• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Systematization of data-driven optimum design incorporating a deep generative model

Research Project

Project/Area Number 20KK0329
Research Category

Fund for the Promotion of Joint International Research (Fostering Joint International Research (A))

Allocation TypeMulti-year Fund
Review Section Basic Section 18030:Design engineering-related
Research InstitutionOsaka University

Principal Investigator

Yaji Kentaro  大阪大学, 大学院工学研究科, 助教 (90779373)

Project Period (FY) 2020 – 2022
Project Status Completed (Fiscal Year 2022)
Budget Amount *help
¥14,040,000 (Direct Cost: ¥10,800,000、Indirect Cost: ¥3,240,000)
Keywordsトポロジー最適化 / 深層学習 / 深層生成モデル / データ駆動型設計 / 機械学習 / データ科学 / 設計工学
Outline of Research at the Start

トポロジー最適化は物理場の数値シミュレーションと数理最適化を駆使することで、構造物の最適な形状と形態を導き出すことができる。しかしその反面、設計自由度が高すぎるが故に生じる強い多峰性により、その適用範囲は限定的と言える。この課題を克服すべく、本研究では高い設計自由度を確保しつつも、強い非線形性を有する複雑な最適化問題を解くことを目指し、深層学習を利用した新たな最適設計の枠組みについて研究を行う。

Outline of Final Research Achievements

In the fundamental research project (Research Project: 20H02054), we are aiming to develop a new framework for optimal designs using topology optimization for flow batteries, which have gained attention as next-generation energy storage systems. Furthermore, this framework has the potential to be applied to design problems that are difficult to solve directly, not limited to flow batteries, and can be systematized as a general-purpose framework. The key lies in the introduction of deep generative models into this framework and the establishment of a solid mathematical foundation. Therefore, we conducted an international collaborative research project that spans the fields of design engineering and data science, in collaboration with researchers from the University of Texas at Austin, specializing in the field of data science. This collaboration aims to achieve a significant advancement in the fundamental research project.

Academic Significance and Societal Importance of the Research Achievements

一部の研究者によってトポロジー最適化に深層学習を組み込むことで最適構造を推定する取り組みが報告されているものの、いずれの先行研究も計算時間の短縮に主眼が置かれており、「深層学習を用いたからこそ解ける」という例は未だ数少ないのが現状である。また、深層学習は瞬時に最適構造を推定する可能性を秘めているものの、学習にある程度の時間を要するため、一概に高速化を実現できるわけではない。このような背景を踏まえ、「従来のトポロジー最適化では解くことができない問題を解く」ことを目的とし、汎用的なデータ駆動型最適設計法の構築を目指すところに本研究の学術的新規性があることを強調しておきたい。

Report

(4 results)
  • 2022 Annual Research Report   Final Research Report ( PDF )
  • 2021 Research-status Report
  • 2020 Research-status Report
  • Research Products

    (4 results)

All 2023 2022 2021 Other

All Int'l Joint Research (1 results) Presentation (2 results) (of which Int'l Joint Research: 2 results) Remarks (1 results)

  • [Int'l Joint Research] テキサス大学オースティン校(米国)2021

    • Year and Date
      2021-04-08
    • Related Report
      2022 Annual Research Report
  • [Presentation] Multifidelity Topology Design on the Probabilistic Principal Component Analysis2023

    • Author(s)
      YAJI Kentaro、BUI-THANH Tan
    • Organizer
      15th World Congress on Structural and Multidisciplinary Optimization
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Accelerating Multifidelity Topology Design Using Neural Networks2022

    • Author(s)
      Kentaro Yaji, Tan Bui
    • Organizer
      Asian Congress of Structural and Multidisciplinary Optimization 2022
    • Related Report
      2021 Research-status Report
    • Int'l Joint Research
  • [Remarks] 研究紹介ホームページ

    • URL

      https://sites.google.com/view/yajiken

    • Related Report
      2022 Annual Research Report

URL: 

Published: 2021-03-19   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi