There exist many approaches for reducing performance variation and thereby improving the quality of products and processes. Among these, Taguchi’s parameter design is most popular among engineers due to its simplicity and effectiveness. Most of the parameter design research in the literature focuses on problems with static characteristics although dynamic parameter design plays an important role in technology development.
This thesis deals with dynamic characteristic problems in which both the performance characteristic and the signal parameter are continuous. First, it is investigated what assumption is needed for Taguchi’s SN ratio to be a proper performance measure in terms of the expected loss after adjustment. It is shown that the Taguchi’s SN ratio is appropriate only for the system in which the performance characteristic is homoscedastic and non-linearity does not exist between the signal parameter and performance characteristic. A proper performance measure is developed using the least squares method for the case where the performance characteristic is homoscedastic and non-linearity exists between the signal parameter and performance characteristic. Another performance measure is developed using the weighted least square method when the variance of the performance characteristic is proportional to the square of the signal parameter and non-linearity exists. Finally, Taguchi’s SN ratio and the proposed performance measure are compared with generated data.
