SPC Statistical
Process Control
By: Amy Ee
Table of Contents
1.
Common cause & special cause
11. Defect vs Defective
2.
Troubleshoot
12. Moving Range
3.
Cpk
13. I Chart
4.
Lean vs Six Sigma
14. MR Chart
5.
Histogram
15. X Bar Chart
6.
Exercise
7.
Control Chart
16. R Chart
8.
Control Chart Data Types
9.
How do you establish Control
Limits
17. Types of Variation
18. Proportion Chart
19. Constant Proportion Chart
10. SPC Rules
20. More Cpk
11. Data Types
21. U Chart and C Chart
Common cause & special cause
Common cause is a natural part of the process.
Minor effect on the output (e.g. minor differences in raw
material, change in ambient temperature)
Special cause not normally present in the process.
Larger impact on process variation. Catastrophic result
(e.g. miscalibrated equipment, major raw material
differences, excursion caused by operator)
Special cause makes output difficult to predict
Control chart is used to point out the presence of special
cause variation
Troubleshoot
1. Special or common cause?
2. Action:
If common cause reduce it
If special cause eliminate it
CPK
[Link]
process-capability-cp-cpk-and-process-performance-pp-ppk-what-difference/
In the Six Sigma quality methodology, process
performance is reported to the organization as a sigma
level. The higher the sigma level, the better the process
is performing.
Another way to report process capability and process
performance is through the statistical measurements
ofCp,Cpk,Pp, andPpk. Click on the link to view
definitions, interpretations and calculations
forCpkandPpkthough the use of forum quotations.
Process has to be STABLE before Cpk and Ppk being
calculated
Lean vs Six Sigma
Lean low hanging fruit
Six Sigma breakthrough (6 sigma variation)
Spec limit +- tolerance (process window)
Variation - range between +- tolerance
Histogram
Summary of variation in a set of data
Normal distribution bell curve, system leverage by
normal cases
Powerhouse distribution system leverage by extreme
cases
SKEW
Standard deviation = variation from average/mean
value
Low standard deviation = robust
High standard deviation = manufacture smaller
electronics part
Exercise
Special cause/common cause example
Common cause - minor differences in raw material
Special cause - miscalibrated equipment
What is variation deviation from standard result
Normal distribution how many % will hit +- 3 sigma:
99.73%
Control chart
A type of run chart
Average/Min/Target = X-bar
Spec limits upper & lower limit
Control limits upper & lower control limits (absolute < spec
limit)
Some control limit can be more than spec limit due to large
variation (cannot be used, control limit has to be < spec limit)
Spec
Control
limit
Average
limit
Control
Spec
limit
limit
Control Chart Data Types
In Range (Within control limit)
Outliers (Out of Control)
Shift (x or more in a row either above or below average)
Trend (x or more in a row either ascending or
descending)
STABLE if all within control limits
How do you establish Control Limits
1. When to monitor/collect data (time shifts by day,
hour, shift etc.)
2. If no past data, collect minimum 25 data points based
on IPC standards (if by shifts, 25 shifts)
Collect 3 subgroups per data point and plot the average
Remove out of control limit data due to special cause
Only release to production if process is stable (all data within
control limits)
3. Control limits = 3sigma
SPC Rules
1. Nelson Test for Special Causes
Data Type
Variable
Measureable
Type of chart: X bar-R, X bar-S, I-MR, X bar-MR
Attribute
Not measureable
Used to decide e.g. Yes/No, True/False
Type of chart: p chart, np chart, u chart, c chart
Defect vs Defective
Defects are the subset of defectives. There may be n
no. of defects to have one defective product.
Moving Range
Range = diff between upper and lower limit
Moving range = data difference (variation) in between
time (xn xn-1)
Useful for equipment monitoring
I Chart
20
15.7
15
10
5.4
5 3.4
13.04
9.9
8.9
7.4
3.1
4.2
5.3 4.8
5.2
2.1
3.4
6.6
4.3 4.7
2.5
5.2 4.9
3.2
3.8
4.8
6.1 6
3.5
4.2
6
5.00
3.1
1
2.3
0
-3.04
-5
Data
Average
UCL
LCL
MR Chart
14.00
11.50
12.00
10.40
9.88
10.00
8.00
6.506.80
6.00
4.00
2.00
2.00
5.104.90
4.90
2.70
1.10
2.70
1.80
1.30
3.70
3.40
3.10
2.50
1.50
0.50
5.00
1.40
0.40
1.10
0.30
0.00
-2.00
1.30
0.80
0.10
2.103.02
1.10
0.70
41915 41917 41919 41921 41923 41925 41927 41929 41931 41933 41935 41937 41939 41941 41943
41914 41916 41918 41920 41922 41924 41926 41928 41930 41932 41934 41936 41938 41940 41942
MR
MRAverage
MRLCL
X Bar Chart
40.00
38.00
35.44
36.00
34.30
33.48
34.00
32.40 32.24
32.00
30.00
35.13
34.42
34.24
33.90
32.90
31.70
29.64
34.24
29.98
29.92
31.20
30.44
31.44
31.00
29.94
28.50
28.26
28.00
25.84
26.00
24.00
22.00
One of the sample
is only 6 thus
bringing the
average down
because data was
keyed in
wrongly etc.
20.00
Average of X
30.80
31.21
Average of Ave X
UCL
LCL
27.28
22.74
R Chart
30
The range is too
big.
In X Bar chart, the data is
within control. This is the
danger of mean, it makes the
data behave like normal.
(data in X Bar cannot be
concluded as stable, R chart
shows this fact)
25
20
24.2
16.5
14.38
15
11.7
9.4
10
9.4
9.1
8
7.9
5.9
5.7
4.3
5
4.2
1
5.2
4.8
5.3
4.1
2.1
1.2
Average of R
6.8
4.7
2.1
0
Range
6.4
RLCL
Types of Variation
Range
Moving Range (< 10 sample size per data point)
Standard Deviation (> 10 sample size per data point)
p Chart
0.00250
0.00217
0.00200
0.00150
0.00150
0.00133
0.00113
0.00100
0.00100
0.00078
0.00050
0.00067
0.00056
0.00050
0.00090
0.00100
0.00089
0.00075
0.00086
0.00064
0.00057
0.00063
0.00057
0.00120
0.00033
0.00014
0.00000
1
6
UCL
Center
10
LCL
11
12
13
14
p (Proportion)
15
16
17
18
19
20
np Chart
16
15.09
14
13
12
12
10
12
9
8
87
7.10
5
4
2
0
1
8
7
6
4
12
1
0.00
2
No. of Defective
10
11
Center
12
13
UCL
14
15
LCL
16
17
18
19
20
More Cpk and Ppk
Control chart Stability
Cpk Capability
Cp = mean/6*sigma
//old method
Cpk = (USL mean/3*sigma); (mean LSL/3*sigma)
//whichever lower
The bigger Cpk is, the more capable is the process, i.e.
one straight line at mean ONLY. When Cpk = 2.0, we
achieved 6 sigma.
Ppk for long term (3 months down the road)
U Chart & C Chart
Area of opportunity of defect
Bigger area, bigger chance
C = constant area
