Takenori Takahashi, Visiting Professor, Graduate School of Keio University
Comprehensive design is a set of everything needed for design. There are two types of comprehensive design: narrow comprehensive design (consisting of three stages: identification, modeling, and optimization) and broad comprehensive design (five stages, including realization check and regression adjustment). A good approximation model is essential for scientific design. To obtain this, factor selection and structure identification (determining the presence or absence of product terms and quadratic terms) are necessary, which is the identification stage. In this study, identification is performed by adding structure identification to screening to select factors. The key to this is ingenious experiments (plans that add identification points to screening plans) and subsequent statistical processing. The modeling stage is where the coefficients of the necessary terms are estimated based on the identification results and the model is determined. The key here is to utilize custom plans and extended plans. The last stage is the optimization stage, which determines the desired level to achieve the design objective. The key here is multi-objective optimization. By the way, since the optimization solution is a hypothesis, it is important to check its realization. If the solution is difficult to accept, adjustments are made using regression. By adding these two stages, the design in practice can be made more reliable.
