1. A New Theme for Design of Experiments or How to Un-Geek the DOE Process
Bill Hooper
Sr. Statistical Engineer
Elkay Manufacturing
ASQ – CQE, CRE, CSSBB, CMQ/OE
May 24, 2010

2. Overview of Presentation
The status-quo of a DOE
Five steps to changing the status-quo for the DOE process
An example DOE with attribute and ordinal data

3. So what is the status quo?
Computer generated designs
Low operator involvement/commitment
Statistics! Statistics! Statistics!
4-6 months to get results
30 days for the rebound

4. So what is not the status quo?

5. #1 – Shut the computer off and design freehand

6. Fraction Factorial Design
Factors: Resolution: IV
Runs: Replicates: Fraction: 1/8
Blocks: Center pts 3
Design Generators: F = ABC, G = ABD, H = BCDE

8. Available Factor Designs and their Resolution
Factors
Runs 2 3 4 5 6 7 8 9 10
Full IV
III 16 V 32 VI 64
VII
128
full
VIII
III
Do only if necessary
Memorize!!

9. #2: Understand Statistics without talking statistics
What is an Alpha error?
What is a Beta error?

10. What the jury found What really happened guilty Not guilty He did it
He did not do it

11. #3 Eliminate the killer Alpha errors!
So what is an example of an alpha error?

12. #4 - Speed does matter (2-4 days tops)

13. Typical fast results

14. #5 – make it visual

15. The defects
Bowl Break
Bowl wrinkle

16. Ordinal/Attribute Data Run Analysis
Breaks
Bowl Wrinkle
Deck Wrinkles
Possible sweet spot

17. Follow-up Runs – making it work
Additional runs to define sweet spot

18. …and it does work

19. Un-geeking the DOE process
Shut the computer off
Understand statistics without talking statistics
Stop the Nelson Mandela errors
Speed
Make it obvious – use graphics

20. …..the simplest DOE

21. Bill Hooper Elkay Manufacturing Company