6 8 10 12 14 16 18 20
Sample

1.6 U C L=1.593

1.2
Sample Range

_
0.8 R=0.753

0.4

0.0 LC L=0
2 4 6 8 10 12 14 16 18 20
Sample

Superior Essex, Feb 2014 93 W2-7 Process Behavior_Inst.ppt

 Case Study #1: Welding Process – Measure Phase
Capability Study determines process not capable:

Process Capability of Diam-Ctr (using pooled Std. Dev.)

LSL
Process Data Within
LSL
Target
4.00000
*
Overall
Not Capable
Potential (Within) Capability
USL
Sample Mean
*
4.95491 Cp * (Cpk=.99).
Sample N 100 C PL 0.99
StDev (Within) 0.32282 C PU *
StDev (O verall) 0.31461
Cpk 0.99
C C pk 0.99
O v erall C apability
Pp *
PPL 1.01
PPU *
Ppk 1.01
C pm *

4.2 4.5 4.8 5.1 5.4 5.7

One-sided
O bserved Performance Exp. Within Performance Exp. O verall Performance
One-sided
specification
PPM < LSL 0.00 PPM < LSL 1547.99 PPM < LSL 1201.79 specification
(LSL = 4 mm)
PPM > USL * PPM > U SL * PPM > USL *
PPM Total 0.00 PPM Total 1547.99 PPM Total 1201.79 (LSL = 4 mm)

The team must determine: How to improve Process Capability?

Reduce variation? Shift average upward?
Superior Essex, Feb 2014 94 W2-7 Process Behavior_Inst.ppt
 Case Study #1: Welding Process – Analyze and Improve
Using DOE, critical factors (Xs) that affect variation were identified, and
the variation in weld nugget diameter was reduced.

Process Capability of Diam-Ctr ( using pooled Std. Dev.)

LSL
Process Data Within
LSL 4.00000 Ov erall
Target *
USL * Potential (Within) Capability
Sample Mean 4.95491 Cp *
Sample N 100 CPL 0.99
StDev (Within) 0.32282 CPU *
StDev (O v erall) 0.31461 Cpk 0.99
CCpk 0.99
O v erall Capability
Pp *
PPL 1.01
PPU *
New Ppk 1.01
Cpm *
Cpk = 2
Process is
4.2 4.5 4.8 5.1 5.4 5.7
O bserv ed Perform ance Exp. Within Performance Exp. O v erall Performance
now Capable
PPM < LSL 0.00
PPM > USL *
PPM < LSL
PPM > USL
1547.99
*
PPM < LSL 1201.79
PPM > USL *
(Cpk=2)
PPM Total 0.00 PPM Total 1547.99 PPM Total 1201.79

 For characteristics with 1-sided spec, we then shift average to most

efficient location that assures high process capability. (For
characteristics with 2-sided specs, we shift average to target).
Superior Essex, Feb 2014 95 W2-7 Process Behavior_Inst.ppt
 Case Study #1: Welding Process – Control Phase
The team determines “Flags” for Preventive Maintenance:
Xbar-R Chart of Diameter-Long Term
1
6.0
1

1
Sample M ean

5.5
U C L=5.439

_
Change
_
5.0 X=5.01 Weld Tip
LC L=4.581
4.5
3 6 9 12 15
Sample
18 21 24 27 30
 This point at which the tip
1.6 U C L=1.574 must be changed was
1.2 determined to be 6 mm.
Sample Range

_
0.8 R=0.744

0.4  Although the print called out

0.0 LC L=0 only a lower specification
3 6 9 12 15
Sample
18 21 24 27 30
limit of 4 mm, an upper
specification limit of 6 mm
should also be used.
Understanding the Physics:
 Over time, the weld tip mushrooms, causing the nugget diameter
to increase, and current density to decrease.
 A “stepper” is used to automatically step up the current over the
life of the tip. But a point is reached at which the current density is
not adequate; the weld weakens, and the tip must be changed.

