Detection of multiple interacting features of different strength in compressible flow fields

Abstract

Compressible flows frequently exhibit multiple features of disparate length scales and orientation which interact with each other. Detection of the local flow phenomena is of significant importance to several areas of flow studies including computational grid adaptation, formulation of numerical schemes (limiters for upwinding, sensors that control local addition of artificial dissipation), interrogation of the physics of a computed or experimentally measured flow field, as well as training and evaluation of neural networks. The present work focuses on effective detection of both individual and multiple flow phenomena, as well as their interaction. Detection of the weak features of a flow field in the presence of strong ones is also an aim of the present study. An appropriate “intensity index” for each of the flow features is proposed. Detection functions (sensors), based on local flow field variation are employed. A new type of sensor, which combines elementary sensors, is proposed and applied. A quantitative approach for sensors evaluation and comparison is presented. The tool for the study is a numerical flow solver and the phenomena encountered include boundary layers, flow separation regions, vortices, jets, wakes, as well as supersonic flow phenomena (normal and oblique shock waves, compressions/expansions, contact discontinuities). Statistical analysis of the shapes of the sensor distribution curves is performed. An analytic expression for setting the threshold for automatic detection is proposed and evaluated.