high velocity for natural convection
 

hi mates,
i modeled natural convection of hot cylinder.
everything looks fine, except high velocity magnitude!
solution is fully converged, 2nd order. also i have done it with CFX and results were the same.
Mesh in dependency has been checked out.
this high velocity seems wrong to me..also we have the experimental device, and we do not sense that high velocity!
i used incompressible ideal gas. in addition boussineque model gave almost same results.
cylinder wall has 383K temperature and farfield temp was set to 300K.
This is residual picture:
http://i.imgur.com/mlWfBdK.png
This is contour of velocity:
http://i.imgur.com/D2mhzlE.png
and pressure:
http://i.imgur.com/7kNXs0L.png
and at last, temperature:
http://i.imgur.com/SbNAZbS.png
Also i searched the forum and some one observed such thing but there were no answers.
Does any one have any opinion??

What are the boundary conditions and there locations? Turbulence model? Apparently, you have reached numerical convergence so it is most likely an issue of appropriate modelling, BCs and selected equation & flow models..

bottom is considered as insulated wall.
right and left walls have been set to pressure inlet (with normal to boundary direction method.)
up boundary is pressure outlet with "from neighboring cell" backflow specification.
Also in another simulation i set all sides (except bottom) to pressure outlet with same backflow specification as previous, but results were the same.
Also i set all sides as wall, but no change in velocity magnitude.:confused::confused:
Turbulence model was k-epsilon | k-omega SST.
sides are too away from cylinder (at least 15times of diameter)
pressure based, coupled.
Also "body force weighted" for pressure interpolation scheme.

Is Reynolds number sub-critical or super-critical? I think you be using a laminar viscous model and not a turbulence model. I also can't think of what type of velocity is too high if it is turbulent.

Other than that, I would check that settings are correct (scale, length, properties, etc).

well thanks for your time.
i checked Gr number. it's more than 1e9, and based on reports, it should be turbulent flow.
i also checked that with laminar approach, but no improvement in results:confused:
i double checked properties and scales. no problem!
i'm going to model it with Openfoam and StarCCM, but i don't think i have any chance!