Taguchi Methods

from the Perspective of Competitive Advantage

Edwin B. Dean


[NASA Logo]
Taguchi methods were developed by Genichi Taguchi to improve the implementation of total quality control in Japan. They are based on the design of experiments to provide near optimal quality characteristics for a specific objective. They are often demeaned by academia for technical deficiencies which can be improved by using response surface methodology. Unfortunately, most of those who demean Taguchi methods have missed the whole point. Taguchi methods are not just a statistical application of design of experiments. Taguchi methods include the integration of statistical design of experiments into a powerful engineering process.

The true power of Taguchi methods comes from their simplicity of implementation. They are often applied on the Japanese manufacturing floor by the technicians to improve their product and their processes. The goal is not just to optimize an arbitrary objective function, as they are so often used in the USA. The goal is to reduce the sensitivity of engineering designs to uncontrollable factors or noise. The objective function used is the signal to noise ratio which is maximized. This moves design targets toward the middle of the design space so that external variation effects the behavior of the design as little as possible. This permits large reductions in both part and assembly tolerances which are major drivers of manufacturing cost. Linking quality characteristics to cost through the Taguchi loss function (Taguch and Yokoyama, 1994) was a major advance in quality engineering, as well as in the ability to design for cost.

Taguchi methods are claimed to have provided as much as 80% of Japanese quality gains. This is no small feat considering that Japanese quality gains have brought a large number of industries in the USA to their knees (Dertouzos, Lester, and Solow, 1989).

Taguchi methods are also called robust design in the U.S.A (Phadke, 1989).

Nair (1992) provides many perspectives of Taguchi's parameter design.

A full understanding of Taguchi methods requires reading Taguchi (1987). However, the beginner should start with Peace (1993) and follow that with either Taguchi (1986) or Phadke (1989). Another good choice is to read Fowlkes and Creveling (1995). Taguchi and Yokoyama (1994) provides a throrough coverage by the master in the style of a typical western design of experiments text. Song, Mathur, and Pattipati (1995) provides a notable extension of robust design into the realm of multicriteria optimization.





Taguchi Methods Bibliography


Surfing the Web

Taguchi Loss Function
Taguchi Methods at Linköping University


Table of Contents | Quality Technologies | Use