Taguchi Experimental Design

What’s the difference betweeen Taguchi and Classical DOE?

Taguchi experimental designs are based on orthogonal arrays and were made popular by the Japanese engineer Genichi Taguchi. They are usually identified with a name such as L8, to indicate an array with 8 runs. Classical experimental designs are also based on orthogonal arrays, but are identified with a superscript to indicate the number of variables – thus, a 23 classical experimental design also has 8 runs. The book “Statistics for Experimenters” by Box, Hunter and Hunter (sometimes called BH2) outlines the methodology of classical DOE, as applied to industry.

Thus, designs generated by the two methods appear to be similar – and they often are! Some industries and organizations swear by Taguchi methods, while others prefer classical methods.

Taguchi methodology emphasizes:

  • Robust design – searching for the set of conditions to achieve optimum behaviour
  • Minimization of the Loss function – minimizing economic loss due to running at non-optimum conditions
  • Maximization of signal-to-noise ratio – achieving best process targets under all uncontrolled conditions (noise)
  • Selection of experimental design from examination of Linear Graphs, which allows investigation of desired interaction effects, based on process knowledge.

Classical methodology emphasizes:

  • Sequential experimentation to model process behaviour (ie develop empirical process models, including modeling the effect of “noise” factors)
  • Prediction of future process behaviour, including optimal settings – from empirical models
  • Investigation and isolation of factors affecting mean and variation independently
  • Selection of experimental design from consideration of the trade-offs in running a fraction of a full factorial design – eg a 2 8-4 design investigates the effects of 8 factors in 16 runs, and the tradeoffs are known before running the experiment. Additional experimentation may be required to clearly identify the effects of interactions.

So, both of these methods use designed experiments to improve processes and products!

Compare Taguchi and Classical Designs, and Why learn classical DOE?

Learn how to set up, run, analyse and present designed experiments, at our intensive hands on Design of Experiments Workshop.