Recent advances in feature extraction techniques for high-speed flowfields

Abstract

Space–time scale-resolved diagnostic and computational campaigns routinely produce high-fidelity multi-disciplinary truth-model quality datasets for complex configurations. The extraction of the primary features of engineering or scientific interest and modeling of potential low-rank dynamics has proven challenging because of the massive sizes of the databases. One approach to overcome these challenges has been through system identification based decomposition techniques. In the present work, we build on comprehensive reviews on the subject by elucidating recent advances and applications for aerodynamic flow problems. Through a succinct but panoramic treatment exemplified with relevant applications, we expect to inform the reader of method capabilities in a manner that can guide selection strategies promising critical insights into complex fluid dynamics problems. The methods are broadly classified into modal-based and physics-based. Major advances in the former are extensions of the linear framework to non-homogeneous flowfields and Floquet analysis of secondary instability, applicable to broad ranges of complexity in basic states and speed regimes. Forced response analysis has aided our understanding of non-modal instability mechanisms which extend in some ways analogous to those in the global stability literature; applications to three-dimensional flows and operator-free concepts have been particularly illustrative. Advances in modal techniques for nonlinear flowfields have sharpened focus on prescribed spectral and interaction characteristics, expanded applicability to large-scale databases through streaming approaches, and integrated multi-physics into analyzed data. Physics-based techniques, motivated by the fundamental splitting theorem of Kovasznay, have proven particularly valuable in educing mechanisms sustaining multi-modal dynamics with unique physical aspects. Helmholtz decomposition combined with signal processing procedures have provided insights into the behavior of wall-bounded and free-shear turbulence, emphasizing the effects of compressibility on energy dynamics, coherent structures, and acoustics. The generalization of physics-based eduction techniques using momentum potential theory has improved our understanding of aeroacoustics of a broad class of flowfields, and further provided direction for flow control of shear-layer noise and hypersonic boundary layer dynamics.