Sie sind auf Seite 1von 54

Process Capability

Using control charts,


distribution analysis, rational
sampling and Ppk to better
understand manufacturing
process capability
1

Process Capability

PPAP
Ongoing
Production

Standard
Ppk > 1.67
Ppk > 1.33

What is the problem that needs to be addressed?

First we will examine the components of


variation that are measured by Cpk and Ppk

Components of Variation
90

20
Subgroup Number

Total Variation
4

Components of Variation
90

20
Subgroup Number

Between Subgroup Variation


5

Components of Variation
90

20
Subgroup Number

Within Subgroup Variation


6

Components of Variation
90
Total
Within

Between

20
Subgroup Number

Within

Between

Total
7

Ppk and Cpk

Within Subgroup
Variation

2within
Cpk measures within
Subgroup variation

Between Subgroup
Variation

2between

Total Variation

2total
Ppk measures
Total variation

Ppk and Cpk

Now for a look at the formulas

Ppk vs. Cpk Formulas


Cpk (Process Potential Capability)

(USL X )
(X LSL)
Cpk = min(
or
)
3 R / d 2
3 R / d 2
Ppk (Process Performance)

(USL X )
( X LSL)
Ppk = min(
or
)
3S
3S
What is the difference above?

10

Ppk vs. Cpk Formulas

Standard Deviation
Uses min and max of each
subgroup to estimate sigma.

Cpk : R / d 2 = R / d 2
n

Ppk : S =

i =1

For subgroup size of 5, this amounts


to only 40% of the data.

( Xi X )
n 1

Uses all of the data


to estimate sigma!
11

Cpk vs. Ppk Training


90

Ppk

80
70

Ppk standard deviation uses the


difference between each reading and
the mean

Total Variation

60
50

40
30
20
1
90

Cpk

3
S u b g r o u p Nu m b e r

Cpk standard deviation uses the


average range divided by D2.
Rave. = (R1+R2+R3+R4+R5)/5
D2 is a constant.

80
70

R3

Within Subgroup Variation

60
50

R5

R2

40
30

R1
R4

20
1

3
S u b g r o u p Nu m b e r

12

Risks of using only Cpk

Cpk ignores between subgroup variation


(It is possible to have out of specification parts and have a Cpk > 1.67)

Note: OK to use Pp, Cpk, Cp, etc. with Ppk for


investigation

13

Results Cpk vs. Ppk

Now consider the following example

14

USL = 100, LSL = 10


The subgroups have the same ranges

Ppk vs. Cpk Results


Run Chart with Less Between Variation

Cpk = 2.53
Ppk = 1.65

Cpk = 2.53
Ppk = 0.52

Dynamic Rate

80
60
40
20
0
2

10

12
14
Sample

16

18

20

22

24

Run Chart with More Between Variation


100
Dynamic Rate

In the bottom
example Cpk = 2.53
even though parts
are out of
specification,
because Cpk
measures only
within subgroup
variation.

100

80
60
40
20
0

15
2

10

12
14
Sample

16

18

20

22

24

Rational Sampling
How do I sample parts so the control
charts show process shifts?

16

Rational Subgroup
Includes only short term variation
(No shifts in the process)

Consecutive parts
Typically 3-7 consecutive parts
Typical 300 piece PPAP rational sampling plan (Measure 5,
skip 7)

Measure parts 1- 5,
Skip parts 6-12,
Measure parts 13 17,
Skip parts 18-24)

Options for 1000 piece run when measuring 125 samples


1. subgroups of 5 evenly spaced: Measure 5 parts, skip 36 parts
2. subgroups of 3 evenly spaced: Measure 3 parts, skip 21 parts..

Why would you use a subgroup of 5 vs. 3 or 3 vs. 5?


17

Rational Subgroups Example


Is this an acceptable rational sampling plan?
Incorrect Sampling
Single subgroup includes
multiple shifts in the process.

Good Rational Sampling


Subgroups are parts built
consecutively and most include
only short term variation.
Run Chart of Profile

Run Chart of Profile


0.007

0.007

0.006

0.006

0.005

0.005

Subgroup 3

Subgroup 4

Subgroup 1

0.004

0.004

0.003

0.003

Subgroup 6

Subgroup 5

0.002

0.002
1

10

20

30

40

50

60

70

80

90 100 110 120 130 140 150

10

20

Subgroup 2
30 40 50

60

70

80

90

100 110 120 130 140 150

Order of Production

Order of Production

18

What are the consequences of subgroups including parts from longer term variation?
Includes parts that span process shifts
Run Chart of Profile
0.007
0.006
0.005
0.004
0.003
0.002
1

10

20

30

40

50

60

70

80

90 100 110 120 130 140 150

Order of Production

Process Shifts are


NOT Detected!

Sample Mean

Xbar-R Chart of Subgroup with Process Shifts


0.006

UCL=0.006095

0.005

_
_
X=0.004425

0.004
0.003

LCL=0.002756
1

10

13

16
Sample

19

22

25

28

UCL=0.006120

Sample Range

0.0060
0.0045

Control Limits are


Inflated

_
R=0.002894

0.0030
0.0015

LCL=0

0.0000
1

10

13

16
Sample

19

22

25

28

19

What happens when consecutive parts are chosen for subgroups?


Run Chart of Profile
0.007

Subgroup

Subgroup

0.006
0.005

Subgroup

Subgroup

0.004
0.003

Subgroup
Subgroup

0.002
1

10

20

30

40

50

60

70

80

90

100 110 120 130 140 150

Order of Production

Xbar-R Chart of Subgroup - Consecutive Parts


Sample Mean

0.006

1 1

1
1 1 1
UCL=0.005093
_
_
X=0.004453

0.005
0.004
0.003

LCL=0.003813
1 1 1
1

1
1
4

1 1

1
1
7

1
10

0.003
Sample Range

1 1 1 1 1 1 1 1

13

16
Sample

19

1 1 1
1
22

25

28

1
UCL=0.002346

0.002
_
R=0.001109

0.001
0.000

LCL=0
1

10

13

16
Sample

19

22

25

28

Process Shifts
ARE Detected!
20

Subgroups Charts below are from same process

UCL=0.006095

0.005

_
_
X=0.004425

0.004
0.003
4

10

13

16
Sample

19

22

25

0.0045
_
R=0.002894

0.0030
0.0015
0.0000
4

10

13

16
Sample

19

22

25

LCL=0.003813
1 1 1

1
1

1 1

1
7

28

1
10

0.003

LCL=0
1

1
1 1 1
UCL=0.005093
_
_
X=0.004453

UCL=0.006120

1 1 1 1 1 1 1 1

0.004

28

0.0060

1 1

0.005

0.003

LCL=0.002756
1

Sample Range

0.006
Sample Mean

0.006

Xbar-R Chart (Subgroups with Consecutive Parts)

Sample Range

Sample Mean

Xbar-R Chart (Subgroups include Process Shifts)

13

16
Sample

19

1 1 1
1
22

25

28

1
UCL=0.002346

0.002
_
R=0.001109

0.001

LCL=0

0.000
1

10

13

16
Sample

19

22

25

28

Subgroups that include


Process Shifts

Rational Subgroups
(Consecutive Samples)

Process Shifts Detected

NO

Yes

X bar Control Limits

0.0027 - 0.006

Range Upper Control


Limit

0.006 (Inflated)

(Inflated) 0.0038 - 0.005


0.002

21

Rational Subgroup Take Away


If parts in a subgroup span process shifts then
control limits are inflated and process shifts are
NOT detected.

Use 3-7 consecutive parts to minimize risk of


including process shifts inside a given subgroup

22

Sampling Plan Initial Capability Study


What parts should I measure and what variables
do I need to keep track of?

23

Sampling Plan - Exhaust Manifold


500 manifolds are being produced for phase 0 PPAP.
MFG Equipment: One milling machine with two spindles.
One broaching bar is used to make the spline.
Time constraints and resources only allow for the
measurement of roughly 150 samples.
How should samples be collected for the capability study of the
SCs?
Increase number of subgroups (Use 3 parts instead of 5 parts)
25 subgroups of 3 from spindle 1, 25 subgroups of 3 from spindle 2
Evenly space subgroups (e.g. spindle 1 measure parts 1-3, skip 4- 10,
measure 11-13, skip 14-20. Same process for spindle 2.

Sampling Plan - Rear Differential Side Gear

Supplier can measure all SCs on 250 parts


1040 gears are being manufactured
1. Two CNC lathes.
2. Two spindles per CNC lathe
3. After the CNC operation all parts enter the process to machine
the spline. This process uses two broaching bars.
4. After broaching the parts are heat treated in one of the two furnaces
each of which hold 520 parts.
5. Each furnace is known to have a risk for variation between top and
bottom and left and right.
6. Parts cannot be etched and any mark put on parts prior to heat treat
will be burned off during the heat treatment.

Put together a sampling plan for the PPAP run.

21 subgroups of 3 for each of the 4 spindles


Keep track of which broach is used on each part and order of broaching
Divide parts between two furnaces and area of furnaces (top, bottom, left, right)

Sampling Plan Cast Aluminum Engine Block


Process: Molten aluminum forms around 14 sand cores to make an
engine block. Sand is then removed leaving a cast block with
passage ways. Block is then cubed (Sides are machined)
Seven molds make a total of the 14 sand cores
There are 6 sets of the seven molds each of which makes a set of
the 14 needed sand cores.
Sets 1-3 are run on production line 1 and sets 4-6 are run on
production line 2.
One machining cell machines all of the parts.
Part has one SC (cast locator to machined surface), and eight HICs
(as cast bore locations to cast datum)
Put together a sampling plan for the PPAP run.
Run 300 pieces for each of the 6 core sets (1800 pieces total)
Rational sampling of 25 subgroups of 5 for each 6 core set (Total of 750 measured parts).
Measure parts 1-5, skip 6-12, measure parts 13-17..

Sampling Plan Summary


Number samples and divide subgroups evenly over
production run (e.g. measure 5, skip 7)
Allocate subgroups across important variables
(e.g. machines, spindles, mold cavities, furnace area etc.)

Keep track of important variables for each measured


part so their affect can be analyzed
If possible divide subgroups so control charts can be
done separately for key variables
(e.g. 25 subgroups of 3 for spindle 1 and 25 subgroups of 3 for spindle 2.

Think of the initial PPAP runs as passive DOEs that can


be used to help quantify affect of critical variables

Good sampling strategy can lead to quicker improvements!


27

Normality, Control, & Stability

Capability numbers (Ppk, Pp, Cpk, Cp) must be


ignored pending a review of the Control, Stability
and the Distribution.

28

Control and Stability

Out of control points = Special Cause


(Do

NOT use Ppk, Pp, Cpk, Cp to predict future performance)


Xbar-R Chart of C15

Within
Overall

O v erall C apability
1.59
1.67
1.50
1.50
*

Sample M ean

3 sigma
limits

U C L=177.19

175

_
_
X=171.59

170
165
160

Not In Control

1
1 1 1

1 1

1 1

11

16

21

26
Sample

31

24

36

41

1
1

LC L=166.00

46

Not Stable

1
1

Sample Range

1 1 1 1
1 1

P otential (Within) C apability


Cp
2.68
C P L 2.82
C P U 2.54
C pk
2.54
Pp
PPL
PPU
P pk
C pm

1 1

180

18

U C L=17.51

12
_
R=7.68

6
0

LC L=0
1

11

16

21

26
Sample

31

36

41

46

29

Is this process in control?


Does the answer change
if the source of variation
is from known Non-Random
variation such as tool wear?
UCL

Bore
Diameter
LCL

Time
30

Demonstrating evidence of stability and


control with Non-Random variation
UCL

One solution is to use


Modified Control Limits

LCL

Bore
Diameter

Time
31

Demonstrating evidence of stability and


control with Non-Random variation
A second solution is
to use endpoint control
Variation from best fit line is common cause.

UCL
Control limits
centered on
nominal
LCL

Dotted line is the desired


Saw Tooth Pattern

A five piece sub-group to ensuring we


ended the tool in the right spot
First piece inspection to get in the ball park
add 4 more pieces to form a sub-group

Modified control charts are sometimes difficult to


manage. An alternative is endpoint control.
Insuring the tool begins and ends where it should.
Any variation from the expected outcome would be
analogous to a trend failure for random systems. 32

Control and Stability with Non-Random


Variation

Some processes have known causes of variation (NonRandom) that can cause a process to shift or drift (e.g. batch
to batch variation, machining tool wear)

Non-Random variation may be acceptable if there are controls


in place for the shifts and drifts

Reference the Ford STA SPC training and see your site STA or
Master Black Belt for further guidance
33

Control and Stability Take Away


1. The chances of a subgroup falling outside the 3 sigma limits is
extremely small (1 subgroup out of 370) thus any out of control
points indicate special causes.
2. For future production the special causes may come back and trigger
the process to make wild swings and out of specification parts.
3. When special causes exist Ppk does not predict the future
performance
4. Non-Random variation may be acceptable if there are controls in
place for the shifts and drifts
5. If special causes exist they must be fixed or controlled
prior to using any capability calculations to predict the
future or for approval of PPAP

Reference the Ford STA SPC training and see your


Master Black Belt for help.

34

Distribution and Normality

Ppk & Cpk equations assume


Normal Distribution (Bell Curve).
Calculations are often extremely
inaccurate if data is not normal.

Skewed NonNormal
P value < .05

Normal Bell Curve


P value > .05
4 .8

5 .6

6 .4

7 .2

8 .0

68
0.

8 .8

Probability Plot of C1

68
0.

22

6
0.

82

8
6
0.

6.962

StDev
N
AD

0.4923
125
0.316

99

0.536

90

95
90

80
70
60
50
40
30
20

80
70
60
50
40
30
20

99
95

Percent

P-Value

10

10

Mean
StDev
N
AD
P-Value

0.1

0.1

6.0

6.5

7.0
C1

6
0.

84

68
0.

99.9

Mean

5.5

Normal

99.9

5.0

83

Probability Plot of Pilot OD

Normal

Percent

16

7.5

8.0

8.5

0.6820

0.6825

0.6830
0.6835
Pilot OD

0.6840

0.6845

0.6830
0.0004121
125
3.361
<0.005

35

46

Distribution and Normality

If process experts or a review of historical data


indicates that the process output is not normally
distributed (and should not be), then the data can be
Fitted to the correct distribution (you should
be working with a Master Black Belt)
This will ensure a more accurate estimate of Ppk.

See example on next page

36

Is this process capable?


Ppk
Normal Distribution Model

1.73

Weibull Distribution Model

0.77

Process Capability of Dimension

Process Capability of Dimension


Calculations Based on Normal Distribution Model
LB

LB

USL

0
*
7
0.901049
5000
1.17469
1.17466

Overall Capability
Pp
*
PPL
*
PPU
0.77
Ppk
0.77

P otential (Within) C apability

CPU
Cpk

1.73
1.73

O v erall C apability

PPU
Ppk

USL

P rocess D ata
LB
0
Target
*
USL
7
S ample M ean 0.901049
S ample N
5000
S hape
0.769151
S cale
0.77244

P rocess D ata

LB
Target
USL
Sample Mean
Sample N
StDev(Within)
StDev(Overall)

Calculations Based on Weibull Distribution Model

1.73
1.73

-2

10

12

0.0 1.6 3.2 4.8 6.4 8.0 9.6 11.2

37

Should the data shown below be fit to the lognormal or the


normal distribution?
Lognormal - 95% CI

Normal - 95% CI

Probability Plot for Hole Diameter

Probability Plot for Hole Diameter

99.9

99.9

P value > 0.05

99

Percent

Percent

50

Lognormal
AD = 0.604
P-Value = 0.114

20
5

80

Goodness of Fit Test

50
20

Normal
AD = 0.820
P-Value = 0.033

1
0.1

95

Goodness of Fit Test

80

P value < 0.05

99

95

1
0.1
4.0

4.5

5.0
5.5
6.0
Hole Diameter

6.5

7.0

4.0

4.5

5.0
5.5
6.0
Hole Diameter

6.5

7.0

Xbar-R Chart of Hole Diameter


Sample Mean

6.0
1

1 1

1
1 1
UCL=5.588
_
_
X=5.261

5.5

5.0
1

1
3

1
5

LCL=4.934

1
7

11

13
15
Sample

17

19

21

23

25

Sample Range

1.2

UCL=1.199

0.9
_
R=0.567

0.6
0.3

LCL=0

0.0
1

11

13
15
Sample

Based on the control chart does


your answer change?

17

19

21

23

25

Looking at the control chart what


is a potential cause for the data
being non-normal?
38

What if the data is grouped by pre/post process shift?


Normal

Normal

Pre-tool Change - Probability Plot of Hole Diameter

Post Tool Change - Probability Plot of Hole Diameter

99.9

99

99

95
90

Percent

80

80
70
60
50
40
30
20

Mean
5.042
StDev
0.2607
N
85
AD
0.128
P-Value 0.984

10
5
1

Mean
5.726
StDev
0.2296
N
40
AD
0.391
P-Value 0.366

70
60
50
40
30
20
10
5
1

0.1

4.0

4.5
5.0
5.5
Hole Diameter (Pre-tool change)

6.0

5.2

5.4
5.6
5.8
6.0
6.2
Hole Diameter(Post-tool change)

Xbar-R Chart of Hole Diameter


Sample Mean

Pre-tool change
5.5
5.0
3

11

13
15
Subgroup

17

Pre-tool change

19

21

23

6.4

Analyzing the pre & post tool


change data separately shows
the process is normally
distributed.

Post-tool change

6.0

Sample Range

Percent

95
90

25

Post-tool change

1.2
0.9
0.6

Always check the control chart


to see if a process shift or out of
control point is causing the data
to be non-normal.

0.3
0.0
1

11

13
15
Subgroup

17

19

21

23

25

39

If the process is expected to have a normal distribution and robust controls


are in place for tool changes is this data acceptable for PPAP?

Normal

Xbar-R Chart of Hole Diameter


Post-tool change

99

Pre-tool change data is on control

5.0
1

11

13
15
Subgroup

Pre-tool change
data is normal

95
90

17

Post-tool change
data
is in 23control
19
21
25

80
70
60
50
40
30
20
10
5
1

95
90
80

Mean
5.042
StDev
0.2607
N
85
AD
0.128
P-Value 0.984

0.1

Sample Range

1.2
0.9

Post-tool change

Post-tool change
data is normal
Mean
5.726
StDev
0.2296
N
40
AD
0.391
P-Value 0.366

70
60
50
40
30
20
10
5
1

4.0

Pre-tool change

Percent

99

6.0
5.5

Post Tool Change - Probability Plot of Hole Diameter

99.9

Percent

Sample Mean

Pre-tool change

Normal

Pre-tool Change - Probability Plot of Hole Diameter

4.5
5.0
5.5
Hole Diameter (Pre-tool change)

6.0

5.2

5.4
5.6
5.8
6.0
6.2
Hole Diameter(Post-tool change)

6.4

Pre-tool change data is stable

0.6
0.3
0.0
1

11

13
15
Subgroup

17

Post-tool
change
19
21
23
25
data is stable

Process Capability of Hole Diameter


LSL

USL

Within
Overall

Ppk > 1.67

Potential (Within) Capability


Cp
3.08
CPL
3.09
CPU
3.06
Cpk
3.06

Yes

Overall Capability

3.0 3.6 4.2 4.8 5.4 6.0 6.6 7.2

Pp
PPL
PPU
Ppk
Cpm

1.85
1.85
1.84
1.84
*

40

Resolution and Normal Distributions


Why is this data not normally
distributed (p-value < 0.05?)

Why is this data from the same process


normally distributed? (p-value < 0.05?)

Probability Plot of Length


Normal

99
95
80
50
20

99
95
80
50
20

0.01

0.01

4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2 5.3 5.4

Length

Gauge resolution = 0.1


(Data is Stacked)

Mean
4.999
StDev
0.09985
N
5000
AD
0.227
P-Value
0.816

99.99

Percent

Mean
4.950
StDev
0.1042
N
5000
AD
188.900
P-Value <0.005

99.99

Percent

Probability Plot of Length


Normal

4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2 5.3 5.4

Length

Gauge resolution = 0.001


(Data is continuous and NOT
Stacked)
41

Distribution and Normality Take Away


1. Use process experts and historical data to determine
expected distribution. This is a physics discussion.
(Do NOT blindly fit data to a distribution)
2. If the data does not match the expected distribution determine the
special causes and fix the process.
3. If the data matches the expected distribution but is not normal
(Anderson-Darling P value < .05) fit the data to the correct
distribution and calculate Ppk. (If process is stable and in
control)
4. If data matches the expected distribution and is normal or
symmetrical calculate Ppk. (If process is stable and in control)
See following slide for list of non-normal distributions

Reference the Ford STA SPC training and see your


42
Master Black Belt for help.

Non-normal distributions
Manufacturing Process
Description

Expected Process Distribute type

Bilateral Process

Normal

Grinding process with auto correction

Uniform

Tapered roller bearing Fatigue

Weibull

Output signal Voltage

Log Normal

Flatness/Roundness

Log Normal

Plating Thickness

Family of Weibull Distributions

FTT - Non Normal - piled up near 100%

Log Normal

Sensor output transient voltage

Exponential
43

PPAP vs. Ongoing Production


When does the process change from a PPAP (Ppk
> 1.67) to ongoing Production (Ppk > 1.33) ?
With many commodities the initial Phase 2 PPAP run will not
experience all of the sources of variation (e.g.: different shifts,
lots of components, machine wear, etc.)
The ongoing production requirement of 1.33 is used after
PPAP once the process and data contains the expected
sources of variation that occur over time.
The PPAP Ppk requirement of 1.67 is higher than the ongoing
production requirement of 1.33 because it is recognized that
there will be a degradation as all of the expected sources of
variation are experienced in ongoing production.
44

Summary
Process must be in control and stable before looking at
capability index values
The data must be fit to the expected
distribution
(Do NOT blindly fit the data to a distribution.

45

Exercise 1
Process Capability of C1
USL
P rocess D ata
LS L
*
Target
*
USL
10
S ample M ean
1.64497
S ample N
10000
S tDev (O v erall) 2.11492

O v erall C apability
Pp
*
PPL
*
P P U 1.32
P pk
1.32
C pm
*

0
O bserv ed P erformance
P P M < LS L
*
P P M > U S L 9700.00
P P M Total
9700.00

12

18

24

30

36

42

E xp. O v erall P erformance


P P M < LS L
*
P P M > U S L 38.99
P P M Total
38.99

Ppk = 1.32 for ongoing production. Assuming the Engineer is


willing to modify the specification so the Ppk is greater than
1.33 is this process capable?
46

Exercise 1 Answer

No

We dont know if process is stable and in control. We also need historical


data or process expert to help determine the expected distribution.
Ppk
Original data

1.32

Lognormal Distribution
Model

0.46

Answers are drastically


different depending upon
the distribution that is used.
Process Capability of C1
Calculations Based on Lognormal Distribution Model
USL
O v erall C apability
Pp
*
PPL
*
PPU
0.46
P pk
0.46
Exp. O v erall P erformance
P P M < LS L
*
P P M > U S L 10940.34
P P M Total
10940.34

47
0

12

18

24

30

36

42

Exercise 2
Given that the Ppk = 1.87 can the PPAP be signed?

No:

There is an out of control point which is likely from a


special cause. Do NOT look at capability numbers until the process
is stable and in control.
Xbar-R Chart of Rule 1 Data
1

Sample Mean

32

28

24

_
_ L=22.87
UC
X=22.17
LC L=21.47

20
1

13

19

25

31
Sample

37

43

49

55

UC L=2.559

Sample Range

2.4
1.8

_
R=1.210

1.2
0.6
0.0

LC L=0
1

13

19

25

31
Sample

37

43

49

55

48

Exercise 3
With a Ppk = 1.66 can the PPAP be signed?
Xbar-R Chart of Pilot OD
0.68350

Out of control condition

0.68325

Sample M ean

Answer: No

U C L=0.683489

_
_
X=0.68296

0.68300
0.68275

Rule # 1

0.68250

LC L=0.682431
1

11

13
Sample

15

17

19

21

23

25

Sample Range

0.0020

U C L=0.001941

0.0015
_
R=0.000918

0.0010
0.0005
0.0000

LC L=0
1

11

13
Sample

15

17

19

21

23

25

Process Capability of Pilot OD

Probability Plot of Pilot OD


Normal

LSL
99.9
Mean
StDev
N
AD
P-Value

99
95

0.6830
0.0004121
125
3.361
<0.005

Percent

90
80
70
60
50
40
30
20

USL

P rocess Data
LS L
0.681
Target
*
USL
0.685
S ample M ean
0.68296
S ample N
125
S tDev (Within)
0.000394584
S tDev (O v erall) 0.000412947

W ithin
Overall
P otential (Within) C apability
Cp
1.69
C P L 1.66
C P U 1.72
C pk 1.66

X
X

O v erall C apability
Pp
PPL
PPU
P pk
C pm

Data is not
normal
P value < .05

10
5

6
0.

1
0.1

0.6820

0.6825

0.6830
0.6835
Pilot OD

0.6840

0.6845

O bserv ed P erformance
P P M < LS L 0.00
P P M > U S L 0.00
P P M Total
0.00

81

0
6
0.

81

68
0.

E xp. Within P erformance


P P M < LS L 0.34
P P M > U S L 0.12
P P M T otal
0.46

22

6
0.

8
82

6
0.

83

4
6
0.

84

E xp. O v erall P erformance


P P M < LS L 1.04
P P M > U S L 0.39
P P M Total
1.43

6
0.

84

1.61
1.58
1.65
1.58
*

49

Exercise 4
With a Ppk = 1.76 can this acceptable for PPAP?

No

Only 10 pieces used in study are not enough to


allow a robust assessment of normality, stability or control.
The PPAP requirement is 25 subgroups with typical subgroup size of 5
Process Capability Sixpack of 10 Parts
Capability H istogr am

Individual Value

I C har t
UCL=0.388

LSL

S pecifications
LS L -2
USL 2

0.0
_
X=-0.596

-0.8

LCL=-1.580

-1.6
1

10

-1.8 -1.2 -0.6

M oving Range C har t


Moving Range

0.0

0.6

1.2

1.8

Nor mal P r ob P lot


A D : 0.234, P : 0.723

UCL=1.209
1.0
__
MR=0.37

0.5

0.0

LCL=0
1

10

Last 1 0 O bser vations

-1.5

-1.0

-0.5

Within
S tD ev 0.328014
Cp
2.03
C pk
1.43

-0.5

-1.0

Within

O v erall
S pecs

6
Observation

0.0

C apability P lot

0.0
Values

USL

10

O v erall
S tD ev 0.26604
Pp
2.51
P pk
1.76
C pm
*

50

Exercise 5
With a process that has an expected normal distribution are these
results acceptable for PPAP?

Yes

25 subgroups with 5 parts each were used per Ford


Specific PPAP

Process Capability Sixpack of 125 Parts

Process is inXbar
control
C har t

C apability H istogr am

Sample Mean

UCL=-0.2586

LSL

USL

Specifications
LS L -2
U SL 2

-0.4
_
_
X=-0.5726

-0.6
-0.8

Process is normal
P value > .05

LCL=-0.8865
1

11

13

15

17

19

21

23

25

-1.8 -1.2 -0.6

UCL=1.151
Sample Range

0.6

1.2

1.8

Nor mal P r ob P lot


A D : 0.665, P : 0.081

R C har t
1.0
_
R=0.544

0.5

0.0

Process is stable
1

11

13

15

LCL=0
17

19

21

23

25

Last 2 5 Subgr oups

-1.5

-1.0

-0.5

Within
S tDev 0.234026
Cp
2.85
C pk
2.03

-0.5

-1.0

Within

O v erall
S pecs

10

15
Sample

0.0

Ppk > 1.67

C apability P lot

0.0
Values

0.0

20

25

O v erall
StD ev 0.238404
Pp
2.8
P pk
2
C pm *

51

Exercise 6
Is this a rational sampling plan?
Subgroup

No subgroup
includes process shifts

Run Chart of Profile


0.007
0.006
0.005
0.004
0.003
0.002
1

10

20

30

40

50

60

70

80

90 100 110 120 130 140 150

Order of Production

52

Exercise 7

Yes subgroups are


consecutive pieces and are less
likely to include process shifts

Is this a rational sampling plan?

Run Chart of Profile


0.007
Subgroup 3

Subgroup 6

0.006
0.005
Subgroup 4

Subgroup 1

0.004
0.003
Subgroup 5
Subgroup 2

0.002
1

10

20

30

40

50

60

70

80

90 100 110 120 130 140 150

Order of Production

53

Exercise 8
Subgroup

Using the subgroup sampling


plan shown on the left results
in the control chart shown
below

Run Chart of Profile


0.007
0.006
0.005
0.004
0.003
0.002
1

10

20

30

40

50

60

70

80

90 100 110 120 130 140 150

Order of Production

Sample Mean

Xbar-R Chart of Subgroup with Process Shifts


0.006

UCL=0.006095

0.005

_
_
X=0.004425

0.004
0.003

LCL=0.002756
1

10

13

16
Sample

19

22

25

28

UCL=0.006120

Sample Range

0.0060
0.0045

_
R=0.002894

0.0030
0.0015
0.0000

LCL=0
1

10

13

16
Sample

19

22

25

28

What is wrong with this


sampling plan and what is the
effect on the X bar R chart?
1. Subgroups are not consecutive parts
and include process shifts
2. Process shifts will NOT be detected
Control limits are inflated (too large)
54

Das könnte Ihnen auch gefallen