Development of parameter design methodologies for dynamic characteristics

Abstract

The objective of the Taguchi parameter design is to improve the performance of a product or process by determining the levels of its design parameters such that the performance characteristic becomes robust against various causes of variation. Parameter design problems are broadly classified into static and dynamic ones depending on whether a signal parameter exists or not. Most of the existing works on parameter design focus on static problems although dynamic parameter design plays an important role in obtaining generalizable results and expediting technology development. For dynamic parameter design, Taguchi proposed a two-step procedure which consists of maximization of the signal-to-noise (SN) ratio and then adjustment of the slope of the relationship between the signal parameter and characteristic to the ideal slope using the so-called adjustment parameter. This two-step procedure separates the design parameters into the dispersion parameters which have a significant effect on the SN ratio and adjustment parameters which have a significant effect only on the slope. This two-step procedure enables certain changes in product or process requirements to be accommodated by changing only the setting of the adjustment parameter. The philosophy and objective of the Taguchi parameter design have been recognized to be very important for improving the quality of a product or a process. However, statistical methods and optimization procedures originally proposed by Taguchi have much room for improvement. First, Taguchi approach assumes that the signal parameter follows a uniform distribution, while in practice its distribution is more adequately described by other distributions than uniform. Secondly, the adjustment parameter in the second step of the Taguchi approach may not exist, and therefore, the Taguchi two-step procedure may fail to identify an optimum design condition. This thesis deals with parameter design for dynamic characteristics. New dynamic p...