Taguchi methods use patten information of each member of population such as photographs of many person or handwritten letters. This is quite different from diagnosis system or prediction of an earthquake. Because the latter has only one unit space and the S/N ratio is calculated from outside of unit. space called signals. Identification needs many unit space corresponding to each person or each letter. The method, Taguchi method for identification, uses distance within unit space to classify any other member outside the unit space. By using the new method to calculate the standerd S/N ratio and magnitude β for each member of unit space are calculated. Then final distance value of each member of unit space is calculated using Y₁ and Y₂

　　　Y₁=β

　　　Y₂=1/√η

Mean square of the distance is used to identify same similar members.

The two-step design strategy of quality engineering was used to develop a high speed and high accuracy paper feeding system for a printer, first by using the standard signal-to-noise ratio to improve the stability of the basic mechanical driving system, then tuning the control system. Specifically, the stability of the mechanical driving system was evaluated by using the current waveform obtained when a step response waveform was supplied to the DC motor as the characteristic value. Experiments and analysis were carried out with friction load torque as a noise factor and with control factors assigned to an L18 orthogonal array. The results showed an improvement in the robustness of the mechanical driving system under the optimum conditions, compared to other conditions. Then a control unit was connected and control stability was evaluated,showing that control had become more stable and variations in linefeed time and linefeed accuracy had been reduced. Using the standard signal-to-noise ratio to improve system functionality led to substantial functional improvement and a reduction in development time.

As for a centrifugal compressor design, one factor experiment was the main idea to check the influence of each shape parameter, and a systematic examination wasn't done enough. Therefbre, by letting the quality engineering apply to the simulation used for the centrifugal compressor design, and by optimizing repeatedly, conditions of optimum design were sought. When the air mass flow for input and the pressure ratio for output were chosen, the optimum shape showed a high-pressure ratio. In this case, high-pressure ratio led to high efficiency. But as an assignment from now, it showed that some measures would be needed to deal with the case of when any convergence answers were not given. Also through this parameter design the points to be paid attention to when quality engineering is applied to the simulation are mentioned.

Using quality engineering methodology, we successfully developed a concurrent ROM-RAM disk. The concurrent ROM-RAM disk is a new type of optical disk that enables synchronous writing and synchronous reading of ROM and RAM information. In order to obtain functionally sufficient read and write performance from concurrent ROM-RAM media, many control factors were extracted and optimized, related to pit shape, the substrate, the recording film, and the optical disk as a whole. To speed up the development process, parameter design was iterated using evaluations of functionality as a yardstick. Cross-talk noise from ROM pits during the reading of RAM data, which was the major error factor,was successfully suppressed in a short time.

Increasing public awareness of environmental issues has led to reconsideration of the use of large amounts of cutting lubricant for machining, and focused attention on dry or semi-dry machining. The authors' firm has been using a semi-dry drilling process to machine zinc die-cast parts for automobiles. Replacement of semi-dry drilling by dry drilling, which requires no cutting lubricant, would eliminate a cleaning step and improve the plant environment. In a previous optimization experiment the dry drilling process was evaluated through its use of electric power, but this proved problematic because of the long time needed for data analysis and the weak effect of noise factors. In a new experiment, the authors succeeded in obtaining stronger noise factor effects by modifying the drilling fixtures. As a result, it became possible to simplify the electric power evaluation and to complete optimization in only four days.

Experimental regression analysis, which can efficiently derive unknown parameters in computational models (in this case, a commercially available optical analysis program), was used in an attempt to obtain a good fit to nonlinear data relating wavelength to transmittance, obtained from measurements of optical components coated with optical thin films by evaporation. The following results were obtained: (1)Experimental regression analysis of successive-approximation simulation data using orthogonal arrays can fit factor values with a fair degree of precision in less time than would be required for a study based only on specialized empirical rules. (2)While Dr. Taguchi, the initiator of this approach, has suggested that successive-approximation simulation should be repeated until no statistically significant difference is found among the factor levels, and the goodness of fit between measured data and calculated data should be represented graphically, to proceed even more efficiently, it is worth considering also displaying the raw data and convergence status at each successive approximation.