It is well known that today we have three options to use "DOE phylosophies":

Traditional DOE
Taguchi Methods
Shainin Techniques

And there are tremendous disscussions about which is best!
but what happen if we combine the best of them in an integrated methodology?

For example:

1. Definition of the problem (Pareto Chart, Brainstorming, etc)
2. Use the quality loss function to determine the non-quality cost.
3. Apply Multi-vari Charts to define type of variation in the problem
4. Perform a Orthogonal Array to find the most important factors
5. With the 2 or 3 most important factors found, validate the solution using DOE Traditional
6. Then we use again Quality Loss function to measure the solution or improvement.

in other words...

1. Pareto Charts, Brainstorming, Ishikawa fishbone diagram
2. Taguchi. Quality Loss Function
3. Shainin. Multi-Vari Charts
4. Taguchi. Orthogonal Arrays
5. DOE 2 or 3 Full Factorial to Validate the solution
6. Taguchi. Quality Loss Function.

In your opinion this methodology will work?

which are the advantages or disadvantages in the methodology, considering time, cost, efforts involved?

Ideas, suggestions, comments, are welcome!

thanks for your help!

The methodology certainly works well. I do it all the time. Rather than belonging to any specific school of thought, and spending valuable time defending it (and slandering others), I prefer to play for all the teams.

As you rightly pointed out, whichever technique is best to address a problem has to be used to our best advantage. After all Taguchi, Shainin and Sir Ronald Fisher (traditional DOE) have given us so much! All the knowledge that comes is welcome! Doesn't matter where it comes from.

Does anyone share this sentiment?

I share the sentiment above. Horses for courses I say.

However, isn’t step “5. DOE 2 or 3 Full Factorial to Validate the solution” encompassed within the Taguchi DoE confirmation run in step 4?

Please do correct me if I am wrong...

I believe that a Full Factorial experiment will look at all the interactions, whereas the Taguchi philosophy does not take cognizence of interactions (by confounding interactions by more factors).

Does a confirmation run avoid confounding?

I would use the following good things of Taguchi methods.

1. Concurrent statistic of S/N ratio to make a trade off evaluation between being closer to the target and having a high variance.

2. Use of orthogonal arrays for conducting initial screening experiments.

3. Parameter Design, and the use of design and noise arrays to evaluate which levels of the factors are robust to tolerance level variations. I would do this with caution, and only when necessary, since the number of runs can be very large.

4. Tolerance design using sensitivity analysis. I find this method of Taguchi very similar to the classical ANOVA.

My point of view about Geoff post:

However, isn’t step “5. DOE 2 or 3 Full Factorial to Validate the solution” encompassed within the Taguchi DoE confirmation run in step 4?

It can be!

But if you do not perform Taguchi DOE confirmation run, Traditional DOE helps you test all possible combinations of factors and levels, allowing for the systematic separation and quantification of all main effects, as well as all interaction effects.

We can use DOE full factorials according Shainin: Full factorials are ideal for quantifying interaction effects.

Many people think Taguchi orthogonal arrays are weak for finding interaction effects.

In Step 3 using Multivari charts help us to reduce a large number of unmanageable possible causes or factors of variation to a much smaller family of factors containing the main effects.

After that, using orthogonal arrays we reduce even more the unmanageable possible causes (usually L4, L8 or L12 are enough)