1. I.5 Taguchi’s Philosophy
Some Important Aspects
Loss Functions
Exploiting Nonlinearities
Examples
Taguchi - Comments and Criticisms

2. Some Important Aspects
Uses DOE to Make Rugged Products and Processes
DOE Is Used As a Tool “For Reducing the Effects of Variation”
Traditional DOE Had “Focused More on Optimizing Average Product Performance Than on Considering the Effects of Variation”
Taguchi’s approach really focused on variation

3. Some Important Aspects Loss Functions
For Squared Error Loss, Loss = Variance + (Bias)2
Minimizing This Loss Involves
Reducing Variation
Targeting The Process
Loss functions are an old idea

4. Some Important Aspects Squared Error Loss
Minimizing This Loss
May Involve Conflicting Goals
You may not be able to simultaneously
Optimally target the process and reduce variation
Taguchi tries to resolve the conflict through signal to noise performance measures
Simultaneous goals are particularly difficult when optimize means “maximize”

5. Some Important Aspects Examples of Loss Functions
Just Meeting Specs versus Squared Error Loss
Specs are almost incidental for squared error loss.

6. Some Important Aspects Loss Functions Sony USA vs Sony JAPAN
Sony Japan uses squared error loss; Sony USA uses notch loss

7. Some Important Aspects Quality Management
The competitive race is never ending
Deming/Shewhart PDSA Cycle
Juran “Managerial Breakthrough”
Kaizen
Continual Improvement
Improvement Occurs When Variation Is Reduced (Mostly Effected at The Product and Process Design Stage)
Skim. Note that some slogans (which Deming abhorred) have a bad name: paradigm shift, interrupted equilibrium

8. Some Important Aspects Reduce The Effects Of Variation! How?
“By Exploiting The Nonlinear Effects of Product Parameters On The Performance Characteristics”
Use DOE
To Search For Interactions Between Control Factors and Noise Factors. If There Is An Interaction, It May Be Useful For Mitigating The Effect Of The Noise Factor
To Identify The Design Parameters That Have The Most Effect On Product Performance.
We will look at how variation propagates through a system. The first sub-bullet will be our focus.

9. Some Important Aspects Example 3: Improving a Process
Which Factors Affect
Accuracy?
Precision?
Skip (we’ve done this before)

10. Some Important Aspects Exploiting Nonlinearities
To Understand This Concept Let’s Consider an Example On Estimating Angles
Ladder Exercise in Minitab (Class Exercise 2). Show schematic (y=hsin(theta)) and review exercise.

11. Some Important Aspects Exploiting Nonlinearities - Other Examples
Electric Circuit
INA Tile
Plasticity of Caramel
Draw single resistor, then series/parallel resistors. The latter has A LOT less variability, but how reliable is it? Good simulation or hands-on exercise. INA Tile was an early Taguchi breakthrough. Draw graph, with X=Kiln heat profile (Uniform to Irregular), y=% shrinkage. Plasticity of caramel—show graphic on webpage.

12. Some Important Aspects Exploiting Nonlinearities To Fix This Idea Let’s
See How To Keep a Hubcap From Falling Off!
Webpage graphic. Stiffer clips are less forgiving. Softer clips can be made with more variation and still meet performance criteria.

13. Taguchi Comments
Developed A Comprehensive Model of Quality Engineering
Quality Engineering Philosophy Is Fundamentally Sound
Exploiting Nonlinearities To Mitigate Noise Factors Is Novel
Loss Functions
Good contributions

14. Taguchi Criticisms
There Is Room For Improvement In His Methodology By The Use Of More Sound Statistical Ideas
Better Designs May Be Available
S/N Ratio Application May Be Better Analyzed If Viewed As A Bivariate Response (S,N) Problem
S/N Can Mask Factor Effects
Ignores Sequential Experimentation
EVOP and Response Surface Techniques
Adaptive design
Don’t explain all this yet. Squared error loss explains second bullet

15. Taguchi Criticisms of Terminology
Traditional DOE Terminology and Methodology Is Modified Which Leads To Unnecessarily Complications
Linear Graphs rather than Alias Structure for Choosing Designs
Skim

16. Taguchi Criticisms of trademark
The Term “Taguchi Methodology”*
Is Objectionable
Ignores the Major Contribution of Others to This Endeavor
Skim

17. Taguchi Critique revisionism
The Term “Taguchi Methodology”*
”Taguchi himself has said that he does not like the use of that term, which to his embarrassment has been used by others, ignorant of statistical history, to include such tools as analysis of variance, fractional factorials, orthogonal arrays, and so forth.” Box et al (1988)
Skim

18. Part I References
G.E.P. Box, W.G. Hunter and J.S.Hunter (2005). Statistics for Experimenters, 2nd ed, John Wiley & Sons, N.Y.
G.E.P. Box, S. Bisgaard and C. Fung (1988). “An Explanation and Critique of Taguchi's Contributions to Quality Engineering,”University of Wisconsin Center for Quality and Productivity Improvement, Report #28.
C. Daniel (1976). Applications of Statistics to Industrial Experimentation, John Wiley & Sons, N.Y.
H. Karatsu (1988). TQC Wisdom of Japan, Productivity Press, Cambridge, MA.
R. Snee (1990). Statistical Thinking and Its Contribution to Total Quality, The American Statistician, 44,