Level: Advanced
【Presenter】
Takenori Takahashi, Visiting Professor, Graduate School of Health Management, Keio University
[Announcement summary]
Superdesign is a general-purpose and analytical design based on superstructural functions. Two extensions are made to further expand this versatility and further enhance the analyticity. One is an excerpt super design that uses only the necessary parts of the function, and the other is a multi-head robust design that uses an assignment function (a function in which the value of a super factor is assigned to a superstructure function) as a column direction function. It is a complex type robust design.
The superstructure function is a function of double estimation structure using the data of the direct product experiment as a polynomial of the superfactor (specially specified design factor). By using this, highly analytical design that utilizes the extreme values (maximum value, minimum value) and inflection points of the function becomes possible. However, if the approximation of the function is not high enough, there is a risk in the design. In this case, it is possible to design with a sufficiently high degree of approximation by extracting the vicinity of the required extrema and inflection and using the superstructure functions in combination. This is an excerpt super design.
Even if the approximation of the superstructure function is high, there is a trade-off problem in optimization when there are many parameters. For example, in the case of robust design, it is not possible to make all parameters robust at the same time, and given a target (target value, target function), it is not possible to bring all parameters close to the target at the same time. Even more difficult is the issue of multiple trade-offs in the case of a design that is both robust and close to the goal. This research proposes the theory of complex superdesign and its application to solve these problems.