Superior Essex, Feb 2014 96 W2-7 Process Behavior_Inst.ppt

 Case Study #1: Welding Process – Ongoing Control
Xbar-R Chart of Diameter-Long Term
1
6.0
1
1

Sample M ean
5.5
U C L=5.439

_
_
5.0 X=5.01

USL = 6 mm
LC L=4.581

LSL = 4 mm
4.5
3 6 9 12 15 18 21 24 27 30

Change
Sample

1.6 U C L=1.574

1.2
Weld Tip

Sample Range
_
0.8 R=0.744

0.4

0.0 LC L=0
3 6 9 12 15 18 21 24 27 30
Sample

To maintain Cpk = 1.33,

change weld tip when
average nugget diameter is
4 standard deviations
away from USL.
5.7
Important:
• The reduced variation enables less frequent weld tip changes!
• In addition, if we just moved the distribution to the right, then
we’re just cranking it up … wasting energy!
Superior Essex, Feb 2014 97 W2-7 Process Behavior_Inst.ppt
 Case Study #1: Welding Process – Ongoing Control
Xbar-R Chart of Diameter-Long Term
1
6.0
1
1

Sample M ean
5.5
U C L=5.439

_
_
5.0 X=5.01

USL = 6 mm
LC L=4.581

LSL = 4 mm
4.5
3 6 9 12 15 18 21 24 27 30

Change
Sample

1.6 U C L=1.574

1.2
Weld Tip

Sample Range
_
0.8 R=0.744

0.4

0.0 LC L=0
3 6 9 12 15 18 21 24 27 30
Sample

NEXT STEPS
 The team should determine what
level of “electric current”
corresponds to the nugget diameter
of 5.7 mm so that the current can
serve as the flag, rather than
5.7 nugget diameter.
 Nugget diameter is a destructive
test, and cannot provide immediate
feedback.

Superior Essex, Feb 2014 98 W2-7 Process Behavior_Inst.ppt

 Workbook: Exercise 7.7

Process Capability Case Study #2:

Six Cavity Mold – You Decide

Superior Essex, Feb 2014 99 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Case Study #2: Six Cavity Mold - You
Decide

A team is evaluating capability:

Allotted Time Molding Process with Six
10 MINUTES Cavities: YP = Force
(Assume process is in statistical
control.)

ASSIGNMENT:
ASSIGNMENT:
1.
1. Review
Review the
the three
three capability
capability studies
studies that
that follow
follow (three
(three alternative
alternative
approaches
approaches to to the
the same
same situation):
situation):
 FIRST
FIRST APPROACH:
APPROACH: DataData from
from all
all six
six cavities
cavities inin one
one graph,
graph,
using
using Normal
Normal Distribution
Distribution
 SECOND
SECOND APPROACH:
APPROACH: DataData from
from all
all six
six cavities
cavities inin one
one graph,
graph,
using
using Largest
Largest Extreme
Extreme Value
Value Distribution..
Distribution..
 THIRD
THIRD APPROACH:
APPROACH: Separate
Separate graphs
graphs (studies)
(studies) onon each
each cavity.
cavity.
2.
2. Answer
Answer inin your
your workbook:
workbook: Which
Which approach
approach is is the
the proper
proper way
way to
to
conduct
conduct the
the capability
capability study?
study? Why?
Why?

Superior Essex, Feb 2014 100 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide

 FIRST APPROACH: Data from all six cavities. Normal distribution was used.
 Is this analysis a proper way to estimate Ppk and PPM? Why or why not?

Process Capability of Force

LSL USL

LS L
P rocess D ata
68.6
O v erall C apability
Pp 2.10
Ppk = 1.79
Target * PPL 2.41
USL 107.8 P P U 1.79
S ample M ean 91.125 P pk 1.79
S ample N 180 C pm *
S tD ev (O v erall) 3.11279

LSL= 68.6
USL=107.8

72 78 84 90 96 102 108
O bserv ed P erformance E xp. O v erall P erformance Anderson-Darling test:
P P M < LS L 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.04 P-value < 0.005.
P P M Total 0.00 P P M Total 0.04

PPM = 0.04

Superior Essex, Feb 2014 101 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide

 SECOND APPROACH: Data from all six cavities, using Largest

Extreme Value Distribution.
 The Largest Extreme Value Distribution gives a good fit to the data
(P-value =0.386).
Process Capability of Force
Calculations Based on Largest Extreme Value Distribution Model

