機械学習による抵抗低減乱流境界層流れの大規模流れ構造の解析手法の検討

抄録

We performed stereoscopic PIV measurements for the drag-reducing turbulent boundary layer flows by injecting surfactant aqueous solution. For the case of the high drag reduction ratio, large-scale turbulence structures were often observed in the fluctuating velocity field. In this study, we discussed two kinds of machine learning methods for the quantitative evaluation of such large-scale flow structures. One is the identification by using the second invariant of the velocity gradient tensor whose differential values are calculated by machine learning. The other is the cluster analysis method by using experimented data on streamwise fluctuation velocity.