Robust Design Methodology - A Framework and Supportive Methods

Doktorsavhandling, 2005

Robust designs are more insensitive to the influence of variation in e.g. assembly, components and use than other comparable designs. If a design is not robust, there will be an unacceptable deviation from the desired target of product characteristics. A literature review conducted as part of this research shows that research on robust design has ma?nly been focused on developing statistical tools; less attention has been given to the basics of variation reduction and to implementation of robust design. The purpose of this thesis is to elaborate on robust design as a concept and to investigate the use of robust design and other comparable frameworks. A further purpose is to develop and explore tools that explain some aspects of design of experiments in a way that clarifies its role in robust design. The research methods applied to fulfil this purpose have been surveys, interviews, a literature review and the application of mathematical statistics. This research shows that only 17% of Swedish manufacturing companies use robust design and that 28% recognize robust design as a concept. However, 80% of the companies work actively to reduce variation between samples of the same product. Further, many companies apply methods that e.g. could be of use in robust design, but probably not for the purpose of variation reduction. Even companies working with robust design, or a comparable framework, have difficulties in applying methods that could support variation reduction. In summary, there seems to be industrial interest in, and a need for, variation reduction, but it has not been met by the research on robust design. Moreover, not even the quite novel framework of Design for Six Sigma, aimed at reducing variation when designing products, appears to have met this need. One means of facilitating industrial applications of robust design is to elaborate on its underlying principles rather than to focus solely on supportive methods. Further, a definition of robust design methodology is a way to highlight these underlying principles. Another means to facilitate applications of robust design is to develop and explore the association between specific methods and robust design. In this research two such associations between design of experiments and robust design have been investigated, one concerning restrictions in complete randomisation and one concerning an interpretation of dispersion effects.