LSL USL
P rocess D ata O v erall C apability
LS L 68.6 Pp 1.80 Ppk = 1.07
Target * PPL 3.80
USL 107.8 PPU 1.07
S ample M ean 91.125 P pk 1.07
S ample N 180
E xp. O v erall P erformance
Location 89.6749
P P M < LS L 0.00
S cale 2.56905
P P M > U S L 862.54
O bserv ed P erformance P P M Total 862.54
P P M < LS L 0.00
P P M > U S L 0.00
P P M Total 0.00 PPM = 863

72 78 84 90 96 102 108

Superior Essex, Feb 2014 102 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.

Histogram ( with Normal Curve) of Force by Cavity

0
. 0 .5 .0 .5 .0 .5 0.
85 87 90 92 95 97 10
Cavity 1 C avity 2 Cqvity 3
20

15

10

5
Frequency

0
Cavity 4 C avity 5 Cavity 6
20

15

10

0
.0 .5 .0 .5 .0 .5 .0 . 0 .5 .0 .5 .0 .5 .0
85 87 90 92 95 97 100 85 87 90 92 95 97 100

Normality Test was passed for each cavity.

Superior Essex, Feb 2014 103 W2-7 Process Behavior_Inst.ppt
 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.
Process Capability of Force-1
Process Capability for Force, Cavity # 1
LSL USL
P rocess D ata O v erall C apability
LS L 68.6 Pp 3.98
Target * P P L 4.76
USL 107.8 P P U 3.20
S ample M ean 92.0541 P pk 3.20
S ample N 30 C pm *
S tD ev (O v erall) 1.64136

72 78 84 90 96 102 108
O bserv ed P erformance E xp. O v erall P erformance
P P M < LS L 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.00
P P M Total 0.00 P P M Total 0.00

Superior Essex, Feb 2014 104 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.

Process
Process Capability
Capability of Force-2
for Force, Cavity # 2

LSL USL
P rocess D ata O v erall C apability
LS L 68.6 Pp 3.91
Target * P P L 4.16
USL 107.8 P P U 3.67
S ample M ean 89.4422 P pk 3.67
S ample N 30 C pm *
S tD ev (O v erall) 1.66951

72 78 84 90 96 102 108
O bserv ed P erformance E xp. O v erall P erformance
P P M < LS L 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.00
P P M Total 0.00 P P M Total 0.00

Superior Essex, Feb 2014 105 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.
Process Capability of Force-3
Process Capability for Force, Cavity # 3
LSL USL
P rocess Data O v erall C apability
LS L 68.6 Pp 3.26
Target * P P L 3.46
USL 107.8 P P U 3.06
S ample M ean 89.4237 P pk 3.06
S ample N 30 C pm *
S tDev (O v erall) 2.00433

72 78 84 90 96 102 108
O bserv ed P erformance Exp. O v erall P erformance
P P M < LS L 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.00
P P M Total 0.00 P P M Total 0.00

Superior Essex, Feb 2014 106 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.

Process
Process Capability
Capability forof Force-4
Force, Cavity # 4
LSL USL
P rocess Data O v erall C apability
LS L 68.6 Pp 3.48
Target * P P L 3.83
USL 107.8 P P U 3.13
S ample M ean 90.1803 P pk 3.13
S ample N 30 C pm *
S tDev (O v erall) 1.87734

72 78 84 90 96 102 108
O bserv ed P erformance Exp. O v erall P erformance
P P M < LS L 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.00
P P M Total 0.00 P P M Total 0.00

Superior Essex, Feb 2014 107 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.
Process
Process Capability
Capability for of Force-5
Force, Cavity # 5
LSL USL
P rocess Data O v erall C apability
LS L 68.6 Pp 3.21
Target * P P L 4.52
USL 107.8 P P U 1.91
S ample M ean 96.1841 P pk 1.91
S ample N 30 C pm *
S tDev (O v erall) 2.03242

72 78 84 90 96 102 108
O bserv ed P erformance Exp. O v erall P erformance
P P M < LS L 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.01
P P M Total 0.00 P P M Total 0.01

Superior Essex, Feb 2014 108 W2-7 Process Behavior_Inst.ppt

 Exercise 7.7: Six Cavity Mold: You Decide
 THIRD APPROACH: Evaluating each cavity separately.

