東北大学情報科学研究科 Graduate School of Information Sciences, Tohoku University

Japanese

東北大学 大学院情報科学研究科 Graduate School of Information Sciences, Tohoku University

Applied Information Sciences

Flow System Informatics D09

  • Prof. Yuji Hattori      
  • Assoc. Prof. Makoto Hirota      
Keywordscomputational fluid dynamics, turbulence, vortex dynamics, fluid dynamics by mathematical approach, hydrodynamic stability, magnetohydrodynamics

Computational and Physical Fluid Dynamics

We are doing fundamental research in fluid dynamics. Fluid motions are ubiquitous in many areas ranging from biological to astronomical scale and in many applications including aeronautical engineering, environmental studies and energy technologies of next generation.  Thanks to the rapid growth of computational power, numerical simulation of fluid motions has acquired a wide range of applications.  There are increasing needs for highly accurate simulation as well as novel methods of obtaining useful knowledge from huge data.  In our laboratory, we are studying fluid motions by numerical simulation and theoretical analysis.

(1) Computational fluid dynamics

  • Development of numerical methods for direct numerical simulation of flows which include complex geometries and/or moving objects
  • Numerical study of reduction of aeroacoustic noise by porous materials

(2) Statistical properties of turbulence and development of new turbulence models

  • Development of new turbulence models using machine learning
  • Enhancement of laminar region of boundary layer on a swept wing

(3) Vortex dynamics and hydrodynamic stability

  • Instability, nonlinear dynamics and transition to turbulence of vortex rings
  • Theory of hybrid instability generated by hyperbolic instability and phase shift due to waves

  • Control of boundary layer transition by wall roughness (top: no control, bottom: with control). The laminar region is expanded by suppressing turbulence transition.

  • Prediction of SGS strain tensor using machine learning. (Left) direct numerical simulation, (right) prediction by neural network. The two distributions are in good agreement.

Japanese