Hi there:

I've been running a simple 3d straight pipe flow using Fluent. The length/diamter ratio is 5 and the diamter of the pipe is chosen as 1. The flow is steady and laminar and Re number is around 1000 based on diamter of the pipe. I generated mesh using Gambit. Two methods are used. One is that I splitted the edges of both end faces into 4 different edges so that map scheme can be used. The results look fine. Then I tried to mesh the cylinder using so-called radial-pattern mesh. Basically what I did is I found central points on each of the end faces and set both of them as trielement vertex. Hex/wedge/cooper scheme is used through the whole volume. The fluent results based on this mesh are really weird: I got almost zero axial velocity in a small central core region of the pipe. I did a plot about axial velocity profile at the outlet from (x=5,y=0,z=0.5) to (x=5,y=0,z=-0.5), where x is the axial direction. The results actually have two peaks. The residual criteria for the simulation is 10^(-5). I also checked the mesh quality of the second mesh. The equalsize skew in the small core region is about 0.8~0.9.

I really have no idea how this happened. So if anybody here has any thoughts, please let me know. Thank you in advance.

Ning

try to construct a o-grid in gambit and then rerun your simulation in fluent

Bogdan:

I tried two methods:

1. I contructed o-grid in the whole cylinder. 2. I made a small square in the center, so that the rest of the flow domain is covered with the o-grid.

I still got the same convergence problem. It seems that it is really hard to get results converged, especially in the central core region. Anyboday knows any way to walk around this? Thanks a lot.

Ning