Process
Process Capability
Capability forofForce,
Force-6Cavity # 6

LSL USL
P rocess Data O v erall C apability
LS L 68.6 Pp 2.76
Target * PPL 2.94
USL 107.8 P P U 2.59
S ample M ean 89.4657 P pk 2.59
S ample N 30 C pm *
S tDev (O v erall) 2.36402

72 78 84 90 96 102 108
O bserv ed P erformance Exp. O v erall P erformance
P P M < LSL 0.00 P P M < LS L 0.00
P P M > U S L 0.00 P P M > U S L 0.00
P P M Total 0.00 P P M Total 0.00

Superior Essex, Feb 2014 109 W2-7 Process Behavior_Inst.ppt

 PART FOUR

Evaluating Process Capability

using Attribute Data

Superior Essex, Feb 2014 110 W2-7 Process Behavior_Inst.ppt

 Process Capability with Attribute Data

The centerline of the stable p-chart shows the process capability:

NUMBER OF INVOICES WITH ERRORS.

P Chart
0.06
UCL=0.05590
0.05
P r o p o rtio n

0.04

0.03 P=0.03195
0.02

0.01
LCL=0.007991

0.00
0 10 20

Sample Number

What is the average defective

rate for this process?

Superior Essex, Feb 2014 111 W2-7 Process Behavior_Inst.ppt

 Process Capability with Attribute Data

The centerline of the stable p-chart shows the process capability:

NUMBER OF INVOICES WITH ERRORS.

P Chart
0.06
UCL=0.05590
0.05
P r o p o rtio n

0.04

0.03 P=0.03195
0.02

0.01
LCL=0.007991
This process
0.00
0 10 20
runs at an
Sample Number
average
defective rate
of 3.2%

What is the average defective

rate for this process?

Superior Essex, Feb 2014 112 W2-7 Process Behavior_Inst.ppt

 PART FIVE

What is Process Sigma Level?

Superior Essex, Feb 2014 113 W2-7 Process Behavior_Inst.ppt

 What is Process Sigma Level?

 Like process capability indices, the Process Sigma Level

measures how well a process meets customer specifications.
 The Process Sigma Level is based on short-term studies
(period of time in which nothing was changed …. no material
changes, no personnel changes, no adjustments, etc.)
 The Process Sigma Level provides an estimate of how good
the process could be.

Important:
Johnson Controls Six Sigma teams use If you need to
standard capability metrics instead of compute Sigma
Process Sigma Level. Level, see your
Knowing how to compute Process Sigma MBB.
Level is sometimes necessary for
communication with customers and others.

Superior Essex, Feb 2014 114 W2-7 Process Behavior_Inst.ppt

 Tale of Two Sigmas

USL
LSL
STANDARD
DEVIATION SIGMA LEVEL

s = 0.08 3 s process
As the Standard
3 3 Deviation gets
USL smaller, the Sigma
LSL

Level gets higher.

s = 0.06 4 s process

4 4
USL
LSL

s = 0.04 6 s process

6 6

Superior Essex, Feb 2014 115 W2-7 Process Behavior_Inst.ppt

 Process Sigma Level - Table

PPM Sigma Level

3.4 6 Remember:
233 5 If you need to
compute Sigma
6210 4 Level, see your
MBB.
66807 3
308538 2
691462 1

 Six Sigma “thinking” employs a 1.5 sigma shift.

 For example: if a process exhibits 4 Sigma capability in
the short term, it would probably exhibit 2.5 capability in
the long term.

Superior Essex, Feb 2014 116 W2-7 Process Behavior_Inst.ppt

 Johnson Controls Project Summary Table

Description Project
( wit h u n it o f m e a s u re ) Before Target After % Improve
YC
YB
Avg: Avg: Avg:
YP
S t.De v: S t.De v: S t.De v:

C pk o r P pk: C pk o r P pk: C pk o r P pk:

YC, YB: show in terms of PPM, % or DPU

Add rows to table if additional Ys

This table summarizes the performance metrics that are to

be used throughout your Six Sigma project.

