Closed-cavity laminar flows at moderate Reynolds numbers

Abstract

The paradox reported by Brady & Acrivos (1981) of the non-existence of similarity solutions in the Reynolds-number range 10·25 < R < 147 for the flow in a tube with an accelerated surface velocity is resolved. It is shown that the source of the difficulty lies in the assumption that the tube is infinite in extent. For a finite tube, it is demonstrated that the presence of the closed end, even though far removed from the origin, affects in a fundamental way the structure of the flow throughout the entire tube. The change in the flow structure that occurs in a finite tube at R = 10·25 is caused by the fluid which is returning from the downstream end; it is shown further that the problem of determining the motion in a long finite tube is equivalent to that of selecting the initial condition for the boundary-layer equations that properly takes into account the presence of the reverse flow. By applying a method originally developed by Klemp & Acrivos (1976) for selecting this condition, the flow in a finite tube is determined numerically for Reynolds numbers up to 70. In addition, it is shown that the same change in structure brought about by the returning fluid occurs in a finite two-dimensional channel at R = 57, even though the corresponding similarity solutions exist for all values of R. The results suggest that similarity solutions should be viewed with caution because they may not represent a real flow once a critical Reynolds number is exceeded.