2004 年 12 巻 4 号 p. 5-13
The role of engineers is to create products/processes to perform a certain function;new products/processes never exist in nature. They can choose any system and set parameter levels of the system as they wish. These are called control factors. Such activities require cost and functionality predictions. When limited experiments are conducted in R&D laboratories it is impossible to predict exact output responses because laboratory conditions are quite different from downstream conditions for example, from large scale manufacturing condition or customer requirements. There are two strategies related to design of experiment for dealing with the reproducibility of design optimization: (1) Find some good functionality measure so that the design has enough robustness at customers hands. Signal to noise was introduced to measure the robustness. (2) Use OA, an orthogonal array, to inspect the downstream optimality by showing no significant interactions among control factor on the SNR.
This chapter tells a short history up to 1984 when the following two books were published by Japanese Standards Association. 1: Measurement study group. Parameter Design for New Product Development. 2: Taguchi, G.(ed.). Reliability Design for New Product Development.