Superior Essex, Feb 2014 117 W2-7 Process Behavior_Inst.ppt

 PART SIX

Set Improvement Targets

Superior Essex, Feb 2014 118 W2-7 Process Behavior_Inst.ppt

 Set Improvement Targets
 After establishing the process baseline, the Six Sigma team sets
improvement targets:
– How much change in the Process Y (reduced variation,
shifting mean) is needed to assure zero defects?
– Note: the goal for Business Ys such as “% Defective” is
always ZERO!

EXAMPLE: CUT HEADLINERS PROCESS (WATERJET)

Baseline for Yp:

Business Y: Downtime
Cut Dimension (mm) Initial State:
Histogram of Dimension B
How much variation
25
Poor
20
Capability reduction in “Dimension”
Cpk= 0.04 is needed to assure no
Frequency

15

10

5 downtime?
0
0.0 0.5 1.0 1.5 2.0 2.5

Given that these specifications truly represent functional

needs of downstream processes and final customer, the
team set a goal to achieve minimum Cpk of 1.33
Superior Essex, Feb 2014 119 W2-7 Process Behavior_Inst.ppt
 Summary: Process Capability Analysis
 Process capability analysis is performed on Process Y’s (not Business Y’s).
 A process must be in statistical control before reliable estimates of
capability can be obtained.
 Process capability is an estimate of how well the process meets
specification limits.
 The Six Sigma team must assure that the specification limits used to do the
analysis truly reflect customer satisfaction.
 To baseline the process, capability analysis is conducted in Measure Phase.
 The same analysis is conducted after counteractions have been
implemented in Improve Phase, to quantify the improvement.
 To assure that the capability estimates reflect reality, the appropriate
method of estimating capability must be used.
 The method used depends on type of data (variable or attribute), and, if
variable, whether the data is normal or non-normal.
 After baselining the process, the team sets improvement targets for the
Process Y and Customer/Business Ys. Remember: ZERO defects is always
the goal.

Superior Essex, Feb 2014 120 W2-7 Process Behavior_Inst.ppt

 PART SEVEN

Example of Measure Phase

Superior Essex, Feb 2014 121 W2-7 Process Behavior_Inst.ppt

 Example of Completed Measure Phase:
CUT STRIPS

 IPO:
Vinyl
INPUT PROCESS

CUT
OUTPUT
Strips
sheets STRIPS cut to
length


 Customer Y: % Rejected at next process (poor fit) = 11%

 Intended Function: To cut strip to specified length

 Process Y: Cut Length (mm)

 Gage R&R: Passed (to measure length)

Process not
stable. What

 Data collection: 17.2
X-barXbar-R
& RChart
Chart - Length
of Length

1
1
could cause
frequent
Over period when
17.1 U C L=17.1083
Sample Mean

_
_

high and low

17.0 X=16.9989

batches of raw 16.9

16.8
1
1
LC L=16.8894

points?
material changed,
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

and operators
0.4 U C L=0.4012
Sample Range

0.3

changed.
_
0.2 R=0.1897

0.1

0.0 LC L=0
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

Superior Essex, Feb 2014 122 W2-7 Process Behavior_Inst.ppt

 Example of Completed Measure Phase:
CUT STRIPS
X-barXbar-R
& RChart
Chart - Length
of Length
17.2
1
1 Special Cause:
Team discovered that
17.1 U C L=17.1083
Sample Mean

_
_
17.0 X=16.9989

16.9
1
LC L=16.8894
operators frequently
16.8
1 3 5 7 9 11 13
Sample
15
1

17 19 21 23 25
adjusted the process.
0.4
1
U C L=0.4012
This over-adjustment
caused frequent high &
Sample Range

0.3

low points (excess

_
0.2 R=0.1897

0.1

0.0
1 3 5 7 9 11 13 15 17 19 21 23 25
LC L=0 variability) to occur.
Sample

Histogram of Length

50
Anderson Darling:
40
P-value < 0.005
LSL = 16.92 USL = 17.08 Data not normal
Frequency

30

20
because process
not stable!
10

0
16.6 16.7 16.8 16.9 17.0 17.1 17.2
Length

Counter-act special cause: Stop adjusting the process!

Superior Essex, Feb 2014 123 W2-7 Process Behavior_Inst.ppt
 Example of Completed Measure Phase:
CUT STRIPS

 New data-collection: Repeat: over period when batches of raw
material changed, and operators changed.
X-bar & R Chart – Length (No adjustments)
Xbar-R Chart of Length-No Adjustments
17.050
U C L=17.0415
17.025
Sample M ean

17.000

16.975
_
_
X=16.9907 Process is now stable. Capability
16.950
LC L=16.9398
analysis can be conducted.
1 3 5 7 9 11 13 15 17 19 21 23 25
Sample

0.20
U C L=0.1864
Process Capability of Length-No Adjustments
0.15
Sample Range

0.10 _
R=0.0881 LSL USL
0.05 Process D ata Within
LS L 16.92 Ov erall
0.00 LC L=0 Target *
1 3 5 7 9 11 13 15 17 19 21 23 25 U SL 17.08 Potential (Within) C apability
Sample
S ample M ean 16.9907 Cp 0.70
S ample N 125 C PL 0.62
S tDev (Within) 0.0378931 C PU 0.79
S tDev (O v erall) 0.0382218 C pk 0.62
O v erall C apability
Anderson Darling: Pp
PPL
0.70
0.62

P-value=0.333 PPU
P pk
0.78
0.62
C pm *

“Normal Process Capability

Analysis” can be used. 16.88 16.92 16.96 17.00 17.04 17.08 17.12
Cpk =
O bserv ed P erformance
P P M < LS L 24000.00
E xp. Within P erformance
P P M < LS L 31119.61
Exp. O v erall Performance
PP M < LS L 32261.48
0.62
P P M > U S L 16000.00 PPM > USL 9191.31 PP M > U SL 9705.60
P P M Total 40000.00 P P M Total 40310.92 PP M T otal 41967.09

 Set Improvement Target: Target for Process Y:

% defective = 4.2

Minimum Cpk=1.33 (with spec limits shown).

This will assure next process

Superior Essex, Feb 2014
always gets strips that fit well
124 W2-7 Process Behavior_Inst.ppt
 Workbook: Exercise 7.8

MEASURE Phase Workshop

Superior Essex, Feb 2014 125 W2-7 Process Behavior_Inst.ppt

 Exercise 7.8: MEASURE Phase Workshop

Assign
timekeeper
to keep team
on track!

ASSIGNMENT:
ASSIGNMENT:
Conduct
Conduct complete
complete Measure
Measure Phase
Phase onon the
the process
process ofof Firing
Firing Catapult:
Catapult:
1.
1. Run
Run the the process
process as
as specified.
specified. Y = Travel Distance
2.
2. Conduct
Conduct Gage Gage R&R.
R&R.
3.
3. Collect
Collect baseline
baseline data,
data, make
make control
control chart
chart and
and initial
initial capability
capability
picture
picture (histogram
(histogram with
with specification
specification limits).
limits).
4.
4. IfIf process
process stable,
stable, estimate
estimate capability.
capability.
5.
5. Use
Use Minitab
Minitab Report
Report Pad
Pad to
to store
store your
your Gage
Gage R&R
R&R analysis,
analysis,
Control
Control Chart
Chart and
and capability
capability picture/analysis.
picture/analysis.

Note: Save this workshop output!

This is the first part of your in-class D-MAIC project.
You will add Analyze Phase & Improve Phase during future training weeks.
Superior Essex, Feb 2014 126 W2-7 Process Behavior_Inst.ppt
 Fire Catapult Process

Process currently runs as follows:

 The catapult sets on the table, and is not fastened down.
 Ball is loaded.
 Rubber band is set to second pull-back position, then
released.
 The distance the ball traveled is measured.
 The pull-back position is not adjusted during the entire data
collection period.

Target distance = 15 feet

USL= 17 feet
LSL= 13 feet

Superior Essex, Feb 2014 127 W2-7 Process Behavior_Inst.ppt