[gnwt4a]

hello,

i am looking for a shortcut to my search for a
numerical method for dns in a straight round pipe
as papers go back decades.

low order fd methods (ie 4th order or less) are out as
they are only good for very low Res.

there is also the problem of the centreline singularity.
fourier in the homogeneous direction is standard and
so for the azimuthal.

so, what would be the best (yes, it is vague) method for
higher Re pipe dns and is suitable for gpu acceleration.

has anyone any ideas (and references)?
--

[FMDenaro]

Originally Posted by gnwt4a

hello,

i am looking for a shortcut to my search for a
numerical method for dns in a straight round pipe
as papers go back decades.

low order fd methods (ie 4th order or less) are out as
they are only good for very low Res.

there is also the problem of the centreline singularity.
fourier in the homogeneous direction is standard and
so for the azimuthal.

so, what would be the best (yes, it is vague) method for
higher Re pipe dns and is suitable for gpu acceleration.

has anyone any ideas (and references)?
--

This flow problem is largely presented in the literature (just google about) and is one of the problems considered also in the well-known AGARD report https://www.google.com/url?sa=t&rct=...FApPb9USUQbkCY


Generally a pseudo-spectral formulation is used, a radial spatial discretization is used. The problem of the singularity at r=0 is only a false problem.