Taguchi

Taguchi L18 DOE

By Ron Updated on March 23rd, 2008

Design of Experiments (DOE) are perhaps the single most powerful tool I have found to screen, characterize, and optimize a process. For an introduction to what DOE’s are please read this post.

Taguchi L18

Tonight I want to focus in a bit more on one of my favorite types of DOE – the Taguchi L18. There are many different types of Taguchi designs but I have found, through personal experience, the L18 to be the best.

You can use the L18 for many purposes. It can serve as an excellent screening DOE allowing you to narrow down a list of several inputs to a more manageable range. You can also use the L18 to characterize a process, meaning you begin to understand how factors interact with one other, etc. Finally, you can use the L18 to optimize predict future performance.

Below is what a standard L18 design looks like. Notice that we can test 2 levels for factor 1 and 3 levels for factors 2 – 7. With this many factors, 18 runs does not allow us to test every possible combination thus the hard core statistics gurus get a bit nervous. However, compared to the more traditional “fractional factorial” design the L18 is far more powerful and orthogonal (balanced).

There are many ways to analyze a L18 design. Taguchi purists will lead you to the “signal to noise” ratio. I have used this method before and it works. But I tend to analyze an L18 in a more traditional manner. I like to take at least 10 repetitions per run and then use regression to determine levels of significance. Depending on the strength of the model I might even complete “multiple response prediction.”

11th Commandment

Lastly, but definitely not least I want to leave you with some Biblical history related to Six Sigma.

When God gave Moses the Ten Commandments he slipped an additional one in. God told Moses to only speak about this commandment to us Six Sigma folks.  In case you haven’t heard yet, this 11th commandment is “thou shall confirm.”

This means no matter what type of DOE you do you must confirm the results before implementing them. If you don’t do this you are taking a big risk… I mean those commandments are serious business!

Subscribe to LSS Academy

If you enjoyed this article please consider subscribing to our full feed RSS. You can also subscribe by email and have new articles sent directly to your inbox.


Have something to say?

Leave your comment and let's talk!

Start your Lean & Six Sigma training today.