1. 1
Who is Epsilon?
www.epsilonplus.dk
Carsten Lund & EPSILON
•Injection Moulding
•Process analysis
•Six Sigma Projects
•GMP assistance and process validation.
•Associate partner RISMA (http://rismasystems.com/)
Background: 15+ years in medical device
•Injection moulding optimizations and validation
•Project management
•Six Sigma and other applied statistics
•Optimization and documentation for work flow
•Operational management; Quality, Moulding, Metrology
3. Validation
MUST VALIDATE MUST NOT
The process to the left does not provide a guarantee that all bad (red) parts
are removed.
The process on the right can only “produce” good (green) parts – it may reject
some good parts but it will never approve bad parts.
Injection Moulding is like the process to the left
It has a lot of variables; process parameters, material variation, wear, ambient
temperature…. You name it
5. Process window
Var = √ dT2 + ds2 + d2 +…..
Just a few variables are controlable
But the output; Variation, can be “forced” to max by those few
Var = √ dT2 + ds2 + d2 +…..
6. DOE
Design Of Experiment
= Planning by applying statistical methods
7. 7
DOE
PLANNING
What is the purpose?
Screening
Details
What kind of output
Any ”specials” to consider?
Two days, two shifts, three machines,
Process knowledge
Select type and extent of the (first) experiment
8. 8
DOE OUPUT
7
6
5
4
3
2
1
A Pareto diagram tells
what paremeters are
significant, but just as
interesting:
– which one are
INsignificant
9. DOE
0,2 0,4 0,6 0,8 1,0 1,2
25,49
25,48
25,47
25,46
Packing time
Q-Measure
75,0 77,5 80,0 82,5 85,0 87,5 90,0
25,49
25,48
25,47
25,46
Mold temp.
Q-Measure
Scatterplot of Q-Measure vs Packing time
Scatterplot of Q-Measure vs Mold temp.
Linearity or not
Check the Data
Samples are just that: samples
Statistical analysis accepts a level
of chance called confidence
10. CONFIDENCE
Based on a sample we can say; this is the result we got and this is what
exactly these 15 parts are like
But if we tried it again we could get another result
LSL USL
I am reasonably (80%) sure that we
But I am very certain (99%) that itw ould get something in this interval
will not fall outside this range
11. Sample size
Signal to noise ratio.
The signal is the “answer” searched for
and the noise is what “hides” the answer
12. MSA
MSA: Measurement System Analysis
The measurement system (not only in the CMM) will add variation to the data
NOTICE it is “variation” not ”error”
Will cause doubt about quality
LSL USL
REJECT REJECT
ACCEPT
DOUBT DOUBT
13. 13
Gauge R&R
What difference do we need to detect?
Gage R&R Study for Pos.2
Small or large part-varitation?
Summary Report
0%10% 30% 100%
Yes No
60,6% of all process variation can be attributed to the
Gage R&R Study for Pos.2
measurement system. The process variation is estimated from
the parts in the study.
60,6%
Summary Report
0%10% 30% 100%
0%10% 30% 100%
Yes No
Yes No
6,5%
7,0%
Measurement system variation equals 6,5% of the tolerance.
60
Yes No
40
7,0% of all process variation can be attributed to the
measurement system. A historical standard deviation is used to
0%10% 30% 100%
30
% of Process
% of Tolerance
Study Information
Number of parts in study 5
Number of operators in study 2
Number of replicates 3
General rules used to determine the capability of the system:
<10%: acceptable
Study Information
10% - 30%: marginal
Number of parts in study 5
>30%: unacceptable
Number of operators in study 2
Number of replicates 3
Examine the bar chart showing the component contributions,
(Replicates: Number of times each operator measured each part)
and use this information to guide improvements:
-- Test-Retest component (Repeatability): The variation that
occurs when the same person measures the same item multiple
Comments
times. This accounts for 58,2% of the measurement variation. It
is 35,2% of the total variation in the process.
General rules used to determine the capability of the system:
-- Operator component (Reproducibility): The variation that
<10%: acceptable
occurs when different people measure the same item. This
10% - 30%: marginal
accounts for 81,4% of the measurement variation. It is 49,3%
>30%: unacceptable
of the total variation in the process.
Variation Breakdown
Is there a problem with repeatability or
reproducibility?
(Replicates: Number of times each operator measured each part)
Comments
Can you adequately assess process performance?
Can you sort good parts from bad?
estimate the process variation.
6,5%
Measurement system variation equals 6,5% of the tolerance.
Examine the bar chart showing the component contributions,
and use this information to guide improvements:
Can you adequately assess process performance?
Can you sort good parts from bad?
Apply historical Standard deviation:
σ = TOLERANCE / 8
14. 14
CAPABILITY
What Cpk do you need?
0.2% x 0.2% = 4 PPM
(Two parts interfacing each with Cpk = 1)
Capability PPM
1,00 2025
1,33 50
1,66 0,5
2,00 0,001
3,00 0,0000
15. 15
CAPABILITY
Samples size do not need to be big !
7
6
5
4
3
2
1
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Capability indeks
Sample size
5,00
4,50
4,00
3,50
3,00
2,50
2,00
1,50
1,00
0,50
0,00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Capability indeks
Sample size
It needs to be adequate !
16. 16
CAPABILITY
Applying Confidence
2,00
1,80
1,60
1,40
1,20
1,00
0,80
0,60
0,40
0,20
0,00
90% large variation
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
4,50
4,00
3,50
3,00
2,50
2,00
1,50
1,00
0,50
0,00
90% small variation
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
4,50
4,00
3,50
3,00
2,50
2,00
1,50
1,00
0,50
0,00
80% small
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
4,50
4,00
3,50
3,00
2,50
2,00
1,50
1,00
0,50
0,00
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
Any mistake is now on the safe side
AND
About 10-15 samples is adequate for both
70% small
17. 17
WRAP UP
CONSIDER THE RISK AND FOCUS ON THE PRODUCT