Inferring secondary flows from smoke or dye flow visualization: Two case studies

Abstract

Spectral element simulations of the steady, incompressible, parabolized Navier–Stokes equations are used to compare numerically simulated smoke or dye (tracer) patterns with numerically calculated spatially developing flow patterns in the following two geometries: curved channel flow and twisted square duct flow (which consists of three joined 90° square curved ducts with perpendicular planes of curvature). Secondary flows in these two geometries are caused by streamwise-oriented vortices, which have been visualized in previous experiments by viewing smoke or dye patterns in cross-sectional planes perpendicular to the streamwise direction. Simulations of tracer patterns (obtained by tracking weightless particles) show that only when there is little streamwise variation of the secondary flow do tracer patterns provide a correct qualitative indication of the secondary flow patterns. For example, tracer patterns misrepresent merging of curved channel vortices and the locations and shapes of the twisted duct vortices. These results highlight the difficulty in obtaining consistent interpretations of tracer patterns in flows with significant streamwise variation, and in obtaining a priori predictions of the validity of inferring secondary flow patterns from tracer patterns. For the two case studies examined, it is found that unless there is little streamwise variation of the secondary flow structure, inferences of secondary flow patterns from experimental tracer patterns should be made only when well validated by other methods.