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

2020 Fiscal Year Final Research Report

Construction of feature extraction method for turbulence big data by machine learning

Research Project

  • PDF
Project/Area Number 18H03758
Research Category

Grant-in-Aid for Scientific Research (A)

Allocation TypeSingle-year Grants
Section一般
Review Section Medium-sized Section 19:Fluid engineering, thermal engineering, and related fields
Research InstitutionKeio University

Principal Investigator

Fukagata Koji  慶應義塾大学, 理工学部(矢上), 教授 (80361517)

Co-Investigator(Kenkyū-buntansha) 山本 誠  東京理科大学, 工学部機械工学科, 教授 (20230584)
岩本 薫  東京農工大学, 工学(系)研究科(研究院), 教授 (50408712)
長谷川 洋介  東京大学, 生産技術研究所, 准教授 (30396783)
塚原 隆裕  東京理科大学, 理工学部機械工学科, 准教授 (60516186)
福島 直哉  東海大学, 工学部, 講師 (80585240)
守 裕也  電気通信大学, 大学院情報理工学研究科, 准教授 (80706383)
Project Period (FY) 2018-04-01 – 2021-03-31
Keywords流体力学 / 乱流 / ビッグデータ / 機械学習 / 低次元モデル
Outline of Final Research Achievements

The purpose of this study is to apply the machine learning technology to "turbulent big data" to extract the nonlinear mode, which is the essence of the self-generation maintenance mechanism of turbulence and cannot be extracted by the conventional linear theory, and to detive its time evolution equation to construct a new nonlinear feature extraction method. In this study, we use an autoencoder based on convolutional neural networks to extract the features of flow fields by compressing high-dimensional flow field information into low-dimensional latent variables, and by using a sparse regression method to derive the equations that govern the time evolution of the latent variables. While this method can extract features with sufficient accuracy for unsteady flows around a cylinder, it was suggested that further reduction in dimension is necessary for turbulent flows.

Free Research Field

流体工学

Academic Significance and Societal Importance of the Research Achievements

本研究では、完全な流れ場データの低次元化による物理的理解にとどまらず、未知の物体周りの流れの予測や不十分なデータからの予測など、流体力学の諸問題への機械学習の応用が大きな可能性を有していることを示した。本研究の成果は、理論、実験、数値シミュションに続く「第4の流体力学」である「データ駆動流体力学」の基盤整備に貢献し、支配方程式・構成方程式が確立されていない流れ場データへの応用の可能性を示唆するものである。

URL: 

Published: 2022-01-27  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi