Previous Page

|

Next Page

Chapter Topics

Taguchi (1978) and Taguchi and Wu (1980) proposed a class of experimental designs for quality improvement, which are known as Taguchi-type designs, crossed-array designs, or product designs. Because these designs involve a crossproduct of two designs, one based on control factors and one based on noise factors, they tend to have more runs than fractional factorial and response surface designs. The control factor designs are typically small orthogonal arrays. These are not intended to (nor are they able to) investigate interactions. Refer to Dehnad (1989) and Phadke (1989) for discussions of product designs.

In ADX, these small orthogonal arrays are referred to as "mixed-level designs" because they usually involve factors with varying numbers of levels. However, if all your factors have two levels, you should use a two-level design because more tools are available.

Control factors are factors whose levels can easily be fixed in a laboratory setting and in the field. Noise factors are factors whose levels can be fixed for purposes of experimentation but not in the field. Noise factors are assumed to account for the variation in the quality characteristics of interest. The design for the control factors is called the inner array, and the design for the noise factors is called the outer array.

The Taguchi approach also involves signal-to-noise ratios, which combine deviation from the target with variation around the target into a single measure. These quality characteristics reflect the idea that not only should the expected outcome be optimized, but the variation around the optimum should be minimized as well.

Covered in This Chapter