The Recognition Taguchi (RT) method, which is one of the techniques used in the MT system, is a method in which sensitivity and S/N ratio are obtained from multivariate data using the standard S/N ratio, and then pattern distances are calculated usillg 2-row-by-2-column variance and covariance matrixes with the sensitivity and the S/N ratio used as items. In the RT method, as the S/N ratio and sensitivity obtained from the original data are used to calculate distance, the result is not affected by the quality of the data, and there are few constraints on the analysis. In this study, distances were calculated by this method from component pattern data of the Atractylodes lancea rhizome, which is one of the raw materials of Kampo medicine. Results obtained from the conventional calculation of the distances from two items were compared with results when distances were calculated using additional statistics.

In this study, we defined the generic function of a piston seal, developed an experimental method to evaluate the generic function, and then performed a furtctionality assessment and parameter design to optimize the piston seal. Since in conventional piston seals air leakage is inversely related to sliding resistance, both could not be reduced at the same time. Also, taking error caused by different usage conditions into consideration in the design stage was considered difficult. In this study, we defined the generic function of the piston seal as converting air pressure energy to piston work. We performed a functionality assessment using this generic function, selected a method best suited for the usage conditions from various sealing methods, and then performed a parameter design for that method to determine the detailed specifications of the seal part. Piston seals made to these specifications performed well in real machine conditions. Thus we were able to specify a piston seal which has excellent energy conversion efficiency and high robustness.

With the development of feature-packed products and the broadening of their user base, the usability of such products has become an important issue. A conventional approach to this issue employs monitors who operate the products while observers take note of any problems that occur. This method has the disadvantage,however, that the results tend to depend on the traits of the monitors;it has not yielded a reliable quantitative index of usability. In this study, function evaluation was used to evaluate usability and an efficient and accurate evaluation method was established.

Because of rising gasoline prices and the problem of global warming due to emissions of carbon dioxide, getting good mileage is an issue of increasing concern for all drivers. These circumstances prompted the application of parameter design to find driver-centered driving conditions for good mileage. With the usual methods, the study would have taken more than a year to complete,including a verification experiment, so a key consideration was how to proceed efficiently without any false starts. Preliminary experiments were carried out to get some general ideas, then provisional optimum conditions were set by separate analysis and a verification experiment was carried out in parallel with the orthogonal array experiment. As a result, there were no failures and the study period was significantly reduced. Most of the optilnal conditions obtained matched everyday driving experience, but unexpected new findings also emerged.

Reduction of calculation time is a major problem in simulation-based quality engineering. In this study the orthogonal function method, which has been proposed as a fast calculation method, was applied to the optimal design of an automobile engine mounting system, and the results were compared to the results of the conventional use of direct product orthogonal arrays. Although the sensitivity results calculated by both methods coincided, the factorial effects of the S/N ratio were different, especially for some control factors, and reproducibility of the S/N ratio gain was not good, apparently because the orthogonal function method does not take the interactions of control factors and noise factors into consideration. A new confounding function method that takes account of these interactions was developed and applied to the same design problem. The new method combines three-level fractional factorial design with the response surface method. Both the S/N ratio and the sensitivity results of the new method agreed well with those of the conventional method. The number of simulation runs was reduced to 25% of the conventional number and the calculation time was reduced to 40% of the conventional time.