Design of Experiments

This is the story of how, with the best of intentions, trials can go badly wrong.

My boss at Tinplate R&D had heard of the Taguchi methods for design of experiments, and since we had just acquired StatGraphics that could help with analysing the results, we decided to have a go at finding out what made the various properties of tinplate, especially the various properties of the resulting cans. I think we decided to produce 4 tinplate coils, which would combine the following parameters :

  • Heavy versus light gauge;
  • High strength versus low strength;
  • Batch annealed versus continuous annealed,

and each coil would have one half tinned with a heavy coating and the other half with a thin coating of tin.

So we went ahead and had our coils produced in Ebbw Vale, cut to sheet, and have them packed into packs with the eight different combinations of properties. We then performed all sorts of tests, starting with the mechanical properties. We then made food cans in our Soudronic welder, gave them standard beads, and carried out can performance tests.

So far so good. When we tried to use StatGraphics for the analysis, nothing much came out of it, so we decided to call in the help of the specialists in Swinden Labs near Rotherham. The result ? Again nothing much.

After some discussion with my counterpart in Swinden Labs he asked me whether he could use my data in a course he was compiling on data analysis. Saying that I had no objection, I felt rather flattered, that is, until he gave me the reason why he wanted to use my data set. “Because it’s a perfect example of how with the best of intentions, a design of experiments can go wrong”.

It turns out that for your design to be valid all parameters under consideration have to be independent of one another, which ours weren’t. Apart from the tin coating, strength, gauge and annealing method are enmeshed to such a degree that for all intents and purposes they can’t be considered to be independent.

Maybe we should have taken advice from the experts at the design stage BEFORE we carried out the experiment.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s