Self-similarity for statistical properties in low-order representations of a large-scale turbulent round jet based on the proper orthogonal decomposition

Abstract This study is an experimental investigation into the self-similarity behavior of first and second order statistical quantities derived from a jet flow based on a) the original data and b) its low-order representations derived from the Proper Orthogonal Decomposition (POD) and c) a comparison of both. The flow under investigation is an air-helium turbulent round jet with Re ≈ 15 400 emerging from a tube into an ambient containing identical gas mass fraction and temperature as the jet at a constant pressure. Instantaneous two-dimensional velocity field measurements were obtained for downstream distances of 5.5 d to 17.4 d in the plane of the axis of the jet, via Particle Image Velocimetry. The snapshot POD algorithm was then applied to this data set to generate low-order representations with rank approximations 1, 5, 10 and 50. These then serve as the basis to derive the respective (rank truncated) statistical properties. All properties are non-dimensionalized with a self-similar framework as obtained from the original jet data. It is found that the statistical properties obtained from the low-order representations a) resemble in shape the asymptotic outline of the original jet and b) that the maximum values (for a given low-order representation) exhibit asymptotic states with increasing downstream distances. This is a strong indication that i) self-similar behavior is equally found in the low-order representations and that ii) this finding is mainly controlled by the large-scale vortices. The sole exception is the axial velocity root-mean-square values, where a distinct dip in the center line of the flow is found. This dip is successively filled up by smaller-scale turbulence for higher order truncations. Additionally, a new criterion – based on the maximum cross-correlation obtained through successive time traces of the temporal POD modes – is suggested to distinguish physically relevant modes from the POD basis in a more quantitative and explicit manner compared to traditional energy-based criteria.