10 Process Capability
10 Process Capability
10 Process Capability
Process Capability
© 2008 ASQ
Slide 2
Learning Objectives
© 2008 ASQ
Slide 3
Basic Principle
Process width
Specification
width
© 2008 ASQ
Slide 5
Process Sigma Yield
1 0.6826895
2 0.9544997
3 0.9973002
4 0.9999367
5 0.9999994
6 1
LSL USL
© 2008 ASQ
Slide 6
Calculating Yield
We know that Z is the number of units of standard deviation on a standard
normal curve which has a mean of zero and a standard deviation of one. We
also know any normal distribution can be converted to the standard normal
using the Z equation
equation.
© 2008 ASQ
Slide 7
LSL-X
LSL X A i using
Again i ttables
bl or software
ft tto look
l k up th
the area
ZLSL = under the curve when Z=-1 we find .1586. This
σ̂
means that 15.86% of the product has a value less
-1-0
ZLSL = = −1 than the lower specification limit.
1
© 2008 ASQ
Slide 8
Calculating Yield Example Cont.
The yield is the product that is less than or equal to the ULS and greater
than or equal to the LSL. Thus the yield is .8413 - .1586 or .6827.
Let’s
L t’ ddo thi
this example
l iin E
Excell
1. Create a Process Sigma column numbered from 1 to 6 starting in
cell A1
2. In cell B2 enter the command = Normsdist(A2)-Normsdist(-A2)
3. Copy that formulae down through A7 and your spread sheet
should look like
Process Sigma Yield
1 0.6826895
2 0.9544997
3 0.9973002
4 0.9999367
5 0.9999994
6 1
© 2008 ASQ
Slide 9
-4 -3 -2 -1 0 1
© 2008 ASQ
Slide 10
Calculating Shifted Process Yield
USL-X Using tables or software to look up the area under
ZUSL = the curve when Z=-.5 we find .3085
σ̂
1015
1-0-1.5
ZUSL = = −.5
1
LSL-X Again using tables or software to look up the area
ZLSL =
σ̂ under the curve when Z=-2.5 we find .0062.
-1-0-1.5
ZLSL = = −2.5 Subtracting the two we obtain .3023
1
© 2008 ASQ
Slide 11
1 ± .06 Inches
© 2008 ASQ
Slide 13
Your Sketch/Calculation
© 2008 ASQ
Slide 15
© 2008 ASQ
Slide 17
USL-LSL
Process Sigma Short Term =
2σ ^ Short Term
USL-LSL
Process Sigma Long Term =
^ Long Term
2σ
However, if these conditions are not met the yield is assumed to be in
the left tail of a standard normal distribution and process sigma is the
number of units of standard deviation (positive or negative) from 50%.
For example consider a process with a short term yield of 95%. Using
t bl or software
tables ft we find
fi d the
th Z value
l off 1 1.64.
64 If the
th yield
i ld is
i llong tterm
the process is assumed to have shifted 1.5 standard deviations thus 1.5
is added to the Z value.
© 2008 ASQ
Slide 18
Process Capability Using Minitab
Consider an injection molding machine with 5 cavities that is producing
plastic bottle caps with a target value of 1 inch and a tolerance of .005.
The data is in Bottle Caps.MTW
© 2008 ASQ
Slide 19
Shift double
Defective Defects click to go to
(Attribute) (Count) P≥.05 Data Normal? P<.05 Control Chart
Stat >Basic Selection
Stat >Normalityy
Yes No
test
No
Stat> Quality Raw
Tools Data
>Process Normal Non-Normal
Capability Yes
Binomial
St at>Quality
Tools>Process
Capability
Poisson Stat>Quality Yes No Is there some sudden shift in the data?
Tools>Process Stratified Do a time series, then brush the top data to see if there is any pattern to the "outliers."
Capability>
Normal
1. Yield can be modeled using e^(-dpu)
2. Process yield can be converted to process Investigate Check for a
sigma using the standard normal distribution. causes and distribution that
Calc>Probability Distributions>Normal> take action. fits
Inverse Stat>Quality
3. To convert from long term to short term add Tools>Individual
1.5 sigma Distribution
Identification
Return
Shift double
click
© 2008 ASQ
Slide 20
Using Minitab
The data is continuous so test for normality
Stat>Basic Statistics>Normality Test
© 2008 ASQ
Slide 21
80
Percent
0.01
0.992 0.994 0.996 0.998 1.000 1.002 1.004 1.006 1.008
Caps
© 2008 ASQ
Slide 22
Using Minitab to Calculate Process
Capability
© 2008 ASQ
Slide 23
Minitab Results
Process Capability of Caps
LSL
Process Data
0.995
Within
Target 1 O
Overall
ll
USL 1.005
Sample Mean 1.00006 Potential (Within) Capability
Sample N 750 Z.Bench 2.26
StDev(Within) 0.00198431 Z.LSL 2.55
StDev(Overall) 0.00198636 Z.USL 2.49
Cpk 0.83
Overall Capability
Z.Bench 2.26
Z.LSL 2.55
Z.USL 2.48
Ppk 0.83
Cpm 0.84
45 60 75 90 05 20 35 50
99 .99 .99 .99 .00 .00 .00 .00
0. 0 0 0 1 1 1 1
Observed Performance Exp. Within Performance Exp. Overall Performance
% < LSL 0.40 % < LSL 0.54 % < LSL 0.54
Let’s examine this in detail
% > USL 0.67 % > USL 0.64 % > USL 0.65
% Total 1.07 % Total 1.18 % Total 1.19
© 2008 ASQ
Slide 24
Process Data
Process Capability of Caps
LSL
Process Data
0.995
Calculated Within
Target 1 directly from Overall
USL 1.005
Sample Mean 1.00006 the data. The Potential (Within) Capability
Sample N 750 Z.Bench 2.26
StDev(Within) 0.00198431
within Z.LSL 2.55
Z.USL 2.49
StDev(Overall) 0.00198636 standard Cpk 0.83
deviation is Overall Capability
© 2008 ASQ
Slide 25
Observed Performance
Process Capability of Caps
45 60 75 90 05 20 35 50
99 .99 .99 .99 .00 .00 .00 .00
0. 0 0 0 1 1 1 1
Observed Performance Exp. Within Performance Exp. Overall Performance
% < LSL 0.40 % < LSL 0.54 % < LSL 0.54
% > USL 0.67 % > USL 0.64 % > USL 0.65
% Total 1.07 % Total 1.18 % Total 1.19
© 2008 ASQ
Slide 26
Expected Within Performance
Process Capability of Caps
The percent of
LSL product thatUSL
Target is
LSL
Process Data
0.995 expected to be Within
Target 1
outside of the Overall
USL 1 005
1.005
Sample Mean 1.00006 upper and lower Potential (Within) Capability
Sample N 750 Z.Bench 2.26
StDev(Within) 0.00198431 specification Z.LSL 2.55
StDev(Overall) 0.00198636 Z.USL 2.49
limits on an short Cpk 0.83
© 2008 ASQ
Slide 27
LSL
Process Data
product that is
0.995
Within
Target 1 Overall
USL expected to be
1 005
1.005
Sample Mean 1.00006 Potential (Within) Capability
outside of the
Sample N 750 Z.Bench 2.26
upper and lower
StDev(Within) 0.00198431 Z.LSL 2.55
StDev(Overall) 0.00198636 Z.USL 2.49
specification Cpk 0.83
Overall Capability
limits on an long Z.Bench 2.26
term basis. This Z.LSL 2.55
Z.USL 2.48
projection is Ppk 0.83
Cpm 0.84
based on the
45 60 75 90 05 20 35 50
overall standard 99 .99 .99 .99 .00 .00 .00 .00
0. 0 0 0 1 1 1 1
deviation and the
Observed Performance Exp. Within Performance Exp. Overall Performance
process
% < LSL 0.40 mean.
% < LSL 0.54 % < LSL 0.54
% > USL 0.67 % > USL 0.64 % > USL 0.65
% Total 1.07 % Total 1.18 % Total 1.19
© 2008 ASQ
Slide 28
Potential Within Capability
Process Capability of Caps
Z USL and Z LSL are
calculated usingLSLthe Target USL
LSL
Process Data
process 0.995
mean, the Within
Target respective 1 specification Overall
USL 1 005
1.005
Samplelimits
Mean and the within
1.00006 Potential (Within) Capability
Sample N 750 Z.Bench 2.26
standard
StDev(Within)
deviation.
0.00198431 Z.LSL 2.55
Z.USL 2.49
Z bench0.00198636
StDev(Overall) is calculated by Cpk 0.83
putting the projected Overall Capability
Z.Bench 2.26
within yield (.982) in the Z.LSL 2.55
left hand tail of a standard Z.USL 2.48
Ppk 0.83
normal then looking up Cpm 0.84
lesser
Observed of ZExp.
Performance or
Within
USL Z
Performance
LSL . Exp. Overall Performance
Note: Minitab uses within to
% < LSL 0.40 % < LSL 0.54 % < LSL 0.54
% > USL 0.67 % > USL 0.64 % > USL 0.65
% Total 1.07 % Total 1.18 % Total 1.19
describe short term variation
and overall to describe long
Worksheet: Bottle Caps.MTW term variation.
© 2008 ASQ
Slide 29
Overall Capability
Process Capability of Caps
Z USL and Z LSL are
calculated using LSL the Target USL
LSL
Process Data
process 0.995 mean, the Within
Target 1
respective specification O
Overall
ll
USL 1.005
limits
Sample Mean and the overall
1.00006 Potential (Within) Capability
Sample N 750 Z.Bench 2.26
standard
StDev(Within) 0.00198431deviation. Z.LSL 2.55
StDev(Overall) 0.00198636 Z.USL 2.49
Z bench is calculated by Cpk 0.83
© 2008 ASQ
Slide 30
Within - Between
Process Capability of Caps
45 60 75 90 05 20 35 50
99 .99 .99 .99 .00 .00 .00 .00
0. 0 0 0 1 1 1 1
Observed Performance Exp. Within Performance Exp. Overall Performance
% < LSL 0.40 % < LSL 0.54 % < LSL 0.54
% > USL 0.67 % > USL 0.64 % > USL 0.65
% Total 1.07 % Total 1.18 % Total 1.19
© 2008 ASQ
Slide 31
© 2008 ASQ
Slide 32
Calculating Cpk and Ppk
CPL and CPU are
Process Capability of Caps
calculated using the
process mean, the LSL Target USL
respective
Process Data specification Within
LSL 0 995
0.995
Targetlimits 1and the within Overall
USL 1.005
standard
Sample Mean 1.00006
deviation. Potential (Within) Capability
Cp
Sample N is calculated
750 by Cp 0.84
StDev(Within) 0.00198431 CPL 0.85
putting
StDev(Overall) the projected
0.00198636 CPU 0.83
Cpk 0.83
within yield (.982) in the Overall Capability
left hand tail of a standard Pp 0.84
PPL 0.85
normal then looking up PPU 0.83
Ppk 0.83
the respective Z score. Cpm 0.84
Then dividing by 45 33.60Cpk
75 is90 05 20 35 50
99 .99 .99 .99 .00 .00 .00 .00
the lesser of CPL 0. 0 or 0CPU.0 1 1 1 1
Observed Performance Exp. Within Performance Exp. Overall Performance
PPM < LSL 4000.00 PPM < LSL 5355.08 PPM < LSL 5395.59
PPM > USL 6666.67 PPM > USL 6432.04 PPM > USL 6478.48
PPM Total 10666.67 PPM Total 11787.12 PPM Total 11874.07
© 2008 ASQ
Slide 33
© 2008 ASQ
Slide 34
Using the Assistant
© 2008 ASQ
Slide 35
Statistical Process
Assistant Result Control chart
Capability Analysis for Caps
indicates the
Diagnostic Report process is stable.
Xbar-R Chart Note: stability and
Confirm that the process is stable.
capability are
1.002
evaluated
an
Mea
1.000
independently.
0.998
0.010
Range
0.005
0.000
1 16 31 46 61 76 91 106 121 136
Normality Plot
Same
The points should be close to the line.
Normality Test result,
(Anderson-Darling)
passes
Results Pass
P-value 0.107 normality
test
© 2008 ASQ
Slide 36
Assistant Result
Capability Analysis for Caps
Summary Report
Process Characterization
Does the process mean differ from 1? Mean 1.0001
0 0.05 0.1 > 0.5 Standard deviation 0.0019864
Yes No
Actual (overall) capability
P = 0.378 Pp 0.84
Ppk 0.83
Z.Bench 2.26
Actual (overall) Capability % Out of spec 1.19
Are the data inside the limits and close to the target?
LSL Target USL
Comments
Conclusions
-- The process mean does not differ significantly from the
target (p > 0.05).
-- The defect rate is 1.19%, which estimates the
percentage of parts from the process that are outside the
spec limits.
© 2008 ASQ
Slide 37
Datafile/Caps Drift.MTW.
© 2008 ASQ
Slide 38
Normality Test
Probability Plot of Caps Drift
Normal
99.99
Mean 1.000
StDev 0.001291
N 720
99 AD 0.326
95 P-Value 0.520
80
Percent
50
0.01
0.9950 0.9975 1.0000 1.0025 1.0050
Caps Drift
© 2008 ASQ
Slide 39
Capability Analysis
© 2008 ASQ
Slide 40
Capability Analysis
Process Capability of Caps Drift
overLSLtimeProcess
asData
indicated by
0.995
Within
the difference
Target 1 between the Overall
USL 1.005
Within
Sampleand
Mean Between
0.999998 lines. Potential (Within) Capability
Sample N 720 Z.Bench 4.76
Plot StDev(Within)
this data0.00102066
using a Z.LSL 4.90
StDev(Overall) Z.USL 4.90
control chart.0.00129055 Cpk 1.63
Overall Capability
Z.Bench 3.70
Z.LSL 3.87
Z.USL 3.88
Ppk 1.29
Cpm 1.29
© 2008 ASQ
Slide 41
Control Chart
Stat>Control Charts>Variable Charts for Subgroups
© 2008 ASQ
Slide 42
Control Charts Shows Process Drift
Xbar-R Chart of Caps Drift
1
1.002 1
1 1
2
555 22222 2222 222 22 55
55555 UCL=1.001368
22 2 22 2 2
2 6
Mean
2 2 _
2 _
Sample M
1.000 X=0.999998
2 22
8 2 2 2 22 2
6 622 6 2
55 6 22 555 2 22222 LCL=0.998629
2222 5 1
0.998 1 111
1 1
1 15 29 43 57 71 85 99 113 127 141
Sample
1
UCL=0.005020
0.0045
e Range
0.0030 _
R=0.002374
Sample
0.0015
2
0.0000 LCL=0
1 15 29 43 57 71 85 99 113 127 141
Sample
© 2008 ASQ
Slide 43
Assistant Setup
© 2008 ASQ
Slide 44
Statistical Process
Assistant Result Control chart indicates
the process is not
Capability Analysis for Caps Drift
Diagnostic Report stable. Note: stability
Xbar-R Chart
and capability are
Confirm that the process is stable.
1.002
evaluated
independently.
Mean
1.000
0.998
0.004
Range
0.002
0.000
1 15 29 43 57 71 85 99 113 127 141
Results Pass
passes
P-value 0.520
normality
test
© 2008 ASQ
Slide 45
Data = Compare
Suppliers.mtw
© 2008 ASQ
Slide 46
Assistant Dialog
© 2008 ASQ
Slide 47
Assistant Output
Before/After Capability Comparison for Supplier 1 vs Supplier 2
Summary Report
Reduction in % Out of Spec Customer Requirements
100% Lower Spec Target Upper Spec
% Out of spec was reduced by 100% from 69.36%
to 0.01%. 14.8 15 15.2
Was the process standard deviation reduced? Statistics Before After Change
0 0.05 0.1 > 0.5 Mean 14.980 15.001 0.020757
Yes No Standard deviation 0.50724 0.050492 -0.45674
Capability
P = 0.000
Pp 0.13 1.32 1.19
Did the process mean change? Ppk 0.12 1.31 1.20
Z.Bench -0.51 3.79 4.30
0 0.05 0.1 > 0.5
% Out of spec 69.36 0.01 -69.35
Yes No
P = 0.265
© 2008 ASQ
Slide 48
Assistant Output
Before/After Capability Comparison for Supplier 1 vs Supplier 2
Process Performance Report
Capability Statistics
Before After Change
Actual (overall)
After Pp 0.13 1.32 1.19
Ppk 0.12 1.31 1.20
Z.Bench -0.51 3.79 4.30
% Out of spec (obs) 69.87 0.00 -69.87
% Out of spec (exp) 69.36 0.01 -69.35
Potential (within)
Cp 0.13 1.34 1.21
Cpk 0.12 1.34 1.22
13.5 14.0 14.5 15.0 15.5 16.0 Z.Bench -0.50 3.86 4.36
% Out of spec (exp) 69.12 0.01 -69.12
© 2008 ASQ
Slide 49
Exercise 2
Assume a process owner has asked you analyze the data in Process
Capability Exercise 1.MTW. Parts A and B are made on different machines
in lots (subgroups) of 5.
The customer has established specification limits of 10 ± .1 and requires a
Ppk of 1.33.
© 2008 ASQ
Slide 50
Exercise 1 Visualizing the Answer
-.06 -.04 -.02 +.02 +.04 +.06
Process average
Design
D i width
idth
(.12 Inches)
There are 3 units of standard deviation between the process average and the
specification limits therefore this is a 3 sigma process (short term).
© 2008 ASQ
Slide 51
© 2008 ASQ
Slide 52
Exercise 1 Calculating the Shifted Yield
.94-1-1.5
ZLSL = = −4.5 Subtracting the two we find a yield of .9332
.02
02
© 2008 ASQ
Slide 53
80
nt
Percen
50
20
5
1
0.01
Probability Plot of Part B
9.8 9.9 10.0 10.1 10.2 Normal
Part A
99.99
Worksheet: Process Capability Exercise 1.MTW Mean 10.00
StDev 0.05647
N 750
99 AD 0.248
95 P-Value 0.752
80
Perrcent
5
either part. 1
0.01
9.8 9.9 10.0 10.1 10.2
Part B
Worksheet: Process Capability Exercise 1.MTW
© 2008 ASQ
Slide 54
Exercise 2 Process Capabilities
© 2008 ASQ
Slide 55
LSL
Process Data
9.9
Within
Target 10 Overall
USL 10.1
Sample Mean 10.0009 Potential (Within) Capability
Sample N 750 Cp 0.67
StDev(Within) 0.0496334 CPL 0.68
StDev(Overall) 0.0504922 CPU 0.67
Cpk 0.67
Overall Capability
Pp 0.66
PPL 0.67
PPU 0.65
Ppk 0.65
Cpm 0.66
© 2008 ASQ
Slide 56
Exercise 2 Prepare the Control Charts
© 2008 ASQ
Slide 57
_
_
Sample M
10.00 X=10.0009
9.95
LCL=9.9343
1 16 31 46 61 76 91 106 121 136
Sample
UCL=0.2441
0.2
Sample Range
4 _
R 0 1154
R=0.1154
0.1 44
4
0.0 LCL=0
1 16 31 46 61 76 91 106 121 136
Sample
© 2008 ASQ
Slide 58
Process Capability Part B
Process Capability of Part B
LSL
Process Data
9.9
Within
Target
g 10 Overall
USL 10.1
Sample Mean 9.99961 Potential (Within) Capability
Sample N 750 Z.Bench 1.71
StDev(Within) 0.0496334 Z.LSL 2.01
StDev(Overall) 0.0564727 Z.USL 2.02
Cpk 0.67
Overall Capability
Z.Bench 1.43
Z.LSL 1.76
Z.USL 1.78
Ppk 0.59
Cpm 0.59
© 2008 ASQ
Slide 59
_
_
Sample Me
10.00 X=9.9996
9.95
5 5 LCL=9.9330
1 1
9.90 1 1 1
1 16 31 46 61 76 91 106 121 136
Sample
UCL=0.2441
0.2
Range
4 _
Sample R
R=0.1154
0.1 44
4
0.0 LCL=0
1 16 31 46 61 76 91 106 121 136
Sample
© 2008 ASQ
Slide 60
Exercise 2 Using Assistant
© 2008 ASQ
Slide 61
Assistant Output
Before/After Capability Comparison for Part A vs Part B
Summary Report
Reduction in % Out of Spec
Customer Requirements
61% This showsLowernegative
Spec improvement.
Target Upper Spec
% Out of spec increased by 61% from 4.77% to
7.66%. Let’s reverse9.9 the order 10 to make10.1
interpretation easier.
Was the process standard deviation reduced? Statistics Before After Change
0 0.05 0.1 > 0.5 Mean 10.001 9.9996 -1.31E-03
Yes No Standard deviation 0.050492 0.056473 0.0059805
Capability
P = 0.999
Pp 0.66 0.59 -0.07
Did the process mean change? Ppk 0.65 0.59 -0.07
Z.Bench 1.67 1.43 -0.24
0 0.05 0.1 > 0.5
% Out of spec 4.77 7.66 2.89
Yes No
P = 0.636
© 2008 ASQ
Slide 62
Exercise 2 Assistant
© 2008 ASQ
Slide 63
Was the process standard deviation reduced? Statistics Before After Change
0 0.05 0.1 > 0.5 Mean 9.9996 10.001 0.0013091
Yes No Standard deviation 0.056473 0.050492 -5.98E-03
Capability
P = 0.001
Pp 0.59 0.66 0.07
Did the process mean change? Ppk 0.59 0.65 0.07
0 0.05 0.1 > 0.5 Z.Bench 1.43 1.67 0.24
% Out of spec 7.66 4.77 -2.89
Yes No
P = 0.636
© 2008 ASQ
Slide 64
Exercise 2 Assistant Output
Capability Comparison for Part B vs Part A
Diagnostic Report Note: “Before” and
Xbar-R Charts
Confirm that the Before and After process conditions are stable.
“After” were changed
Part B Part A
to “Part B” and “Part
10.05 A” by double clicking
A
Mean
0.1
0.0
1 16 31 46 61 76 91 106 121 136 1 16 31 46 61 76 91 106 121 136
Normality Plots
The points should be close to the line.
Part B Part A
Normality Test
(Anderson-Darling)
Part B Part A
© 2008 ASQ
Slide 65
What We Learned
© 2008 ASQ
Slide 66