9501217 Nair This research is concerned with the development of efficient methods for parameter design, an approach for variation reduction introduced by Taguchi. In parameter design, the goal is to achieve robustness by reducing the sensitivity of the product or process to uncontrollable disturbances (noise variables). The research will exploit the structure inherent in parameter design experiments to obtain efficient techniques. Both static problems with fixed target as well as situations with dynamic characteristics will be studied. The applicability of parameter design will also be broadened by developing methods for more complex split-plot experiments and for studies where the noise variables cannot be controlled, as in on-line investigations. Models and methods for analyzing non-standard data, such as count or categorical responses or data from mixture models, will be developed. Finally, the case of multiple quality characteristics will be considered and a systematic data analysis strategy will be developed. Software will be developed to facilitate the application of the models and the analysis techniques provided. The importance of quality improvement methods in today's competitive market cannot be over emphasized. Industrial experiments are expensive and time consuming to conduct. Therefore, it is important that efficient methods of design and analysis are used to maximize the quality of information from such studies. The outcome of this research will enable product and process engineers to conduct parameter design studies more efficiently, especially in dynamic systems. If successful, the methods developed as part of this research will broaden the scope of engineering applications to include more complex experimental situations, on-line studies, and the analysis of different types of data. The research will also yield a comprehensive, systematic strategy for analyzing multiple quality characteristics as commonly found in practice.
