Originally Posted by Sergei
All right.
For dt, the smaller dt the more accurate the result is, but the movie becomes slower as dt decreases. Use the default value for a start.
For Re, there are two answers, one easy, one hard.
The easy one: Flow Illustrator is not a predictive tool. Whatever parameters you submit, you cannot rely on the results to be valid. If you know what the correct answer should look like, you can play with the parameters until the movie looks right. As I wrote in the info section: treat it as artist impression, the artist is you. The only hard part of this is to know what is right :-)
The hard one:
Flow Illustrator uses the bitmap as a grid. The grids corresponding to bitmaps of the size for which calculation time is reasonable are nowhere near what is needed to resolve turbulence, and anyway 2D calculations cannot do it in principle. Flow Illustrator does not use any tubulence modeling, the Reynolds number corresponds to the laminar viscosity. Qualitatively, in the calculation the diffusion effects (read viscosity) is the sum of the laminar viscosity (determined by Re parameter) and the viscous-like effects of errors caused by the grid being not fine enough. Reynolds number based on grid viscosity has an order of magnitude of the number of the bitmap points across the picture (since the finite difference scheme is of the first order). Say, one has 800x600 bitmap, then the Reynolds number based on the grid viscosity is of order 600. So, once the Re one prescribes is much greater than this value, almost all the viscosity is the grid viscosity: the movies for Re=100,000 and Re=2000,000 will most likely look the same because the flow behaviour will be determined by the grid viscosity (there is, however, an issue of numerical stability into which I would not go at this point: one can recognise instability by relatively fine wiggles and other small-scale irregularities of the picture).
Still, as turbulence is not resolved, one has to do something about it. Turbulence is usually modeled by adding turbulent viscosity to laminar viscosity. The problem is, turbulent viscosity varies from point to point, while laminar viscosity is the same everywhere. In particular, turbulent viscosity is larger in mixing layers and smaller in boundary layers. It also strongly varies across boundary layers. So, what can one do with the Flow Illustrator? Select Re in the region 100-1000 if you want to model better the mixing layers, or in the region of 1000-10,000 if you are more after boundary layers, and hope for the best: anyway this will not be accurate. Why these values? -Too long a story, and, anyway, one would have to start from the very beginning.
One nice thing is, however, that flows modeled by many people here are separated. Many features of separated flows are determined primarily by the position of the separation points, provided the rest is modeled not too wrongly. In real flows and in the Flow Illustrator alike, if Re is large enough, the flow always separates from sharp corners. Note, it may also separate from a smooth wall, but from a corner it always separates. So, if the shape of the body is such that there is no separation from smooth walls but there are corners, even Flow Illustrator might serve as a predictive tool. One only needs to know a lot of aerodynamics to say if the body is such that there will be no separation from the smooth walls ...
Enough for this post, I think.
|