二段階最適化法による混合水準過飽和計画の構成  [in Japanese] CONSTRUCTING MIXED-LEVEL SUPERSATURATED DESIGNS BY TWO-STAGE OPTIMIZATION TECHNIQUE  [in Japanese]

Access this Article

Search this Article

Author(s)

Abstract

過飽和計画は,実験回数よりも因子数が多い場合に用いられる実験計画であり,製品開発の初期段階でコンピュータシミュレーションを行う際,有効な少数因子の特定に利用されている.過飽和計画の研究は,水準配置に関する構成法とその計画を用いたデータの解析法とに分かれる.本研究では,混合水準過飽和計画の構成問題を扱う.第一ステップとして,列間の非直交度を最小化するように乱択によって列追加を行うことで初期計画を生成し,第二ステップとして,得られた初期計画の列の要素を入れ替えることで直交性の改善を行うという二段階最適化法の提案を行う.二段階最適化法の性能を評価するため,複数の論文で有用性が示されている既存の構成法との比較を行う.比較実験した全ての場合において,既存の構成法と同等かもしくはより良い計画が得られることを示し,二段階最適化法の有用性を示す.

Supersaturated designs are fractional factorial designs in which the number of columns is greater than or equal to the number of runs. They are helpful when the number of factors to be assigned in an experiment is large, for example, when screening to find a few active factors from many candidates in the primary stage of a scientific investigation and product development. There are several theoretical and computational approaches to construct supersaturated designs, such as random balance design, application of BIBD, algorithmic approach including permutation of rows and columns and so forth. This paper constructs supersaturated designs by an application of random generation and swap method as a two-stage optimization technique. The first stage intends to minimize the total dependency of all paired columns by applying the random generation. The second stage swaps some elements in columns based on the coincidence numbers of rows. The constructed designs are compared with the previous studies by numerical examples. It is demonstrated that the proposed designs have advantages in terms of low-dependency measured by χ^2 values comparing with the previous studies.

Journal

  • Bulletin of the Computational Statistics of Japan

    Bulletin of the Computational Statistics of Japan 24(1), 23-35, 2011

    Japanese Society of Computational Statistics

References:  33

Codes

  • NII Article ID (NAID)
    110008762269
  • NII NACSIS-CAT ID (NCID)
    AN10195854
  • Text Lang
    JPN
  • Article Type
    ART
  • ISSN
    0914-8930
  • NDL Article ID
    11286981
  • NDL Source Classification
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL Call No.
    Z14-1382
  • Data Source
    CJP  NDL  NII-ELS  J-STAGE 
Page Top